Merge pull request #493 from jdmaia/hip_stencil_padding

Fix alignment on StencilEntry struct for HIP

BLAS_benchmark/BatchBlasBench.cc

@@ -120,7 +120,7 @@ inline void acceleratorMemSet(void *base,int value,size_t bytes) { cudaMemset(ba
 	     cudaGetErrorString( err ));				\
       printf("File %s Line %d\n",__FILE__,__LINE__);			\
       fflush(stdout);							\
-      if (acceleratorAbortOnGpuError) assert(err==cudaSuccess);		\
+      if (acceleratorAbortOnGpuError) GRID_ASSERT(err==cudaSuccess);		\
     }									\
   }
 
@@ -168,7 +168,7 @@ public:
     if ( (_Tp*)ptr == (_Tp *) NULL ) {
       printf("Grid Device Allocator got NULL for %lu bytes\n",(unsigned long) bytes );
     }
-    assert( ( (_Tp*)ptr != (_Tp *)NULL ) );
+    GRID_ASSERT( ( (_Tp*)ptr != (_Tp *)NULL ) );
     return ptr;
   }
 
@@ -276,11 +276,11 @@ public:
   {
 #ifdef GRID_HIP
     auto err = hipDeviceSynchronize();
-    assert(err==hipSuccess);
+    GRID_ASSERT(err==hipSuccess);
 #endif
 #ifdef GRID_CUDA
     auto err = cudaDeviceSynchronize();
-    assert(err==cudaSuccess);
+    GRID_ASSERT(err==cudaSuccess);
 #endif
 #ifdef GRID_SYCL
     accelerator_barrier();
@@ -305,8 +305,8 @@ public:
   {
     RealD t2=usecond();
 
-    assert(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
-    assert(OpB!=GridBLAS_OP_T);
+    GRID_ASSERT(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
+    GRID_ASSERT(OpB!=GridBLAS_OP_T);
 
     int lda = m; // m x k column major
     int ldb = k; // k x n column major
@@ -341,7 +341,7 @@ public:
 			    (hipblasDoubleComplex *) Bkn, ldb,
 			    (hipblasDoubleComplex *) &beta_p[0],
 			    (hipblasDoubleComplex *) Cmn, ldc);
-    assert(err==HIPBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==HIPBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_CUDA
     cublasOperation_t hOpA;
@@ -361,7 +361,7 @@ public:
 			   (cuDoubleComplex *) Bkn, ldb,
 			   (cuDoubleComplex *) &beta_p[0],
 			   (cuDoubleComplex *) Cmn, ldc);
-    assert(err==CUBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==CUBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_SYCL
       int64_t m64=m;
@@ -433,8 +433,8 @@ public:
   {
     RealD t2=usecond();
 
-    assert(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
-    assert(OpB!=GridBLAS_OP_T);
+    GRID_ASSERT(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
+    GRID_ASSERT(OpB!=GridBLAS_OP_T);
 
     int lda = m; // m x k column major
     int ldb = k; // k x n column major
@@ -469,7 +469,7 @@ public:
 			    (hipblasComplex *) Bkn, ldb,
 			    (hipblasComplex *) &beta_p[0],
 			    (hipblasComplex *) Cmn, ldc);
-    assert(err==HIPBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==HIPBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_CUDA
     cublasOperation_t hOpA;
@@ -489,7 +489,7 @@ public:
 			   (cuComplex *) Bkn, ldb,
 			   (cuComplex *) &beta_p[0],
 			   (cuComplex *) Cmn, ldc);
-    assert(err==CUBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==CUBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_SYCL
       int64_t m64=m;
@@ -595,11 +595,11 @@ public:
   {
     RealD t2=usecond();
     int32_t batchCount = Amk.size();
-    assert(Bkn.size()==batchCount);
-    assert(Cmn.size()==batchCount);
+    GRID_ASSERT(Bkn.size()==batchCount);
+    GRID_ASSERT(Cmn.size()==batchCount);
 
-    assert(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
-    assert(OpB!=GridBLAS_OP_T);
+    GRID_ASSERT(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
+    GRID_ASSERT(OpB!=GridBLAS_OP_T);
 
     int lda = m; // m x k column major
     int ldb = k; // k x n column major
@@ -636,7 +636,7 @@ public:
 				   (hipblasDoubleComplex **)&Cmn[0], ldc,
 				   batchCount);
     //	 std::cout << " hipblas return code " <<(int)err<<std::endl;
-    assert(err==HIPBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==HIPBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_CUDA
     cublasOperation_t hOpA;
@@ -657,7 +657,7 @@ public:
 				  (cuDoubleComplex *) &beta_p[0],
 				  (cuDoubleComplex **)&Cmn[0], ldc,
 				  batchCount);
-    assert(err==CUBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==CUBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_SYCL
       int64_t m64=m;
@@ -804,8 +804,8 @@ public:
     RealD t2=usecond();
     int32_t batchCount = Amk.size();
 
-    assert(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
-    assert(OpB!=GridBLAS_OP_T);
+    GRID_ASSERT(OpA!=GridBLAS_OP_T); // Complex case expect no transpose
+    GRID_ASSERT(OpB!=GridBLAS_OP_T);
 
     int lda = m; // m x k column major
     int ldb = k; // k x n column major
@@ -821,8 +821,8 @@ public:
     acceleratorCopyToDevice((void *)&beta ,(void *)&beta_p[0],sizeof(ComplexF));
     RealD t0=usecond();
 
-    assert(Bkn.size()==batchCount);
-    assert(Cmn.size()==batchCount);
+    GRID_ASSERT(Bkn.size()==batchCount);
+    GRID_ASSERT(Cmn.size()==batchCount);
 #ifdef GRID_HIP
     hipblasOperation_t hOpA;
     hipblasOperation_t hOpB;
@@ -843,7 +843,7 @@ public:
 				   (hipblasComplex **)&Cmn[0], ldc,
 				   batchCount);
 
-    assert(err==HIPBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==HIPBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_CUDA
     cublasOperation_t hOpA;
@@ -864,7 +864,7 @@ public:
 				  (cuComplex *) &beta_p[0],
 				  (cuComplex **)&Cmn[0], ldc,
 				  batchCount);
-    assert(err==CUBLAS_STATUS_SUCCESS);
+    GRID_ASSERT(err==CUBLAS_STATUS_SUCCESS);
 #endif
 #ifdef GRID_SYCL
       int64_t m64=m;

CLAUDE.md

@@ -0,0 +1,185 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## What This Is
+
+Grid is a data-parallel C++ library for lattice QCD. It provides SIMD-vectorised lattice containers, MPI-based domain decomposition, GPU acceleration (CUDA/HIP/SYCL), and a full suite of QCD algorithms including HMC.
+
+## Build
+
+Uses GNU Autotools. The bootstrap step only needs to run once (or after `configure.ac` changes).
+
+```bash
+./bootstrap.sh                    # downloads Eigen 3.4.0, generates configure
+mkdir build && cd build
+../configure [options]
+make -j$(nproc)
+make check                        # run root-level tests
+make install
+```
+
+Key configure options:
+
+| Option | Common values |
+|--------|---------------|
+| `--enable-simd=` | `AVX2`, `AVX512`, `KNL`, `A64FX`, `NEONv8`, `GPU` |
+| `--enable-comms=` | `mpi-auto`, `mpi3-auto`, `none` |
+| `--enable-accelerator=` | `cuda`, `hip`, `sycl` |
+| `--enable-shm=` | `shmopen`, `hugetlbfs`, `nvlink` |
+| `--enable-Nc=` | `3` (default), `2`, `4`, `5` |
+| `--with-gmp=`, `--with-mpfr=`, `--with-fftw=`, `--with-lime=` | paths to libs |
+| `--enable-hdf5`, `--enable-mkl`, `--enable-lapack` | optional features |
+
+GPU builds additionally need `--enable-gen-simd-width=64` (sets 512-bit SIMD width for GPU warp/wavefront sizing) and `--enable-unified=no --enable-shm=nvlink` for multi-GPU runs.
+
+To speed up compilation, `--disable-fermion-reps --disable-gparity` skips instantiating G-parity and higher-representation fermion operators.
+
+Platform recipes from `README.md`:
+- **KNL**: `--enable-simd=KNL --enable-comms=mpi3-auto --enable-mkl`
+- **Skylake/Haswell**: `--enable-simd=AVX512` or `AVX2` + `--enable-comms=mpi3-auto`
+- **AMD EPYC**: `--enable-simd=AVX2 --enable-comms=mpi3`
+- **A64FX (Fugaku)**: `--enable-simd=A64FX --enable-comms=mpi3 --enable-shm=shmget` (see `SVE_README.txt`)
+
+Complete, working `configure` invocations for specific HPC systems (Frontier/ROCm, Perlmutter/CUDA, Summit, SDCC-A100, etc.) live in `systems/<platform>/config-command`. These are the canonical references for production builds.
+
+Required external libs: GMP, MPFR, OpenSSL, zlib.
+
+### Use `systems/` for real machines
+
+`systems/<machine>/` holds the known-good build for each production platform (`Frontier`, `Aurora`, `Perlmutter`, `Summit`, `Tursa`, `Lumi`, `Booster`, `Crusher`, `SDCC-*`, `mac-arm`, …). Each contains a `config-command` (the exact `../../configure` invocation) and a `sourceme.sh` (module loads and env). **Prefer copying/adapting these over hand-rolling configure flags** — they encode compiler workarounds, `LDFLAGS`, and shared-memory settings that are easy to get wrong. `systems/WorkArounds.txt` records known vendor bugs.
+
+Note the GPU builds use `--enable-simd=GPU --enable-gen-simd-width=64`, so `Nsimd` is *not* 1 on device (it is `64/sizeof(scalar)`).
+
+### Regenerating `Make.inc` — required after adding or deleting source files
+
+`Make.inc` files are generated, not tracked in git (`.gitignore`d). `scripts/filelist` walks `Grid/`, `tests/*`, `benchmarks/`, `examples/`, and `HMC/` and writes the file lists and per-test `bin_PROGRAMS` rules. Every new `.cc`/`.h` in `Grid/`, and every new `Test_*.cc` / `Benchmark_*.cc` / `Example_*.cc`, is invisible to the build until you run:
+
+```bash
+./scripts/filelist    # from the source root, then re-run configure/make
+```
+
+`bootstrap.sh` runs it for you on the first setup.
+
+## Running Tests and Benchmarks
+
+```bash
+# From build directory
+make check                          # root-level tests (Test_simd, Test_cshift, etc.)
+make -C tests/<subdir> tests        # build tests in a subdirectory
+make tests                          # build all tests across all subdirectories
+./tests/core/Test_simd              # run a single test binary directly
+mpirun -n 4 ./tests/core/Test_cshift --grid 16.16.16.16 --mpi 1.1.1.4
+```
+
+`make check` is a thin smoke test — building a subdirectory with `make -C tests/<subdir> tests` and running the relevant binaries directly is the normal development loop. Test binaries take Grid's standard command-line arguments (`--grid`, `--mpi`, `--accelerator-threads`, `--threads`, `--debug-signals`, `--log`); see `Grid/util/Init.cc`.
+
+Test subdirectories and their focus: `core` (SIMD, stencil, comms), `solver` (CG, GMRES, eigensolvers), `hmc` (MD integrators), `forces` (fermion forces), `lanczos`, `IO`, `smearing`, `sp2n`, `debug`.
+
+Tests and benchmarks that need optional fermion representations are guarded by `disable_tests_without_instantiations.h` / `disable_benchmarks_without_instantiations.h`, so a `--disable-fermion-reps --disable-gparity` build silently compiles them to no-ops.
+
+## Architecture
+
+### Layer stack (bottom to top)
+
+1. **SIMD layer** (`Grid/simd/`) — platform-specific intrinsics wrapped into `vRealF`, `vComplexD`, etc. The SIMD width and layout are compile-time constants controlled by `--enable-simd`.
+
+2. **Tensor layer** (`Grid/tensors/`) — Lorentz/colour/spin tensor algebra built on top of SIMD types. `iMatrix`, `iVector`, `iScalar` templates compose into QCD types like `ColourMatrix`, `SpinColourVector`.
+
+3. **Lattice layer** (`Grid/lattice/`) — `Lattice<T>` container: a site-local tensor replicated across a distributed Cartesian grid. All arithmetic is site-parallel and expression-template-fused.
+
+4. **Cartesian/comms layer** (`Grid/cartesian/`, `Grid/communicator/`) — `GridCartesian` holds the MPI topology and local/global geometry. `Grid/cshift/` implements nearest-neighbour halo exchange; `Grid/stencil/` is the optimised multi-hop stencil used by Dirac operators.
+
+5. **Algorithm layer** (`Grid/algorithms/`) — iterative solvers (CG, GMRES, BiCGSTAB, mixed-precision), eigensolvers (Lanczos, LAPACK), FFT, smearing, and multigrid.
+
+6. **QCD layer** (`Grid/qcd/`) — gauge and fermion actions, HMC integrators, observables.
+
+### QCD subsystem (`Grid/qcd/`)
+
+- `action/fermion/` — Wilson, Clover, DWF (Mobius), Staggered, twisted-mass, G-parity variants
+- `action/gauge/` — Wilson gauge, Symanzik, Iwasaki, DBW2, plaquette+rect
+- `representations/` — Fundamental, Adjoint, Two-index, Sp(2n)
+- `hmc/` — Leapfrog, OMF2/OMF4 integrators; pseudofermion refreshment; Metropolis accept/reject
+- `smearing/` — APE, Stout, HEX, gradient flow
+- `observables/` — Polyakov loop, plaquette, topological charge
+
+### GPU acceleration and the view/memory-manager discipline
+### Multigrid (`Grid/algorithms/multigrid/`)
+
+Aggregation-based algebraic multigrid for Wilson-type fermions. Key files: `CoarsenedMatrix.h` (coarse operator), `GeneralCoarsenedMatrix.h` and `GeneralCoarsenedMatrixMultiRHS.h` (general coarsening supporting multi-RHS solves), `Aggregates.h` (near-null vector construction), `Geometry.h` (coarse-grid geometry). `MultiGrid.h` is the top-level include.
+
+### GPU acceleration
+
+GPU support is injected via macros in `Grid/threads/Accelerator.h` — `accelerator_for(i, n, nsimd, {...})`, `accelerator_forNB` (non-blocking, must be followed by `accelerator_barrier()`), `accelerator_for2dNB`, and `accelerator_inline`. On a CPU build these degrade to `thread_for` (OpenMP). Unified virtual memory is on by default (`--enable-unified=yes`); device-aware MPI (`--enable-accelerator-aware-mpi`) avoids device→host copies on transfers.
+
+Lattice data is **not** directly addressable inside a kernel. You must open a view with the correct access mode so `Grid/allocator/MemoryManager.h` can move/mark the data:
+
+```cpp
+autoView(out_v, out, AcceleratorWriteDiscard);   // RAII; closes at end of scope
+autoView(in_v,  in,  AcceleratorRead);
+accelerator_for(ss, grid->oSites(), Nsimd, {
+    coalescedWrite(out_v[ss], coalescedRead(in_v[ss]));
+});
+```
+
+Modes are `AcceleratorRead/Write/WriteDiscard` and `CpuRead/Write/WriteDiscard`. Getting the mode wrong (e.g. `AcceleratorRead` on a field you write) produces stale-data bugs that only appear on GPU builds. Inside kernels use `coalescedRead`/`coalescedWrite` rather than raw `operator[]` — they map the SIMD lane onto `threadIdx.x` so accesses stay coalesced.
+
+### Repo-local debugging skills (`skills/`)
+
+`skills/` contains hard-won, Grid-specific playbooks written as invocable skill files. Consult them before debugging in these areas rather than reasoning from first principles:
+
+| File | Covers |
+|---|---|
+| `gpu-memory-performance.md` | `acceleratorThreads()`, LambdaApply thread mapping, `coalescedRead` idiom, fused vs staged HBM access |
+| `gpu-runtime-correctness.md` | GPU runtime returning early from sync, silent wrong answers |
+| `communication-overlap.md` | 7-phase halo pipeline, per-packet events, host-staging vs GPU-direct RDMA |
+| `mpi-heterogeneous.md` | `MPI_Sendrecv` device-buffer aliasing, deterministic reductions |
+| `compiler-validation.md` | Isolating GPU compiler codegen bugs, minimal reproducers |
+| `correctness-verification.md` | Double-run fingerprinting, per-packet checksums, flight recorder |
+| `hang-diagnosis.md` | Diagnosing MPI/accelerator hangs |
+
+The key loop macros (defined in `Grid/threads/Accelerator.h`) are:
+- `accelerator_for(iter, num, nsimd, {...})` — maps to CUDA/HIP kernel or OpenMP loop; `nsimd` is the innermost SIMD lane count
+- `accelerator_forNB(...)` — non-blocking variant (no implicit barrier)
+- `accelerator_for2dNB(iter1, num1, iter2, num2, nsimd, {...})` — 2D kernel launch
+- `thread_for(iter, num, {...})` — CPU OpenMP loop (never dispatches to GPU)
+
+On CPU builds, `accelerator_for` aliases to `thread_for`.
+
+### Solver patterns
+
+`SchurRedBlack` (`Grid/algorithms/iterative/SchurRedBlack.h`) implements red-black (even/odd) preconditioning for fermion operators. Most production fermion solves use `SchurRedBlackDiagMooeeSolve` or similar wrappers that internally call a `ConjugateGradient` on the Schur complement.
+
+Mixed-precision solvers (`ConjugateGradientMixedPrec`, `BiCGSTABMixedPrec`) drive a double-precision outer loop with single-precision inner solves.
+
+### Memory and I/O
+
+- `Grid/allocator/` — aligned/NUMA-aware allocators; caching allocator via `--enable-alloc-cache`
+- `Grid/parallelIO/` — distributed parallel reader/writer for ILDG (via LIME), SciDAC, and native binary formats
+- `Grid/serialisation/` — text, binary, HDF5, XML/JSON serialisation of arbitrary Grid objects
+
+### Executables
+
+- `HMC/` — production HMC driver programmes (e.g. `Mobius2p1f.cc`, `DWF_plus_DSDR_nf2plus1_Shamir_Gparity.cc`)
+- `benchmarks/` — `Benchmark_dwf`, `Benchmark_ITT`, `Benchmark_comms`, `Benchmark_memory_bandwidth`, … used to qualify a new machine
+- `examples/` — small, readable programmes (`Example_plaquette.cc`, `Example_Mobius_spectrum.cc`) that are the best starting point for learning the API
+
+Each of these directories auto-builds every top-level `.cc` as its own binary via `scripts/filelist`.
+
+Every programme is wrapped in `Grid_init(&argc, &argv)` / `Grid_finalize()` (`Grid/util/Init.h`).
+
+## Key Conventions
+
+- **C++17** is required throughout.
+- Template structure: most classes are templated on `<_FImpl>` (fermion impl) or `<Gimpl>` (gauge impl), which encode the representation and precision. Instantiation is controlled by `--enable-fermion-instantiations`.
+- **Tensor indices are positional, not labelled.** The `Grid/tensors/` arithmetic recurses structurally over the `iScalar`/`iVector`/`iMatrix` nest: each level defines only the {scalar,vector,matrix}² products at its own level, with element types resolved by automatic type deduction, so every colour/spin/lorentz combination composes from ~200 lines (versus the pre-C++11 QDP++/PETE approach of machine-generating every case). An index's meaning derives entirely from its nesting depth counted from the outside; `iScalar` is the identity/broadcast case at every level. Never insert or remove a nesting level casually — the multiplication tables contract by position.
+- **Multigrid coarsening deepens the tensor nest by one level.** A coarse site vector is `iVector<CComplex,nbasis>`, and `innerProduct` on it returns `iScalar<CComplex>` — one level deeper than the fine block scalar. So the block-inner-product scalar type gains one `iScalar` wrapper per MG level (fine: `vTComplex`; level 2: `iScalar<vTComplex>`; see `examples/Example_pvdagm_3level.cc`). When calling `blockInnerProduct`/`blockZAXPY`/`blockOrthogonalise` on coarse fields, the coarse scalar type must match `decltype(innerProduct(siteVector(),siteVector()))` exactly; a wrong depth fails to compile (no viable `operator=` deep in the instantiation chain) rather than mis-contracting.
+- **Grids are borrowed, never owned.** `conformable` is pointer identity, so every object that interoperates must hold the *same* `GridCartesian *`; a class that minted its own grid internally could never conform with anything else. Ownership is therefore not available, and lifetime is managed by scope discipline instead of reference counting: whoever creates a grid retains it beyond every object it handed a reference to. Anything *derived* from a grid inherits this — `~PaddedCell` dereferences its `unpadded_grid`, so a `PaddedCell` cannot even be **destroyed** after its parent grid, only used. Where a consumer must let go early, it offers an explicit hand-back (`MultiGeneralCoarsenedOperatorV2::ReleaseGrid()`) to be called *before* the grid is destroyed.
+
+- The `RealD`/`RealF`/`ComplexD`/`ComplexF` typedefs are used everywhere; avoid raw `double`/`float`.
+- Use `GRID_ASSERT(cond)` (defined in `Grid/GridStd.h`), not bare `assert` — it prints a Grid-formatted message and aborts cleanly under MPI.
+- Logging is stream-based, not macro-based: `std::cout << GridLogMessage << ... << std::endl;`. Channels declared in `Grid/log/Log.h` include `GridLogError`, `GridLogWarning`, `GridLogDebug`, `GridLogPerformance`, `GridLogIterative`, `GridLogSolver`, `GridLogHMC`, `GridLogComms`, `GridLogMemory`, `GridLogDslash`, `GridLogIRL`, `GridLogMG`. A subset is switched on at runtime with e.g. `--log Error,Warning,Message,Performance,Iterative,Integrator,Debug,Colours` (names given without the `GridLog` prefix).
+- Performance-critical paths use `GRID_TRACE(name)` from `Grid/perfmon/Tracing.h` (compiled out unless `--enable-tracing` selects a backend) and the `GridStopWatch` timers in `Grid/perfmon/Timer.h`.
+- Reductions across MPI ranks go through `GridBase::GlobalSum` / `GlobalMax`; never reduce with bare MPI calls inside library code.
+- Everything lives in `NAMESPACE_BEGIN(Grid)` / `NAMESPACE_END(Grid)` macros; follow the surrounding file rather than writing `namespace Grid { }`.
+- British spelling is used in identifiers and comments (`colour`, `neighbour`, `serialisation`).

Grid/DisableWarnings.h

@@ -51,11 +51,13 @@ directory
 #pragma nv_diag_suppress cast_to_qualified_type
  //disables nvcc specific warning in many files
 #pragma nv_diag_suppress esa_on_defaulted_function_ignored
+#pragma nv_diag_suppress declared_but_not_referenced
 #pragma nv_diag_suppress extra_semicolon
 #else
  //disables nvcc specific warning in json.hpp
 #pragma diag_suppress unsigned_compare_with_zero
 #pragma diag_suppress cast_to_qualified_type
+#pragma diag_suppress declared_but_not_referenced
  //disables nvcc specific warning in many files
 #pragma diag_suppress esa_on_defaulted_function_ignored
 #pragma diag_suppress extra_semicolon

Grid/GridCore.h

@@ -54,7 +54,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/serialisation/Serialisation.h>
 #include <Grid/util/Sha.h>
 #include <Grid/communicator/Communicator.h> 
-#include <Grid/cartesian/Cartesian.h>    
+#include <Grid/communicator/RingAllReduce.h>
+#include <Grid/cartesian/Cartesian.h>
 #include <Grid/tensors/Tensors.h>      
 #include <Grid/lattice/Lattice.h>      
 #include <Grid/cshift/Cshift.h>       

Grid/GridStd.h

@@ -1,9 +1,17 @@
 #ifndef GRID_STD_H
 #define GRID_STD_H
 
+///////////////////
+// Grid config
+///////////////////
+#include "Config.h"
+
 ///////////////////
 // Std C++ dependencies
 ///////////////////
+#define _NBACKTRACE (256)
+extern void * Grid_backtrace_buffer[_NBACKTRACE];
+
 #include <cassert>
 #include <complex>
 #include <memory>
@@ -15,7 +23,9 @@
 #include <random>
 #include <functional>
 #include <stdio.h>
+#include <string.h>
 #include <stdlib.h>
+#include <unistd.h>
 #include <strings.h>
 #include <stdio.h>
 #include <signal.h>
@@ -23,11 +33,36 @@
 #include <sys/time.h>
 #include <chrono>
 #include <zlib.h>
+#ifdef HAVE_EXECINFO_H
+#include <execinfo.h>
+#endif
+
+void GridAbort(void);
+
+#define ASSLOG(A) ::write(STDERR_FILENO,A,::strlen(A));
+#ifdef HAVE_EXECINFO_H
+#define GRID_ASSERT(b) if(!(b)) {					\
+    fflush(stdout); \
+    ASSLOG(" GRID_ASSERT failure: ");					\
+    ASSLOG(__FILE__);							\
+    ASSLOG(" : ");							\
+    ASSLOG(#b);								\
+    ASSLOG(" : ");							\
+    int symbols = backtrace(Grid_backtrace_buffer,_NBACKTRACE);		\
+    backtrace_symbols_fd(Grid_backtrace_buffer,symbols,STDERR_FILENO);	\
+    GridAbort();							\
+  };
+#else
+#define GRID_ASSERT(b) if(!(b)) {					\
+    ASSLOG(" GRID_ASSERT failure: ");					\
+    ASSLOG(__FILE__);							\
+    ASSLOG(" : ");							\
+    ASSLOG(#b);								\
+    ASSLOG(" : ");							\
+    GridAbort();							\
+  };
+#endif
 
-///////////////////
-// Grid config
-///////////////////
-#include "Config.h"
 
 #ifdef TOFU
 #undef GRID_COMMS_THREADS

Grid/Makefile.am

@@ -54,21 +54,25 @@ Version.h: version-cache
 include Make.inc
 include Eigen.inc
 
-extra_sources+=$(WILS_FERMION_FILES)
-extra_sources+=$(STAG_FERMION_FILES)
+if BUILD_FERMION_INSTANTIATIONS
+  extra_sources+=$(WILS_FERMION_FILES)
+  extra_sources+=$(STAG_FERMION_FILES)
 if BUILD_ZMOBIUS
-  extra_sources+=$(ZWILS_FERMION_FILES)
+    extra_sources+=$(ZWILS_FERMION_FILES)
 endif
 if BUILD_GPARITY
-  extra_sources+=$(GP_FERMION_FILES)
+    extra_sources+=$(GP_FERMION_FILES)
 endif
 if BUILD_FERMION_REPS
-  extra_sources+=$(ADJ_FERMION_FILES)
-  extra_sources+=$(TWOIND_FERMION_FILES)
+    extra_sources+=$(ADJ_FERMION_FILES)
+    extra_sources+=$(TWOIND_FERMION_FILES)
 endif
 if BUILD_SP
     extra_sources+=$(SP_FERMION_FILES)
-    extra_sources+=$(SP_TWOIND_FERMION_FILES)
+if BUILD_FERMION_REPS
+      extra_sources+=$(SP_TWOIND_FERMION_FILES)
+endif
+endif
 endif
 
 lib_LIBRARIES = libGrid.a

Grid/Namespace.h

@@ -29,8 +29,8 @@ directory
 #pragma once
 
 #include <type_traits>
-#include <cassert>
 #include <exception>
+#include <cassert>
 
 #define NAMESPACE_BEGIN(A) namespace A {
 #define NAMESPACE_END(A)   }

Grid/algorithms/Algorithms.h

@@ -51,6 +51,8 @@ NAMESPACE_CHECK(approx);
 #include <Grid/algorithms/deflation/MultiRHSBlockProject.h>
 #include <Grid/algorithms/deflation/MultiRHSDeflation.h>
 #include <Grid/algorithms/deflation/MultiRHSBlockCGLinalg.h>
+// Not really deflation, but useful
+#include <Grid/algorithms/blas/MomentumProject.h>
 NAMESPACE_CHECK(deflation);
 #include <Grid/algorithms/iterative/ConjugateGradient.h>
 NAMESPACE_CHECK(ConjGrad);

Grid/algorithms/FFT.h

@@ -1,6 +1,6 @@
 /*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+    Grid physics library, www.github.com/paboyle/Grid
 
     Source file: ./lib/Cshift.h
 
@@ -28,6 +28,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifndef _GRID_FFT_H_
 #define _GRID_FFT_H_
 
+#ifdef GRID_CUDA
+#include <cufft.h>
+#endif
+
+#ifdef GRID_HIP
+#include <hipfft/hipfft.h>
+#endif
+
+#if !defined(GRID_CUDA) && !defined(GRID_HIP)
 #ifdef HAVE_FFTW
 #if defined(USE_MKL) || defined(GRID_SYCL)
 #include <fftw/fftw3.h>
@@ -35,266 +44,664 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <fftw3.h>
 #endif
 #endif
+#endif
 
 NAMESPACE_BEGIN(Grid);
 
-template<class scalar> struct FFTW { };
-
-#ifdef HAVE_FFTW	
-template<> struct FFTW<ComplexD> {
-public:
-
-  typedef fftw_complex FFTW_scalar;
-  typedef fftw_plan    FFTW_plan;
-
-  static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
-				      FFTW_scalar *in, const int *inembed,		
-				      int istride, int idist,		
-				      FFTW_scalar *out, const int *onembed,		
-				      int ostride, int odist,		
-				      int sign, unsigned flags) {
-    return ::fftw_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
-  }	  
-    
-  static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
-    ::fftw_flops(p,add,mul,fmas);
-  }
-
-  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
-    ::fftw_execute_dft(p,in,out);
-  }
-  inline static void fftw_destroy_plan(const FFTW_plan p) {
-    ::fftw_destroy_plan(p);
-  }
-};
-
-template<> struct FFTW<ComplexF> {
-public:
-
-  typedef fftwf_complex FFTW_scalar;
-  typedef fftwf_plan    FFTW_plan;
-
-  static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
-				      FFTW_scalar *in, const int *inembed,		
-				      int istride, int idist,		
-				      FFTW_scalar *out, const int *onembed,		
-				      int ostride, int odist,		
-				      int sign, unsigned flags) {
-    return ::fftwf_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
-  }	  
-    
-  static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
-    ::fftwf_flops(p,add,mul,fmas);
-  }
-
-  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
-    ::fftwf_execute_dft(p,in,out);
-  }
-  inline static void fftw_destroy_plan(const FFTW_plan p) {
-    ::fftwf_destroy_plan(p);
-  }
-};
-
-#endif
-
 #ifndef FFTW_FORWARD
 #define FFTW_FORWARD (-1)
 #define FFTW_BACKWARD (+1)
+#define FFTW_ESTIMATE (0)
 #endif
 
-class FFT {
-private:
-    
-  GridCartesian *vgrid;
-  GridCartesian *sgrid;
-    
-  int Nd;
-  double flops;
-  double flops_call;
-  uint64_t usec;
-    
-  Coordinate dimensions;
-  Coordinate processors;
-  Coordinate processor_coor;
-    
+template<class scalar> struct FFTW {
+};
+
+#ifdef GRID_HIP
+template<> struct FFTW<ComplexD> {
 public:
-    
   static const int forward=FFTW_FORWARD;
   static const int backward=FFTW_BACKWARD;
-    
-  double Flops(void) {return flops;}
-  double MFlops(void) {return flops/usec;}
-  double USec(void)   {return (double)usec;}    
-
-  FFT ( GridCartesian * grid ) :
-    vgrid(grid),
-    Nd(grid->_ndimension),
-    dimensions(grid->_fdimensions),
-    processors(grid->_processors),
-    processor_coor(grid->_processor_coor)
-  {
-    flops=0;
-    usec =0;
-    Coordinate layout(Nd,1);
-    sgrid = new GridCartesian(dimensions,layout,processors,*grid);
-  };
-    
-  ~FFT ( void)  {
-    delete sgrid;
-  }
-    
-  template<class vobj>
-  void FFT_dim_mask(Lattice<vobj> &result,const Lattice<vobj> &source,Coordinate mask,int sign){
-
-    conformable(result.Grid(),vgrid);
-    conformable(source.Grid(),vgrid);
-    Lattice<vobj> tmp(vgrid);
-    tmp = source;
-    for(int d=0;d<Nd;d++){
-      if( mask[d] ) {
-	FFT_dim(result,tmp,d,sign);
-	tmp=result;
-      }
-    }
-  }
-
-  template<class vobj>
-  void FFT_all_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int sign){
-    Coordinate mask(Nd,1);
-    FFT_dim_mask(result,source,mask,sign);
-  }
-
-
-  template<class vobj>
-  void FFT_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int dim, int sign){
-#ifndef HAVE_FFTW
-    std::cerr << "FFTW is not compiled but is called"<<std::endl;
-    assert(0);
-#else
-    conformable(result.Grid(),vgrid);
-    conformable(source.Grid(),vgrid);
-
-    int L = vgrid->_ldimensions[dim];
-    int G = vgrid->_fdimensions[dim];
-      
-    Coordinate layout(Nd,1);
-    Coordinate pencil_gd(vgrid->_fdimensions);
-      
-    pencil_gd[dim] = G*processors[dim];
-      
-    // Pencil global vol LxLxGxLxL per node
-    GridCartesian pencil_g(pencil_gd,layout,processors,*vgrid);
-      
-    // Construct pencils
-    typedef typename vobj::scalar_object sobj;
-    typedef typename sobj::scalar_type   scalar;
-      
-    Lattice<sobj> pgbuf(&pencil_g);
-    autoView(pgbuf_v , pgbuf, CpuWrite);
-    std::cout << "CPU view" << std::endl;
-    
-    typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
-    typedef typename FFTW<scalar>::FFTW_plan   FFTW_plan;
-      
-    int Ncomp = sizeof(sobj)/sizeof(scalar);
-    int Nlow  = 1;
-    for(int d=0;d<dim;d++){
-      Nlow*=vgrid->_ldimensions[d];
-    }
-      
-    int rank = 1;  /* 1d transforms */
-    int n[] = {G}; /* 1d transforms of length G */
-    int howmany = Ncomp;
-    int odist,idist,istride,ostride;
-    idist   = odist   = 1;          /* Distance between consecutive FT's */
-    istride = ostride = Ncomp*Nlow; /* distance between two elements in the same FT */
-    int *inembed = n, *onembed = n;
-      
-    scalar div;
-    if ( sign == backward ) div = 1.0/G;
-    else if ( sign == forward ) div = 1.0;
-    else assert(0);
-      
-    std::cout << GridLogPerformance<<"Making FFTW plan" << std::endl;
+  typedef hipfftDoubleComplex FFTW_scalar;
+  typedef hipfftHandle        FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
     FFTW_plan p;
+    auto rv = hipfftPlanMany(&p,rank,n,n,istride,idist,n,ostride,odist,HIPFFT_Z2Z,howmany);
+    GRID_ASSERT(rv==HIPFFT_SUCCESS);
+    return p;
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    hipfftResult rv;
+    if ( sign == forward ) rv =hipfftExecZ2Z(p,in,out,HIPFFT_FORWARD);
+    else                   rv =hipfftExecZ2Z(p,in,out,HIPFFT_BACKWARD);
+    accelerator_barrier();
+    GRID_ASSERT(rv==HIPFFT_SUCCESS);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { hipfftDestroy(p); }
+};
+template<> struct FFTW<ComplexF> {
+public:
+  static const int forward=FFTW_FORWARD;
+  static const int backward=FFTW_BACKWARD;
+  typedef hipfftComplex FFTW_scalar;
+  typedef hipfftHandle  FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
+    FFTW_plan p;
+    auto rv = hipfftPlanMany(&p,rank,n,n,istride,idist,n,ostride,odist,HIPFFT_C2C,howmany);
+    GRID_ASSERT(rv==HIPFFT_SUCCESS);
+    return p;
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    hipfftResult rv;
+    if ( sign == forward ) rv =hipfftExecC2C(p,in,out,HIPFFT_FORWARD);
+    else                   rv =hipfftExecC2C(p,in,out,HIPFFT_BACKWARD);
+    accelerator_barrier();
+    GRID_ASSERT(rv==HIPFFT_SUCCESS);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { hipfftDestroy(p); }
+};
+#endif
+
+#ifdef GRID_CUDA
+template<> struct FFTW<ComplexD> {
+public:
+  static const int forward=FFTW_FORWARD;
+  static const int backward=FFTW_BACKWARD;
+  typedef cufftDoubleComplex FFTW_scalar;
+  typedef cufftHandle        FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
+    FFTW_plan p;
+    cufftPlanMany(&p,rank,n,n,istride,idist,n,ostride,odist,CUFFT_Z2Z,howmany);
+    return p;
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    if ( sign == forward ) cufftExecZ2Z(p,in,out,CUFFT_FORWARD);
+    else                   cufftExecZ2Z(p,in,out,CUFFT_INVERSE);
+    accelerator_barrier();
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { cufftDestroy(p); }
+};
+template<> struct FFTW<ComplexF> {
+public:
+  static const int forward=FFTW_FORWARD;
+  static const int backward=FFTW_BACKWARD;
+  typedef cufftComplex FFTW_scalar;
+  typedef cufftHandle  FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
+    FFTW_plan p;
+    cufftPlanMany(&p,rank,n,n,istride,idist,n,ostride,odist,CUFFT_C2C,howmany);
+    return p;
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    if ( sign == forward ) cufftExecC2C(p,in,out,CUFFT_FORWARD);
+    else                   cufftExecC2C(p,in,out,CUFFT_INVERSE);
+    accelerator_barrier();
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { cufftDestroy(p); }
+};
+#endif
+
+#if !defined(GRID_CUDA) && !defined(GRID_HIP)
+#ifdef HAVE_FFTW
+template<> struct FFTW<ComplexD> {
+public:
+  typedef fftw_complex FFTW_scalar;
+  typedef fftw_plan    FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
+    return ::fftw_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    ::fftw_execute_dft(p,in,out);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { ::fftw_destroy_plan(p); }
+};
+template<> struct FFTW<ComplexF> {
+public:
+  typedef fftwf_complex FFTW_scalar;
+  typedef fftwf_plan    FFTW_plan;
+  static FFTW_plan fftw_plan_many_dft(int rank, int *n,int howmany,
+				      FFTW_scalar *in, int *inembed,
+				      int istride, int idist,
+				      FFTW_scalar *out, int *onembed,
+				      int ostride, int odist,
+				      int sign, unsigned flags) {
+    return ::fftwf_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
+  }
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out, int sign) {
+    ::fftwf_execute_dft(p,in,out);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) { ::fftwf_destroy_plan(p); }
+};
+#endif
+#endif
+
+struct FFTbase {
+  double        flops;
+  double        flops_call;
+  uint64_t      usec;
+  GridCartesian *_grid;
+
+  static const int forward  = FFTW_FORWARD;
+  static const int backward = FFTW_BACKWARD;
+
+  double Flops(void)  { return flops; }
+  double MFlops(void) { return flops / usec; }
+  double USec(void)   { return (double)usec; }
+
+  FFTbase(GridCartesian *grid) : _grid(grid), flops(0), flops_call(0), usec(0) {}
+};
+
+// Barrel-shift gather, FFT execute, and insert. Called by both FFT and PlannedFFT.
+// The caller is responsible for plan acquisition and destruction.
+template<class vobj>
+static void FFT_dim_execute(
+    Lattice<vobj>       &result,
+    const Lattice<vobj> &source,
+    int dim, int sign,
+    typename FFTW<typename vobj::scalar_type>::FFTW_plan p,
+    GridCartesian *grid,
+    double &flops, double &flops_call, uint64_t &usec)
+{
+  typedef typename vobj::scalar_type   scalar;
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type   scalar_type;
+  typedef typename vobj::vector_type   vector_type;
+  typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
+
+#if 0
+  // ======================= ORIGINAL barrel-shift path =======================
+  // Preserved for reference.  Superseded by the transpose / all-to-all path
+  // below (the single active path for ALL P): the barrel is a P-fold redundant
+  // all-gather -- every rank assembles and transforms all Nperp lines of length
+  // G, keeping only its own L points.  The transpose partitions the Nperp
+  // perpendicular lines across the P ranks along dim, so each rank transforms
+  // only ceil(Nperp/P) lines and moves (P-1)/P of its data instead of P-1
+  // redundant copies.  See CartesianRingAllToAll in communicator/RingAllReduce.h.
+  const int Ndim = grid->Nd();
+  int L = grid->_ldimensions[dim];
+  int G = grid->_fdimensions[dim];
+  int Ncomp = sizeof(sobj) / sizeof(scalar);
+  int64_t Nlow = 1, Nhigh = 1;
+  for (int d = 0; d < dim;      d++) Nlow  *= grid->_ldimensions[d];
+  for (int d = dim+1; d < Ndim; d++) Nhigh *= grid->_ldimensions[d];
+  int64_t Nperp = Nlow * Nhigh;
+
+  deviceVector<scalar> pgbuf(Nperp * Ncomp * G);
+  scalar *pgbuf_v = &pgbuf[0];
+  int howmany = Ncomp * Nperp;
+
+  scalar div;
+  if      (sign == FFTW_BACKWARD) div = 1.0 / G;
+  else if (sign == FFTW_FORWARD)  div = 1.0;
+  else GRID_ASSERT(0);
+
+  double t_pencil = 0, t_fft = 0, t_copy = 0, t_shift = 0;
+  double t_total  = -usecond();
+
+  result = source;
+  int pc = grid->_processor_coor[dim];
+
+  const Coordinate ldims      = grid->_ldimensions;
+  const Coordinate rdims      = grid->_rdimensions;
+  const Coordinate sdims      = grid->_simd_layout;
+  const Coordinate processors = grid->_processors;
+
+  Coordinate pgdims(Ndim);
+  pgdims[0] = G;
+  for (int d = 0, dd = 1; d < Ndim; d++)
+    if (d != dim) pgdims[dd++] = ldims[d];
+  int64_t pgvol = 1;
+  for (int d = 0; d < Ndim; d++) pgvol *= pgdims[d];
+
+  const int Nsimd = vobj::Nsimd();
+  t_pencil = -usecond();
+  for (int p_idx = 0; p_idx < processors[dim]; p_idx++) {
+    t_copy -= usecond();
+    autoView(r_v, result, AcceleratorRead);
+    accelerator_for(idx, grid->oSites(), vobj::Nsimd(), {
+#ifdef GRID_SIMT
     {
-      FFTW_scalar *in = (FFTW_scalar *)&pgbuf_v[0];
-      FFTW_scalar *out= (FFTW_scalar *)&pgbuf_v[0];
-      p = FFTW<scalar>::fftw_plan_many_dft(rank,n,howmany,
-					   in,inembed,
-					   istride,idist,
-					   out,onembed,
-					   ostride, odist,
-					   sign,FFTW_ESTIMATE);
-    }
-      
-    // Barrel shift and collect global pencil
-    std::cout << GridLogPerformance<<"Making pencil" << std::endl;
-    Coordinate lcoor(Nd), gcoor(Nd);
-    result = source;
-    int pc = processor_coor[dim];
-    for(int p=0;p<processors[dim];p++) {
-      {
-	autoView(r_v,result,CpuRead);
-	autoView(p_v,pgbuf,CpuWrite);
-	thread_for(idx, sgrid->lSites(),{
-          Coordinate cbuf(Nd);
-          sobj s;
-	  sgrid->LocalIndexToLocalCoor(idx,cbuf);
-	  peekLocalSite(s,r_v,cbuf);
-	  cbuf[dim]+=((pc+p) % processors[dim])*L;
-	  pokeLocalSite(s,p_v,cbuf);
-        });
-      }
-      if (p != processors[dim] - 1) {
-	result = Cshift(result,dim,L);
+      int lane = acceleratorSIMTlane(Nsimd);
+#else
+    for (int lane = 0; lane < Nsimd; lane++) {
+#endif
+      Coordinate icoor, ocoor, pgcoor;
+      Lexicographic::CoorFromIndex(icoor, lane, sdims);
+      Lexicographic::CoorFromIndex(ocoor, idx,  rdims);
+      pgcoor[0] = ocoor[dim] + icoor[dim]*rdims[dim] + ((pc+p_idx)%processors[dim])*L;
+      for (int d = 0, dd = 1; d < Ndim; d++)
+        if (d != dim) { pgcoor[dd] = ocoor[d] + icoor[d]*rdims[d]; dd++; }
+      int64_t pgidx;
+      Lexicographic::IndexFromCoor(pgcoor, pgidx, pgdims);
+      vector_type *from = (vector_type *)&r_v[idx];
+      scalar_type  stmp;
+      for (int w = 0; w < Ncomp; w++) {
+        stmp = getlane(from[w], lane);
+        pgbuf_v[pgidx + w*pgvol] = stmp;
       }
+#ifdef GRID_SIMT
     }
-      
-    std::cout <<GridLogPerformance<< "Looping orthog" << std::endl;
-    // Loop over orthog coords
-    int NN=pencil_g.lSites();
-    GridStopWatch timer;
-    timer.Start();
-    thread_for( idx,NN,{
-        Coordinate cbuf(Nd);
-	pencil_g.LocalIndexToLocalCoor(idx, cbuf);
-	if ( cbuf[dim] == 0 ) {  // restricts loop to plane at lcoor[dim]==0
-	  FFTW_scalar *in = (FFTW_scalar *)&pgbuf_v[idx];
-	  FFTW_scalar *out= (FFTW_scalar *)&pgbuf_v[idx];
-	  FFTW<scalar>::fftw_execute_dft(p,in,out);
-	}
+#else
+    }
+#endif
     });
-    timer.Stop();
-      
-    // performance counting
-    double add,mul,fma;
-    FFTW<scalar>::fftw_flops(p,&add,&mul,&fma);
-    flops_call = add+mul+2.0*fma;
-    usec += timer.useconds();
-    flops+= flops_call*NN;
-      
-    std::cout <<GridLogPerformance<< "Writing back results " << std::endl;
-    // writing out result
+    t_copy += usecond();
+    if (p_idx != processors[dim] - 1) {
+      Lattice<vobj> temp(grid);
+      t_shift -= usecond();
+      temp = Cshift(result, dim, L); result = temp;
+      t_shift += usecond();
+    }
+  }
+  t_pencil += usecond();
+
+  FFTW_scalar *in  = (FFTW_scalar *)pgbuf_v;
+  FFTW_scalar *out = (FFTW_scalar *)pgbuf_v;
+  t_fft = -usecond();
+  FFTW<scalar>::fftw_execute_dft(p, in, out, sign);
+  t_fft += usecond();
+
+  flops_call = 5.0 * howmany * G * log2(G);
+  usec  = t_fft;
+  flops = flops_call;
+
+  result = Zero();
+  double t_insert = -usecond();
+  {
+    autoView(r_v, result, AcceleratorWrite);
+    accelerator_for(idx, grid->oSites(), Nsimd, {
+#ifdef GRID_SIMT
     {
-      autoView(pgbuf_v,pgbuf,CpuRead);
-      autoView(result_v,result,CpuWrite);
-      thread_for(idx,sgrid->lSites(),{
-	Coordinate clbuf(Nd), cgbuf(Nd);
-	sobj s;
-	sgrid->LocalIndexToLocalCoor(idx,clbuf);
-	cgbuf = clbuf;
-	cgbuf[dim] = clbuf[dim]+L*pc;
-	peekLocalSite(s,pgbuf_v,cgbuf);
-	pokeLocalSite(s,result_v,clbuf);
+      int lane = acceleratorSIMTlane(Nsimd);
+#else
+    for (int lane = 0; lane < Nsimd; lane++) {
+#endif
+      Coordinate icoor(Ndim), ocoor(Ndim), pgcoor(Ndim);
+      Lexicographic::CoorFromIndex(icoor, lane, sdims);
+      Lexicographic::CoorFromIndex(ocoor, idx,  rdims);
+      pgcoor[0] = ocoor[dim] + icoor[dim]*rdims[dim] + pc*L;
+      for (int d = 0, dd = 1; d < Ndim; d++)
+        if (d != dim) { pgcoor[dd] = ocoor[d] + icoor[d]*rdims[d]; dd++; }
+      int64_t pgidx;
+      Lexicographic::IndexFromCoor(pgcoor, pgidx, pgdims);
+      vector_type *to  = (vector_type *)&r_v[idx];
+      scalar_type  stmp;
+      for (int w = 0; w < Ncomp; w++) {
+        stmp = pgbuf_v[pgidx + w*pgvol];
+        putlane(to[w], stmp, lane);
+      }
+#ifdef GRID_SIMT
+    }
+#else
+    }
+#endif
+    });
+  }
+  result = result * div;
+  t_insert += usecond();
+  t_total  += usecond();
+
+  std::cout << GridLogPerformance << " FFT took   " << t_total/1.0e6  << " s" << std::endl;
+  std::cout << GridLogPerformance << " FFT pencil " << t_pencil/1.0e6 << " s" << std::endl;
+  std::cout << GridLogPerformance << "  of which copy " << t_copy/1.0e6  << " s" << std::endl;
+  std::cout << GridLogPerformance << "  of which shift" << t_shift/1.0e6 << " s" << std::endl;
+  std::cout << GridLogPerformance << " FFT kernels " << t_fft/1.0e6    << " s" << std::endl;
+  std::cout << GridLogPerformance << " FFT insert  " << t_insert/1.0e6 << " s" << std::endl;
+#endif
+
+  // ==================== transpose / all-to-all pencil FFT ====================
+  //
+  // P ranks lie along dim; each holds a local L-slab (L=ldim) of every global
+  // line (length G=fdim=L*P) and Nperp perpendicular lines.  We partition the
+  // Nperp lines across the P ranks: the rank at coord `own` owns the lines with
+  // olin/Oloc == own, gathers their full G points via a cartesian all-to-all,
+  // transforms only Oloc = ceil(Nperp/P) of them, then scatters the result back
+  // with the inverse all-to-all.
+  //
+  // The owned-line count is CEIL-padded to a multiple of P (Oloc*P >= Nperp) so
+  // the all-to-all stays SYMMETRIC (one uniform chunk) for ANY (P,Nperp) -- in
+  // particular a P carrying a factor absent from Nperp, e.g. P=3 on the T axis
+  // of 128^3x288 whose perpendicular volume is a pure power of two.  This keeps
+  // the transpose as total over decompositions as the barrel it replaces (no
+  // new geometry constraint), at a cost of <= (P-1) padded lines out of Nperp.
+  // Padding slots olin in [Nperp, Oloc*P) are never packed and never unpacked.
+  //
+  // The load-bearing identity: the all-to-all block index == cartesian coord
+  // along dim == which L-slab [c*L, c*L+L) of the global line -- so the block a
+  // rank receives carries the global-x tag needed to order the FFT input.
+  //
+  // Degenerate P: P=1 -> both all-to-alls are self-copies and the reorders are
+  // the identity (G=L), i.e. a pure local FFT.  P=2 -> each all-to-all moves
+  // half a field, two of them one field, equal to the barrel's single Cshift.
+  {
+    const int Ndim = grid->Nd();
+    int     L     = grid->_ldimensions[dim];
+    int     G     = grid->_fdimensions[dim];
+    int     Ncomp = sizeof(sobj) / sizeof(scalar);
+    int     P     = grid->_processors[dim];
+    int64_t Nperp = 1;
+    for (int d = 0; d < Ndim; d++)
+      if (d != dim) Nperp *= grid->_ldimensions[d];
+
+    int64_t Oloc  = (Nperp + P - 1) / P;        // ceil: owned (padded) lines/rank
+    int64_t chunk = (int64_t)L * Oloc * Ncomp;  // one all-to-all block (uniform)
+    int64_t nbuf  = (int64_t)P * chunk;         // == Ncomp*Oloc*G, one field's worth
+    int64_t howmany_local = (int64_t)Ncomp * Oloc;
+
+    scalar div;
+    if      (sign == FFTW_BACKWARD) div = 1.0 / G;
+    else if (sign == FFTW_FORWARD)  div = 1.0;
+    else GRID_ASSERT(0);
+
+    double t_total = -usecond();
+    double t_pack = 0, t_a2a = 0, t_reorder = 0, t_fft = 0, t_unpack = 0;
+
+    deviceVector<scalar> sbuf(nbuf);
+    deviceVector<scalar> rbuf(nbuf);
+    deviceVector<scalar> pgbuf(nbuf);           // FFTW pencil buffer, Ncomp*Oloc lines of G
+    scalar *sbuf_v  = &sbuf[0];
+    scalar *rbuf_v  = &rbuf[0];
+    scalar *pgbuf_v = &pgbuf[0];
+
+    // Pad slots (olin in [Nperp, Oloc*P)) are never packed, so they would carry
+    // garbage device memory into the FFT.  Zero them so the padded lines stay
+    // finite -- but ONLY when padding is actually present.  In the common
+    // Nperp % P == 0 case pack writes every sbuf entry bijectively, so skip the
+    // whole-buffer memset and its device sync entirely.
+    if ( (int64_t)Oloc*P != Nperp )
+      acceleratorMemSet(sbuf_v, 0, nbuf*sizeof(scalar));
+
+    const Coordinate ldims = grid->_ldimensions;
+    const Coordinate rdims = grid->_rdimensions;
+    const Coordinate sdims = grid->_simd_layout;
+    const int Nsimd = vobj::Nsimd();
+
+    // ---- 1. pack: source -> sbuf.  block = owner coord; payload xloc + L*(slot + Oloc*w)
+    t_pack -= usecond();
+    {
+      autoView(s_v, source, AcceleratorRead);
+      accelerator_for(idx, grid->oSites(), Nsimd, {
+#ifdef GRID_SIMT
+      {
+        int lane = acceleratorSIMTlane(Nsimd);
+#else
+      for (int lane = 0; lane < Nsimd; lane++) {
+#endif
+        Coordinate icoor(Ndim), ocoor(Ndim);
+        Lexicographic::CoorFromIndex(icoor, lane, sdims);
+        Lexicographic::CoorFromIndex(ocoor, idx,  rdims);
+        int64_t xloc = ocoor[dim] + icoor[dim]*rdims[dim];
+        int64_t olin = 0, str = 1;
+        for (int d = 0; d < Ndim; d++) {
+          if (d == dim) continue;
+          int64_t c = ocoor[d] + icoor[d]*rdims[d];
+          olin += str * c;
+          str  *= ldims[d];
+        }
+        int64_t own  = olin / Oloc;
+        int64_t slot = olin - own*Oloc;
+        vector_type *from = (vector_type *)&s_v[idx];
+        for (int w = 0; w < Ncomp; w++) {
+          scalar_type stmp = getlane(from[w], lane);
+          sbuf_v[ own*chunk + xloc + L*(slot + Oloc*w) ] = stmp;
+        }
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
       });
     }
-    result = result*div;
-      
-    std::cout <<GridLogPerformance<< "Destroying plan " << std::endl;
-    // destroying plan
-    FFTW<scalar>::fftw_destroy_plan(p);
+    t_pack += usecond();
+
+    // ---- 2. forward all-to-all: gather my owned lines' L-slabs from every rank
+    t_a2a -= usecond();
+    CartesianRingAllToAll(grid, sbuf_v, rbuf_v, (uint64_t)chunk, dim);
+    t_a2a += usecond();
+
+    // ---- 3. reorder rbuf -> pgbuf: contiguous G-lines (w,slot), xpos = src*L + xloc
+    t_reorder -= usecond();
+    accelerator_for(q, nbuf, 1, {
+      int64_t xpos = q % G;
+      int64_t t    = q / G;           // = w*Oloc + slot
+      int64_t slot = t % Oloc;
+      int64_t w    = t / Oloc;
+      int64_t src  = xpos / L;
+      int64_t xloc = xpos % L;
+      pgbuf_v[q] = rbuf_v[ src*chunk + xloc + L*(slot + Oloc*w) ];
+    });
+    t_reorder += usecond();
+
+    // ---- 4. FFT: Ncomp*Oloc contiguous lines of length G (istride 1, idist G)
+    {
+      FFTW_scalar *in  = (FFTW_scalar *)pgbuf_v;
+      FFTW_scalar *out = (FFTW_scalar *)pgbuf_v;
+      t_fft -= usecond();
+      FFTW<scalar>::fftw_execute_dft(p, in, out, sign);
+      t_fft += usecond();
+    }
+    flops_call = 5.0 * (double)howmany_local * G * log2(G);
+    usec  = (uint64_t)t_fft;
+    flops = flops_call;
+
+    // ---- 5a. reorder pgbuf -> sbuf: block = destination coord; xpos = dst*L + xloc
+    t_reorder -= usecond();
+    accelerator_for(j, nbuf, 1, {
+      int64_t dst  = j / chunk;
+      int64_t r    = j % chunk;
+      int64_t xloc = r % L;
+      int64_t u    = r / L;           // = slot + Oloc*w
+      int64_t slot = u % Oloc;
+      int64_t w    = u / Oloc;
+      int64_t xpos = dst*L + xloc;
+      sbuf_v[j] = pgbuf_v[ w*Oloc*G + slot*G + xpos ];
+    });
+    t_reorder += usecond();
+
+    // ---- 5b. inverse all-to-all: scatter transformed L-slabs back
+    t_a2a -= usecond();
+    CartesianRingAllToAll(grid, sbuf_v, rbuf_v, (uint64_t)chunk, dim);
+    t_a2a += usecond();
+
+    // ---- 5c. unpack rbuf -> result (x div); block = owner coord of each line
+    t_unpack -= usecond();
+    {
+      autoView(r_v, result, AcceleratorWrite);
+      accelerator_for(idx, grid->oSites(), Nsimd, {
+#ifdef GRID_SIMT
+      {
+        int lane = acceleratorSIMTlane(Nsimd);
+#else
+      for (int lane = 0; lane < Nsimd; lane++) {
 #endif
+        Coordinate icoor(Ndim), ocoor(Ndim);
+        Lexicographic::CoorFromIndex(icoor, lane, sdims);
+        Lexicographic::CoorFromIndex(ocoor, idx,  rdims);
+        int64_t xloc = ocoor[dim] + icoor[dim]*rdims[dim];
+        int64_t olin = 0, str = 1;
+        for (int d = 0; d < Ndim; d++) {
+          if (d == dim) continue;
+          int64_t c = ocoor[d] + icoor[d]*rdims[d];
+          olin += str * c;
+          str  *= ldims[d];
+        }
+        int64_t own  = olin / Oloc;
+        int64_t slot = olin - own*Oloc;
+        vector_type *to = (vector_type *)&r_v[idx];
+        for (int w = 0; w < Ncomp; w++) {
+          scalar_type stmp = div * rbuf_v[ own*chunk + xloc + L*(slot + Oloc*w) ];
+          putlane(to[w], stmp, lane);
+        }
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
+      });
+    }
+    t_unpack += usecond();
+    t_total  += usecond();
+
+    std::cout << GridLogPerformance << " FFT took     " << t_total/1.0e6   << " s (transpose P=" << P << ")" << std::endl;
+    std::cout << GridLogPerformance << " FFT pack     " << t_pack/1.0e6    << " s" << std::endl;
+    std::cout << GridLogPerformance << " FFT alltoall " << t_a2a/1.0e6     << " s" << std::endl;
+    std::cout << GridLogPerformance << " FFT reorder  " << t_reorder/1.0e6 << " s" << std::endl;
+    std::cout << GridLogPerformance << " FFT kernels  " << t_fft/1.0e6     << " s" << std::endl;
+    std::cout << GridLogPerformance << " FFT unpack   " << t_unpack/1.0e6  << " s" << std::endl;
+  }
+}
+
+class FFT : public FFTbase {
+public:
+  FFT(GridCartesian *grid) : FFTbase(grid) {}
+  ~FFT() {}
+
+  template<class vobj>
+  void FFT_dim_mask(Lattice<vobj> &result, const Lattice<vobj> &source, Coordinate mask, int sign) {
+    const int Ndim = _grid->Nd();
+    Lattice<vobj> tmp = source;
+    for (int d = 0; d < Ndim; d++) {
+      if (mask[d]) {
+        FFT_dim(result, tmp, d, sign);
+        tmp = result;
+      }
+    }
+  }
+
+  template<class vobj>
+  void FFT_all_dim(Lattice<vobj> &result, const Lattice<vobj> &source, int sign) {
+    Coordinate mask(_grid->Nd(), 1);
+    FFT_dim_mask(result, source, mask, sign);
+  }
+
+  template<class vobj>
+  void FFT_dim(Lattice<vobj> &result, const Lattice<vobj> &source, int dim, int sign) {
+    GRID_ASSERT(source.Grid() == _grid);
+    GRID_ASSERT(result.Grid() == _grid);
+    conformable(result.Grid(), source.Grid());
+
+    typedef typename vobj::scalar_type   scalar;
+    typedef typename vobj::scalar_object sobj;
+    typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
+    typedef typename FFTW<scalar>::FFTW_plan   FFTW_plan;
+
+    const int Ndim = _grid->Nd();
+    int G     = _grid->_fdimensions[dim];
+    int Ncomp = sizeof(sobj) / sizeof(scalar);
+    int64_t Nperp = 1;
+    for (int d = 0; d < Ndim; d++)
+      if (d != dim) Nperp *= _grid->_ldimensions[d];
+    int     P     = _grid->_processors[dim];
+    int64_t Oloc  = (Nperp + P - 1) / P;   // ceil-padded owned lines/rank (see FFT_dim_execute)
+    int n[]     = {G};
+    int howmany = Ncomp * (int)Oloc;
+
+    deviceVector<scalar> dummy(2);
+    FFTW_scalar *buf = (FFTW_scalar *)&dummy[0];
+    FFTW_plan p = FFTW<scalar>::fftw_plan_many_dft(1, n, howmany,
+                                                    buf, n, 1, G,
+                                                    buf, n, 1, G,
+                                                    sign, FFTW_ESTIMATE);
+    FFT_dim_execute(result, source, dim, sign, p, _grid, flops, flops_call, usec);
+    FFTW<scalar>::fftw_destroy_plan(p);
+  }
+};
+
+template<class vobj>
+class PlannedFFT : public FFTbase {
+private:
+  typedef typename vobj::scalar_type   scalar;
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::vector_type   vector_type;
+  typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
+  typedef typename FFTW<scalar>::FFTW_plan   FFTW_plan;
+
+  std::vector<FFTW_plan> forward_plans;
+  std::vector<FFTW_plan> backward_plans;
+
+  void PlanCreate() {
+    const int Ndim = _grid->Nd();
+    forward_plans.resize(Ndim);
+    backward_plans.resize(Ndim);
+
+    for (int d = 0; d < Ndim; d++) {
+      int G       = _grid->_fdimensions[d];
+      int Ncomp   = sizeof(sobj) / sizeof(scalar);
+      int64_t Nperp = 1;
+      for (int dd = 0; dd < Ndim; dd++)
+        if (dd != d) Nperp *= _grid->_ldimensions[dd];
+      int     P     = _grid->_processors[d];
+      int64_t Oloc  = (Nperp + P - 1) / P;   // ceil-padded owned lines/rank (see FFT_dim_execute)
+      int howmany = Ncomp * (int)Oloc;
+      int n[]     = {G};
+
+      deviceVector<scalar> dummy(2);
+      FFTW_scalar *buf = (FFTW_scalar *)&dummy[0];
+
+      forward_plans[d]  = FFTW<scalar>::fftw_plan_many_dft(1, n, howmany, buf, n, 1, G, buf, n, 1, G, FFTW_FORWARD,  FFTW_ESTIMATE);
+      backward_plans[d] = FFTW<scalar>::fftw_plan_many_dft(1, n, howmany, buf, n, 1, G, buf, n, 1, G, FFTW_BACKWARD, FFTW_ESTIMATE);
+    }
+  }
+
+  void PlanDestroy() {
+    for (auto p : forward_plans)  FFTW<scalar>::fftw_destroy_plan(p);
+    for (auto p : backward_plans) FFTW<scalar>::fftw_destroy_plan(p);
+    forward_plans.clear();
+    backward_plans.clear();
+  }
+
+public:
+  PlannedFFT(GridCartesian *grid) : FFTbase(grid) { PlanCreate(); }
+  ~PlannedFFT() { PlanDestroy(); }
+
+  void FFT_dim_mask(Lattice<vobj> &result, const Lattice<vobj> &source, Coordinate mask, int sign) {
+    const int Ndim = _grid->Nd();
+    Lattice<vobj> tmp = source;
+    for (int d = 0; d < Ndim; d++) {
+      if (mask[d]) {
+        FFT_dim(result, tmp, d, sign);
+        tmp = result;
+      }
+    }
+  }
+
+  void FFT_all_dim(Lattice<vobj> &result, const Lattice<vobj> &source, int sign) {
+    Coordinate mask(_grid->Nd(), 1);
+    FFT_dim_mask(result, source, mask, sign);
+  }
+
+  void FFT_dim(Lattice<vobj> &result, const Lattice<vobj> &source, int dim, int sign) {
+    GRID_ASSERT(source.Grid() == _grid);
+    GRID_ASSERT(result.Grid() == _grid);
+    GRID_ASSERT((int)forward_plans.size() == _grid->Nd());
+    conformable(result.Grid(), source.Grid());
+    FFTW_plan p = (sign == forward ? forward_plans : backward_plans)[dim];
+    FFT_dim_execute(result, source, dim, sign, p, _grid, flops, flops_call, usec);
   }
 };
 

Grid/algorithms/LinearOperator.h

@@ -64,7 +64,7 @@ public:
 //
 // I'm not entirely happy with implementation; to share the Schur code between herm and non-herm
 // while still having a "OpAndNorm" in the abstract base I had to implement it in both cases
-// with an assert trap in the non-herm. This isn't right; there must be a better C++ way to
+// with an GRID_ASSERT trap in the non-herm. This isn't right; there must be a better C++ way to
 // do it, but I fear it required multiple inheritance and mixed in abstract base classes
 /////////////////////////////////////////////////////////////////////////////////////////////
 
@@ -148,22 +148,22 @@ public:
   // Support for coarsening to a multigrid
   void OpDiag (const Field &in, Field &out) {
     _Mat.Mdiag(in,out);
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDir  (const Field &in, Field &out,int dir,int disp) {
     _Mat.Mdir(in,out,dir,disp);
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDirAll  (const Field &in, std::vector<Field> &out){
-    assert(0);
+    GRID_ASSERT(0);
   };
   void Op     (const Field &in, Field &out){
     _Mat.M(in,out);
-    assert(0);
+    GRID_ASSERT(0);
   }
   void AdjOp     (const Field &in, Field &out){
     _Mat.Mdag(in,out);
-    assert(0);
+    GRID_ASSERT(0);
   }
   void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
     HermOp(in,out);
@@ -188,13 +188,13 @@ public:
   ShiftedHermOpLinearOperator(LinearOperatorBase<Field> &Mat,RealD shift): _Mat(Mat), _shift(shift){};
   // Support for coarsening to a multigrid
   void OpDiag (const Field &in, Field &out) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDir  (const Field &in, Field &out,int dir,int disp) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDirAll  (const Field &in, std::vector<Field> &out){
-    assert(0);
+    GRID_ASSERT(0);
   };
   void Op     (const Field &in, Field &out){
     HermOp(in,out);
@@ -271,10 +271,42 @@ public:
     _Mat.Mdag(in,out);
   }
   void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
-    assert(0);
+    GRID_ASSERT(0);
   }
   void HermOp(const Field &in, Field &out){
-    assert(0);
+    GRID_ASSERT(0);
+  }
+};
+template<class Matrix,class Field>
+class ShiftedNonHermitianLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat;
+  RealD shift;
+public:
+  ShiftedNonHermitianLinearOperator(Matrix &Mat,RealD shft): _Mat(Mat),shift(shft){};
+  // Support for coarsening to a multigrid
+  void OpDiag (const Field &in, Field &out) {
+    _Mat.Mdiag(in,out);
+    out = out + shift*in;
+  }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) {
+    _Mat.Mdir(in,out,dir,disp);
+  }
+  void OpDirAll  (const Field &in, std::vector<Field> &out){
+    _Mat.MdirAll(in,out);
+  };
+  void Op     (const Field &in, Field &out){
+    _Mat.M(in,out);
+    out = out + shift * in;
+  }
+  void AdjOp     (const Field &in, Field &out){
+    _Mat.Mdag(in,out);
+    out = out + shift * in;
+  }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
+    GRID_ASSERT(0);
+  }
+  void HermOp(const Field &in, Field &out){
+    GRID_ASSERT(0);
   }
 };
 
@@ -313,13 +345,13 @@ class SchurOperatorBase :  public LinearOperatorBase<Field> {
   }
   // Support for coarsening to a multigrid
   void OpDiag (const Field &in, Field &out) {
-    assert(0); // must coarsen the unpreconditioned system
+    GRID_ASSERT(0); // must coarsen the unpreconditioned system
   }
   void OpDir  (const Field &in, Field &out,int dir,int disp) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDirAll  (const Field &in, std::vector<Field> &out){
-    assert(0);
+    GRID_ASSERT(0);
   };
 };
 template<class Matrix,class Field>
@@ -415,10 +447,10 @@ class NonHermitianSchurOperatorBase :  public LinearOperatorBase<Field>
     MpcDag(tmp,out);
   }
   virtual void HermOpAndNorm(const Field& in, Field& out, RealD& n1, RealD& n2) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   virtual void HermOp(const Field& in, Field& out) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   void Op(const Field& in, Field& out) {
     Mpc(in, out);
@@ -428,13 +460,13 @@ class NonHermitianSchurOperatorBase :  public LinearOperatorBase<Field>
   }
   // Support for coarsening to a multigrid
   void OpDiag(const Field& in, Field& out) {
-    assert(0); // must coarsen the unpreconditioned system
+    GRID_ASSERT(0); // must coarsen the unpreconditioned system
   }
   void OpDir(const Field& in, Field& out, int dir, int disp) {
-    assert(0);
+    GRID_ASSERT(0);
   }
   void OpDirAll(const Field& in, std::vector<Field>& out){
-    assert(0);
+    GRID_ASSERT(0);
   };
 };
 
@@ -548,7 +580,7 @@ class SchurStaggeredOperator :  public SchurOperatorBase<Field> {
  public:
   SchurStaggeredOperator (Matrix &Mat): _Mat(Mat), tmp(_Mat.RedBlackGrid()) 
   { 
-    assert( _Mat.isTrivialEE() );
+    GRID_ASSERT( _Mat.isTrivialEE() );
     mass = _Mat.Mass();
   }
   virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
@@ -579,7 +611,7 @@ class SchurStaggeredOperator :  public SchurOperatorBase<Field> {
     Mpc(in,out);
   }
   virtual void MpcDagMpc(const Field &in, Field &out) {
-    assert(0);// Never need with staggered
+    GRID_ASSERT(0);// Never need with staggered
   }
 };
 template<class Matrix,class Field> using SchurStagOperator = SchurStaggeredOperator<Matrix,Field>;
@@ -591,7 +623,7 @@ template<class Field> class OperatorFunction {
 public:
   virtual void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) = 0;
   virtual void operator() (LinearOperatorBase<Field> &Linop, const std::vector<Field> &in,std::vector<Field> &out) {
-    assert(in.size()==out.size());
+    GRID_ASSERT(in.size()==out.size());
     for(int k=0;k<in.size();k++){
       (*this)(Linop,in[k],out[k]);
     }
@@ -605,7 +637,7 @@ public:
 
   virtual void operator() (const std::vector<Field> &in, std::vector<Field> &out)
   {
-    assert(in.size() == out.size());
+    GRID_ASSERT(in.size() == out.size());
 
     for (unsigned int i = 0; i < in.size(); ++i)
     {

Grid/algorithms/approx/Chebyshev.h

@@ -269,7 +269,9 @@ public:
     RealD xscale = 2.0/(hi-lo);
     RealD mscale = -(hi+lo)/(hi-lo);
     Linop.HermOp(T0,y);
+    grid->Barrier();
     axpby(T1,xscale,mscale,y,in);
+    grid->Barrier();
 
     // sum = .5 c[0] T0 + c[1] T1
     //    out = ()*T0 + Coeffs[1]*T1;

Grid/algorithms/approx/Remez.cc

@@ -121,7 +121,7 @@ double AlgRemez::generateApprox(int num_degree, int den_degree,
   // Reallocate arrays, since degree has changed
   if (num_degree != n || den_degree != d) allocate(num_degree,den_degree);
 
-  assert(a_len<=SUM_MAX);
+  GRID_ASSERT(a_len<=SUM_MAX);
 
   step = new bigfloat[num_degree+den_degree+2];
 
@@ -151,9 +151,9 @@ double AlgRemez::generateApprox(int num_degree, int den_degree,
     equations();
     if (delta < tolerance) {
       std::cout<<"Delta too small, try increasing precision\n";
-      assert(0);
+      GRID_ASSERT(0);
     };    
-    assert( delta>= tolerance);
+    GRID_ASSERT( delta>= tolerance);
 
     search(step);
   }

Grid/algorithms/approx/Remez.h

@@ -134,7 +134,7 @@ class AlgRemez
   virtual ~AlgRemez();
 
   int getDegree(void){ 
-    assert(n==d);
+    GRID_ASSERT(n==d);
     return n;
   }
   // Reset the bounds of the approximation

Grid/algorithms/approx/RemezGeneral.cc

@@ -28,11 +28,11 @@ void AlgRemezGeneral::setupPolyProperties(int num_degree, int den_degree, PolyTy
   pow_n = num_degree;
   pow_d = den_degree;
 
-  if(pow_n % 2 == 0 && num_type_in == PolyType::Odd) assert(0);
-  if(pow_n % 2 == 1 && num_type_in == PolyType::Even) assert(0);
+  if(pow_n % 2 == 0 && num_type_in == PolyType::Odd) GRID_ASSERT(0);
+  if(pow_n % 2 == 1 && num_type_in == PolyType::Even) GRID_ASSERT(0);
 
-  if(pow_d % 2 == 0 && den_type_in == PolyType::Odd) assert(0);
-  if(pow_d % 2 == 1 && den_type_in == PolyType::Even) assert(0);
+  if(pow_d % 2 == 0 && den_type_in == PolyType::Odd) GRID_ASSERT(0);
+  if(pow_d % 2 == 1 && den_type_in == PolyType::Even) GRID_ASSERT(0);
 
   num_type = num_type_in;
   den_type = den_type_in;
@@ -112,9 +112,9 @@ double AlgRemezGeneral::generateApprox(const int num_degree, const int den_degre
     equations();
     if (delta < tolerance) {
       std::cout<<"Iteration " << iter-1 << " delta too small (" << delta << "<" << tolerance << "), try increasing precision\n";
-      assert(0);
+      GRID_ASSERT(0);
     };    
-    assert( delta>= tolerance );
+    GRID_ASSERT( delta>= tolerance );
 
     search();
   }
@@ -278,7 +278,7 @@ void AlgRemezGeneral::equations(){
       if(num_pows[j] != -1){ *aa++ = z; t++; }
       z *= x;
     }
-    assert(t == n+1);
+    GRID_ASSERT(t == n+1);
 
     z = (bigfloat)1l;
     t = 0;
@@ -286,7 +286,7 @@ void AlgRemezGeneral::equations(){
       if(den_pows[j] != -1){ *aa++ = -y * z; t++; }
       z *= x;
     }
-    assert(t == d);
+    GRID_ASSERT(t == d);
 
     B[i] = y * z;		// Right hand side vector
   }

Grid/algorithms/approx/RemezGeneral.h

@@ -106,7 +106,7 @@ class AlgRemezGeneral{
 		  bigfloat (*f)(bigfloat x, void *data), void *data);
 
   inline int getDegree(void) const{ 
-    assert(n==d);
+    GRID_ASSERT(n==d);
     return n;
   }
   // Reset the bounds of the approximation

Grid/algorithms/approx/ZMobius.cc

@@ -74,7 +74,7 @@ bigfloat epsilonMobius(bigfloat x, void* data){
 void computeZmobiusOmega(std::vector<ComplexD> &omega_out, const int Ls_out,
 			 const std::vector<RealD> &omega_in, const int Ls_in,
 			 const RealD lambda_bound){
-  assert(omega_in.size() == Ls_in);
+  GRID_ASSERT(omega_in.size() == Ls_in);
   omega_out.resize(Ls_out);
 
   //Use the Remez algorithm to generate the appropriate rational polynomial

Grid/algorithms/blas/BatchedInverse.h

@@ -0,0 +1,329 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: BatchedInverse.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/blas/BatchedBlas.h>
+
+#ifdef GRID_HIP
+#include <rocsolver/rocsolver.h>
+#endif
+// GRID_CUDA: batched LU inversion lives in cuBLAS (getrfBatched/getriBatched);
+//            cublas_v2.h already included via BatchedBlas.h.
+// GRID_SYCL: oneapi/mkl.hpp already included via BatchedBlas.h (lapack::getrf/getri).
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// GridBLASInverse: cross-platform batched dense matrix inversion.
+//
+// HIGH LEVEL contract (deliberately NOT a getrf/getrs interface): invert a
+// batch of dense N x N matrices IN PLACE,
+//
+//     A[i] <- A[i]^{-1}       i = 0 .. batchCount-1
+//
+// Layout: column major, lda = N, contiguous per batch element; pointer list
+// exactly as GridBLAS::gemmBatched (deviceVector<T*> of device pointers).
+// Each backend chooses HOW:
+//   HIP  : rocSOLVER getrf_batched + getri_batched
+//   CUDA : cuBLAS getrfBatched + getriBatched (out-of-place getri; workspace
+//          hidden here, result copied back so the surface stays in-place)
+//   SYCL : oneMKL LAPACK getrf + getri per batch element (USM, in-order queue)
+//   CPU  : Eigen PartialPivLU (the correctness oracle for all of the above)
+//
+// The int32 vendor-batched entry points bound N < 2^31 (asserted); the huge
+// single-matrix ILP64 path (getrf_64 + blocked identity-getrs harvest, proven
+// in the dense coarse-coarse setup at N=69120) migrates here as a batch==1
+// large-N dispatch in a follow-up -- the recursive Schur leaves are the
+// batched consumers this surface is shaped for.
+//
+// NB GPU-backend call signatures are written to vendor documentation but the
+// air-gapped development loop compiles only the CPU/Eigen path; verify the
+// rocSOLVER/cuBLAS/oneMKL calls against headers on first device compile.
+// Semantics are locked by the CPU unit test (Test_batched_blas).
+///////////////////////////////////////////////////////////////////////////////
+class GridBLASInverse {
+public:
+
+#ifdef GRID_HIP
+  // rocSOLVER runs on a rocblas_handle (distinct type from hipblasHandle_t)
+  static rocblas_handle & Handle(void) {
+    static rocblas_handle h;
+    static int init = 0;
+    if ( !init ) {
+      auto st = rocblas_create_handle(&h);
+      GRID_ASSERT(st == rocblas_status_success);
+      init = 1;
+    }
+    return h;
+  }
+#endif
+#ifdef GRID_CUDA
+  // cuBLAS batched LU shares the GridBLAS handle
+  static cublasHandle_t & Handle(void) {
+    GridBLAS::Init();
+    return GridBLAS::gridblasHandle;
+  }
+#endif
+#ifdef GRID_SYCL
+  static sycl::queue * & Handle(void) {
+    GridBLAS::Init();
+    return GridBLAS::gridblasHandle;
+  }
+#endif
+
+  GridBLASInverse() {};
+  ~GridBLASInverse() {};
+
+  ///////////////////////////////////////////////////////////////////////////
+  // SINGLE large matrix, in place: blocked LU + identity solve.
+  //   HIP : rocsolver_zgetrf_64 (blocked, GEMM-based) then ONE zgetrs_64 with
+  //         the N x N identity as RHS (trsm-based) into a scratch N x N, copied
+  //         back.  Extra device memory: N*N*16 B (1.2 GB at N=8640).
+  //   other backends: falls through to inverseBatched(batch 1).
+  // Why: rocSOLVER getri_batched is a small-matrix routine -- measured 0.35 s
+  // at N=1920, 0.53 s at 4320, 1.84 s at 8640 (n^1.2-1.8, i.e. overhead
+  // bound) as the big-leaf inverse of the 2D Schur recursion, where 287
+  // ranks wait on it.  The _64 getrf/getrs pair is the path proven in the
+  // 1-rank dense coarse-coarse setup at N=69120 (DenseCoarseMatrix.h).
+  // Same in-place, column-major, lda=N contract as inverseBatched.
+  // NB: written to rocSOLVER's documented z*_64 signatures; not compiled on
+  // HIP in the air-gapped loop -- verify on first hipcc build.
+  ///////////////////////////////////////////////////////////////////////////
+  double lastGetrfUs = 0.0, lastGetrsUs = 0.0;   // split timing of the last inverseLU (HIP path)
+  void inverseLU(int64_t N, ComplexD *A)
+  {
+#ifdef GRID_HIP
+    rocblas_handle handle = Handle();
+    deviceVector<int64_t>  ipiv((uint64_t)N);
+    deviceVector<int64_t>  info(1);
+    double t0 = usecond();
+    auto st1 = rocsolver_zgetrf_64(handle, N, N, (rocblas_double_complex *)A, N, &ipiv[0], &info[0]);
+    GRID_ASSERT(st1 == rocblas_status_success);
+    accelerator_barrier();
+    lastGetrfUs = usecond()-t0;
+    int64_t info_h = -1; acceleratorCopyFromDevice(&info[0], &info_h, sizeof(int64_t));
+    GRID_ASSERT(info_h == 0);
+    deviceVector<ComplexD> X((uint64_t)N*N);
+    { ComplexD *x = &X[0]; const int64_t NN = N;
+      accelerator_for(idx, (uint64_t)N*N, 1, { int64_t j = idx/NN, i = idx - j*NN; x[idx] = (i==j) ? ComplexD(1.0,0.0) : ComplexD(0.0,0.0); });
+      accelerator_barrier(); }
+    auto st2 = rocsolver_zgetrs_64(handle, rocblas_operation_none, N, N,
+                                   (rocblas_double_complex *)A, N, &ipiv[0],
+                                   (rocblas_double_complex *)&X[0], N);
+    GRID_ASSERT(st2 == rocblas_status_success);
+    accelerator_barrier();
+    lastGetrsUs = usecond()-t0-lastGetrfUs;
+    acceleratorCopyDeviceToDevice((void *)&X[0], (void *)A, (uint64_t)N*N*sizeof(ComplexD));
+#else
+    deviceVector<ComplexD*> bp(1); std::vector<ComplexD*> ptr(1); ptr[0] = A;
+    acceleratorCopyToDevice(&ptr[0], &bp[0], sizeof(ComplexD*));
+    inverseBatched(N, bp);
+#endif
+  }
+
+  void inverseBatched(int64_t N, deviceVector<ComplexF*> &Amat)
+  {
+    int32_t batchCount = Amat.size();
+    GRID_ASSERT(batchCount > 0);
+
+#ifdef GRID_HIP
+    GRID_ASSERT( N < 2147483647L );
+    rocblas_int n   = (rocblas_int)N;
+    rocblas_int lda = (rocblas_int)N;
+
+    deviceVector<rocblas_int> ipiv((uint64_t)batchCount*N);
+    deviceVector<rocblas_int> info(batchCount);
+
+    auto st1 = rocsolver_cgetrf_batched(Handle(), n, n,
+					(rocblas_float_complex *const *)&Amat[0], lda,
+					&ipiv[0], (rocblas_stride)N,
+					&info[0], batchCount);
+    GRID_ASSERT(st1 == rocblas_status_success);
+    auto st2 = rocsolver_cgetri_batched(Handle(), n,
+					(rocblas_float_complex *const *)&Amat[0], lda,
+					&ipiv[0], (rocblas_stride)N,
+					&info[0], batchCount);
+    GRID_ASSERT(st2 == rocblas_status_success);
+    accelerator_barrier();
+    std::vector<rocblas_int> info_h(batchCount);
+    acceleratorCopyFromDevice(&info[0],&info_h[0],batchCount*sizeof(rocblas_int));
+    for(int i=0;i<batchCount;i++) GRID_ASSERT(info_h[i]==0); // singular pivot => abort loudly
+#endif
+#ifdef GRID_CUDA
+    GRID_ASSERT( N < 2147483647L );
+    int n = (int)N;
+
+    deviceVector<int> ipiv((uint64_t)batchCount*N);
+    deviceVector<int> info(batchCount);
+
+    auto st1 = cublasCgetrfBatched(Handle(), n,
+				   (cuComplex **)&Amat[0], n,
+				   &ipiv[0], &info[0], batchCount);
+    GRID_ASSERT(st1 == CUBLAS_STATUS_SUCCESS);
+
+    // getri is OUT of place: hidden workspace keeps the surface in-place
+    deviceVector<ComplexF>  work((uint64_t)batchCount*N*N);
+    deviceVector<ComplexF*> Cptr(batchCount);
+    std::vector<ComplexF*>  Cptr_h(batchCount);
+    std::vector<ComplexF*>  Aptr_h(batchCount);
+    for(int i=0;i<batchCount;i++) Cptr_h[i] = &work[(uint64_t)i*N*N];
+    acceleratorCopyToDevice(&Cptr_h[0],&Cptr[0],batchCount*sizeof(ComplexF*));
+    acceleratorCopyFromDevice(&Amat[0],&Aptr_h[0],batchCount*sizeof(ComplexF*));
+
+    auto st2 = cublasCgetriBatched(Handle(), n,
+				   (const cuComplex *const *)&Amat[0], n,
+				   &ipiv[0],
+				   (cuComplex **)&Cptr[0], n,
+				   &info[0], batchCount);
+    GRID_ASSERT(st2 == CUBLAS_STATUS_SUCCESS);
+    accelerator_barrier();
+    std::vector<int> info_h(batchCount);
+    acceleratorCopyFromDevice(&info[0],&info_h[0],batchCount*sizeof(int));
+    for(int i=0;i<batchCount;i++) GRID_ASSERT(info_h[i]==0);
+    for(int i=0;i<batchCount;i++)
+      acceleratorCopyDeviceToDevice(Cptr_h[i],Aptr_h[i],(uint64_t)N*N*sizeof(ComplexF));
+#endif
+#ifdef GRID_SYCL
+    // Per-element oneMKL LAPACK on the in-order queue; group API optimisation later.
+    sycl::queue *q = Handle();
+    std::vector<ComplexF*> Aptr_h(batchCount);
+    acceleratorCopyFromDevice(&Amat[0],&Aptr_h[0],batchCount*sizeof(ComplexF*));
+
+    int64_t lwf = oneapi::mkl::lapack::getrf_scratchpad_size<std::complex<float> >(*q,N,N,N);
+    int64_t lwi = oneapi::mkl::lapack::getri_scratchpad_size<std::complex<float> >(*q,N,N);
+    deviceVector<ComplexF> scratchf(lwf);
+    deviceVector<ComplexF> scratchi(lwi);
+    deviceVector<int64_t>  ipiv(N);
+    for(int i=0;i<batchCount;i++){
+      oneapi::mkl::lapack::getrf(*q,N,N,(std::complex<float>*)Aptr_h[i],N,&ipiv[0],
+				 (std::complex<float>*)&scratchf[0],lwf);
+      oneapi::mkl::lapack::getri(*q,N,  (std::complex<float>*)Aptr_h[i],N,&ipiv[0],
+				 (std::complex<float>*)&scratchi[0],lwi);
+    }
+    q->wait();
+#endif
+#if !defined(GRID_SYCL) && !defined(GRID_CUDA) && !defined(GRID_HIP)
+    // Reference implementation; the oracle the unit test locks semantics with.
+    thread_for (p, batchCount, {
+      Eigen::Map<Eigen::MatrixXcf> eA(Amat[p],N,N);
+      Eigen::PartialPivLU<Eigen::MatrixXcf> lu(eA);
+      eA = lu.inverse();
+      });
+#endif
+  }
+
+  void inverseBatched(int64_t N, deviceVector<ComplexD*> &Amat)
+  {
+    int32_t batchCount = Amat.size();
+    GRID_ASSERT(batchCount > 0);
+
+#ifdef GRID_HIP
+    GRID_ASSERT( N < 2147483647L );
+    rocblas_int n   = (rocblas_int)N;
+    rocblas_int lda = (rocblas_int)N;
+
+    deviceVector<rocblas_int> ipiv((uint64_t)batchCount*N);
+    deviceVector<rocblas_int> info(batchCount);
+
+    auto st1 = rocsolver_zgetrf_batched(Handle(), n, n,
+					(rocblas_double_complex *const *)&Amat[0], lda,
+					&ipiv[0], (rocblas_stride)N,
+					&info[0], batchCount);
+    GRID_ASSERT(st1 == rocblas_status_success);
+    auto st2 = rocsolver_zgetri_batched(Handle(), n,
+					(rocblas_double_complex *const *)&Amat[0], lda,
+					&ipiv[0], (rocblas_stride)N,
+					&info[0], batchCount);
+    GRID_ASSERT(st2 == rocblas_status_success);
+    accelerator_barrier();
+    std::vector<rocblas_int> info_h(batchCount);
+    acceleratorCopyFromDevice(&info[0],&info_h[0],batchCount*sizeof(rocblas_int));
+    for(int i=0;i<batchCount;i++) GRID_ASSERT(info_h[i]==0);
+#endif
+#ifdef GRID_CUDA
+    GRID_ASSERT( N < 2147483647L );
+    int n = (int)N;
+
+    deviceVector<int> ipiv((uint64_t)batchCount*N);
+    deviceVector<int> info(batchCount);
+
+    auto st1 = cublasZgetrfBatched(Handle(), n,
+				   (cuDoubleComplex **)&Amat[0], n,
+				   &ipiv[0], &info[0], batchCount);
+    GRID_ASSERT(st1 == CUBLAS_STATUS_SUCCESS);
+
+    deviceVector<ComplexD>  work((uint64_t)batchCount*N*N);
+    deviceVector<ComplexD*> Cptr(batchCount);
+    std::vector<ComplexD*>  Cptr_h(batchCount);
+    std::vector<ComplexD*>  Aptr_h(batchCount);
+    for(int i=0;i<batchCount;i++) Cptr_h[i] = &work[(uint64_t)i*N*N];
+    acceleratorCopyToDevice(&Cptr_h[0],&Cptr[0],batchCount*sizeof(ComplexD*));
+    acceleratorCopyFromDevice(&Amat[0],&Aptr_h[0],batchCount*sizeof(ComplexD*));
+
+    auto st2 = cublasZgetriBatched(Handle(), n,
+				   (const cuDoubleComplex *const *)&Amat[0], n,
+				   &ipiv[0],
+				   (cuDoubleComplex **)&Cptr[0], n,
+				   &info[0], batchCount);
+    GRID_ASSERT(st2 == CUBLAS_STATUS_SUCCESS);
+    accelerator_barrier();
+    std::vector<int> info_h(batchCount);
+    acceleratorCopyFromDevice(&info[0],&info_h[0],batchCount*sizeof(int));
+    for(int i=0;i<batchCount;i++) GRID_ASSERT(info_h[i]==0);
+    for(int i=0;i<batchCount;i++)
+      acceleratorCopyDeviceToDevice(Cptr_h[i],Aptr_h[i],(uint64_t)N*N*sizeof(ComplexD));
+#endif
+#ifdef GRID_SYCL
+    sycl::queue *q = Handle();
+    std::vector<ComplexD*> Aptr_h(batchCount);
+    acceleratorCopyFromDevice(&Amat[0],&Aptr_h[0],batchCount*sizeof(ComplexD*));
+
+    int64_t lwf = oneapi::mkl::lapack::getrf_scratchpad_size<std::complex<double> >(*q,N,N,N);
+    int64_t lwi = oneapi::mkl::lapack::getri_scratchpad_size<std::complex<double> >(*q,N,N);
+    deviceVector<ComplexD> scratchf(lwf);
+    deviceVector<ComplexD> scratchi(lwi);
+    deviceVector<int64_t>  ipiv(N);
+    for(int i=0;i<batchCount;i++){
+      oneapi::mkl::lapack::getrf(*q,N,N,(std::complex<double>*)Aptr_h[i],N,&ipiv[0],
+				 (std::complex<double>*)&scratchf[0],lwf);
+      oneapi::mkl::lapack::getri(*q,N,  (std::complex<double>*)Aptr_h[i],N,&ipiv[0],
+				 (std::complex<double>*)&scratchi[0],lwi);
+    }
+    q->wait();
+#endif
+#if !defined(GRID_SYCL) && !defined(GRID_CUDA) && !defined(GRID_HIP)
+    thread_for (p, batchCount, {
+      Eigen::Map<Eigen::MatrixXcd> eA(Amat[p],N,N);
+      Eigen::PartialPivLU<Eigen::MatrixXcd> lu(eA);
+      eA = lu.inverse();
+      });
+#endif
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/blas/MomentumProject.h

@@ -0,0 +1,300 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: MomentumProject.h
+
+    Copyright (C) 2025
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+/* 
+   MultiMomProject
+
+   Import vectors -> nxyz x (ncomponent x nt)
+   Import complex phases -> nmom x nxy
+
+   apply = via (possibly batched) GEMM
+*/
+template<class Field, class ComplexField>
+class MomentumProject
+{
+public:
+
+  typedef typename Field::scalar_type   scalar;
+  typedef typename Field::scalar_object scalar_object;
+
+  GridBase *grid;
+  uint64_t nmom;
+  uint64_t nxyz;
+  uint64_t nt;
+  uint64_t nbtw;
+  uint64_t words;
+
+  deviceVector<scalar> BLAS_V;      // 
+  deviceVector<scalar> BLAS_M;      // 
+  deviceVector<scalar> BLAS_P;      // 
+  
+  MomentumProject(){};
+ ~MomentumProject(){ Deallocate(); };
+  
+  void Deallocate(void)
+  {
+    grid=nullptr;
+    nmom=0;
+    nxyz=0;
+    nt=0;
+    nbtw=0;
+    words=0;
+    BLAS_V.resize(0);
+    BLAS_M.resize(0);
+    BLAS_P.resize(0);
+  }
+  void Allocate(int _nmom,GridBase *_grid)
+  {
+    grid=_grid;
+    Coordinate ldims = grid->LocalDimensions();
+
+    nmom=_nmom;
+    nt   = ldims[grid->Nd()-1];
+    nxyz = grid->lSites()/nt;
+    words = sizeof(scalar_object)/sizeof(scalar);
+    nbtw = nt * words;
+
+    BLAS_V.resize (nxyz * nt * words );
+    BLAS_M.resize (nmom * nxyz       );
+    BLAS_P.resize (nmom * nt * words );
+  }
+  void ImportMomenta(const std::vector <ComplexField> &momenta)
+  {
+    GRID_ASSERT(momenta.size()==nmom);
+    //    might as well just make the momenta here
+    typedef typename Field::vector_object vobj;
+
+    int nd = grid->_ndimension;
+
+    uint64_t sz = BLAS_M.size();
+
+    GRID_ASSERT(momenta.size()==nmom)
+    GRID_ASSERT(momenta[0].Grid()==grid);
+    GRID_ASSERT(sz = nxyz * nmom);
+    
+    Coordinate rdimensions = grid->_rdimensions;
+    Coordinate ldims       = grid->LocalDimensions();
+    int64_t osites         = grid->oSites();
+    Coordinate simd        = grid->_simd_layout;
+    const int Nsimd        = vobj::Nsimd();
+    uint64_t lwords        = words; // local variable for copy in to GPU
+    int64_t Nxyz = nxyz;
+    auto blasData_p  = &BLAS_M[0];
+    for(int m=0;m<momenta.size();m++){
+
+      autoView( Data   , momenta[m], AcceleratorRead);
+      auto Data_p  = &Data[0];
+
+      accelerator_for(xyz,nxyz,1,{
+	  //////////////////////////////////////////
+	  // isite -- map lane within buffer to lane within lattice
+	  ////////////////////////////////////////////
+	    Coordinate lcoor(nd,0);
+	    Lexicographic::CoorFromIndex(lcoor,xyz,ldims);
+	    
+	    Coordinate icoor(nd);
+	    Coordinate ocoor(nd);
+	    for (int d = 0; d < nd; d++) {
+	      icoor[d] = lcoor[d]/rdimensions[d];
+	      ocoor[d] = lcoor[d]%rdimensions[d];
+	    }
+	    int64_t osite;
+	    int64_t isite;
+	    Lexicographic::IndexFromCoor(ocoor,osite,rdimensions);
+	    Lexicographic::IndexFromCoor(icoor,isite,simd);
+	    
+	    // BLAS_M[nmom][slice_vol]
+	    // Fortran Column major BLAS layout is M_xyz,mom
+	    scalar data = extractLane(isite,Data[osite]);
+	    uint64_t idx = xyz+m*Nxyz;
+	    blasData_p[idx] = data;
+	});
+    }
+  }
+  void ImportVector(Field &vec)
+  {
+    typedef typename Field::vector_object vobj;
+
+    int nd = grid->_ndimension;
+
+    uint64_t sz = BLAS_V.size();
+
+    GRID_ASSERT(sz = nxyz * words * nt);
+    
+    Coordinate rdimensions = grid->_rdimensions;
+    Coordinate ldims= grid->LocalDimensions();
+    int64_t osites = grid->oSites();
+    Coordinate simd = grid->_simd_layout;
+    const int Nsimd = vobj::Nsimd();
+    uint64_t lwords= words; // local variable for copy in to GPU
+
+    auto blasData_p  = &BLAS_V[0];
+    autoView( Data   , vec, AcceleratorRead);
+    auto Data_p  = &Data[0];
+
+    int64_t nwords = words;// for capture
+    int64_t Nt     = nt;// for capture
+    
+    accelerator_for(sf,osites,Nsimd,{
+#ifdef GRID_SIMT
+        {
+	  int lane=acceleratorSIMTlane(Nsimd); // buffer lane
+#else
+	  for(int lane=0;lane<Nsimd;lane++) {
+#endif
+	  //////////////////////////////////////////
+	  // isite -- map lane within buffer to lane within lattice
+	  ////////////////////////////////////////////
+	    Coordinate lcoor(nd,0);
+	    Coordinate icoor(nd);
+	    Coordinate ocoor(nd);
+	    
+	    Lexicographic::CoorFromIndex(icoor,lane,simd);
+	    Lexicographic::CoorFromIndex(ocoor,sf,rdimensions);
+
+	  
+	    int64_t l_xyz = 0;
+	    for (int d = 0; d < nd; d++) {
+	      lcoor[d] = rdimensions[d]*icoor[d] + ocoor[d];
+	    }
+	    uint64_t l_t   = lcoor[nd-1];
+
+	    Coordinate xyz_coor = lcoor;
+	    xyz_coor[nd-1] =0;
+	    Lexicographic::IndexFromCoor(xyz_coor,l_xyz,ldims);
+
+	    
+	    scalar_object data = extractLane(lane,Data[sf]);
+	    scalar *data_words = (scalar *) &data;
+	    for(int w = 0 ; w < nwords; w++) {
+	      // BLAS_V[slice_vol][nt][words]
+	      // Fortran Column major BLAS layout is V_(t,w)_xyz
+	      uint64_t idx = w+l_t*nwords + l_xyz * nwords * Nt;
+	      blasData_p[idx] = data_words[w];
+	    }
+#ifdef GRID_SIMT
+	}
+#else
+	}
+#endif
+	});
+  }
+  void ExportMomentumProjection(std::vector<typename Field::scalar_object> &projection)
+  {
+    projection.resize(nmom*nt);
+    acceleratorCopyFromDevice(&BLAS_P[0],(scalar *)&projection[0],BLAS_P.size()*sizeof(scalar));
+    // Could decide on a layout late?
+  }
+
+  // Row major layout "C" order:
+  // BLAS_V[slice_vol][nt][words]
+  // BLAS_M[nmom][slice_vol]
+  // BLAS_P[nmom][nt][words]
+  //
+  // Fortran Column major BLAS layout is V_(w,t)_xyz
+  // Fortran Column major BLAS layout is M_xyz,mom
+  // Fortran Column major BLAS layout is P_(w,t),mom
+  //
+  // Projected
+  //
+  // P = (V * M)_(w,t),mom
+  //
+  void Project(Field &data,std::vector< typename Field::scalar_object > & projected_gdata)
+  {
+    double t_import=0;
+    double t_export=0;
+    double t_gemm  =0;
+    double t_allreduce=0;
+    t_import-=usecond();
+    this->ImportVector(data);
+
+    std::vector< typename Field::scalar_object > projected_planes;
+
+    deviceVector<scalar *> Vd(1);
+    deviceVector<scalar *> Md(1);
+    deviceVector<scalar *> Pd(1);
+
+    scalar * Vh = & BLAS_V[0];
+    scalar * Mh = & BLAS_M[0];
+    scalar * Ph = & BLAS_P[0];
+
+    acceleratorPut(Vd[0],Vh);
+    acceleratorPut(Md[0],Mh);
+    acceleratorPut(Pd[0],Ph);
+    t_import+=usecond();
+
+    GridBLAS BLAS;
+
+    /////////////////////////////////////////
+    // P_im = VMmx . Vxi
+    /////////////////////////////////////////
+    t_gemm-=usecond();
+    BLAS.gemmBatched(GridBLAS_OP_N,GridBLAS_OP_N, 
+    		     words*nt,nmom,nxyz,
+		     scalar(1.0),
+		     Vd,
+		     Md,
+		     scalar(0.0),  // wipe out result
+		     Pd);
+    BLAS.synchronise();
+    t_gemm+=usecond();
+
+    t_export-=usecond();
+    ExportMomentumProjection(projected_planes); // resizes
+    t_export+=usecond();
+
+    /////////////////////////////////
+    // Reduce across MPI ranks
+    /////////////////////////////////
+    int nd = grid->Nd();
+    int gt = grid->GlobalDimensions()[nd-1];
+    int lt = grid->LocalDimensions()[nd-1];
+    projected_gdata.resize(gt*nmom);
+    for(int t=0;t<gt*nmom;t++){ // global Nt array with zeroes for stuff not on this node
+      projected_gdata[t]=Zero();
+    }
+    for(int t=0;t<lt;t++){
+    for(int m=0;m<nmom;m++){
+      int st = grid->LocalStarts()[nd-1];
+      projected_gdata[t+st + gt*m] = projected_planes[t+lt*m];
+    }}
+    t_allreduce-=usecond();
+    grid->GlobalSumVector((scalar *)&projected_gdata[0],gt*nmom*words);
+    t_allreduce+=usecond();
+
+    std::cout << GridLogPerformance<<" MomentumProject t_import  "<<t_import<<"us"<<std::endl;
+    std::cout << GridLogPerformance<<" MomentumProject t_export  "<<t_export<<"us"<<std::endl;
+    std::cout << GridLogPerformance<<" MomentumProject t_gemm    "<<t_gemm<<"us"<<std::endl;
+    std::cout << GridLogPerformance<<" MomentumProject t_reduce  "<<t_allreduce<<"us"<<std::endl;
+
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/deflation/Deflation.h

@@ -69,8 +69,8 @@ public:
   DeflatedGuesser(const std::vector<Field> & _evec, const std::vector<RealD> & _eval, const unsigned int _N)
   : evec(_evec), eval(_eval), N(_N)
   {
-    assert(evec.size()==eval.size());
-    assert(N <= evec.size());
+    GRID_ASSERT(evec.size()==eval.size());
+    GRID_ASSERT(N <= evec.size());
   } 
 
   virtual void operator()(const Field &src,Field &guess) {
@@ -141,11 +141,10 @@ public:
     }
     //postprocessing
     std::cout << GridLogMessage << "Start BlockPromote for loop" << std::endl;
-    for (int j=0;j<Nsrc;j++)
-    {
-    std::cout << GridLogMessage << "BlockProject iter: " << j << std::endl;
-    blockPromote(guess_coarse[j],guess[j],subspace);
-    guess[j].Checkerboard() = src[j].Checkerboard();
+    for (int j=0;j<Nsrc;j++) {
+      std::cout << GridLogMessage << "BlockProject iter: " << j << std::endl;
+      blockPromote(guess_coarse[j],guess[j],subspace);
+      guess[j].Checkerboard() = src[j].Checkerboard();
     }
   };
 

Grid/algorithms/deflation/MultiRHSBlockCGLinalg.h

@@ -160,7 +160,7 @@ public:
     uint64_t words;
 
     nrhs = X.size();
-    assert(X.size()==Y.size());
+    GRID_ASSERT(X.size()==Y.size());
     conformable(X[0],Y[0]);
 
     grid  = X[0].Grid();
@@ -259,7 +259,7 @@ public:
     uint64_t words;
 
     nrhs = X.size();
-    assert(X.size()==Y.size());
+    GRID_ASSERT(X.size()==Y.size());
     conformable(X[0],Y[0]);
 
     grid  = X[0].Grid();
@@ -267,7 +267,7 @@ public:
     vol   = grid->oSites()/rd0;
     words = rd0*sizeof(vector_object)/sizeof(scalar);
     int64_t vw = vol * words;
-    assert(vw == grid->lSites()*sizeof(scalar_object)/sizeof(scalar));
+    GRID_ASSERT(vw == grid->lSites()*sizeof(scalar_object)/sizeof(scalar));
 
     RealD t0 = usecond();
     BLAS_X.resize(nrhs * vw); // cost free if size doesn't change

Grid/algorithms/deflation/MultiRHSBlockProject.h

@@ -61,6 +61,17 @@ public:
   uint64_t coarse_vol;
   uint64_t words;
 
+  ////////////////////////////////////////////////////////////////////////////
+  // Blocking geometry in full local coordinates. Addressing in lSites rather
+  // than (lane,oSite) lets the fine and coarse spaces carry different SIMD
+  // layouts, so an unvectorised coarse space may block a vectorised fine one.
+  ////////////////////////////////////////////////////////////////////////////
+  Coordinate fine_ldimensions;
+  Coordinate coarse_ldimensions;
+  Coordinate block_ldimensions;
+  Coordinate fine_simd;
+  Coordinate coarse_simd;
+
   // Row major layout "C" order:
   // BLAS_V[coarse_vol][nbasis][block_vol][words]
   // BLAS_F[coarse_vol][nrhs][block_vol][words]
@@ -120,35 +131,47 @@ public:
     fine_vol   = fine_grid->lSites();
     coarse_vol = coarse_grid->lSites();
     block_vol = fine_vol/coarse_vol;
-    
+
     words = sizeof(scalar_object)/sizeof(scalar);
 
+    int nd = coarse_grid->_ndimension;
+    GRID_ASSERT(fine_grid->_ndimension == nd);
+
+    fine_ldimensions.resize(nd);
+    coarse_ldimensions.resize(nd);
+    block_ldimensions.resize(nd);
+    fine_simd   = fine_grid->_simd_layout;
+    coarse_simd = coarse_grid->_simd_layout;
+    for(int d=0;d<nd;d++){
+      GRID_ASSERT(fine_grid->_processors[d] == coarse_grid->_processors[d]);
+      fine_ldimensions  [d] = fine_grid->_rdimensions  [d]*fine_grid->_simd_layout  [d];
+      coarse_ldimensions[d] = coarse_grid->_rdimensions[d]*coarse_grid->_simd_layout[d];
+      block_ldimensions [d] = fine_ldimensions[d]/coarse_ldimensions[d];
+      GRID_ASSERT(block_ldimensions[d]*coarse_ldimensions[d] == fine_ldimensions[d]);
+    }
+    GRID_ASSERT(block_vol == fine_vol/coarse_vol);
+
     BLAS_V.resize (fine_vol * words * nbasis );
   }
   void ImportFineGridVectors(std::vector <Field > &vecs, deviceVector<scalar> &blas)
   {
+    GRID_TRACE("ImportFineGridVectors");
     int nvec = vecs.size();
     typedef typename Field::vector_object vobj;
     //    std::cout << GridLogMessage <<" BlockProjector importing "<<nvec<< " fine grid vectors" <<std::endl;
 
-    assert(vecs[0].Grid()==fine_grid);
-
-    subdivides(coarse_grid,fine_grid); // require they map
+    GRID_ASSERT(vecs[0].Grid()==fine_grid);
 
     int _ndimension = coarse_grid->_ndimension;
-    assert(block_vol == fine_grid->oSites() / coarse_grid->oSites());
-    
-    Coordinate  block_r      (_ndimension);
-    for(int d=0 ; d<_ndimension;d++){
-      block_r[d] = fine_grid->_rdimensions[d] / coarse_grid->_rdimensions[d];
-    }
 
     uint64_t sz = blas.size();
 
     acceleratorMemSet(&blas[0],0,blas.size()*sizeof(scalar));
 
     Coordinate fine_rdimensions = fine_grid->_rdimensions;
-    Coordinate coarse_rdimensions = coarse_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate block_l  = block_ldimensions;
+    Coordinate fsimd    = fine_simd;
     int64_t bv= block_vol;
     for(int v=0;v<vecs.size();v++){
 
@@ -162,9 +185,7 @@ public:
 
       // loop over fine sites
       const int Nsimd = vobj::Nsimd();
-      //      std::cout << "sz "<<sz<<std::endl;
-      //      std::cout << "prod "<<Nsimd * coarse_grid->oSites() * block_vol * nvec * words<<std::endl;
-      assert(sz == Nsimd * coarse_grid->oSites() * block_vol * nvec * words);
+      GRID_ASSERT(sz == coarse_vol * block_vol * nvec * words);
       uint64_t lwords= words; // local variable for copy in to GPU
       accelerator_for(sf,osites,Nsimd,{
 #ifdef GRID_SIMT
@@ -175,30 +196,30 @@ public:
 #endif
 	  // One thread per fine site
 	  Coordinate coor_f(_ndimension);
+	  Coordinate coor_l(_ndimension);
 	  Coordinate coor_b(_ndimension);
 	  Coordinate coor_c(_ndimension);
 
-	  // Fine site to fine coor
+	  // Fine (oSite,lane) to full local coor
 	  Lexicographic::CoorFromIndex(coor_f,sf,fine_rdimensions);
+	  Lexicographic::CoorFromIndex(coor_l,lane,fsimd);
+	  for(int d=0;d<_ndimension;d++) coor_f[d] += fine_rdimensions[d]*coor_l[d];
+
+	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_l[d];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_l[d];
 
-	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_r[d];
-	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_r[d];
-	  
 	  int sc;// coarse site
 	  int sb;// block site
-	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_rdimensions);
-	  Lexicographic::IndexFromCoor(coor_b,sb,block_r);
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+	  Lexicographic::IndexFromCoor(coor_b,sb,block_l);
 
           scalar_object data = extractLane(lane,fineData[sf]);
 
-	  // BLAS layout address calculation
-	  // words * block_vol * nbasis x coarse_vol
-	  // coarse oSite x block vole x lanes
-	  int64_t site = (lane*osites + sc*bv)*nvec
-   	               + v*bv
+	  // BLAS_F[coarse_vol][nvec][block_vol][words]
+	  int64_t site = (sc*nvec + v)*bv
 	               + sb;
 
-	  //	  assert(site*lwords<sz);
+	  //	  GRID_ASSERT(site*lwords<sz);
 
 	  scalar_object * ptr = (scalar_object *)&blasData_p[site*lwords];
 
@@ -213,25 +234,333 @@ public:
       //      std::cout << " BlockProjector imported vector"<<v<<std::endl;
     }
   }
+  ////////////////////////////////////////////////////////////////////////////
+  // Import direct from multiRHS fine order, avoiding the unpack to a vector
+  // of single RHS fields.
+  //
+  //   fine mrhs grid  : rhs is dimension 0, undistributed, unvectorised
+  //   Grid  order     : F[fine_vol][nrhs][words]
+  //   BLAS  order     : BLAS_F[lane][coarse_vol][nrhs][block_vol][words]
+  //
+  // The gather of block_vol from fine_vol, and the transpose of nrhs against
+  // block_vol, are the irreducible part: this is not the identity.
+  ////////////////////////////////////////////////////////////////////////////
+  void ImportFineGridMrhsVectors(Field &vec_mrhs, deviceVector<scalar> &blas)
+  {
+    typedef typename Field::vector_object vobj;
+
+    GridBase *fine_mrhs_grid = vec_mrhs.Grid();
+    int _ndimension = coarse_grid->_ndimension;
+
+    GRID_ASSERT(fine_mrhs_grid->_ndimension == _ndimension+1);
+    GRID_ASSERT(fine_mrhs_grid->_simd_layout[0] == 1);
+    GRID_ASSERT(fine_mrhs_grid->_processors[0] == 1);
+    for(int d=0;d<_ndimension;d++){
+      GRID_ASSERT(fine_mrhs_grid->_rdimensions[d+1] == fine_grid->_rdimensions[d]);
+      GRID_ASSERT(fine_mrhs_grid->_simd_layout[d+1] == fine_grid->_simd_layout[d]);
+    }
+    int nvec = fine_mrhs_grid->_rdimensions[0];   // nrhs
+
+    uint64_t sz = blas.size();
+    acceleratorMemSet(&blas[0],0,blas.size()*sizeof(scalar));
+
+    Coordinate fine_mrhs_rdimensions = fine_mrhs_grid->_rdimensions;
+    Coordinate fine_rdimensions      = fine_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate block_l  = block_ldimensions;
+    Coordinate fsimd    = fine_simd;
+    int64_t bv= block_vol;
+
+    autoView( fineData   , vec_mrhs, AcceleratorRead);
+    auto blasData_p  = &blas[0];
+    auto fineData_p  = &fineData[0];
+
+    int64_t osites     = fine_grid->oSites();       // D dimensional
+    int64_t osites_hi  = fine_mrhs_grid->oSites();  // nvec * osites
+
+    const int Nsimd = vobj::Nsimd();
+    GRID_ASSERT(sz == coarse_vol * block_vol * nvec * words);
+    uint64_t lwords= words;
+    int64_t lnvec = nvec;
+
+    accelerator_for(sfr,osites_hi,Nsimd,{
+#ifdef GRID_SIMT
+      {
+	int lane=acceleratorSIMTlane(Nsimd); // buffer lane
+#else
+	for(int lane=0;lane<Nsimd;lane++) {
+#endif
+	  Coordinate coor_hi(_ndimension+1);
+	  Coordinate coor_f(_ndimension);
+	  Coordinate coor_l(_ndimension);
+	  Coordinate coor_b(_ndimension);
+	  Coordinate coor_c(_ndimension);
+
+	  // rhs is dimension 0 of the D+1 grid
+	  Lexicographic::CoorFromIndex(coor_hi,sfr,fine_mrhs_rdimensions);
+	  int v = coor_hi[0];
+	  for(int d=0;d<_ndimension;d++) coor_f[d] = coor_hi[d+1];
+
+	  Lexicographic::CoorFromIndex(coor_l,lane,fsimd);
+	  for(int d=0;d<_ndimension;d++) coor_f[d] += fine_rdimensions[d]*coor_l[d];
+
+	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_l[d];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_l[d];
+
+	  int sc;// coarse site
+	  int sb;// block site
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+	  Lexicographic::IndexFromCoor(coor_b,sb,block_l);
+
+	  scalar_object data = extractLane(lane,fineData[sfr]);
+
+	  int64_t site = (sc*lnvec + v)*bv
+	               + sb;
+
+	  scalar_object * ptr = (scalar_object *)&blasData_p[site*lwords];
+	  *ptr = data;
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
+    });
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Export direct to multiRHS coarse order.
+  //
+  //   Grid  order : C[coarse_vol][nrhs][nbasis]
+  //   BLAS  order : BLAS_C[lane][coarse_vol][nrhs][nbasis]
+  //
+  // At Nsimd()==1 these are the same sequence of addresses.
+  ////////////////////////////////////////////////////////////////////////////
+  template<class vobj>
+  void ExportCoarseGridMrhsVectors(Lattice<vobj> &vec_mrhs, deviceVector<scalar> &blas)
+  {
+    typedef typename vobj::scalar_object coarse_scalar_object;
+
+    GridBase *coarse_mrhs_grid = vec_mrhs.Grid();
+    int _ndimension = coarse_grid->_ndimension;
+
+    GRID_ASSERT(coarse_mrhs_grid->_ndimension == _ndimension+1);
+    GRID_ASSERT(coarse_mrhs_grid->_simd_layout[0] == 1);
+    GRID_ASSERT(coarse_mrhs_grid->_processors[0] == 1);
+    for(int d=0;d<_ndimension;d++){
+      GRID_ASSERT(coarse_mrhs_grid->_rdimensions[d+1] == coarse_grid->_rdimensions[d]);
+      GRID_ASSERT(coarse_mrhs_grid->_simd_layout[d+1] == coarse_grid->_simd_layout[d]);
+    }
+    int nvec = coarse_mrhs_grid->_rdimensions[0];   // nrhs
+
+    Coordinate coarse_mrhs_rdimensions = coarse_mrhs_grid->_rdimensions;
+    Coordinate coarse_rdimensions      = coarse_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate csimd    = coarse_simd;
+
+    autoView( coarseData   , vec_mrhs, AcceleratorWrite);
+    auto blasData_p    = &blas[0];
+    auto coarseData_p  = &coarseData[0];
+
+    int64_t osites    = coarse_grid->oSites();       // D dimensional
+    int64_t osites_hi = coarse_mrhs_grid->oSites();  // nvec * osites
+
+    const int Nsimd = vobj::Nsimd();
+    uint64_t cwords=sizeof(typename vobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(cwords==nbasis);
+    int64_t lnvec = nvec;
+
+    accelerator_for(scr,osites_hi,Nsimd,{
+#ifdef GRID_SIMT
+      {
+	int lane=acceleratorSIMTlane(Nsimd); // buffer lane
+#else
+	for(int lane=0;lane<Nsimd;lane++) {
+#endif
+	  Coordinate coor_hi(_ndimension+1);
+	  Coordinate coor_l(_ndimension);
+	  Coordinate coor_c(_ndimension);
+
+	  Lexicographic::CoorFromIndex(coor_hi,scr,coarse_mrhs_rdimensions);
+	  int v = coor_hi[0];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_hi[d+1];
+
+	  Lexicographic::CoorFromIndex(coor_l,lane,csimd);
+	  for(int d=0;d<_ndimension;d++) coor_c[d] += coarse_rdimensions[d]*coor_l[d];
+
+	  int sc;
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+
+	  int64_t blas_site = (sc*lnvec + v)*cwords;
+	  coarse_scalar_object * ptr = (coarse_scalar_object *)&blasData_p[blas_site];
+	  coarse_scalar_object data = *ptr;
+	  insertLane(lane,coarseData[scr],data);
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
+    });
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Reverse directions: BLAS_F -> fine mrhs field, coarse mrhs field -> BLAS_C
+  ////////////////////////////////////////////////////////////////////////////
+  void ExportFineGridMrhsVectors(Field &vec_mrhs, deviceVector<scalar> &blas)
+  {
+    typedef typename Field::vector_object vobj;
+
+    GridBase *fine_mrhs_grid = vec_mrhs.Grid();
+    int _ndimension = coarse_grid->_ndimension;
+
+    GRID_ASSERT(fine_mrhs_grid->_ndimension == _ndimension+1);
+    GRID_ASSERT(fine_mrhs_grid->_simd_layout[0] == 1);
+    for(int d=0;d<_ndimension;d++){
+      GRID_ASSERT(fine_mrhs_grid->_rdimensions[d+1] == fine_grid->_rdimensions[d]);
+      GRID_ASSERT(fine_mrhs_grid->_simd_layout[d+1] == fine_grid->_simd_layout[d]);
+    }
+    int nvec = fine_mrhs_grid->_rdimensions[0];
+
+    Coordinate fine_mrhs_rdimensions = fine_mrhs_grid->_rdimensions;
+    Coordinate fine_rdimensions      = fine_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate block_l  = block_ldimensions;
+    Coordinate fsimd    = fine_simd;
+    int64_t bv= block_vol;
+
+    autoView( fineData   , vec_mrhs, AcceleratorWrite);
+    auto blasData_p  = &blas[0];
+    auto fineData_p  = &fineData[0];
+
+    int64_t osites     = fine_grid->oSites();
+    int64_t osites_hi  = fine_mrhs_grid->oSites();
+
+    const int Nsimd = vobj::Nsimd();
+    uint64_t lwords= words;
+    int64_t lnvec = nvec;
+
+    accelerator_for(sfr,osites_hi,Nsimd,{
+#ifdef GRID_SIMT
+      {
+	int lane=acceleratorSIMTlane(Nsimd);
+#else
+	for(int lane=0;lane<Nsimd;lane++) {
+#endif
+	  Coordinate coor_hi(_ndimension+1);
+	  Coordinate coor_f(_ndimension);
+	  Coordinate coor_l(_ndimension);
+	  Coordinate coor_b(_ndimension);
+	  Coordinate coor_c(_ndimension);
+
+	  Lexicographic::CoorFromIndex(coor_hi,sfr,fine_mrhs_rdimensions);
+	  int v = coor_hi[0];
+	  for(int d=0;d<_ndimension;d++) coor_f[d] = coor_hi[d+1];
+
+	  Lexicographic::CoorFromIndex(coor_l,lane,fsimd);
+	  for(int d=0;d<_ndimension;d++) coor_f[d] += fine_rdimensions[d]*coor_l[d];
+
+	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_l[d];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_l[d];
+
+	  int sc,sb;
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+	  Lexicographic::IndexFromCoor(coor_b,sb,block_l);
+
+	  int64_t site = (sc*lnvec + v)*bv
+	               + sb;
+
+	  scalar_object * ptr = (scalar_object *)&blasData_p[site*lwords];
+	  scalar_object data = *ptr;
+	  insertLane(lane,fineData[sfr],data);
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
+    });
+  }
+
+  template<class vobj>
+  void ImportCoarseGridMrhsVectors(Lattice<vobj> &vec_mrhs, deviceVector<scalar> &blas)
+  {
+    typedef typename vobj::scalar_object coarse_scalar_object;
+
+    GridBase *coarse_mrhs_grid = vec_mrhs.Grid();
+    int _ndimension = coarse_grid->_ndimension;
+
+    GRID_ASSERT(coarse_mrhs_grid->_ndimension == _ndimension+1);
+    GRID_ASSERT(coarse_mrhs_grid->_simd_layout[0] == 1);
+    for(int d=0;d<_ndimension;d++){
+      GRID_ASSERT(coarse_mrhs_grid->_rdimensions[d+1] == coarse_grid->_rdimensions[d]);
+      GRID_ASSERT(coarse_mrhs_grid->_simd_layout[d+1] == coarse_grid->_simd_layout[d]);
+    }
+    int nvec = coarse_mrhs_grid->_rdimensions[0];
+
+    Coordinate coarse_mrhs_rdimensions = coarse_mrhs_grid->_rdimensions;
+    Coordinate coarse_rdimensions      = coarse_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate csimd    = coarse_simd;
+
+    autoView( coarseData   , vec_mrhs, AcceleratorRead);
+    auto blasData_p    = &blas[0];
+    auto coarseData_p  = &coarseData[0];
+
+    int64_t osites    = coarse_grid->oSites();
+    int64_t osites_hi = coarse_mrhs_grid->oSites();
+
+    const int Nsimd = vobj::Nsimd();
+    uint64_t cwords=sizeof(typename vobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(cwords==nbasis);
+    int64_t lnvec = nvec;
+
+    accelerator_for(scr,osites_hi,Nsimd,{
+#ifdef GRID_SIMT
+      {
+	int lane=acceleratorSIMTlane(Nsimd);
+#else
+	for(int lane=0;lane<Nsimd;lane++) {
+#endif
+	  Coordinate coor_hi(_ndimension+1);
+	  Coordinate coor_l(_ndimension);
+	  Coordinate coor_c(_ndimension);
+
+	  Lexicographic::CoorFromIndex(coor_hi,scr,coarse_mrhs_rdimensions);
+	  int v = coor_hi[0];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_hi[d+1];
+
+	  Lexicographic::CoorFromIndex(coor_l,lane,csimd);
+	  for(int d=0;d<_ndimension;d++) coor_c[d] += coarse_rdimensions[d]*coor_l[d];
+
+	  int sc;
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+
+	  coarse_scalar_object data = extractLane(lane,coarseData[scr]);
+
+	  int64_t blas_site = (sc*lnvec + v)*cwords;
+	  coarse_scalar_object * ptr = (coarse_scalar_object *)&blasData_p[blas_site];
+	  *ptr = data;
+#ifdef GRID_SIMT
+      }
+#else
+      }
+#endif
+    });
+  }
+
   void ExportFineGridVectors(std::vector <Field> &vecs, deviceVector<scalar> &blas)
   {
+    GRID_TRACE("ExportFineGridVectors");
     typedef typename Field::vector_object vobj;
 
     int nvec = vecs.size();
 
-    assert(vecs[0].Grid()==fine_grid);
-
-    subdivides(coarse_grid,fine_grid); // require they map
+    GRID_ASSERT(vecs[0].Grid()==fine_grid);
 
     int _ndimension = coarse_grid->_ndimension;
-    assert(block_vol == fine_grid->oSites() / coarse_grid->oSites());
-    
-    Coordinate  block_r      (_ndimension);
-    for(int d=0 ; d<_ndimension;d++){
-      block_r[d] = fine_grid->_rdimensions[d] / coarse_grid->_rdimensions[d];
-    }
+
     Coordinate fine_rdimensions = fine_grid->_rdimensions;
-    Coordinate coarse_rdimensions = coarse_grid->_rdimensions;
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate block_l  = block_ldimensions;
+    Coordinate fsimd    = fine_simd;
 
     //    std::cout << " export fine Blas norm "<<blasNorm2(blas)<<std::endl;
 
@@ -259,23 +588,24 @@ public:
 #endif
 	  // One thread per fine site
 	  Coordinate coor_f(_ndimension);
+	  Coordinate coor_l(_ndimension);
 	  Coordinate coor_b(_ndimension);
 	  Coordinate coor_c(_ndimension);
 
 	  Lexicographic::CoorFromIndex(coor_f,sf,fine_rdimensions);
+	  Lexicographic::CoorFromIndex(coor_l,lane,fsimd);
+	  for(int d=0;d<_ndimension;d++) coor_f[d] += fine_rdimensions[d]*coor_l[d];
+
+	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_l[d];
+	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_l[d];
 
-	  for(int d=0;d<_ndimension;d++) coor_b[d] = coor_f[d]%block_r[d];
-	  for(int d=0;d<_ndimension;d++) coor_c[d] = coor_f[d]/block_r[d];
-	  
 	  int sc;
 	  int sb;
-	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_rdimensions);
-	  Lexicographic::IndexFromCoor(coor_b,sb,block_r);
+	  Lexicographic::IndexFromCoor(coor_c,sc,coarse_l);
+	  Lexicographic::IndexFromCoor(coor_b,sb,block_l);
 
-	  // BLAS layout address calculation
-	  // words * block_vol * nbasis x coarse_vol 	  
-	  int64_t site = (lane*osites + sc*bv)*nvec
-   	               + v*bv
+	  // BLAS_F[coarse_vol][nvec][block_vol][words]
+	  int64_t site = (sc*nvec + v)*bv
 	               + sb;
 
 	  scalar_object * ptr = (scalar_object *)&blasData_p[site*lwords];
@@ -294,19 +624,22 @@ public:
   template<class vobj>
   void ImportCoarseGridVectors(std::vector <Lattice<vobj> > &vecs, deviceVector<scalar> &blas)
   {
+    GRID_TRACE("ImportCoarseGridVectors");
     int nvec = vecs.size();
     typedef typename vobj::scalar_object coarse_scalar_object;
 
     //    std::cout << " BlockProjector importing "<<nvec<< " coarse grid vectors" <<std::endl;
 
-    assert(vecs[0].Grid()==coarse_grid);
+    GRID_ASSERT(vecs[0].Grid()==coarse_grid);
 
     int _ndimension = coarse_grid->_ndimension;
 
     uint64_t sz = blas.size();
 
     Coordinate coarse_rdimensions = coarse_grid->_rdimensions;
-    
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate csimd    = coarse_simd;
+
     for(int v=0;v<vecs.size();v++){
 
       //      std::cout << " BlockProjector importing coarse vector"<<v<<" "<<norm2(vecs[v])<<std::endl;
@@ -320,7 +653,7 @@ public:
       // loop over fine sites
       const int Nsimd = vobj::Nsimd();
       uint64_t cwords=sizeof(typename vobj::scalar_object)/sizeof(scalar);
-      assert(cwords==nbasis);
+      GRID_ASSERT(cwords==nbasis);
       
       accelerator_for(sc,osites,Nsimd,{
 #ifdef GRID_SIMT
@@ -330,8 +663,17 @@ public:
 	  for(int lane=0;lane<Nsimd;lane++) {
 #endif
            // C_br per site
-	    int64_t blas_site = (lane*osites + sc)*nvec*cwords + v*cwords;
-	    
+	    Coordinate coor_c(_ndimension);
+	    Coordinate coor_l(_ndimension);
+	    Lexicographic::CoorFromIndex(coor_c,sc,coarse_rdimensions);
+	    Lexicographic::CoorFromIndex(coor_l,lane,csimd);
+	    for(int d=0;d<_ndimension;d++) coor_c[d] += coarse_rdimensions[d]*coor_l[d];
+
+	    int scl;
+	    Lexicographic::IndexFromCoor(coor_c,scl,coarse_l);
+
+	    int64_t blas_site = (scl*nvec + v)*cwords;
+
 	    coarse_scalar_object data = extractLane(lane,coarseData[sc]);
 
 	    coarse_scalar_object * ptr = (coarse_scalar_object *)&blasData_p[blas_site];
@@ -349,18 +691,21 @@ public:
   template<class vobj>
   void ExportCoarseGridVectors(std::vector <Lattice<vobj> > &vecs, deviceVector<scalar> &blas)
   {
+    GRID_TRACE("ExportCoarseGridVectors");
     int nvec = vecs.size();
     typedef typename vobj::scalar_object coarse_scalar_object;
     //    std::cout << GridLogMessage<<" BlockProjector exporting "<<nvec<< " coarse grid vectors" <<std::endl;
 
-    assert(vecs[0].Grid()==coarse_grid);
+    GRID_ASSERT(vecs[0].Grid()==coarse_grid);
 
     int _ndimension = coarse_grid->_ndimension;
     
     uint64_t sz = blas.size();
 
     Coordinate coarse_rdimensions = coarse_grid->_rdimensions;
-    
+    Coordinate coarse_l = coarse_ldimensions;
+    Coordinate csimd    = coarse_simd;
+
     //    std::cout << " export coarsee Blas norm "<<blasNorm2(blas)<<std::endl;
     for(int v=0;v<vecs.size();v++){
 
@@ -375,7 +720,7 @@ public:
       // loop over fine sites
       const int Nsimd = vobj::Nsimd();
       uint64_t cwords=sizeof(typename vobj::scalar_object)/sizeof(scalar);
-      assert(cwords==nbasis);
+      GRID_ASSERT(cwords==nbasis);
       
       accelerator_for(sc,osites,Nsimd,{
 	  // Wrap in a macro "FOR_ALL_LANES(lane,{ ... });
@@ -385,7 +730,16 @@ public:
 #else
 	  for(int lane=0;lane<Nsimd;lane++) {
 #endif
-	    int64_t blas_site = (lane*osites + sc)*nvec*cwords + v*cwords;
+	    Coordinate coor_c(_ndimension);
+	    Coordinate coor_l(_ndimension);
+	    Lexicographic::CoorFromIndex(coor_c,sc,coarse_rdimensions);
+	    Lexicographic::CoorFromIndex(coor_l,lane,csimd);
+	    for(int d=0;d<_ndimension;d++) coor_c[d] += coarse_rdimensions[d]*coor_l[d];
+
+	    int scl;
+	    Lexicographic::IndexFromCoor(coor_c,scl,coarse_l);
+
+	    int64_t blas_site = (scl*nvec + v)*cwords;
 	    coarse_scalar_object * ptr = (coarse_scalar_object *)&blasData_p[blas_site];
 	    coarse_scalar_object data = *ptr;
 	    insertLane(lane,coarseData[sc],data);
@@ -406,10 +760,11 @@ public:
   template<class cobj>
   void blockProject(std::vector<Field> &fine,std::vector< Lattice<cobj> > & coarse)
   {
+    GRID_TRACE("BlockProject");
     int nrhs=fine.size();
     int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
     //    std::cout << "blockProject nbasis " <<nbasis<<" " << _nbasis<<std::endl;
-    assert(nbasis==_nbasis);
+    GRID_ASSERT(nbasis==_nbasis);
     
     BLAS_F.resize (fine_vol * words * nrhs );
     BLAS_C.resize (coarse_vol * nbasis * nrhs );
@@ -425,17 +780,20 @@ public:
     deviceVector<scalar *> Cd(coarse_vol);
 
     //    std::cout << "BlockProject pointers"<<std::endl;
-    for(int c=0;c<coarse_vol;c++){
-      // BLAS_V[coarse_vol][nbasis][block_vol][words]
-      // BLAS_F[coarse_vol][nrhs][block_vol][words]
-      // BLAS_C[coarse_vol][nrhs][nbasis]
-      scalar * Vh = & BLAS_V[c*nbasis*block_vol*words];
-      scalar * Fh = & BLAS_F[c*nrhs*block_vol*words];
-      scalar * Ch = & BLAS_C[c*nrhs*nbasis];
-
-      acceleratorPut(Vd[c],Vh);
-      acceleratorPut(Fd[c],Fh);
-      acceleratorPut(Cd[c],Ch);
+    // ONE bulk transfer per table.  acceleratorPut is a *synchronous* 8-byte
+    // hipMemcpy, so the elementwise form emitted 3*coarse_vol of them per call:
+    // a traced run showed 272k hipMemcpy calls costing 4.5 s of API time to move
+    // 0.7 s worth of bytes.  Same fix as BatchedBlas.h's staging rewrite.
+    if ( coarse_vol ) {
+      std::vector<scalar *> hVd(coarse_vol), hFd(coarse_vol), hCd(coarse_vol);
+      for(int c=0;c<coarse_vol;c++){
+	hVd[c] = & BLAS_V[c*nbasis*block_vol*words];
+	hFd[c] = & BLAS_F[c*nrhs*block_vol*words];
+	hCd[c] = & BLAS_C[c*nrhs*nbasis];
+      }
+      acceleratorCopyToDevice(&hVd[0],&Vd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hFd[0],&Fd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hCd[0],&Cd[0],coarse_vol*sizeof(scalar *));
     }
 
     GridBLAS BLAS;
@@ -462,9 +820,10 @@ public:
   template<class cobj>
   void blockPromote(std::vector<Field> &fine,std::vector<Lattice<cobj> > & coarse)
   {
+    GRID_TRACE("BlockPromote");
     int nrhs=fine.size();
     int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
-    assert(nbasis==_nbasis);
+    GRID_ASSERT(nbasis==_nbasis);
     
     BLAS_F.resize (fine_vol * words * nrhs );
     BLAS_C.resize (coarse_vol * nbasis * nrhs );
@@ -477,16 +836,20 @@ public:
     deviceVector<scalar *> Fd(coarse_vol);
     deviceVector<scalar *> Cd(coarse_vol);
 
-    for(int c=0;c<coarse_vol;c++){
-      // BLAS_V[coarse_vol][nbasis][block_vol][words]
-      // BLAS_F[coarse_vol][nrhs][block_vol][words]
-      // BLAS_C[coarse_vol][nrhs][nbasis]
-      scalar * Vh = & BLAS_V[c*nbasis*block_vol*words];
-      scalar * Fh = & BLAS_F[c*nrhs*block_vol*words];
-      scalar * Ch = & BLAS_C[c*nrhs*nbasis];
-      acceleratorPut(Vd[c],Vh);
-      acceleratorPut(Fd[c],Fh);
-      acceleratorPut(Cd[c],Ch);
+    // ONE bulk transfer per table.  acceleratorPut is a *synchronous* 8-byte
+    // hipMemcpy, so the elementwise form emitted 3*coarse_vol of them per call:
+    // a traced run showed 272k hipMemcpy calls costing 4.5 s of API time to move
+    // 0.7 s worth of bytes.  Same fix as BatchedBlas.h's staging rewrite.
+    if ( coarse_vol ) {
+      std::vector<scalar *> hVd(coarse_vol), hFd(coarse_vol), hCd(coarse_vol);
+      for(int c=0;c<coarse_vol;c++){
+	hVd[c] = & BLAS_V[c*nbasis*block_vol*words];
+	hFd[c] = & BLAS_F[c*nrhs*block_vol*words];
+	hCd[c] = & BLAS_C[c*nrhs*nbasis];
+      }
+      acceleratorCopyToDevice(&hVd[0],&Vd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hFd[0],&Fd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hCd[0],&Cd[0],coarse_vol*sizeof(scalar *));
     }
 
     /////////////////////////////////////////
@@ -508,6 +871,183 @@ public:
     ExportFineGridVectors(fine, BLAS_F);
     //    std::cout << " exported "<<std::endl;
   }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // multiRHS ordered interfaces.  The GEMM is identical; only the import and
+  // export differ, so those are the whole of the layout question.
+  ////////////////////////////////////////////////////////////////////////////
+  template<class cobj>
+  void blockProject(Field &fine_mrhs,Lattice<cobj> &coarse_mrhs)
+  {
+    GRID_TRACE("BlockProjectMrhs");
+    int nrhs = fine_mrhs.Grid()->_rdimensions[0];
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse_mrhs.Grid()->_rdimensions[0]==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportFineGridMrhsVectors(fine_mrhs,BLAS_F);
+    ProjectBLAS(nrhs);
+    ExportCoarseGridMrhsVectors(coarse_mrhs,BLAS_C);
+  }
+
+  template<class cobj>
+  void blockPromote(Field &fine_mrhs,Lattice<cobj> &coarse_mrhs)
+  {
+    GRID_TRACE("BlockPromoteMrhs");
+    int nrhs = fine_mrhs.Grid()->_rdimensions[0];
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse_mrhs.Grid()->_rdimensions[0]==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportCoarseGridMrhsVectors(coarse_mrhs,BLAS_C);
+    PromoteBLAS(nrhs);
+    ExportFineGridMrhsVectors(fine_mrhs,BLAS_F);
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Mixed orderings. A single RHS fine operator produces a vector of fine
+  // fields with no packing; the coarse side is still wanted in mrhs order.
+  ////////////////////////////////////////////////////////////////////////////
+  template<class cobj>
+  void blockProject(std::vector<Field> &fine,Lattice<cobj> &coarse_mrhs)
+  {
+    GRID_TRACE("BlockProjectMixed");
+    int nrhs = fine.size();
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse_mrhs.Grid()->_rdimensions[0]==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportFineGridVectors(fine,BLAS_F);
+    ProjectBLAS(nrhs);
+    ExportCoarseGridMrhsVectors(coarse_mrhs,BLAS_C);
+  }
+
+  template<class cobj>
+  void blockProject(Field &fine_mrhs,std::vector< Lattice<cobj> > &coarse)
+  {
+    GRID_TRACE("BlockProjectMixed");
+    int nrhs = fine_mrhs.Grid()->_rdimensions[0];
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse.size()==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportFineGridMrhsVectors(fine_mrhs,BLAS_F);
+    ProjectBLAS(nrhs);
+    ExportCoarseGridVectors(coarse,BLAS_C);
+  }
+
+  template<class cobj>
+  void blockPromote(std::vector<Field> &fine,Lattice<cobj> &coarse_mrhs)
+  {
+    GRID_TRACE("BlockPromoteMixed");
+    int nrhs = fine.size();
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse_mrhs.Grid()->_rdimensions[0]==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportCoarseGridMrhsVectors(coarse_mrhs,BLAS_C);
+    PromoteBLAS(nrhs);
+    ExportFineGridVectors(fine,BLAS_F);
+  }
+
+  template<class cobj>
+  void blockPromote(Field &fine_mrhs,std::vector< Lattice<cobj> > &coarse)
+  {
+    GRID_TRACE("BlockPromoteMixed");
+    int nrhs = fine_mrhs.Grid()->_rdimensions[0];
+    int _nbasis = sizeof(typename cobj::scalar_object)/sizeof(scalar);
+    GRID_ASSERT(nbasis==_nbasis);
+    GRID_ASSERT(coarse.size()==nrhs);
+
+    BLAS_F.resize (fine_vol * words * nrhs );
+    BLAS_C.resize (coarse_vol * nbasis * nrhs );
+
+    ImportCoarseGridVectors(coarse,BLAS_C);
+    PromoteBLAS(nrhs);
+    ExportFineGridMrhsVectors(fine_mrhs,BLAS_F);
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Pointer tables and the batched GEMM, shared by both orderings
+  ////////////////////////////////////////////////////////////////////////////
+  void BLASPointers(int nrhs,
+		    deviceVector<scalar *> &Vd,
+		    deviceVector<scalar *> &Fd,
+		    deviceVector<scalar *> &Cd)
+  {
+    // ONE bulk transfer per table.  acceleratorPut is a *synchronous* 8-byte
+    // hipMemcpy, so the elementwise form emitted 3*coarse_vol of them per call:
+    // a traced run showed 272k hipMemcpy calls costing 4.5 s of API time to move
+    // 0.7 s worth of bytes.  Same fix as BatchedBlas.h's staging rewrite.
+    if ( coarse_vol ) {
+      std::vector<scalar *> hVd(coarse_vol), hFd(coarse_vol), hCd(coarse_vol);
+      for(int c=0;c<coarse_vol;c++){
+	hVd[c] = & BLAS_V[c*nbasis*block_vol*words];
+	hFd[c] = & BLAS_F[c*nrhs*block_vol*words];
+	hCd[c] = & BLAS_C[c*nrhs*nbasis];
+      }
+      acceleratorCopyToDevice(&hVd[0],&Vd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hFd[0],&Fd[0],coarse_vol*sizeof(scalar *));
+      acceleratorCopyToDevice(&hCd[0],&Cd[0],coarse_vol*sizeof(scalar *));
+    }
+  }
+
+  // C_br = V^dag F
+  void ProjectBLAS(int nrhs)
+  {
+    GRID_TRACE("ProjectBLAS");
+    deviceVector<scalar *> Vd(coarse_vol);
+    deviceVector<scalar *> Fd(coarse_vol);
+    deviceVector<scalar *> Cd(coarse_vol);
+    BLASPointers(nrhs,Vd,Fd,Cd);
+
+    GridBLAS BLAS;
+    int64_t vw = block_vol * words;
+    BLAS.gemmBatched(GridBLAS_OP_C,GridBLAS_OP_N,
+		     nbasis,nrhs,vw,
+		     scalar(1.0),
+		     Vd,
+		     Fd,
+		     scalar(0.0),  // wipe out C
+		     Cd);
+    BLAS.synchronise();
+  }
+
+  // F_xr = Vxb Cbr
+  void PromoteBLAS(int nrhs)
+  {
+    GRID_TRACE("PromoteBLAS");
+    deviceVector<scalar *> Vd(coarse_vol);
+    deviceVector<scalar *> Fd(coarse_vol);
+    deviceVector<scalar *> Cd(coarse_vol);
+    BLASPointers(nrhs,Vd,Fd,Cd);
+
+    GridBLAS BLAS;
+    int64_t vw = block_vol * words;
+    BLAS.gemmBatched(GridBLAS_OP_N,GridBLAS_OP_N,
+		     vw,nrhs,nbasis,
+		     scalar(1.0),
+		     Vd,
+		     Cd,
+		     scalar(0.0),  // wipe out F
+		     Fd);
+    BLAS.synchronise();
+  }
 };
 
 NAMESPACE_END(Grid);

Grid/algorithms/deflation/MultiRHSDeflation.h

@@ -98,7 +98,7 @@ public:
   void ImportEigenVector(Field &evec,RealD &_eval, int ev)
   {
     //    std::cout << " ev " <<ev<<" eval "<<_eval<< std::endl;
-    assert(ev<eval.size());
+    GRID_ASSERT(ev<eval.size());
     eval[ev] = _eval;
 
     int64_t offset = ev*vol*words;
@@ -113,7 +113,7 @@ public:
   // Could use to import a batch of eigenvectors
   void ImportEigenBasis(std::vector<Field> &evec,std::vector<RealD> &_eval, int _ev0, int _nev)
   {
-    assert(_ev0+_nev<=evec.size());
+    GRID_ASSERT(_ev0+_nev<=evec.size());
 
     Allocate(_nev,evec[0].Grid());
     
@@ -126,8 +126,8 @@ public:
   void DeflateSources(std::vector<Field> &source,std::vector<Field> & guess)
   {
     int nrhs = source.size();
-    assert(source.size()==guess.size());
-    assert(grid == guess[0].Grid());
+    GRID_ASSERT(source.size()==guess.size());
+    GRID_ASSERT(grid == guess[0].Grid());
     conformable(guess[0],source[0]);
 
     int64_t vw = vol * words;
@@ -189,7 +189,7 @@ public:
 		     Cd);
     BLAS.synchronise();
 
-    assert(BLAS_C.size()==nev*nrhs);
+    GRID_ASSERT(BLAS_C.size()==nev*nrhs);
 
     std::vector<scalar> HOST_C(BLAS_C.size());      // nrhs . nev -- the coefficients 
     acceleratorCopyFromDevice(&BLAS_C[0],&HOST_C[0],BLAS_C.size()*sizeof(scalar));

Grid/algorithms/iterative/AdefGeneric.h

@@ -53,7 +53,22 @@ class TwoLevelCG : public LinearFunction<Field>
   // Fine operator, Smoother, CoarseSolver
   LinearOperatorBase<Field>   &_FineLinop;
   LinearFunction<Field>   &_Smoother;
-  
+
+  GridStopWatch ProjectTimer;
+  GridStopWatch PromoteTimer;
+  GridStopWatch CoarseTimer;
+  GridStopWatch SmoothTimer;
+  GridStopWatch MatrixTimer;
+  GridStopWatch M3Timer;
+  GridStopWatch LinalgTimer;
+
+  int64_t M3Calls;
+  int64_t SmoothCalls;
+  int64_t MatrixCalls;
+  int64_t ProjectCalls;
+  int64_t CoarseCalls;
+  int64_t PromoteCalls;
+
   // more most opertor functions
   TwoLevelCG(RealD tol,
 	     Integer maxit,
@@ -103,12 +118,20 @@ class TwoLevelCG : public LinearFunction<Field>
     RealD tn;
     
     GridStopWatch HDCGTimer;
+    ProjectTimer.Reset();
+    PromoteTimer.Reset();
+    CoarseTimer.Reset();
+    SmoothTimer.Reset();
+    MatrixTimer.Reset();
+    M3Timer.Reset();
+    LinalgTimer.Reset();
+    M3Calls = SmoothCalls = MatrixCalls = ProjectCalls = CoarseCalls = PromoteCalls = 0;
     HDCGTimer.Start();
     //////////////////////////
     // x0 = Vstart -- possibly modify guess
     //////////////////////////
     Vstart(x,src);
-    
+
     // r0 = b -A x0
     _FineLinop.HermOp(x,mmp[0]);
     axpy (r, -1.0,mmp[0], src);    // Recomputes r=src-Ax0
@@ -145,33 +168,40 @@ class TwoLevelCG : public LinearFunction<Field>
       int peri_kp = (k+1) % mmax;
 
       rtz=rtzp;
+      M3Timer.Start();
       d= PcgM3(p[peri_k],mmp[peri_k]);
+      M3Timer.Stop();
+      M3Calls++;
       a = rtz/d;
-    
+
       // Memorise this
       pAp[peri_k] = d;
-      
+
+      LinalgTimer.Start();
       axpy(x,a,p[peri_k],x);
       RealD rn = axpy_norm(r,-a,mmp[peri_k],r);
+      LinalgTimer.Stop();
 
       // Compute z = M x
       PcgM1(r,z);
-      
+
       {
 	RealD n1,n2;
 	n1=norm2(r);
 	n2=norm2(z);
 	std::cout << GridLogMessage<<"HDCG::fPcg iteration "<<k<<" : vector r,z "<<n1<<" "<<n2<<"\n";
       }
+      LinalgTimer.Start();
       rtzp =real(innerProduct(r,z));
+      LinalgTimer.Stop();
       std::cout << GridLogMessage<<"HDCG::fPcg iteration "<<k<<" : inner rtzp "<<rtzp<<"\n";
 
       //    PcgM2(z,p[0]);
       PcgM2(z,mu); // ADEF-2 this is identity. Axpy possible to eliminate
-      
+
       p[peri_kp]=mu;
 
-      // Standard search direction  p -> z + b p    
+      // Standard search direction  p -> z + b p
       b = (rtzp)/rtz;
       
       int northog;
@@ -202,8 +232,25 @@ class TwoLevelCG : public LinearFunction<Field>
       if ( rn <= rsq ) { 
 
 	HDCGTimer.Stop();
-	std::cout<<GridLogMessage<<"HDCG: fPcg converged in "<<k<<" iterations and "<<HDCGTimer.Elapsed()<<std::endl;;
-	
+	std::cout<<GridLogMessage<<"HDCG: fPcg converged in "<<k<<" iterations and "<<HDCGTimer.Elapsed()<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg breakdown"<<std::endl;
+	auto mspc = [](GridStopWatch &sw, int64_t n) -> double {
+	  return (n > 0) ? sw.useconds() * 1e-3 / n : 0.0;
+	};
+	std::cout<<GridLogMessage<<"HDCG: fPcg M3 (fine MVM)  "<<M3Timer.Elapsed()
+		 <<"  "<<M3Calls<<" calls  "<<mspc(M3Timer,M3Calls)<<" ms/call"<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg linalg         "<<LinalgTimer.Elapsed()<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg smoother       "<<SmoothTimer.Elapsed()
+		 <<"  "<<SmoothCalls<<" calls  "<<mspc(SmoothTimer,SmoothCalls)<<" ms/call"<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg matrix (in M1) "<<MatrixTimer.Elapsed()
+		 <<"  "<<MatrixCalls<<" calls  "<<mspc(MatrixTimer,MatrixCalls)<<" ms/call"<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg project        "<<ProjectTimer.Elapsed()
+		 <<"  "<<ProjectCalls<<" calls  "<<mspc(ProjectTimer,ProjectCalls)<<" ms/call"<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg coarse         "<<CoarseTimer.Elapsed()
+		 <<"  "<<CoarseCalls<<" calls  "<<mspc(CoarseTimer,CoarseCalls)<<" ms/call"<<std::endl;
+	std::cout<<GridLogMessage<<"HDCG: fPcg promote        "<<PromoteTimer.Elapsed()
+		 <<"  "<<PromoteCalls<<" calls  "<<mspc(PromoteTimer,PromoteCalls)<<" ms/call"<<std::endl;
+
 	_FineLinop.HermOp(x,mmp[0]);			  
 	axpy(tmp,-1.0,src,mmp[0]);
 	
@@ -270,7 +317,7 @@ class TwoLevelCG : public LinearFunction<Field>
     std::vector<RealD> src_nrm(nrhs);
     for(int rhs=0;rhs<nrhs;rhs++) {
       src_nrm[rhs]=norm2(src[rhs]);
-      assert(src_nrm[rhs]!=0.0);
+      GRID_ASSERT(src_nrm[rhs]!=0.0);
     }
     std::vector<RealD> tn(nrhs);
 
@@ -475,35 +522,29 @@ class TwoLevelADEF2 : public TwoLevelCG<Field>
     CoarseField PleftProj(this->coarsegrid);
     CoarseField PleftMss_proj(this->coarsegrid);
 
-    GridStopWatch SmootherTimer;
-    GridStopWatch MatrixTimer;
-    SmootherTimer.Start();
+    this->SmoothTimer.Start();
     this->_Smoother(in,Min);
-    SmootherTimer.Stop();
+    this->SmoothTimer.Stop();
+    this->SmoothCalls++;
 
-    MatrixTimer.Start();
+    this->MatrixTimer.Start();
     this->_FineLinop.HermOp(Min,out);
-    MatrixTimer.Stop();
+    this->MatrixTimer.Stop();
+    this->MatrixCalls++;
     axpy(tmp,-1.0,out,in);          // tmp  = in - A Min
 
-    GridStopWatch ProjTimer;
-    GridStopWatch CoarseTimer;
-    GridStopWatch PromTimer;
-    ProjTimer.Start();
-    this->_Aggregates.ProjectToSubspace(PleftProj,tmp);     
-    ProjTimer.Stop();
-    CoarseTimer.Start();
+    this->ProjectTimer.Start();
+    this->_Aggregates.ProjectToSubspace(PleftProj,tmp);
+    this->ProjectTimer.Stop();
+    this->ProjectCalls++;
+    this->CoarseTimer.Start();
     this->_CoarseSolver(PleftProj,PleftMss_proj); // Ass^{-1} [in - A Min]_s
-    CoarseTimer.Stop();
-    PromTimer.Start();
-    this->_Aggregates.PromoteFromSubspace(PleftMss_proj,tmp);// tmp = Q[in - A Min]  
-    PromTimer.Stop();
-    std::cout << GridLogPerformance << "PcgM1 breakdown "<<std::endl;
-    std::cout << GridLogPerformance << "\tSmoother   " << SmootherTimer.Elapsed() <<std::endl;
-    std::cout << GridLogPerformance << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
-    std::cout << GridLogPerformance << "\tProj       " << ProjTimer.Elapsed() <<std::endl;
-    std::cout << GridLogPerformance << "\tCoarse     " << CoarseTimer.Elapsed() <<std::endl;
-    std::cout << GridLogPerformance << "\tProm       " << PromTimer.Elapsed() <<std::endl;
+    this->CoarseTimer.Stop();
+    this->CoarseCalls++;
+    this->PromoteTimer.Start();
+    this->_Aggregates.PromoteFromSubspace(PleftMss_proj,tmp);// tmp = Q[in - A Min]
+    this->PromoteTimer.Stop();
+    this->PromoteCalls++;
 
     axpy(out,1.0,Min,tmp); // Min+tmp
   }

Grid/algorithms/iterative/AdefMrhs.h

@@ -92,8 +92,8 @@ class TwoLevelCGmrhs
   // Vector case
   virtual void operator() (std::vector<Field> &src, std::vector<Field> &x)
   {
-    //    SolveSingleSystem(src,x);
-    SolvePrecBlockCG(src,x);
+    SolveSingleSystem(src,x);
+	// SolvePrecBlockCG(src,x);
   }
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -161,7 +161,7 @@ class TwoLevelCGmrhs
     ////////////////////////////////////////////
     std::vector<RealD> ssq(nrhs);
     for(int rhs=0;rhs<nrhs;rhs++){
-      ssq[rhs]=norm2(src[rhs]); assert(ssq[rhs]!=0.0);
+      ssq[rhs]=norm2(src[rhs]); GRID_ASSERT(ssq[rhs]!=0.0);
     }      
 
     ///////////////////////////
@@ -382,7 +382,7 @@ class TwoLevelCGmrhs
     }
     HDCGTimer.Stop();
     std::cout<<GridLogMessage<<"HDCG: PrecBlockCGrQ not converged "<<HDCGTimer.Elapsed()<<std::endl;
-    assert(0);
+    GRID_ASSERT(0);
   }
 
   virtual void SolveSingleSystem (std::vector<Field> &src, std::vector<Field> &x)
@@ -415,7 +415,7 @@ class TwoLevelCGmrhs
     std::vector<RealD> src_nrm(nrhs);
     for(int rhs=0;rhs<nrhs;rhs++) {
       src_nrm[rhs]=norm2(src[rhs]);
-      assert(src_nrm[rhs]!=0.0);
+      GRID_ASSERT(src_nrm[rhs]!=0.0);
     }
     std::vector<RealD> tn(nrhs);
 

Grid/algorithms/iterative/BiCGSTAB.h

@@ -47,7 +47,7 @@ class BiCGSTAB : public OperatorFunction<Field>
   public:
     using OperatorFunction<Field>::operator();
     
-    bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
+    bool ErrorOnNoConverge;  // throw an GRID_ASSERT when the CG fails to converge.
                              // Defaults true.
     RealD Tolerance;
     Integer MaxIterations;
@@ -77,7 +77,7 @@ class BiCGSTAB : public OperatorFunction<Field>
 
       // Initial residual computation & set up
       RealD guess = norm2(psi);
-      assert(std::isnan(guess) == 0);
+      GRID_ASSERT(std::isnan(guess) == 0);
     
       Linop.Op(psi, v);
       b = norm2(v);
@@ -214,7 +214,7 @@ class BiCGSTAB : public OperatorFunction<Field>
           std::cout << GridLogMessage << "\tAxpyNorm   " << AxpyNormTimer.Elapsed() << std::endl;
           std::cout << GridLogMessage << "\tLinearComb " << LinearCombTimer.Elapsed() << std::endl;
 
-          if(ErrorOnNoConverge){ assert(true_residual / Tolerance < 10000.0); }
+          if(ErrorOnNoConverge){ GRID_ASSERT(true_residual / Tolerance < 10000.0); }
 
           IterationsToComplete = k;	
 
@@ -224,7 +224,7 @@ class BiCGSTAB : public OperatorFunction<Field>
       
       std::cout << GridLogMessage << "BiCGSTAB did NOT converge" << std::endl;
 
-      if(ErrorOnNoConverge){ assert(0); }
+      if(ErrorOnNoConverge){ GRID_ASSERT(0); }
       IterationsToComplete = k;
     }
 };

Grid/algorithms/iterative/BlockConjugateGradient.h

@@ -98,7 +98,7 @@ class BlockConjugateGradient : public OperatorFunction<Field> {
   int Nblock;
 
   BlockCGtype CGtype;
-  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
+  bool ErrorOnNoConverge;  // throw an GRID_ASSERT when the CG fails to converge.
                            // Defaults true.
   RealD Tolerance;
   Integer MaxIterations;
@@ -201,7 +201,7 @@ void operator()(LinearOperatorBase<Field> &Linop, const Field &Src, Field &Psi)
   } else if (CGtype == CGmultiRHS ) {
     CGmultiRHSsolve(Linop,Src,Psi);
   } else {
-    assert(0);
+    GRID_ASSERT(0);
   }
 }
 virtual void operator()(LinearOperatorBase<Field> &Linop, const std::vector<Field> &Src, std::vector<Field> &Psi) 
@@ -209,7 +209,7 @@ virtual void operator()(LinearOperatorBase<Field> &Linop, const std::vector<Fiel
   if ( CGtype == BlockCGrQVec ) {
     BlockCGrQsolveVec(Linop,Src,Psi);
   } else {
-    assert(0);
+    GRID_ASSERT(0);
   }
 }
 
@@ -259,10 +259,10 @@ void BlockCGrQsolve(LinearOperatorBase<Field> &Linop, const Field &B, Field &X)
   for(int b=0;b<Nblock;b++) std::cout << "src["<<b<<"]" << ssq[b] <<std::endl;
 
   sliceNorm(residuals,B,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   sliceNorm(residuals,X,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   /************************************************************************
    * Block conjugate gradient rQ (Sebastien Birk Thesis, after Dubrulle 2001)
@@ -402,7 +402,7 @@ void BlockCGrQsolve(LinearOperatorBase<Field> &Linop, const Field &B, Field &X)
   std::cout << GridLogMessage << "BlockConjugateGradient(rQ) did NOT converge "<<k<<" / "<<MaxIterations
 	    <<" residual "<< std::sqrt(max_resid)<< std::endl;
 
-  if (ErrorOnNoConverge) assert(0);
+  if (ErrorOnNoConverge) GRID_ASSERT(0);
   IterationsToComplete = k;
 }
 //////////////////////////////////////////////////////////////////////////
@@ -438,10 +438,10 @@ void CGmultiRHSsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &
   for(int b=0;b<Nblock;b++) sssum+=ssq[b];
 
   sliceNorm(residuals,Src,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   sliceNorm(residuals,Psi,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   // Initial search dir is guess
   Linop.HermOp(Psi, AP);
@@ -540,7 +540,7 @@ void CGmultiRHSsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &
   }
   std::cout << GridLogMessage << "MultiRHSConjugateGradient did NOT converge" << std::endl;
 
-  if (ErrorOnNoConverge) assert(0);
+  if (ErrorOnNoConverge) GRID_ASSERT(0);
   IterationsToComplete = k;
 }
 
@@ -554,7 +554,7 @@ void CGmultiRHSsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &
 void BlockCGrQsolveVec(LinearOperatorBase<Field> &Linop, const std::vector<Field> &B, std::vector<Field> &X) 
 {
   Nblock = B.size();
-  assert(Nblock == X.size());
+  GRID_ASSERT(Nblock == X.size());
 
   std::cout<<GridLogMessage<<" Block Conjugate Gradient Vec rQ : Nblock "<<Nblock<<std::endl;
 
@@ -594,10 +594,10 @@ void BlockCGrQsolveVec(LinearOperatorBase<Field> &Linop, const std::vector<Field
   for(int b=0;b<Nblock;b++) sssum+=ssq[b];
 
   for(int b=0;b<Nblock;b++){ residuals[b] = norm2(B[b]);}
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   for(int b=0;b<Nblock;b++){ residuals[b] = norm2(X[b]);}
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
+  for(int b=0;b<Nblock;b++){ GRID_ASSERT(std::isnan(residuals[b])==0); }
 
   /************************************************************************
    * Block conjugate gradient rQ (Sebastien Birk Thesis, after Dubrulle 2001)
@@ -731,7 +731,7 @@ void BlockCGrQsolveVec(LinearOperatorBase<Field> &Linop, const std::vector<Field
   }
   std::cout << GridLogMessage << "BlockConjugateGradient(rQ) did NOT converge" << std::endl;
 
-  if (ErrorOnNoConverge) assert(0);
+  if (ErrorOnNoConverge) GRID_ASSERT(0);
   IterationsToComplete = k;
 }
 

Grid/algorithms/iterative/CommunicationAvoidingGeneralisedMinimalResidual.h

@@ -36,7 +36,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
  public:
   using OperatorFunction<Field>::operator();
 
-  bool ErrorOnNoConverge; // Throw an assert when CAGMRES fails to converge,
+  bool ErrorOnNoConverge; // Throw an GRID_ASSERT when CAGMRES fails to converge,
                           // defaults to true
 
   RealD   Tolerance;
@@ -82,7 +82,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     conformable(psi, src);
 
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD cp;
     RealD ssq = norm2(src);
@@ -137,7 +137,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     std::cout << GridLogMessage << "CommunicationAvoidingGeneralisedMinimalResidual did NOT converge" << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
   }
 
   RealD outerLoopBody(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi, RealD rsq) {
@@ -185,7 +185,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
       }
     }
 
-    assert(0); // Never reached
+    GRID_ASSERT(0); // Never reached
     return cp;
   }
 

Grid/algorithms/iterative/ConjugateGradient.h

@@ -45,7 +45,7 @@ public:
 
   using OperatorFunction<Field>::operator();
   
-  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
+  bool ErrorOnNoConverge;  // throw an GRID_ASSERT when the CG fails to converge.
                            // Defaults true.
   RealD Tolerance;
   Integer MaxIterations;
@@ -94,7 +94,7 @@ public:
     ssq = norm2(src);
     RealD guess = norm2(psi);
     NormTimer.Stop();
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
     AssignTimer.Start();
     if ( guess == 0.0 ) {
       r = src;
@@ -212,17 +212,21 @@ public:
 		  << "\tTarget " << Tolerance << std::endl;
 
 	//	std::cout << GridLogMessage << "\tPreamble   " << PreambleTimer.Elapsed() <<std::endl;
-	std::cout << GridLogMessage << "\tSolver Elapsed    " << SolverTimer.Elapsed() <<std::endl;
-        std::cout << GridLogPerformance << "Time breakdown "<<std::endl;
-	std::cout << GridLogPerformance << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
-	std::cout << GridLogPerformance << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
-	std::cout << GridLogPerformance << "\t\tInner      " << InnerTimer.Elapsed() <<std::endl;
-	std::cout << GridLogPerformance << "\t\tAxpyNorm   " << AxpyNormTimer.Elapsed() <<std::endl;
-	std::cout << GridLogPerformance << "\t\tLinearComb " << LinearCombTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tPreamble   " << PreambleTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tConstruct  " << ConstructTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tNorm       " << NormTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tAssign     " << AssignTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tSolver     " << SolverTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "Solver breakdown "<<std::endl;
+	std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\t\tInner      " << InnerTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\t\tAxpyNorm   " << AxpyNormTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\t\tLinearComb " << LinearCombTimer.Elapsed() <<std::endl;
 
 	std::cout << GridLogDebug << "\tMobius flop rate " << DwfFlops/ usecs<< " Gflops " <<std::endl;
 
-        if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
+        if (ErrorOnNoConverge) GRID_ASSERT(true_residual / Tolerance < 10000.0);
 
 	IterationsToComplete = k;	
 	TrueResidual = true_residual;
@@ -251,7 +255,7 @@ public:
     std::cout << GridLogPerformance << "\t\tAxpyNorm   " << AxpyNormTimer.Elapsed() <<std::endl;
     std::cout << GridLogPerformance << "\t\tLinearComb " << LinearCombTimer.Elapsed() <<std::endl;
 
-    if (ErrorOnNoConverge) assert(0);
+    if (ErrorOnNoConverge) GRID_ASSERT(0);
     IterationsToComplete = k;
 
   }

Grid/algorithms/iterative/ConjugateGradientMixedPrecBatched.h

@@ -77,7 +77,7 @@ public:
   }
 
   void operator() (const std::vector<FieldD> &src_d_in, std::vector<FieldD> &sol_d){
-    assert(src_d_in.size() == sol_d.size());
+    GRID_ASSERT(src_d_in.size() == sol_d.size());
     int NBatch = src_d_in.size();
 
     std::cout << GridLogMessage << "NBatch = " << NBatch << std::endl;

Grid/algorithms/iterative/ConjugateGradientMultiShift.h

@@ -98,9 +98,9 @@ public:
     std::vector<RealD> alpha(nshift,1.0);
     std::vector<Field>   ps(nshift,grid);// Search directions
 
-    assert(psi.size()==nshift);
-    assert(mass.size()==nshift);
-    assert(mresidual.size()==nshift);
+    GRID_ASSERT(psi.size()==nshift);
+    GRID_ASSERT(mass.size()==nshift);
+    GRID_ASSERT(mresidual.size()==nshift);
   
     // remove dynamic sized arrays on stack; 2d is a pain with vector
     std::vector<RealD>  bs(nshift);
@@ -122,7 +122,7 @@ public:
   
     // Check lightest mass
     for(int s=0;s<nshift;s++){
-      assert( mass[s]>= mass[primary] );
+      GRID_ASSERT( mass[s]>= mass[primary] );
       converged[s]=0;
     }
   
@@ -338,7 +338,7 @@ public:
     }
     // ugly hack
     std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
-    //  assert(0);
+    //  GRID_ASSERT(0);
   }
 
 };

Grid/algorithms/iterative/ConjugateGradientMultiShiftCleanup.h

@@ -118,9 +118,9 @@ public:
     FieldF r_f(SinglePrecGrid);
     FieldD mmp_d(DoublePrecGrid);
 
-    assert(psi_d.size()==nshift);
-    assert(mass.size()==nshift);
-    assert(mresidual.size()==nshift);
+    GRID_ASSERT(psi_d.size()==nshift);
+    GRID_ASSERT(mass.size()==nshift);
+    GRID_ASSERT(mresidual.size()==nshift);
   
     // dynamic sized arrays on stack; 2d is a pain with vector
     std::vector<RealD>  bs(nshift);
@@ -141,7 +141,7 @@ public:
 
     // Check lightest mass
     for(int s=0;s<nshift;s++){
-      assert( mass[s]>= mass[primary] );
+      GRID_ASSERT( mass[s]>= mass[primary] );
       converged[s]=0;
     }
   
@@ -179,7 +179,7 @@ public:
     Linop_d.HermOpAndNorm(p_d,mmp_d,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
     tmp_d = tmp_d - mmp_d;
     std::cout << " Testing operators match "<<norm2(mmp_d)<<" f "<<norm2(mmp_f)<<" diff "<< norm2(tmp_d)<<std::endl;
-    //    assert(norm2(tmp_d)< 1.0e-4);
+    //    GRID_ASSERT(norm2(tmp_d)< 1.0e-4);
 
     axpy(mmp_d,mass[0],p_d,mmp_d);
     RealD rn = norm2(p_d);
@@ -365,7 +365,7 @@ public:
    
     }
     std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
-    assert(0);
+    GRID_ASSERT(0);
   }
 
 };

Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h

@@ -48,12 +48,12 @@ public:
 
   ShiftedLinop(LinearOperatorBase<Field> &_linop_base, RealD _shift): linop_base(_linop_base), shift(_shift){}
 
-  void OpDiag (const Field &in, Field &out){ assert(0); }
-  void OpDir  (const Field &in, Field &out,int dir,int disp){ assert(0); }
-  void OpDirAll  (const Field &in, std::vector<Field> &out){ assert(0); }
+  void OpDiag (const Field &in, Field &out){ GRID_ASSERT(0); }
+  void OpDir  (const Field &in, Field &out,int dir,int disp){ GRID_ASSERT(0); }
+  void OpDirAll  (const Field &in, std::vector<Field> &out){ GRID_ASSERT(0); }
   
-  void Op     (const Field &in, Field &out){ assert(0); }
-  void AdjOp  (const Field &in, Field &out){ assert(0); }
+  void Op     (const Field &in, Field &out){ GRID_ASSERT(0); }
+  void AdjOp  (const Field &in, Field &out){ GRID_ASSERT(0); }
 
   void HermOp(const Field &in, Field &out){
     linop_base.HermOp(in, out);
@@ -151,9 +151,9 @@ public:
     FieldD r_d(DoublePrecGrid);
     FieldD mmp_d(DoublePrecGrid);
 
-    assert(psi_d.size()==nshift);
-    assert(mass.size()==nshift);
-    assert(mresidual.size()==nshift);
+    GRID_ASSERT(psi_d.size()==nshift);
+    GRID_ASSERT(mass.size()==nshift);
+    GRID_ASSERT(mresidual.size()==nshift);
   
     // dynamic sized arrays on stack; 2d is a pain with vector
     std::vector<RealD>  bs(nshift);
@@ -174,7 +174,7 @@ public:
 
     // Check lightest mass
     for(int s=0;s<nshift;s++){
-      assert( mass[s]>= mass[primary] );
+      GRID_ASSERT( mass[s]>= mass[primary] );
       converged[s]=0;
     }
   
@@ -211,7 +211,7 @@ public:
     Linop_d.HermOpAndNorm(p_d,mmp_d,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
     tmp_d = tmp_d - mmp_d;
     std::cout << " Testing operators match "<<norm2(mmp_d)<<" f "<<norm2(mmp_f)<<" diff "<< norm2(tmp_d)<<std::endl;
-    assert(norm2(tmp_d)< 1.0);
+    GRID_ASSERT(norm2(tmp_d)< 1.0);
 
     axpy(mmp_d,mass[0],p_d,mmp_d);
     RealD rn = norm2(p_d);
@@ -408,7 +408,7 @@ public:
    
     }
     std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
-    assert(0);
+    GRID_ASSERT(0);
   }
 
 };

Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h

@@ -35,7 +35,7 @@ template<class FieldD,class FieldF,
 	 typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0> 
 class ConjugateGradientReliableUpdate : public LinearFunction<FieldD> {
 public:
-  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
+  bool ErrorOnNoConverge;  // throw an GRID_ASSERT when the CG fails to converge.
   // Defaults true.
   RealD Tolerance;
   Integer MaxIterations;
@@ -66,7 +66,7 @@ public:
       DoFinalCleanup(true),
       Linop_fallback(NULL)
   {
-    assert(Delta > 0. && Delta < 1. && "Expect  0 < Delta < 1");
+    GRID_ASSERT(Delta > 0. && Delta < 1. && "Expect  0 < Delta < 1");
   };
 
   void setFallbackLinop(LinearOperatorBase<FieldF> &_Linop_fallback, const RealD _fallback_transition_tol){
@@ -90,7 +90,7 @@ public:
 
     // Initial residual computation & set up
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
     
     Linop_d.HermOpAndNorm(psi, mmp, d, b);
     
@@ -217,7 +217,7 @@ public:
 	  CG(Linop_d,src,psi);
 	  IterationsToCleanup = CG.IterationsToComplete;
 	}
-	else if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
+	else if (ErrorOnNoConverge) GRID_ASSERT(true_residual / Tolerance < 10000.0);
 
 	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate complete.\n";
 	return;
@@ -263,7 +263,7 @@ public:
     std::cout << GridLogMessage << "ConjugateGradientReliableUpdate did NOT converge"
 	      << std::endl;
       
-    if (ErrorOnNoConverge) assert(0);
+    if (ErrorOnNoConverge) GRID_ASSERT(0);
     IterationsToComplete = k;
     ReliableUpdatesPerformed = l;      
   }    

Grid/algorithms/iterative/ConjugateResidual.h

@@ -106,7 +106,7 @@ public:
     }
 
     std::cout<<GridLogMessage<<"ConjugateResidual did NOT converge"<<std::endl;
-    assert(0);
+    GRID_ASSERT(0);
   }
 };
 NAMESPACE_END(Grid);

Grid/algorithms/iterative/FlexibleCommunicationAvoidingGeneralisedMinimalResidual.h

@@ -36,7 +36,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
  public:
   using OperatorFunction<Field>::operator();
 
-  bool ErrorOnNoConverge; // Throw an assert when FCAGMRES fails to converge,
+  bool ErrorOnNoConverge; // Throw an GRID_ASSERT when FCAGMRES fails to converge,
                           // defaults to true
 
   RealD   Tolerance;
@@ -87,7 +87,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     conformable(psi, src);
 
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD cp;
     RealD ssq = norm2(src);
@@ -144,7 +144,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     std::cout << GridLogMessage << "FlexibleCommunicationAvoidingGeneralisedMinimalResidual did NOT converge" << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
   }
 
   RealD outerLoopBody(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi, RealD rsq) {
@@ -191,7 +191,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
       }
     }
 
-    assert(0); // Never reached
+    GRID_ASSERT(0); // Never reached
     return cp;
   }
 

Grid/algorithms/iterative/FlexibleGeneralisedMinimalResidual.h

@@ -36,7 +36,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
   using OperatorFunction<Field>::operator();
 
-  bool ErrorOnNoConverge; // Throw an assert when FGMRES fails to converge,
+  bool ErrorOnNoConverge; // Throw an GRID_ASSERT when FGMRES fails to converge,
                           // defaults to true
 
   RealD   Tolerance;
@@ -85,7 +85,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     conformable(psi, src);
 
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD cp;
     RealD ssq = norm2(src);
@@ -142,7 +142,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     std::cout << GridLogMessage << "FlexibleGeneralisedMinimalResidual did NOT converge" << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
   }
 
   RealD outerLoopBody(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi, RealD rsq) {
@@ -189,7 +189,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
       }
     }
 
-    assert(0); // Never reached
+    GRID_ASSERT(0); // Never reached
     return cp;
   }
 

Grid/algorithms/iterative/GCRCoefficients.h

@@ -0,0 +1,87 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/iterative/GCRCoefficients.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////////////
+// Recorded GCR coefficients: per-step means over calls of the step length
+// a_k and the orthogonalisation coefficients b_kj (already scaled and
+// signed as applied: p_{k+1} = r + sum_j b_kj p_{k-j}).
+//////////////////////////////////////////////////////////////////////////////
+struct GCRCoefficients {
+  int mmax = 0;
+  // Every recorded call is kept: calls[c] = list of (a_k, [b_kj]) per step.
+  // A(k)/B(k,j) return the coefficients of the SELECTED call: by default the
+  // last complete one.  Selection "mean" averages coefficients over calls --
+  // kept for comparison only: the mean of the coefficients of a nonlinear
+  // recurrence is not the mean of the polynomials, and in practice (Frontier
+  // M3, 2026-08-26) it was worse than every individual call.
+  enum Select { Last=0, First=1, Index=2, Mean=3 };
+  Select select = Last;
+  int    index  = 0;
+  typedef std::vector<std::pair<ComplexD,std::vector<ComplexD> > > Call;
+  std::vector<Call> calls;
+  Call current;
+  void RecordA(int k, ComplexD a){
+    if ( k==0 && current.size() ) { calls.push_back(current); current.clear(); }
+    if ( (int)current.size() <= k ) current.resize(k+1);
+    current[k].first = a;
+  }
+  void RecordB(int k, const std::vector<ComplexD> &b){
+    if ( (int)current.size() <= k ) current.resize(k+1);
+    current[k].second = b;
+  }
+  void Flush(void){ if ( current.size() ) { calls.push_back(current); current.clear(); } }
+  int  Calls(void) const { return calls.size() + (current.size() ? 1 : 0); }
+  const Call & Chosen(void) const {
+    GRID_ASSERT( calls.size() || current.size() );
+    if ( calls.empty() ) return current;
+    if ( select==First ) return calls.front();
+    if ( select==Index ) { GRID_ASSERT(index>=0 && index<(int)calls.size()); return calls[index]; }
+    return calls.back();
+  }
+  int  Steps(void) const { return select==Mean ? MeanSteps() : Chosen().size(); }
+  int  NB(int k) const   { return select==Mean ? MeanNB(k)   : Chosen()[k].second.size(); }
+  ComplexD A(int k) const    { return select==Mean ? MeanA(k)   : Chosen()[k].first; }
+  ComplexD B(int k,int j) const { return select==Mean ? MeanB(k,j) : Chosen()[k].second[j]; }
+  // mean over calls (comparison only)
+  int MeanSteps(void) const { int m=0; for(auto &c:calls) m = std::max(m,(int)c.size()); return m; }
+  int MeanNB(int k) const { int m=0; for(auto &c:calls) if(k<(int)c.size()) m = std::max(m,(int)c[k].second.size()); return m; }
+  ComplexD MeanA(int k) const { ComplexD s(0.0); int n=0; for(auto &c:calls) if(k<(int)c.size()){ s+=c[k].first; n++; } return s/(double)n; }
+  ComplexD MeanB(int k,int j) const { ComplexD s(0.0); int n=0; for(auto &c:calls) if(k<(int)c.size() && j<(int)c[k].second.size()){ s+=c[k].second[j]; n++; } return s/(double)n; }
+  void Report(const std::string &name) const {
+    const char *sel[4]={"last","first","index","mean"};
+    std::cout << GridLogMessage << "GCRCoefficients " << name << ": " << Calls() << " calls, " << Steps() << " steps, mmax " << mmax
+              << ", selection " << sel[select] << std::endl;
+    for(int k=0;k<Steps();k++){
+      std::cout << GridLogMessage << "  step " << k << " a=(" << real(A(k)) << "," << imag(A(k)) << ")";
+      for(int j=0;j<NB(k);j++) std::cout << " b[" << j << "]=(" << real(B(k,j)) << "," << imag(B(k,j)) << ")";
+      // spread of a_k across calls: how different the individual polynomials are
+      if ( calls.size()>1 ) {
+        RealD lo=1e300, hi=0; for(auto &c:calls) if(k<(int)c.size()){ RealD x=real(c[k].first), y=imag(c[k].first); RealD m=std::sqrt(x*x+y*y); lo=std::min(lo,m); hi=std::max(hi,m); }
+        std::cout << "   |a| over calls [" << lo << "," << hi << "]";
+      }
+      std::cout << std::endl;
+    }
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/iterative/GeneralisedMinimalResidual.h

@@ -36,7 +36,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
   using OperatorFunction<Field>::operator();
 
-  bool ErrorOnNoConverge; // Throw an assert when GMRES fails to converge,
+  bool ErrorOnNoConverge; // Throw an GRID_ASSERT when GMRES fails to converge,
                           // defaults to true
 
   RealD   Tolerance;
@@ -80,7 +80,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     conformable(psi, src);
 
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD cp;
     RealD ssq = norm2(src);
@@ -135,7 +135,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     std::cout << GridLogMessage << "GeneralisedMinimalResidual did NOT converge" << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
   }
 
   RealD outerLoopBody(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi, RealD rsq) {
@@ -181,7 +181,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
       }
     }
 
-    assert(0); // Never reached
+    GRID_ASSERT(0); // Never reached
     return cp;
   }
 

Grid/algorithms/iterative/ImplicitlyRestartedBlockLanczos.h

@@ -175,7 +175,7 @@ public:
       eresid(_eresid),  MaxIter(_MaxIter),
       diagonalisation(_diagonalisation),split_test(0),
       Nevec_acc(_Nu)
-  { assert( (Nk%Nu==0) && (Nm%Nu==0) ); };
+  { GRID_ASSERT( (Nk%Nu==0) && (Nm%Nu==0) ); };
 
   ////////////////////////////////
   // Helpers
@@ -206,7 +206,7 @@ public:
           Glog<<"orthogonalize after: "<<j<<" of "<<k<<" "<< ip <<std::endl;
       }
     }
-    assert(normalize(w,if_print) != 0);
+    GRID_ASSERT(normalize(w,if_print) != 0);
   }
   void reorthogonalize(Field& w, std::vector<Field>& evec, int k)
   {
@@ -225,7 +225,7 @@ public:
       w[i] = w[i] - ip * evec[j];
     }}
     for(int i=0; i<_Nu; ++i)
-    assert(normalize(w[i],if_print) !=0);
+    GRID_ASSERT(normalize(w[i],if_print) !=0);
   }
 
 
@@ -244,7 +244,7 @@ public:
     const uint64_t sites = grid->lSites();
 
     int Nbatch = R/Nevec_acc;
-    assert( R%Nevec_acc == 0 );
+    GRID_ASSERT( R%Nevec_acc == 0 );
 //    Glog << "nBatch, Nevec_acc, R, Nu = " 
 //         << Nbatch << "," << Nevec_acc << "," << R << "," << Nu << std::endl;
     
@@ -302,7 +302,7 @@ public:
       }
     }
     for (int i=0; i<Nu; ++i) {
-      assert(normalize(w[i],do_print)!=0);
+      GRID_ASSERT(normalize(w[i],do_print)!=0);
     }
     
     Glog << "cuBLAS Zgemm done"<< std::endl;
@@ -374,8 +374,8 @@ cudaStat = cudaMallocManaged((void **)&evec_acc, Nevec_acc*sites*12*sizeof(CUDA_
   {
     std::string fname = std::string(cname+"::calc_irbl()"); 
     GridBase *grid = evec[0].Grid();
-    assert(grid == src[0].Grid());
-    assert( Nu = src.size() );
+    GRID_ASSERT(grid == src[0].Grid());
+    GRID_ASSERT( Nu = src.size() );
     
     Glog << std::string(74,'*') << std::endl;
     Glog << fname + " starting iteration 0 /  "<< MaxIter<< std::endl;
@@ -396,7 +396,7 @@ cudaStat = cudaMallocManaged((void **)&evec_acc, Nevec_acc*sites*12*sizeof(CUDA_
     }
     Glog << std::string(74,'*') << std::endl;
     
-    assert(Nm == evec.size() && Nm == eval.size());
+    GRID_ASSERT(Nm == evec.size() && Nm == eval.size());
 
     std::vector<std::vector<ComplexD>> lmd(Nu,std::vector<ComplexD>(Nm,0.0));  
     std::vector<std::vector<ComplexD>> lme(Nu,std::vector<ComplexD>(Nm,0.0));  
@@ -579,8 +579,8 @@ cudaStat = cudaMallocManaged((void **)&evec_acc, Nevec_acc*sites*12*sizeof(CUDA_
   {
     std::string fname = std::string(cname+"::calc_rbl()"); 
     GridBase *grid = evec[0].Grid();
-    assert(grid == src[0].Grid());
-    assert( Nu = src.size() );
+    GRID_ASSERT(grid == src[0].Grid());
+    GRID_ASSERT( Nu = src.size() );
 
     int Np = (Nm-Nk);
     if (Np > 0 && MaxIter > 1) Np /= MaxIter;
@@ -607,7 +607,7 @@ cudaStat = cudaMallocManaged((void **)&evec_acc, Nevec_acc*sites*12*sizeof(CUDA_
     }
     Glog << std::string(74,'*') << std::endl;
     
-    assert(Nm == evec.size() && Nm == eval.size());
+    GRID_ASSERT(Nm == evec.size() && Nm == eval.size());
 	
     std::vector<std::vector<ComplexD>> lmd(Nu,std::vector<ComplexD>(Nm,0.0));  
     std::vector<std::vector<ComplexD>> lme(Nu,std::vector<ComplexD>(Nm,0.0));  
@@ -785,7 +785,7 @@ private:
     
     int Nu = w.size();
     int Nm = evec.size();
-    assert( b < Nm/Nu );
+    GRID_ASSERT( b < Nm/Nu );
 //    GridCartesian *grid = evec[0]._grid;
     
     // converts block index to full indicies for an interval [L,R)
@@ -796,7 +796,7 @@ private:
 
     Glog << "Using split grid"<< std::endl;
 //   LatticeGaugeField s_Umu(SGrid);
-   assert((Nu%mrhs)==0);
+   GRID_ASSERT((Nu%mrhs)==0);
    std::vector<Field>   in(mrhs,f_grid);
      
     Field s_in(sf_grid);
@@ -906,7 +906,7 @@ if(split_test){
     
     for (int u=0; u<Nu; ++u) {
 //      Glog << "norm2(w[" << u << "])= "<< norm2(w[u]) << std::endl;
-      assert (!isnan(norm2(w[u])));
+      GRID_ASSERT (!isnan(norm2(w[u])));
       for (int k=L+u; k<R; ++k) {
         Glog <<" In block "<< b << "," <<" beta[" << u << "," << k-L << "] = " << lme[u][k] << std::endl;
       }
@@ -929,8 +929,8 @@ if(split_test){
 			 Eigen::MatrixXcd & Qt, // Nm x Nm
 			 GridBase *grid)
   {
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     Eigen::MatrixXcd BlockTriDiag = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     for ( int u=0; u<Nu; ++u ) {
@@ -970,8 +970,8 @@ if(split_test){
 			 GridBase *grid)
   {
     Glog << "diagonalize_lapack: Nu= "<<Nu<<" Nk= "<<Nk<<" Nm= "<<std::endl;
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     Eigen::MatrixXcd BlockTriDiag = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     for ( int u=0; u<Nu; ++u ) {
@@ -1119,7 +1119,7 @@ if (1){
       diagonalize_lapack(eval,lmd,lme,Nu,Nk,Nm,Qt,grid);
 #endif
     } else { 
-      assert(0);
+      GRID_ASSERT(0);
     }
   }
   
@@ -1131,8 +1131,8 @@ if (1){
          Eigen::MatrixXcd& M)
   {
     //Glog << "unpackHermitBlockTriDiagMatToEigen() begin" << '\n'; 
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     M = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     // rearrange 
@@ -1159,8 +1159,8 @@ if (1){
          Eigen::MatrixXcd& M)
   {
     //Glog << "packHermitBlockTriDiagMatfromEigen() begin" << '\n'; 
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     
     // rearrange 
     for ( int u=0; u<Nu; ++u ) {

Grid/algorithms/iterative/ImplicitlyRestartedBlockLanczosCoarse.h

@@ -121,7 +121,7 @@ public:
       eresid(_eresid),  MaxIter(_MaxIter),
       diagonalisation(_diagonalisation),
       Nevec_acc(_Nu)
-  { assert( (Nk%Nu==0) && (Nm%Nu==0) ); };
+  { GRID_ASSERT( (Nk%Nu==0) && (Nm%Nu==0) ); };
 
   ////////////////////////////////
   // Helpers
@@ -151,7 +151,7 @@ public:
           Glog<<"orthogonalize after: "<<j<<" of "<<k<<" "<< ip <<std::endl;
       }
     }
-    assert(normalize(w,if_print) != 0);
+    GRID_ASSERT(normalize(w,if_print) != 0);
   }
   void reorthogonalize(Field& w, std::vector<Field>& evec, int k)
   {
@@ -169,7 +169,7 @@ public:
       w[i] = w[i] - ip * evec[j];
     }}
     for(int i=0; i<_Nu; ++i)
-    assert(normalize(w[i],if_print) !=0);
+    GRID_ASSERT(normalize(w[i],if_print) !=0);
   }
   
   void orthogonalize_blockhead(Field& w, std::vector<Field>& evec, int k, int Nu)
@@ -205,8 +205,8 @@ public:
   {
     std::string fname = std::string(cname+"::calc_irbl()"); 
     GridBase *grid = evec[0].Grid();
-    assert(grid == src[0].Grid());
-    assert( Nu = src.size() );
+    GRID_ASSERT(grid == src[0].Grid());
+    GRID_ASSERT( Nu = src.size() );
     
     Glog << std::string(74,'*') << std::endl;
     Glog << fname + " starting iteration 0 /  "<< MaxIter<< std::endl;
@@ -227,7 +227,7 @@ public:
     }
     Glog << std::string(74,'*') << std::endl;
     
-    assert(Nm == evec.size() && Nm == eval.size());
+    GRID_ASSERT(Nm == evec.size() && Nm == eval.size());
 
     std::vector<std::vector<ComplexD>> lmd(Nu,std::vector<ComplexD>(Nm,0.0));  
     std::vector<std::vector<ComplexD>> lme(Nu,std::vector<ComplexD>(Nm,0.0));  
@@ -413,8 +413,8 @@ public:
   {
     std::string fname = std::string(cname+"::calc_rbl()"); 
     GridBase *grid = evec[0].Grid();
-    assert(grid == src[0].Grid());
-    assert( Nu = src.size() );
+    GRID_ASSERT(grid == src[0].Grid());
+    GRID_ASSERT( Nu = src.size() );
 
     int Np = (Nm-Nk);
     if (Np > 0 && MaxIter > 1) Np /= MaxIter;
@@ -441,7 +441,7 @@ public:
     }
     Glog << std::string(74,'*') << std::endl;
     
-    assert(Nm == evec.size() && Nm == eval.size());
+    GRID_ASSERT(Nm == evec.size() && Nm == eval.size());
 	
     std::vector<std::vector<ComplexD>> lmd(Nu,std::vector<ComplexD>(Nm,0.0));  
     std::vector<std::vector<ComplexD>> lme(Nu,std::vector<ComplexD>(Nm,0.0));  
@@ -622,7 +622,7 @@ private:
     
     int Nu = w.size();
     int Nm = evec.size();
-    assert( b < Nm/Nu );
+    GRID_ASSERT( b < Nm/Nu );
     
     // converts block index to full indicies for an interval [L,R)
     int L = Nu*b;
@@ -630,7 +630,7 @@ private:
 
     Real beta;
 
-    assert((Nu%mrhs)==0);
+    GRID_ASSERT((Nu%mrhs)==0);
     std::vector<Field>   in(mrhs,f_grid);
     std::vector<Field>   out(mrhs,f_grid);
 
@@ -711,7 +711,7 @@ private:
     
     for (int u=0; u<Nu; ++u) {
       //      Glog << "norm2(w[" << u << "])= "<< norm2(w[u]) << std::endl;
-      assert (!isnan(norm2(w[u])));
+      GRID_ASSERT (!isnan(norm2(w[u])));
       for (int k=L+u; k<R; ++k) {
 	//        Glog <<" In block "<< b << "," <<" beta[" << u << "," << k-L << "] = " << lme[u][k] << std::endl;
       }
@@ -734,8 +734,8 @@ private:
 			 Eigen::MatrixXcd & Qt, // Nm x Nm
 			 GridBase *grid)
   {
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     Eigen::MatrixXcd BlockTriDiag = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     for ( int u=0; u<Nu; ++u ) {
@@ -775,8 +775,8 @@ private:
 			 GridBase *grid)
   {
     Glog << "diagonalize_lapack: Nu= "<<Nu<<" Nk= "<<Nk<<" Nm= "<<std::endl;
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     Eigen::MatrixXcd BlockTriDiag = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     for ( int u=0; u<Nu; ++u ) {
@@ -924,7 +924,7 @@ if (1){
       diagonalize_lapack(eval,lmd,lme,Nu,Nk,Nm,Qt,grid);
 #endif
     } else { 
-      assert(0);
+      GRID_ASSERT(0);
     }
   }
   
@@ -936,8 +936,8 @@ if (1){
          Eigen::MatrixXcd& M)
   {
     //    Glog << "unpackHermitBlockTriDiagMatToEigen() begin" << '\n'; 
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     M = Eigen::MatrixXcd::Zero(Nk,Nk);
     
     // rearrange 
@@ -964,8 +964,8 @@ if (1){
          Eigen::MatrixXcd& M)
   {
     //    Glog << "packHermitBlockTriDiagMatfromEigen() begin" << '\n'; 
-    assert( Nk%Nu == 0 && Nm%Nu == 0 );
-    assert( Nk <= Nm );
+    GRID_ASSERT( Nk%Nu == 0 && Nm%Nu == 0 );
+    GRID_ASSERT( Nk <= Nm );
     
     // rearrange 
     for ( int u=0; u<Nu; ++u ) {

Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h

@@ -211,7 +211,7 @@ until convergence
   void calc(std::vector<RealD>& eval, std::vector<Field>& evec,  const Field& src, int& Nconv, bool reverse=false)
   {
     GridBase *grid = src.Grid();
-    assert(grid == evec[0].Grid());
+    GRID_ASSERT(grid == evec[0].Grid());
     
     //    GridLogIRL.TimingMode(1);
     std::cout << GridLogIRL <<"**************************************************************************"<< std::endl;
@@ -231,7 +231,7 @@ until convergence
     }
     std::cout << GridLogIRL <<"**************************************************************************"<< std::endl;
 	
-    assert(Nm <= evec.size() && Nm <= eval.size());
+    GRID_ASSERT(Nm <= evec.size() && Nm <= eval.size());
     
     // quickly get an idea of the largest eigenvalue to more properly normalize the residuum
     RealD evalMaxApprox = 0.0;
@@ -245,9 +245,10 @@ until convergence
 	_HermOp(src_n,tmp);
 	//	std::cout << GridLogMessage<< tmp<<std::endl; exit(0);
 	//	std::cout << GridLogIRL << " _HermOp " << norm2(tmp) << std::endl;
-	RealD vnum = real(innerProduct(src_n,tmp)); // HermOp.
+//	RealD vnum = real(innerProduct(src_n,tmp)); // HermOp.
+	RealD vnum = real(innerProduct(tmp,tmp)); // HermOp^2.
 	RealD vden = norm2(src_n);
-	RealD na = vnum/vden;
+	RealD na = std::sqrt(vnum/vden);
 	if (fabs(evalMaxApprox/na - 1.0) < 0.0001)
 	  i=_MAX_ITER_IRL_MEVAPP_;
 	evalMaxApprox = na;
@@ -255,6 +256,7 @@ until convergence
 	src_n = tmp;
       }
     }
+    std::cout << GridLogIRL << " Final evalMaxApprox  " << evalMaxApprox << std::endl;
 	
     std::vector<RealD> lme(Nm);  
     std::vector<RealD> lme2(Nm);
@@ -335,7 +337,7 @@ until convergence
       }
       std::cout<<GridLogIRL <<"QR decomposed "<<std::endl;
 
-      assert(k2<Nm);      assert(k2<Nm);      assert(k1>0);
+      GRID_ASSERT(k2<Nm);      GRID_ASSERT(k2<Nm);      GRID_ASSERT(k1>0);
 
       basisRotate(evec,Qt,k1-1,k2+1,0,Nm,Nm); /// big constraint on the basis
       std::cout<<GridLogIRL <<"basisRotated  by Qt *"<<k1-1<<","<<k2+1<<")"<<std::endl;
@@ -461,7 +463,7 @@ until convergence
   {
     std::cout<<GridLogDebug << "Lanczos step " <<k<<std::endl;
     const RealD tiny = 1.0e-20;
-    assert( k< Nm );
+    GRID_ASSERT( k< Nm );
 
     GridStopWatch gsw_op,gsw_o;
 
@@ -595,7 +597,7 @@ until convergence
     }  else if ( diagonalisation == IRLdiagonaliseWithEigen ) { 
       diagonalize_Eigen(lmd,lme,Nk,Nm,Qt,grid);
     } else { 
-      assert(0);
+      GRID_ASSERT(0);
     }
   }
 
@@ -685,7 +687,7 @@ void diagonalize_lapack(std::vector<RealD>& lmd,
     }
   }
 #else 
-  assert(0);
+  GRID_ASSERT(0);
 #endif
 }
 

Grid/algorithms/iterative/LocalCoherenceLanczos.h

@@ -80,7 +80,7 @@ public:
   ProjectedHermOp(LinearOperatorBase<FineField>& linop, std::vector<FineField> & _subspace) : 
     _Linop(linop), subspace(_subspace)
   {  
-    assert(subspace.size() >0);
+    GRID_ASSERT(subspace.size() >0);
   };
 
   void operator()(const CoarseField& in, CoarseField& out) {
@@ -346,12 +346,12 @@ public:
 
   void testFine(RealD resid) 
   {
-    assert(evals_fine.size() == nbasis);
-    assert(subspace.size() == nbasis);
+    GRID_ASSERT(evals_fine.size() == nbasis);
+    GRID_ASSERT(subspace.size() == nbasis);
     PlainHermOp<FineField>    Op(_FineOp);
     ImplicitlyRestartedLanczosHermOpTester<FineField> SimpleTester(Op);
     for(int k=0;k<nbasis;k++){
-      assert(SimpleTester.ReconstructEval(k,resid,subspace[k],evals_fine[k],1.0)==1);
+      GRID_ASSERT(SimpleTester.ReconstructEval(k,resid,subspace[k],evals_fine[k],1.0)==1);
     }
   }
 
@@ -359,8 +359,8 @@ public:
   //hence the smoother can be tuned after running the coarse Lanczos by using a different smoother here
   void testCoarse(RealD resid,ChebyParams cheby_smooth,RealD relax) 
   {
-    assert(evals_fine.size() == nbasis);
-    assert(subspace.size() == nbasis);
+    GRID_ASSERT(evals_fine.size() == nbasis);
+    GRID_ASSERT(subspace.size() == nbasis);
     //////////////////////////////////////////////////////////////////////////////////////////////////
     // create a smoother and see if we can get a cheap convergence test and smooth inside the IRL
     //////////////////////////////////////////////////////////////////////////////////////////////////
@@ -380,7 +380,7 @@ public:
   void calcFine(ChebyParams cheby_parms,int Nstop,int Nk,int Nm,RealD resid, 
 		RealD MaxIt, RealD betastp, int MinRes)
   {
-    assert(nbasis<=Nm);
+    GRID_ASSERT(nbasis<=Nm);
     Chebyshev<FineField>      Cheby(cheby_parms);
     FunctionHermOp<FineField> ChebyOp(Cheby,_FineOp);
     PlainHermOp<FineField>    Op(_FineOp);
@@ -400,8 +400,8 @@ public:
     IRL.calc(evals_fine,subspace,src,Nconv,false);
     
     // Shrink down to number saved
-    assert(Nstop>=nbasis);
-    assert(Nconv>=nbasis);
+    GRID_ASSERT(Nstop>=nbasis);
+    GRID_ASSERT(Nconv>=nbasis);
     evals_fine.resize(nbasis);
     subspace.resize(nbasis,_FineGrid);
   }
@@ -433,7 +433,7 @@ public:
     ImplicitlyRestartedLanczos<CoarseField> IRL(ChebyOp,ChebyOp,ChebySmoothTester,Nstop,Nk,Nm,resid,MaxIt,betastp,MinRes);
     int Nconv=0;
     IRL.calc(evals_coarse,evec_coarse,src,Nconv,false);
-    assert(Nconv>=Nstop);
+    GRID_ASSERT(Nconv>=Nstop);
     evals_coarse.resize(Nstop);
     evec_coarse.resize (Nstop,_CoarseGrid);
     for (int i=0;i<Nstop;i++){

Grid/algorithms/iterative/MinimalResidual.h

@@ -35,7 +35,7 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
  public:
   using OperatorFunction<Field>::operator();
 
-  bool ErrorOnNoConverge; // throw an assert when the MR fails to converge.
+  bool ErrorOnNoConverge; // throw an GRID_ASSERT when the MR fails to converge.
                           // Defaults true.
   RealD   Tolerance;
   Integer MaxIterations;
@@ -59,7 +59,7 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
 
     // Initial residual computation & set up
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
@@ -136,7 +136,7 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
         std::cout << GridLogMessage << "MR Time elapsed: Linalg  " << LinalgTimer.Elapsed() << std::endl;
 
         if (ErrorOnNoConverge)
-          assert(true_residual / Tolerance < 10000.0);
+          GRID_ASSERT(true_residual / Tolerance < 10000.0);
 
         IterationsToComplete = k;
 
@@ -148,7 +148,7 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
               << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
 
     IterationsToComplete = k;
   }

Grid/algorithms/iterative/MixedPrecisionFlexibleGeneralisedMinimalResidual.h

@@ -37,7 +37,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
   using OperatorFunction<FieldD>::operator();
 
-  bool ErrorOnNoConverge; // Throw an assert when MPFGMRES fails to converge,
+  bool ErrorOnNoConverge; // Throw an GRID_ASSERT when MPFGMRES fails to converge,
                           // defaults to true
 
   RealD   Tolerance;
@@ -91,7 +91,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     conformable(psi, src);
 
     RealD guess = norm2(psi);
-    assert(std::isnan(guess) == 0);
+    GRID_ASSERT(std::isnan(guess) == 0);
 
     RealD cp;
     RealD ssq = norm2(src);
@@ -150,7 +150,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     std::cout << GridLogMessage << "MPFGMRES did NOT converge" << std::endl;
 
     if (ErrorOnNoConverge)
-      assert(0);
+      GRID_ASSERT(0);
   }
 
   RealD outerLoopBody(LinearOperatorBase<FieldD> &LinOp, const FieldD &src, FieldD &psi, RealD rsq) {
@@ -197,7 +197,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
       }
     }
 
-    assert(0); // Never reached
+    GRID_ASSERT(0); // Never reached
     return cp;
   }
 

Grid/algorithms/iterative/PowerMethod.h

@@ -43,4 +43,69 @@ template<class Field> class PowerMethod
     return evalMaxApprox;
   }
 };
+
+// Non-Hermitian sibling of PowerMethod.  Drives Op() -- the operator the
+// caller actually applies -- rather than HermOp(), so it returns |lambda_max|
+// of a NON-Hermitian operator and reports the spectral-edge diagnostics the
+// Hermitian PowerMethod cannot:
+//   step 0 : |A v|/|v| on the (random) start src -- a one-sample lower bound
+//            on sigma_max(A).
+//   step k : |A v_k| -> |lambda_max| as v_k -> the dominant eigenvector; the
+//            complex Rayleigh quotient <v,Av> gives its phase (real => on the
+//            axis, a non-converging oscillation => a conjugate pair of equal
+//            modulus at the top).
+//   return : |lambda_max|, with a SUMMARY line flagging non-normality when the
+//            step-0 sigma_max lower bound sits well above the converged
+//            |lambda_max| -- the case where the numerical range extends beyond
+//            the spectrum and a spectrum-based smoother/Chebyshev bound is
+//            unsafe (the field of values must be used instead).
+template<class Field> class PowerMethodNonHermitian
+{
+ public:
+
+  template<typename T> static RealD normalise(T& v)
+  {
+    RealD nn = sqrt(norm2(v));
+    v = v * (1.0/nn);
+    return nn;
+  }
+
+  RealD operator()(LinearOperatorBase<Field> &Op, const Field &src)
+  {
+    GridBase *grid = src.Grid();
+    Field v(grid), Av(grid);
+    v = src;
+
+    RealD ratio = 0.0, ratio0 = 0.0;
+    ComplexD rq(0.0);
+    const int _MAX_ITER_EST_ = 200;
+
+    for (int i=0;i<_MAX_ITER_EST_;i++) {
+      normalise(v);                       // v is now unit
+      Op.Op(v,Av);
+      ratio = sqrt(norm2(Av));            // |A v| = |lambda_max| in the limit
+      rq    = innerProduct(v,Av);         // complex Rayleigh quotient
+
+      if ( i==0 ) {
+        ratio0 = ratio;
+        std::cout << GridLogMessage << "PowerMethodNonHermitian: step 0 (random v): |Av|/|v| = "
+                  << ratio << "   [lower bound on sigma_max]" << std::endl;
+      }
+      if ( (i%10==0) || (i==_MAX_ITER_EST_-1) )
+        std::cout << GridLogMessage << "PowerMethodNonHermitian: step " << i << " |Av|/|v| = " << ratio
+                  << "  Rayleigh (" << real(rq) << "," << imag(rq) << ")" << std::endl;
+
+      v = Av;
+    }
+
+    std::cout << GridLogMessage << "PowerMethodNonHermitian: |lambda_max| ~ " << ratio
+              << "  Rayleigh (" << real(rq) << "," << imag(rq) << ")"
+              << "  phase " << atan2(imag(rq),real(rq)) << " rad"
+              << "  step-0/converged = " << ratio0/ratio
+              << (ratio0/ratio > 1.2 ? "   ** non-normal: sigma_max well above |lambda_max| **"
+                                     : "   (near-normal)")
+              << std::endl;
+    return ratio;
+  }
+};
 }

Grid/algorithms/iterative/PrecConjugateResidual.h

@@ -112,7 +112,7 @@ public:
     }
 
     std::cout<<GridLogMessage<<"PrecConjugateResidual did NOT converge"<<std::endl;
-    assert(0);
+    GRID_ASSERT(0);
   }
 };
 NAMESPACE_END(Grid);

Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h

@@ -118,7 +118,7 @@ public:
 
     }
     GCRLogLevel<<"Variable Preconditioned GCR did not converge"<<std::endl;
-    //    assert(0);
+    //    GRID_ASSERT(0);
   }
 
   RealD GCRnStep(const Field &src, Field &psi,RealD rsq){
@@ -221,7 +221,7 @@ public:
       int northog = ((kp)>(mmax-1))?(mmax-1):(kp);  // if more than mmax done, we orthog all mmax history.
       for(int back=0;back<northog;back++){
 
-	int peri_back=(k-back)%mmax;   	  assert((k-back)>=0);
+	int peri_back=(k-back)%mmax;   	  GRID_ASSERT((k-back)>=0);
 
 	b=-real(innerProduct(q[peri_back],Az))/qq[peri_back];
 	p[peri_kp]=p[peri_kp]+b*p[peri_back];
@@ -231,9 +231,10 @@ public:
       qq[peri_kp]=norm2(q[peri_kp]); // could use axpy_norm
       LinalgTimer.Stop();
     }
-    assert(0); // never reached
+    GRID_ASSERT(0); // never reached
     return cp;
   }
 };
 NAMESPACE_END(Grid);
+#undef GCRLogLevel
 #endif

Grid/algorithms/iterative/PrecGeneralisedConjugateResidualNonHermitian.h

@@ -28,6 +28,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 /*  END LEGAL */
 #ifndef GRID_PREC_GCR_NON_HERM_H
 #define GRID_PREC_GCR_NON_HERM_H
+#include <Grid/algorithms/iterative/GCRCoefficients.h>
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 //VPGCR Abe and Zhang, 2005.
@@ -38,13 +39,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
 NAMESPACE_BEGIN(Grid);
 
-#define GCRLogLevel std::cout << GridLogMessage <<std::string(level,'\t')<< " Level "<<level<<" " 
+#define GCRLogLevel std::cout << GridLogMessage <<std::string(level,'\t')<< name<<" "
 
 template<class Field>
 class PrecGeneralisedConjugateResidualNonHermitian : public LinearFunction<Field> {
 public:                                                
   using LinearFunction<Field>::operator();
   RealD   Tolerance;
+  RealD   SSQ;
   Integer MaxIterations;
   int verbose;
   int mmax;
@@ -54,11 +56,38 @@ public:
   GridStopWatch PrecTimer;
   GridStopWatch MatTimer;
   GridStopWatch LinalgTimer;
+  std::string name;
+  int ZeroGuess  = 0;  // caller contract: guess is always zero => first-cycle r0 = src, skip the apply
+  // persistent GCR history (see GCRnStep)
+  GridBase           *hist_grid = nullptr;
+  std::vector<Field>  q;
+  std::vector<Field>  p;
+  std::vector<RealD>  qq;
+  int FirstCycle = 0;
 
   LinearFunction<Field>     &Preconditioner;
   LinearOperatorBase<Field> &Linop;
 
-  void Level(int lv) { level=lv; };
+  void Name(std::string _name) { name = _name; };
+
+  void Level(int n) { Name("Level " + std::to_string(n)); level = n; }
+
+  void SetZeroGuess(int z) { ZeroGuess = z; };
+  // Coefficient logging: one line per step with the step length a_k and the
+  // orthogonalisation coefficients b_{k,j}.  These are the data from which a
+  // FIXED polynomial smoother can be harvested: if they are stable from call
+  // to call, the adaptive GCR can be replaced by a stationary p(A) with the
+  // same applies and no reductions.  Off by default; boss rank prints.
+  int  LogCoeffs = 0;
+  void LogCoefficients(int l) { LogCoeffs = l; };
+  // Optional recorder of the per-step coefficients (means over calls), for
+  // replay by GCRReplaySmoother (Smoothers.h).  Records only; no effect on
+  // the iteration.
+  GCRCoefficients *Recorder = nullptr;
+  void SetCoefficientRecorder(GCRCoefficients *r) { Recorder = r; if(r) r->mmax = mmax; };
+  // Free the persistent history (e.g. when this solver is replaced by a
+  // replayed polynomial): re-made on the next call if ever needed again.
+  void ReleaseHistory(void) { q.clear(); p.clear(); qq.clear(); hist_grid = nullptr; };
 
   PrecGeneralisedConjugateResidualNonHermitian(RealD tol,Integer maxit,LinearOperatorBase<Field> &_Linop,LinearFunction<Field> &Prec,int _mmax,int _nstep) : 
     Tolerance(tol), 
@@ -67,16 +96,17 @@ public:
     Preconditioner(Prec),
     mmax(_mmax),
     nstep(_nstep)
-  { 
-    level=1;
+  {
+    Level(1);
     verbose=1;
   };
 
   void operator() (const Field &src, Field &psi){
 
-    psi=Zero();
+    //    psi=Zero();
     RealD cp, ssq,rsq;
     ssq=norm2(src);
+    SSQ=ssq;
     rsq=Tolerance*Tolerance*ssq;
       
     Field r(src.Grid());
@@ -89,11 +119,12 @@ public:
     SolverTimer.Start();
 
     steps=0;
+    FirstCycle=1;
     for(int k=0;k<MaxIterations;k++){
 
       cp=GCRnStep(src,psi,rsq);
 
-      GCRLogLevel <<"PGCR("<<mmax<<","<<nstep<<") "<< steps <<" steps cp = "<<cp<<" target "<<rsq <<std::endl;
+      GCRLogLevel <<"PGCR("<<mmax<<","<<nstep<<") "<< steps <<" steps cp = "<<sqrt(cp/ssq)<<" target "<<sqrt(rsq/ssq) <<std::endl;
 
       if(cp<rsq) {
 
@@ -113,77 +144,85 @@ public:
 
     }
     GCRLogLevel<<"Variable Preconditioned GCR did not converge"<<std::endl;
-    //    assert(0);
+    //    GRID_ASSERT(0);
   }
 
   RealD GCRnStep(const Field &src, Field &psi,RealD rsq){
 
     RealD cp;
     ComplexD a, b;
-    //    ComplexD zAz;
-    RealD zAAz;
     ComplexD rq;
 
     GridBase *grid = src.Grid();
 
+    // Only r and one scratch for the restart residual; the new p/q directions
+    // are produced directly in their persistent history slots.
     Field r(grid);
-    Field z(grid);
-    Field tmp(grid);
-    Field ttmp(grid);
     Field Az(grid);
 
     ////////////////////////////////
     // history for flexible orthog
     ////////////////////////////////
-    std::vector<Field> q(mmax,grid);
-    std::vector<Field> p(mmax,grid);
-    std::vector<RealD> qq(mmax);
+    // History arrays are PERSISTENT across calls (allocated once per grid,
+    // re-made only if the grid or mmax changes).  The per-call form
+    // std::vector<Field>(mmax,grid) built a temporary and copy-constructed it
+    // mmax times on every restart cycle -- measured ~5 ms per fine-smoother
+    // call.  Safe: every entry is written before it is read within a cycle
+    // (q[kp],p[kp] assigned before the northog loop can reach them), so no
+    // stale content is ever consumed.
+    if ( hist_grid != grid || (int)q.size() != mmax ) {
+      q.clear(); p.clear();
+      q.reserve(mmax); p.reserve(mmax);
+      for(int i=0;i<mmax;i++){ q.emplace_back(grid); p.emplace_back(grid); }
+      qq.assign(mmax, 0.0);
+      hist_grid = grid;
+    }
       
     GCRLogLevel<< "PGCR nStep("<<nstep<<")"<<std::endl;
 
     //////////////////////////////////
-    // initial guess x0 is taken as nonzero.
-    // r0=src-A x0 = src
+    // r0 = src - A x0.  ZeroGuess: on the first cycle x0==0 by caller
+    // contract (enforced here), so r0 = src exactly; skip the apply.
+    // Restart cycles (psi!=0) always do the full computation.
     //////////////////////////////////
-    MatTimer.Start();
-    Linop.Op(psi,Az);
-    //    zAz = innerProduct(Az,psi);
-    zAAz= norm2(Az);
-    MatTimer.Stop();
-    
+    if (ZeroGuess && FirstCycle) {
+      psi = Zero();
+      LinalgTimer.Start();
+      r = src;
+      LinalgTimer.Stop();
+    } else {
+      MatTimer.Start();
+      Linop.Op(psi,Az);
+      MatTimer.Stop();
+      LinalgTimer.Start();
+      r=src-Az;
+      LinalgTimer.Stop();
+    }
+    FirstCycle=0;
 
-    LinalgTimer.Start();
-    r=src-Az;
-    LinalgTimer.Stop();
-    GCRLogLevel<< "PGCR true residual r = src - A psi   "<<norm2(r) <<std::endl;
-    
     /////////////////////
     // p = Prec(r)
     /////////////////////
 
+    // p[0] = Prec(r), q[0] = A p[0], written straight into the history slots
     PrecTimer.Start();
-    Preconditioner(r,z);
+    Preconditioner(r,p[0]);
     PrecTimer.Stop();
 
     MatTimer.Start();
-    Linop.Op(z,Az);
+    Linop.Op(p[0],q[0]);
     MatTimer.Stop();
 
     LinalgTimer.Start();
 
-    //    zAz = innerProduct(Az,psi);
-    zAAz= norm2(Az);
+    qq[0]= norm2(q[0]);
 
-    //p[0],q[0],qq[0] 
-    p[0]= z;
-    q[0]= Az;
-    qq[0]= zAAz;
-    
     cp =norm2(r);
     LinalgTimer.Stop();
+    GCRLogLevel<< "PGCR true residual "<< sqrt(cp/SSQ)     <<std::endl;
 
     for(int k=0;k<nstep;k++){
-
+      GRID_TRACE("PGCR_step");
       steps++;
 
       int kp     = k+1;
@@ -193,50 +232,76 @@ public:
       LinalgTimer.Start();
       rq= innerProduct(q[peri_k],r); // what if rAr not real?
       a = rq/qq[peri_k];
+      if ( Recorder ) Recorder->RecordA(k,a);
 
       axpy(psi,a,p[peri_k],psi);         
 
       cp = axpy_norm(r,-a,q[peri_k],r);
       LinalgTimer.Stop();
+      if ( LogCoeffs ) {
+        GCRLogLevel<<"coeff["<<k<<"] a=("<<real(a)<<","<<imag(a)<<")"
+                   <<" |r|/|r0|="<<sqrt(cp/SSQ)<<std::endl;
+      }
 
-      GCRLogLevel<< "PGCR step["<<steps<<"]  resid " << cp << " target " <<rsq<<std::endl; 
+      GCRLogLevel<< "PGCR step["<<steps<<"]  resid " << sqrt(cp/SSQ)<<std::endl;
 
       if((k==nstep-1)||(cp<rsq)){
 	return cp;
       }
 
-
+      // New direction written straight into its history slot: p = Prec(r), q = A p.
       PrecTimer.Start();
-      Preconditioner(r,z);// solve Az = r
+      Preconditioner(r,p[peri_kp]);
       PrecTimer.Stop();
 
       MatTimer.Start();
-      Linop.Op(z,Az);
+      Linop.Op(p[peri_kp],q[peri_kp]);
       MatTimer.Stop();
-      //      zAz = innerProduct(Az,psi);
-      zAAz= norm2(Az);
 
       LinalgTimer.Start();
 
-      q[peri_kp]=Az;
-      p[peri_kp]=z;
-
       int northog = ((kp)>(mmax-1))?(mmax-1):(kp);  // if more than mmax done, we orthog all mmax history.
+      std::ostringstream bs;
+
+      // Classical Gram-Schmidt: every coefficient is taken against the
+      // UN-updated new q (so all northog inner products are independent and
+      // batchable), then the window is applied.  The coefficient is complex:
+      // for a non-Hermitian operator <q_j,Aq> is complex and keeping only the
+      // real part left the q's non-orthogonal.
+      // Batched: one fused kernel + one reduction for all coefficients,
+      // one fused pass per update (independent of mmax).
+      std::vector<const Field*> qwin(northog), pwin(northog);
       for(int back=0;back<northog;back++){
-
-	int peri_back=(k-back)%mmax;   	  assert((k-back)>=0);
-
-	b=-real(innerProduct(q[peri_back],Az))/qq[peri_back];
-	p[peri_kp]=p[peri_kp]+b*p[peri_back];
-	q[peri_kp]=q[peri_kp]+b*q[peri_back];
-
+	int peri_back=(k-back)%mmax;   	  GRID_ASSERT((k-back)>=0);
+	GRID_ASSERT(peri_back!=peri_kp);
+	qwin[back] = &q[peri_back];
+	pwin[back] = &p[peri_back];
+      }
+      std::vector<ComplexD> bcoef;
+      innerProductMulti(bcoef,qwin,q[peri_kp]);
+      for(int back=0;back<northog;back++){
+	int peri_back=(k-back)%mmax;
+	bcoef[back] = -bcoef[back]/qq[peri_back];
+        if ( LogCoeffs ) bs<<" b["<<back<<"]="<<bcoef[back];
+      }
+      if ( Recorder && northog ) Recorder->RecordB(k,bcoef);
+      if ( northog ) {
+	axpyMulti(p[peri_kp],bcoef,pwin);
+	qq[peri_kp]=axpyMultiNorm(q[peri_kp],bcoef,qwin);
+      } else {
+	qq[peri_kp]=norm2(q[peri_kp]);
+      }
+      if ( LogCoeffs && northog ) {
+        GCRLogLevel<<"coeff["<<k<<"]"<<bs.str()<<std::endl;
       }
       qq[peri_kp]=norm2(q[peri_kp]); // could use axpy_norm
       LinalgTimer.Stop();
     }
-    assert(0); // never reached
+    GRID_ASSERT(0); // never reached
     return cp;
   }
 };
 NAMESPACE_END(Grid);
+
+#undef GCRLogLevel
 #endif

Grid/algorithms/iterative/QuasiMinimalResidual.h

@@ -79,7 +79,7 @@ class QuasiMinimalResidual : public OperatorFunction<Field> {
 
     LinOp.Op(x,r); r = b - r;
 
-    assert(normb> 0.0);
+    GRID_ASSERT(normb> 0.0);
 
     resid = norm2(r)/normb;
     if (resid <= Tolerance) {
@@ -105,8 +105,8 @@ class QuasiMinimalResidual : public OperatorFunction<Field> {
     for (int i = 1; i <= MaxIterations; i++) {
 
       // Breakdown tests
-      assert( rho != 0.0);
-      assert( xi  != 0.0);
+      GRID_ASSERT( rho != 0.0);
+      GRID_ASSERT( xi  != 0.0);
 
       v = (1. / rho) * v_tld;
       y = (1. / rho) * y;
@@ -134,10 +134,10 @@ class QuasiMinimalResidual : public OperatorFunction<Field> {
       ep=Zep.real();
       std::cout << "Zep "<<Zep <<std::endl;
       // Complex Audit
-      assert(abs(ep)>0);
+      GRID_ASSERT(abs(ep)>0);
 
       beta = ep / delta;
-      assert(abs(beta)>0);
+      GRID_ASSERT(abs(beta)>0);
 
       v_tld = p_tld - beta * v;
       y = v_tld;
@@ -158,7 +158,7 @@ class QuasiMinimalResidual : public OperatorFunction<Field> {
       std::cout << "theta "<<theta<<std::endl;
       std::cout << "gamma "<<gamma<<std::endl;
 
-      assert(abs(gamma)> 0.0);
+      GRID_ASSERT(abs(gamma)> 0.0);
 
       eta = -eta * rho_1 * gamma* gamma / (beta * gamma_1 * gamma_1);
 
@@ -178,7 +178,7 @@ class QuasiMinimalResidual : public OperatorFunction<Field> {
       }
       std::cout << "Iteration "<<i<<" resid " << resid<<std::endl;
     }
-    assert(0);
+    GRID_ASSERT(0);
     return;                            // no convergence
   }
 #else

Grid/algorithms/iterative/SchurRedBlack.h

@@ -327,9 +327,9 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     GRID_ASSERT(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      GRID_ASSERT( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  GRID_ASSERT(  tmp.Checkerboard() ==Odd);     
 
       _Matrix.Mooee(tmp,src_o); // Extra factor of "m" in source from dumb choice of matrix norm.
     }
@@ -347,17 +347,17 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o,tmp);        assert(  tmp.Checkerboard()   ==Even);
-      src_e = src_e-tmp;               assert(  src_e.Checkerboard() ==Even);
-      _Matrix.MooeeInv(src_e,sol_e);   assert(  sol_e.Checkerboard() ==Even);
+      _Matrix.Meooe(sol_o,tmp);        GRID_ASSERT(  tmp.Checkerboard()   ==Even);
+      src_e = src_e-tmp;               GRID_ASSERT(  src_e.Checkerboard() ==Even);
+      _Matrix.MooeeInv(src_e,sol_e);   GRID_ASSERT(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e); assert(  sol_e.Checkerboard() ==Even);
-      setCheckerboard(sol,sol_o); assert(  sol_o.Checkerboard() ==Odd );
+      setCheckerboard(sol,sol_e); GRID_ASSERT(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o); GRID_ASSERT(  sol_o.Checkerboard() ==Odd );
     }
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
     {
       SchurStaggeredOperator<Matrix,Field> _HermOpEO(_Matrix);
-      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.Checkerboard()==Odd);
+      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  GRID_ASSERT(sol_o.Checkerboard()==Odd);
     };
     virtual void RedBlackSolve   (Matrix & _Matrix,const std::vector<Field> &src_o,  std::vector<Field> &sol_o)
     {
@@ -396,13 +396,13 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = Mdag * (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     GRID_ASSERT(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      GRID_ASSERT( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  GRID_ASSERT(  tmp.Checkerboard() ==Odd);     
 
       // get the right MpcDag
       SchurDiagMooeeOperator<Matrix,Field> _HermOpEO(_Matrix);
-      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.Checkerboard() ==Odd);       
+      _HermOpEO.MpcDag(tmp,src_o);     GRID_ASSERT(src_o.Checkerboard() ==Odd);       
 
     }
     virtual void RedBlackSolution(Matrix & _Matrix,const Field &sol_o, const Field &src_e,Field &sol)
@@ -416,17 +416,17 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o,tmp);          assert(  tmp.Checkerboard()   ==Even);
-      src_e_i = src_e-tmp;               assert(  src_e_i.Checkerboard() ==Even);
-      _Matrix.MooeeInv(src_e_i,sol_e);   assert(  sol_e.Checkerboard() ==Even);
+      _Matrix.Meooe(sol_o,tmp);          GRID_ASSERT(  tmp.Checkerboard()   ==Even);
+      src_e_i = src_e-tmp;               GRID_ASSERT(  src_e_i.Checkerboard() ==Even);
+      _Matrix.MooeeInv(src_e_i,sol_e);   GRID_ASSERT(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e); assert(  sol_e.Checkerboard() ==Even);
-      setCheckerboard(sol,sol_o); assert(  sol_o.Checkerboard() ==Odd );
+      setCheckerboard(sol,sol_e); GRID_ASSERT(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o); GRID_ASSERT(  sol_o.Checkerboard() ==Odd );
     }
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
     {
       SchurDiagMooeeOperator<Matrix,Field> _HermOpEO(_Matrix);
-      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.Checkerboard()==Odd);
+      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  GRID_ASSERT(sol_o.Checkerboard()==Odd);
     };
     virtual void RedBlackSolve   (Matrix & _Matrix,const std::vector<Field> &src_o,  std::vector<Field> &sol_o)
     {
@@ -461,9 +461,9 @@ namespace Grid {
         /////////////////////////////////////////////////////
         // src_o = Mdag * (source_o - Moe MeeInv source_e)
         /////////////////////////////////////////////////////
-        _Matrix.MooeeInv(src_e, tmp);   assert(   tmp.Checkerboard() == Even );
-        _Matrix.Meooe   (tmp, Mtmp);    assert(  Mtmp.Checkerboard() == Odd  );     
-        src_o -= Mtmp;                  assert( src_o.Checkerboard() == Odd  );     
+        _Matrix.MooeeInv(src_e, tmp);   GRID_ASSERT(   tmp.Checkerboard() == Even );
+        _Matrix.Meooe   (tmp, Mtmp);    GRID_ASSERT(  Mtmp.Checkerboard() == Odd  );     
+        src_o -= Mtmp;                  GRID_ASSERT( src_o.Checkerboard() == Odd  );     
       }
       
       virtual void RedBlackSolution(Matrix& _Matrix, const Field& sol_o, const Field& src_e, Field& sol)
@@ -478,18 +478,18 @@ namespace Grid {
         ///////////////////////////////////////////////////
         // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
         ///////////////////////////////////////////////////
-        _Matrix.Meooe(sol_o, tmp);         assert(     tmp.Checkerboard() == Even );
-        src_e_i = src_e - tmp;             assert( src_e_i.Checkerboard() == Even );
-        _Matrix.MooeeInv(src_e_i, sol_e);  assert(   sol_e.Checkerboard() == Even );
+        _Matrix.Meooe(sol_o, tmp);         GRID_ASSERT(     tmp.Checkerboard() == Even );
+        src_e_i = src_e - tmp;             GRID_ASSERT( src_e_i.Checkerboard() == Even );
+        _Matrix.MooeeInv(src_e_i, sol_e);  GRID_ASSERT(   sol_e.Checkerboard() == Even );
        
-        setCheckerboard(sol, sol_e); assert( sol_e.Checkerboard() == Even );
-        setCheckerboard(sol, sol_o); assert( sol_o.Checkerboard() == Odd  );
+        setCheckerboard(sol, sol_e); GRID_ASSERT( sol_e.Checkerboard() == Even );
+        setCheckerboard(sol, sol_o); GRID_ASSERT( sol_o.Checkerboard() == Odd  );
       }
 
       virtual void RedBlackSolve(Matrix& _Matrix, const Field& src_o, Field& sol_o)
       {
         NonHermitianSchurDiagMooeeOperator<Matrix,Field> _OpEO(_Matrix);
-        this->_HermitianRBSolver(_OpEO, src_o, sol_o);  assert(sol_o.Checkerboard() == Odd);
+        this->_HermitianRBSolver(_OpEO, src_o, sol_o);  GRID_ASSERT(sol_o.Checkerboard() == Odd);
       }
 
       virtual void RedBlackSolve(Matrix& _Matrix, const std::vector<Field>& src_o, std::vector<Field>& sol_o)
@@ -539,13 +539,13 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = Mpcdag *MooeeInv * (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     GRID_ASSERT(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      GRID_ASSERT( Mtmp.Checkerboard() ==Odd);     
       Mtmp=src_o-Mtmp;                 
-      _Matrix.MooeeInv(Mtmp,tmp);      assert( tmp.Checkerboard() ==Odd);     
+      _Matrix.MooeeInv(Mtmp,tmp);      GRID_ASSERT( tmp.Checkerboard() ==Odd);     
       
       // get the right MpcDag
-      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.Checkerboard() ==Odd);       
+      _HermOpEO.MpcDag(tmp,src_o);     GRID_ASSERT(src_o.Checkerboard() ==Odd);       
     }
 
     virtual void RedBlackSolution(Matrix & _Matrix,const Field &sol_o, const Field &src_e,Field &sol)
@@ -560,12 +560,12 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o,tmp);    assert(  tmp.Checkerboard()   ==Even);
-      tmp = src_e-tmp;             assert(  src_e.Checkerboard() ==Even);
-      _Matrix.MooeeInv(tmp,sol_e); assert(  sol_e.Checkerboard() ==Even);
+      _Matrix.Meooe(sol_o,tmp);    GRID_ASSERT(  tmp.Checkerboard()   ==Even);
+      tmp = src_e-tmp;             GRID_ASSERT(  src_e.Checkerboard() ==Even);
+      _Matrix.MooeeInv(tmp,sol_e); GRID_ASSERT(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e);  assert(  sol_e.Checkerboard() ==Even);
-      setCheckerboard(sol,sol_o);  assert(  sol_o.Checkerboard() ==Odd );
+      setCheckerboard(sol,sol_e);  GRID_ASSERT(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o);  GRID_ASSERT(  sol_o.Checkerboard() ==Odd );
     };
 
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
@@ -612,12 +612,12 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = Mdag * (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     GRID_ASSERT(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      GRID_ASSERT( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  GRID_ASSERT(  tmp.Checkerboard() ==Odd);     
 
       // get the right MpcDag
-      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.Checkerboard() ==Odd);       
+      _HermOpEO.MpcDag(tmp,src_o);     GRID_ASSERT(src_o.Checkerboard() ==Odd);       
     }
 
     virtual void RedBlackSolution(Matrix & _Matrix,const Field &sol_o, const Field &src_e,Field &sol)
@@ -638,12 +638,12 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o_i,tmp);    assert(  tmp.Checkerboard()   ==Even);
-      tmp = src_e-tmp;               assert(  src_e.Checkerboard() ==Even);
-      _Matrix.MooeeInv(tmp,sol_e);   assert(  sol_e.Checkerboard() ==Even);
+      _Matrix.Meooe(sol_o_i,tmp);    GRID_ASSERT(  tmp.Checkerboard()   ==Even);
+      tmp = src_e-tmp;               GRID_ASSERT(  src_e.Checkerboard() ==Even);
+      _Matrix.MooeeInv(tmp,sol_e);   GRID_ASSERT(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e);    assert(  sol_e.Checkerboard() ==Even);
-      setCheckerboard(sol,sol_o_i);  assert(  sol_o_i.Checkerboard() ==Odd );
+      setCheckerboard(sol,sol_e);    GRID_ASSERT(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o_i);  GRID_ASSERT(  sol_o_i.Checkerboard() ==Odd );
     };
 
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
@@ -684,9 +684,9 @@ namespace Grid {
         /////////////////////////////////////////////////////
         // src_o = Mdag * (source_o - Moe MeeInv source_e)
         /////////////////////////////////////////////////////
-        _Matrix.MooeeInv(src_e, tmp);   assert(   tmp.Checkerboard() == Even );
-        _Matrix.Meooe   (tmp, Mtmp);    assert(  Mtmp.Checkerboard() == Odd  );     
-        src_o -= Mtmp;                  assert( src_o.Checkerboard() == Odd  );     
+        _Matrix.MooeeInv(src_e, tmp);   GRID_ASSERT(   tmp.Checkerboard() == Even );
+        _Matrix.Meooe   (tmp, Mtmp);    GRID_ASSERT(  Mtmp.Checkerboard() == Odd  );     
+        src_o -= Mtmp;                  GRID_ASSERT( src_o.Checkerboard() == Odd  );     
       }
 
       virtual void RedBlackSolution(Matrix& _Matrix, const Field& sol_o, const Field& src_e, Field& sol)
@@ -707,12 +707,12 @@ namespace Grid {
         ///////////////////////////////////////////////////
         // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
         ///////////////////////////////////////////////////
-        _Matrix.Meooe(sol_o_i, tmp);    assert(   tmp.Checkerboard() == Even );
-        tmp = src_e - tmp;              assert( src_e.Checkerboard() == Even );
-        _Matrix.MooeeInv(tmp, sol_e);   assert( sol_e.Checkerboard() == Even );
+        _Matrix.Meooe(sol_o_i, tmp);    GRID_ASSERT(   tmp.Checkerboard() == Even );
+        tmp = src_e - tmp;              GRID_ASSERT( src_e.Checkerboard() == Even );
+        _Matrix.MooeeInv(tmp, sol_e);   GRID_ASSERT( sol_e.Checkerboard() == Even );
        
-        setCheckerboard(sol, sol_e);    assert(   sol_e.Checkerboard() == Even );
-        setCheckerboard(sol, sol_o_i);  assert( sol_o_i.Checkerboard() == Odd  );
+        setCheckerboard(sol, sol_e);    GRID_ASSERT(   sol_e.Checkerboard() == Even );
+        setCheckerboard(sol, sol_o_i);  GRID_ASSERT( sol_o_i.Checkerboard() == Odd  );
       };
 
       virtual void RedBlackSolve(Matrix& _Matrix, const Field& src_o, Field& sol_o)

Grid/algorithms/multigrid/Aggregates.h

@@ -30,6 +30,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 /*  END LEGAL */
 #pragma once
 
+#include <Grid/algorithms/iterative/PrecGeneralisedConjugateResidualNonHermitian.h>
+
 NAMESPACE_BEGIN(Grid);
 
 inline RealD AggregatePowerLaw(RealD x)
@@ -64,7 +66,21 @@ public:
   {
   };
   
-  
+  void GlobalOrthonormalise(void)
+  {
+    // Normalise all vectors
+    for(int i=0;i<nbasis; i++){
+      RealD scale = std::pow(norm2(subspace[i]),-0.5); 
+      subspace[i] = subspace[i]*scale;
+    }
+    for(int i=0;i<nbasis; i++){
+      for(int j=0;j<i; j++){
+	basisOrthogonalize(subspace,subspace[i],j);
+      }
+      RealD scale = std::pow(norm2(subspace[i]),-0.5); 
+      subspace[i] = subspace[i]*scale;
+    }
+  }  
   void Orthogonalise(void){
     CoarseScalar InnerProd(CoarseGrid); 
     //    std::cout << GridLogMessage <<" Block Gramm-Schmidt pass 1"<<std::endl;
@@ -95,7 +111,7 @@ public:
 
     RealD scale;
 
-    ConjugateGradient<FineField> CG(1.0e-2,100,false);
+    ConjugateGradient<FineField> CG(1.0e-4,2000,false);
     FineField noise(FineGrid);
     FineField Mn(FineGrid);
 
@@ -108,14 +124,16 @@ public:
       
       hermop.Op(noise,Mn); std::cout<<GridLogMessage << "noise   ["<<b<<"] <n|MdagM|n> "<<norm2(Mn)<<std::endl;
 
-      for(int i=0;i<1;i++){
+      for(int i=0;i<2;i++){
 
 	CG(hermop,noise,subspace[b]);
 
 	noise = subspace[b];
 	scale = std::pow(norm2(noise),-0.5); 
 	noise=noise*scale;
-
+	
+	hermop.Op(noise,Mn); std::cout<<GridLogMessage << "intermediate["<<i<<"] <i|MdagM|i> "<<norm2(Mn)<<std::endl;
+	
       }
 
       hermop.Op(noise,Mn); std::cout<<GridLogMessage << "filtered["<<b<<"] <f|MdagM|f> "<<norm2(Mn)<<std::endl;
@@ -124,6 +142,60 @@ public:
     }
   }
 
+  virtual void CreateSubspaceGCR(GridParallelRNG  &RNG,LinearOperatorBase<FineField> &DiracOp,int nn=nbasis)
+  {
+    RealD scale;
+
+    TrivialPrecon<FineField> simple_fine;
+    //    PrecGeneralisedConjugateResidualNonHermitian<FineField> GCR(0.001,10,DiracOp,simple_fine,30,30);
+    //    PrecGeneralisedConjugateResidualNonHermitian<FineField> GCR(0.001,10,DiracOp,simple_fine,12,12);
+    //    PrecGeneralisedConjugateResidualNonHermitian<FineField> GCR(0.001,30,DiracOp,simple_fine,12,12);
+    //    PrecGeneralisedConjugateResidualNonHermitian<FineField> GCR(0.0005,30,DiracOp,simple_fine,20,20);
+    PrecGeneralisedConjugateResidualNonHermitian<FineField> GCR(0.0005,30,DiracOp,simple_fine,10,10);
+    FineField noise(FineGrid);
+    FineField src(FineGrid);
+    FineField guess(FineGrid);
+    FineField Mn(FineGrid);
+
+    for(int b=0;b<nn;b++){
+      
+      subspace[b] = Zero();
+      gaussian(RNG,noise);
+      scale = std::pow(norm2(noise),-0.5); 
+      noise=noise*scale;
+      
+      DiracOp.Op(noise,Mn); std::cout<<GridLogMessage << "noise   ["<<b<<"] <n|Op|n> "<<innerProduct(noise,Mn)<<std::endl;
+
+      for(int i=0;i<3;i++){
+	//  void operator() (const Field &src, Field &psi){
+#if 1
+	if (i==0)std::cout << GridLogMessage << " inverting on noise "<<std::endl;
+	src = noise;
+	guess=Zero();
+	GCR(src,guess);
+	subspace[b] = guess;
+#else
+	if (i==0)std::cout << GridLogMessage << " inverting on zero "<<std::endl;
+	src=Zero();
+	guess = noise;
+	GCR(src,guess);
+	subspace[b] = guess;
+#endif
+	noise = subspace[b];
+	scale = std::pow(norm2(noise),-0.5); 
+	noise=noise*scale;
+	
+	DiracOp.Op(noise,Mn); std::cout<<GridLogMessage << "intermediate["<<i<<"] <f|Op|f> "<<innerProduct(noise,Mn)<<" <f|OpDagOp|f>"<<norm2(Mn)<<std::endl;
+
+      }
+
+      DiracOp.Op(noise,Mn); std::cout<<GridLogMessage << "filtered["<<b<<"] <f|Op|f> "<<innerProduct(noise,Mn)<<" <f|OpDagOp|f>"<<norm2(Mn)<<std::endl;
+      subspace[b]   = noise;
+
+    }
+    GlobalOrthonormalise();
+  }
+
   ////////////////////////////////////////////////////////////////////////////////////////////////
   // World of possibilities here. But have tried quite a lot of experiments (250+ jobs run on Summit)
   // and this is the best I found
@@ -160,14 +232,21 @@ public:
 
     int b =0;
     {
+      ComplexD ip;
       // Filter
       Chebyshev<FineField> Cheb(lo,hi,orderfilter);
       Cheb(hermop,noise,Mn);
       // normalise
       scale = std::pow(norm2(Mn),-0.5); 	Mn=Mn*scale;
       subspace[b]   = Mn;
-      hermop.Op(Mn,tmp); 
-      std::cout<<GridLogMessage << "filt ["<<b<<"] <n|MdagM|n> "<<norm2(tmp)<<std::endl;
+
+      hermop.Op(Mn,tmp);
+      ip= innerProduct(Mn,tmp); 
+      std::cout<<GridLogMessage << "filt ["<<b<<"] <n|Op|n> "<<norm2(tmp)<<" "<<ip<<std::endl;
+
+      hermop.AdjOp(Mn,tmp); 
+      ip = innerProduct(Mn,tmp); 
+      std::cout<<GridLogMessage << "filt ["<<b<<"] <n|AdjOp|n> "<<norm2(tmp)<<" "<<ip<<std::endl;
       b++;
     }
 
@@ -213,8 +292,18 @@ public:
 	  Mn=*Tnp;
 	  scale = std::pow(norm2(Mn),-0.5);         Mn=Mn*scale;
 	  subspace[b] = Mn;
-	  hermop.Op(Mn,tmp); 
-	  std::cout<<GridLogMessage << n<<" filt ["<<b<<"] <n|MdagM|n> "<<norm2(tmp)<<std::endl;
+
+
+	  ComplexD ip;
+
+	  hermop.Op(Mn,tmp);
+	  ip= innerProduct(Mn,tmp); 
+	  std::cout<<GridLogMessage << "filt ["<<b<<"] <n|Op|n> "<<norm2(tmp)<<" "<<ip<<std::endl;
+
+	  hermop.AdjOp(Mn,tmp); 
+	  ip = innerProduct(Mn,tmp); 
+	  std::cout<<GridLogMessage << "filt ["<<b<<"] <n|AdjOp|n> "<<norm2(tmp)<<" "<<ip<<std::endl;
+	  
 	  b++;
 	}
 
@@ -226,7 +315,7 @@ public:
 	  
       }
     }
-    assert(b==nn);
+    GRID_ASSERT(b==nn);
   }
 
 

Grid/algorithms/multigrid/BlockCyclic.h

@@ -0,0 +1,221 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/BlockCyclic.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// BlockCyclicLayout: the index arithmetic of a 2D block-cyclic distribution
+// of an N x N matrix over a Pr x Pc logical process grid with block size nb.
+//
+// This is stage 1 of the 2D distributed dense inverse
+// (documentation/DistributedDenseInverse2D.tex).  It is deliberately
+// COMMUNICATOR-FREE: every mapping is a static pure function of
+// (N, nb, Pr, Pc), so the whole layout is exhaustively unit-testable on one
+// rank with no MPI in the loop (Test_blockcyclic).  A thin instance layer
+// binds a world rank to a grid coordinate and caches local extents.
+//
+// Conventions (fixed here, relied on by every later stage):
+//
+//  * Global block b of a dimension with Pg processes is owned by process
+//    coordinate  b % Pg  (ScaLAPACK csrc=0), and is that process's local
+//    block  b / Pg.
+//  * Rank <-> grid coordinate is ROW MAJOR over the process grid:
+//        rank = p*Pc + q ,  p = rank/Pc ,  q = rank%Pc .
+//    The eventual ring transport must construct its neighbour tables with
+//    the same convention.
+//  * Local storage is COLUMN MAJOR with ld = mloc, matching BlockRows:
+//    local element (i,j) lives at  data[i + j*mloc].
+//  * The trailing partial block (N % nb != 0) belongs to the owner of the
+//    last full-size block position; only that one block is short.
+//
+// Element (gi,gj) therefore lives on grid coordinate
+//    ( (gi/nb) % Pr , (gj/nb) % Pc )
+// at local coordinate
+//    ( ((gi/nb)/Pr)*nb + gi%nb , ((gj/nb)/Pc)*nb + gj%nb ).
+//
+// Everything here is host-side integer arithmetic; nothing allocates.
+///////////////////////////////////////////////////////////////////////////////
+
+class BlockCyclicLayout
+{
+public:
+  ///////////////////////////////////////////////////////////////////////////
+  // Closest-to-square factorisation Pr*Pc == P with Pr <= Pc.
+  // For fixed P the per-rank SUMMA volume  N^2 (1/Pr + 1/Pc)  is minimised
+  // at the most square grid.  P=288 -> 16 x 18.
+  ///////////////////////////////////////////////////////////////////////////
+  // Nearest-to-square factorisation with Pr >= Pc.  The orientation matters
+  // for SUMMA: B panels travel along process columns in slots of
+  // S = ceil(Pc/Pr) panels, padded; with Pr >= Pc, S == 1 and there is no
+  // padding.  (288 as 16x18 had S=2 -- half the B-ring bytes were zeros;
+  // 18x16 moves ~30% fewer bytes for the same inverse.)
+  static void ChooseProcessGrid(int P, int &Pr, int &Pc)
+  {
+    GRID_ASSERT(P >= 1);
+    int r = 1;
+    for(int f=1; (int64_t)f*f <= (int64_t)P; f++)
+      if ( P % f == 0 ) r = f;
+    Pc = r;
+    Pr = P / r;
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Number of rows (or columns) of a dimension of global extent N, block nb,
+  // owned by process coordinate p of Pg.  ScaLAPACK "numroc", csrc=0.
+  ///////////////////////////////////////////////////////////////////////////
+  static int64_t NumLocal(int64_t N, int64_t nb, int p, int Pg)
+  {
+    GRID_ASSERT(N  >= 0);
+    GRID_ASSERT(nb >= 1);
+    GRID_ASSERT(p  >= 0);
+    GRID_ASSERT(p  <  Pg);
+    int64_t nblocks = N / nb;            // full blocks
+    int64_t extra   = N % nb;            // trailing partial block
+    int64_t full    = nblocks / Pg;      // full blocks everyone owns
+    int64_t rem     = nblocks % Pg;      // coords [0,rem) own one more
+    int64_t n = full*nb;
+    if ( p <  (int)rem ) n += nb;        // an extra full block
+    if ( p == (int)rem ) n += extra;     // the partial block, if any
+    return n;
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Global index -> (owner coordinate, local index) in one dimension.
+  ///////////////////////////////////////////////////////////////////////////
+  static void GlobalToLocal(int64_t g, int64_t nb, int Pg,
+                            int &owner, int64_t &loc)
+  {
+    GRID_ASSERT(g >= 0);
+    int64_t b = g / nb;                  // global block
+    owner = (int)(b % Pg);
+    loc   = (b / Pg)*nb + (g % nb);
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // (process coordinate, local index) -> global index in one dimension.
+  // Inverse of GlobalToLocal on the owned set.
+  ///////////////////////////////////////////////////////////////////////////
+  static int64_t LocalToGlobal(int64_t l, int64_t nb, int p, int Pg)
+  {
+    GRID_ASSERT(l >= 0);
+    int64_t lb = l / nb;                 // local block
+    int64_t b  = lb*Pg + p;              // global block
+    return b*nb + (l % nb);
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Instance layer: bind a rank of a Pr x Pc grid.
+  ///////////////////////////////////////////////////////////////////////////
+  int64_t N;         // global matrix dimension (square)
+  int64_t nb;        // block size
+  int     Pr, Pc;    // process grid
+  int     me;        // world rank within the grid, row major
+  int     prow, pcol;// my grid coordinate
+  int64_t mloc,nloc; // my local extents; storage column major, ld = mloc
+
+  BlockCyclicLayout(int64_t N_, int64_t nb_, int Pr_, int Pc_, int me_)
+  {
+    N  = N_;
+    nb = nb_;
+    Pr = Pr_;
+    Pc = Pc_;
+    me = me_;
+    GRID_ASSERT( N  >= 0 );
+    GRID_ASSERT( nb >= 1 );
+    GRID_ASSERT( Pr >= 1 );
+    GRID_ASSERT( Pc >= 1 );
+    GRID_ASSERT( me >= 0 );
+    GRID_ASSERT( me <  Pr*Pc );
+    prow = me / Pc;                      // ROW MAJOR rank convention
+    pcol = me % Pc;
+    mloc = NumLocal(N, nb, prow, Pr);
+    nloc = NumLocal(N, nb, pcol, Pc);
+  }
+
+  // Owning rank of global element (gi,gj), row-major rank convention.
+  int OwnerRank(int64_t gi, int64_t gj) const
+  {
+    int pr,pc; int64_t li,lj;
+    GlobalToLocal(gi, nb, Pr, pr, li);
+    GlobalToLocal(gj, nb, Pc, pc, lj);
+    return pr*Pc + pc;
+  }
+
+  // My local storage offset of global element (gi,gj).
+  // The caller must know I own it; asserted, not assumed.
+  int64_t LocalOffset(int64_t gi, int64_t gj) const
+  {
+    int pr,pc; int64_t li,lj;
+    GlobalToLocal(gi, nb, Pr, pr, li);
+    GlobalToLocal(gj, nb, Pc, pc, lj);
+    GRID_ASSERT( pr == prow );
+    GRID_ASSERT( pc == pcol );
+    return li + lj*mloc;                 // column major, ld = mloc
+  }
+
+  // Do I own global element (gi,gj)?
+  int Owns(int64_t gi, int64_t gj) const
+  {
+    return OwnerRank(gi,gj) == me;
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Block-aligned global range [g0,g1) -> my contiguous local range [l0,l1).
+  //
+  // For fixed owner p the local index is monotone in the global index, so a
+  // coordinate's owned elements of ANY global range are contiguous in local
+  // storage; and for a BLOCK-ALIGNED range the bounds are exactly
+  // NumLocal(g0) and NumLocal(g1), because NumLocal(g,...) counts the owned
+  // elements below g.  This is what lets a windowed product view the local
+  // sub-matrix of a global window as &data[l0 + c0*mloc] with the SAME ld --
+  // no gather, no copy.  Verified exhaustively in Test_blockcyclic T7.
+  //
+  // g0 must be a block multiple; g1 a block multiple or N itself.
+  ///////////////////////////////////////////////////////////////////////////
+  static void RangeToLocal(int64_t g0, int64_t g1,
+                           int64_t N, int64_t nb, int p, int Pg,
+                           int64_t &l0, int64_t &l1)
+  {
+    GRID_ASSERT( 0 <= g0 );
+    GRID_ASSERT( g0 <= g1 );
+    GRID_ASSERT( g1 <= N );
+    GRID_ASSERT( g0 % nb == 0 );
+    GRID_ASSERT( (g1 % nb == 0) || (g1 == N) );
+    l0 = NumLocal(g0, nb, p, Pg);
+    l1 = NumLocal(g1, nb, p, Pg);
+  }
+
+  // Instance forms, rows and columns of my own coordinate.
+  void RowRange(int64_t g0, int64_t g1, int64_t &l0, int64_t &l1) const
+  { RangeToLocal(g0,g1,N,nb,prow,Pr,l0,l1); }
+  void ColRange(int64_t g0, int64_t g1, int64_t &l0, int64_t &l1) const
+  { RangeToLocal(g0,g1,N,nb,pcol,Pc,l0,l1); }
+
+  // Size of global block b (the trailing block may be short).
+  int64_t BlockSize(int64_t b) const
+  {
+    int64_t lo = b*nb;
+    GRID_ASSERT( lo < N );
+    return std::min(N, lo+nb) - lo;
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/BlockCyclicRedistribute.h

@@ -0,0 +1,283 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/BlockCyclicRedistribute.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/multigrid/BlockCyclicSumma.h>
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// My rows of a distributed dense matrix in the 1D rank-major row layout:
+// rank r owns contiguous global rows [rowStart[r], rowStart[r+1]) of an
+// N x N matrix, stored rows x cols column major, ld = rows; element (i,j)
+// at data[i + j*ld].  This is the layout the stencil->dense import
+// produces and the apply slab consumes; the 2D inverse slots between them
+// through the redistribution below.
+///////////////////////////////////////////////////////////////////////////////
+class BlockRows
+{
+public:
+  deviceVector<ComplexD> data;
+  int64_t                rows;
+  int64_t                cols;
+  int64_t                ld;
+
+  BlockRows()
+  {
+    rows = 0;
+    cols = 0;
+    ld   = 0;
+  }
+  void Resize(int64_t r, int64_t c)
+  {
+    rows = r;
+    cols = c;
+    ld   = r;
+    data.resize((uint64_t)r*c);
+  }
+  ComplexD *ColumnWindow(int64_t col0)
+  {
+    GRID_ASSERT( col0 >= 0 );
+    GRID_ASSERT( col0 <= cols );
+    return &data[(uint64_t)col0*ld];
+  }
+};
+
+///////////////////////////////////////////////////////////////////////////////
+// Stage 4 of the 2D distributed dense inverse: redistribution between the
+// 1D rank-major row layout (BlockRows above) and the 2D block-cyclic
+// layout.
+//
+// This is what lets the EXISTING stencil->dense import, its certificate,
+// the fp32 slab conversion and the apply path all remain byte-for-byte
+// untouched: the 2D inverse slots between them as
+//
+//    RowsToCyclic -> BlockCyclicSchurInverse::Invert -> CyclicToRows
+//
+// Volume is one matrix pass each way -- N^2/P elements per rank (~1 GB at
+// production), trivial against the inversion itself.
+//
+// Transport: PURE POINT-TO-POINT, like everything else in this stack.
+// Ranks exchange in a round-robin TOURNAMENT (the circle method, on an odd
+// modulus M so it covers every pair exactly once for any P, with byes):
+// at round r, ranks x and y are partners iff x+y == r (mod M).  Each
+// meeting handles both directed edges of the pair in ONE SendToRecvFrom,
+// padded to the larger of the two edge sizes -- SendToRecvFrom carries a
+// single byte count for both directions, and both endpoints compute the
+// same max from the shared descriptors, so there is no asymmetric-size
+// case and no zero-count shape.  Pairs with nothing to exchange skip the
+// round, decided identically at both ends.
+//
+// Element enumeration within an edge is canonical -- ascending global
+// column outer, ascending global row inner -- and each endpoint builds its
+// OWN local offset tables from the shared descriptors, so no index data is
+// ever transmitted.  The round trip is BITWISE exact (pure data movement,
+// no arithmetic): Test_schur2d_redist proves it.
+///////////////////////////////////////////////////////////////////////////////
+
+class BlockCyclicRedistribute
+{
+public:
+  /////////////////////////////////////////////////////////////////////////
+  // The directed edge (1D rank r1, 2D rank r2): global rows of r1's range
+  // whose row-block coordinate is r2's prow; ALL global columns whose
+  // column-block coordinate is r2's pcol.  Every rank can enumerate any
+  // edge from (rowStart, layout) alone.
+  /////////////////////////////////////////////////////////////////////////
+  static void EdgeRows(const std::vector<int64_t> &rowStart, int r1,
+                       const BlockCyclicLayout &L, int r2,
+                       std::vector<int64_t> &rows)
+  {
+    rows.clear();
+    int p = r2 / L.Pc;                       // row-major rank convention
+    for(int64_t i=rowStart[r1]; i<rowStart[r1+1]; i++)
+      if ( (int)((i/L.nb) % L.Pr) == p ) rows.push_back(i);
+  }
+  static void EdgeCols(const BlockCyclicLayout &L, int r2,
+                       std::vector<int64_t> &cols)
+  {
+    cols.clear();
+    int q = r2 % L.Pc;
+    for(int64_t b=0; b*L.nb<L.N; b++){
+      if ( (int)(b % L.Pc) != q ) continue;
+      int64_t g0=b*L.nb, g1=std::min(L.N,(b+1)*L.nb);
+      for(int64_t j=g0;j<g1;j++) cols.push_back(j);
+    }
+  }
+
+  /////////////////////////////////////////////////////////////////////////
+  // Gather/scatter one edge between a matrix (device, column major, ld)
+  // and a dense edge buffer, through device offset tables.
+  //   buffer(a,b) = elem(rows[a], cols[b]),  a fastest.
+  /////////////////////////////////////////////////////////////////////////
+  static void MoveEdge(int toBuffer,
+                       ComplexD *mat, int64_t ld,
+                       const std::vector<int64_t> &roff,  // per-row offset in mat
+                       const std::vector<int64_t> &coff,  // per-col offset in mat
+                       ComplexD *buf)
+  {
+    int64_t nr = roff.size();
+    int64_t nc = coff.size();
+    if ( !(nr && nc) ) return;
+    deviceVector<int64_t> dro(nr), dco(nc);
+    acceleratorCopyToDevice((void *)&roff[0], (void *)&dro[0], nr*sizeof(int64_t));
+    acceleratorCopyToDevice((void *)&coff[0], (void *)&dco[0], nc*sizeof(int64_t));
+    int64_t *ro = &dro[0];
+    int64_t *co = &dco[0];
+    if ( toBuffer ) {
+      accelerator_for(idx, (uint64_t)(nr*nc), 1, {
+        int64_t b = idx / nr;
+        int64_t a = idx - b*nr;
+        buf[a + b*nr] = mat[ ro[a] + co[b]*ld ];
+      });
+    } else {
+      accelerator_for(idx, (uint64_t)(nr*nc), 1, {
+        int64_t b = idx / nr;
+        int64_t a = idx - b*nr;
+        mat[ ro[a] + co[b]*ld ] = buf[a + b*nr];
+      });
+    }
+  }
+
+  /////////////////////////////////////////////////////////////////////////
+  // My offset tables for an edge, on whichever side I am.
+  /////////////////////////////////////////////////////////////////////////
+  static void Offsets1D(const std::vector<int64_t> &rows,
+                        const std::vector<int64_t> &cols,
+                        int64_t row0,
+                        std::vector<int64_t> &roff, std::vector<int64_t> &coff)
+  {
+    roff.resize(rows.size());  coff.resize(cols.size());
+    for(uint64_t a=0;a<rows.size();a++) roff[a] = rows[a]-row0;   // local row
+    for(uint64_t b=0;b<cols.size();b++) coff[b] = cols[b];        // global col
+  }
+  static void Offsets2D(const BlockCyclicLayout &L,
+                        const std::vector<int64_t> &rows,
+                        const std::vector<int64_t> &cols,
+                        std::vector<int64_t> &roff, std::vector<int64_t> &coff)
+  {
+    roff.resize(rows.size());  coff.resize(cols.size());
+    for(uint64_t a=0;a<rows.size();a++){
+      int p; int64_t l;
+      BlockCyclicLayout::GlobalToLocal(rows[a], L.nb, L.Pr, p, l);
+      GRID_ASSERT( p == L.prow );
+      roff[a] = l;
+    }
+    for(uint64_t b=0;b<cols.size();b++){
+      int q; int64_t l;
+      BlockCyclicLayout::GlobalToLocal(cols[b], L.nb, L.Pc, q, l);
+      GRID_ASSERT( q == L.pcol );
+      coff[b] = l;
+    }
+  }
+
+  /////////////////////////////////////////////////////////////////////////
+  // The worker.  dir=+1 : 1D rows -> block cyclic ;  dir=-1 : back.
+  /////////////////////////////////////////////////////////////////////////
+  static void Redistribute(int dir, GridBase *grid,
+                           const std::vector<int64_t> &rowStart,
+                           ComplexD *rows1d, int64_t myrows,
+                           BlockCyclicMatrix &A)
+  {
+    BlockCyclicLayout &L = A.layout;
+    int P  = grid->ProcessorCount();
+    int me = grid->ThisRank();
+    GRID_ASSERT( (int)rowStart.size() == P+1 );
+    GRID_ASSERT( rowStart[P] == L.N );
+    GRID_ASSERT( rowStart[me+1]-rowStart[me] == myrows );
+    int64_t row0 = rowStart[me];
+    int64_t ld1  = myrows ? myrows : 1;
+
+    std::vector<int64_t> rows, cols, roff, coff;
+    deviceVector<ComplexD> sbuf(1), rbuf(1);
+
+    ///////////////////////////////////////////////////////////////////////
+    // Self edge: purely local, via a bounce buffer (shares all the code).
+    ///////////////////////////////////////////////////////////////////////
+    EdgeRows(rowStart, me, L, me, rows);
+    EdgeCols(L, me, cols);
+    if ( rows.size() && cols.size() ){
+      uint64_t ne = rows.size()*cols.size();
+      if ( sbuf.size() < ne ) sbuf.resize(ne);
+      std::vector<int64_t> roff2, coff2;
+      Offsets1D(rows, cols, row0, roff, coff);
+      Offsets2D(L,  rows, cols, roff2, coff2);
+      if ( dir > 0 ) {
+        MoveEdge(1, rows1d,     ld1,     roff,  coff,  &sbuf[0]);
+        MoveEdge(0, &A.data[0], L.mloc,  roff2, coff2, &sbuf[0]);
+      } else {
+        MoveEdge(1, &A.data[0], L.mloc,  roff2, coff2, &sbuf[0]);
+        MoveEdge(0, rows1d,     ld1,     roff,  coff,  &sbuf[0]);
+      }
+    }
+
+    ///////////////////////////////////////////////////////////////////////
+    // Tournament over all pairs: odd modulus M, partner = (r - me) mod M.
+    // Every unordered pair meets exactly once; partner==me or >=P is a bye.
+    ///////////////////////////////////////////////////////////////////////
+    int M = (P%2) ? P : P+1;
+    for(int r=0;r<M;r++){
+      int partner = (int)(((int64_t)r - me + 2L*M) % M);
+      if ( partner == me || partner >= P ) continue;
+
+      // outbound edge: my (dir>0 ? 1D rows : 2D data) -> partner
+      // inbound  edge: partner -> my (dir>0 ? 2D data : 1D rows)
+      std::vector<int64_t> orow, ocol, irow, icol;
+      if ( dir > 0 ) { EdgeRows(rowStart, me,      L, partner, orow); EdgeCols(L, partner, ocol);
+                       EdgeRows(rowStart, partner, L, me,      irow); EdgeCols(L, me,      icol); }
+      else           { EdgeRows(rowStart, partner, L, me,      orow); EdgeCols(L, me,      ocol);
+                       EdgeRows(rowStart, me,      L, partner, irow); EdgeCols(L, partner, icol); }
+
+      uint64_t nout = orow.size()*ocol.size();
+      uint64_t nin  = irow.size()*icol.size();
+      if ( !(nout || nin) ) continue;              // both ends compute this identically
+
+      uint64_t nmax = std::max(nout,nin);          // symmetric padded transfer
+      if ( sbuf.size() < nmax ) sbuf.resize(nmax);
+      if ( rbuf.size() < nmax ) rbuf.resize(nmax);
+
+      if ( nout ){
+        if ( dir > 0 ) { Offsets1D(orow, ocol, row0, roff, coff);
+                         MoveEdge(1, rows1d, ld1, roff, coff, &sbuf[0]); }
+        else           { Offsets2D(L, orow, ocol, roff, coff);
+                         MoveEdge(1, &A.data[0], L.mloc, roff, coff, &sbuf[0]); }
+      }
+      grid->SendToRecvFrom((void *)&sbuf[0], partner,
+                           (void *)&rbuf[0], partner,
+                           nmax*sizeof(ComplexD));
+      if ( nin ){
+        if ( dir > 0 ) { Offsets2D(L, irow, icol, roff, coff);
+                         MoveEdge(0, &A.data[0], L.mloc, roff, coff, &rbuf[0]); }
+        else           { Offsets1D(irow, icol, row0, roff, coff);
+                         MoveEdge(0, rows1d, ld1, roff, coff, &rbuf[0]); }
+      }
+    }
+  }
+
+  static void RowsToCyclic(GridBase *grid, const std::vector<int64_t> &rowStart,
+                           ComplexD *rows1d, int64_t myrows, BlockCyclicMatrix &A)
+  { Redistribute(+1, grid, rowStart, rows1d, myrows, A); }
+
+  static void CyclicToRows(GridBase *grid, const std::vector<int64_t> &rowStart,
+                           BlockCyclicMatrix &A, ComplexD *rows1d, int64_t myrows)
+  { Redistribute(-1, grid, rowStart, rows1d, myrows, A); }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/BlockCyclicSchurInverse.h

@@ -0,0 +1,446 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/BlockCyclicSchurInverse.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/blas/BatchedInverse.h>
+#include <Grid/algorithms/multigrid/BlockCyclicSumma.h>
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// Stage 3 of the 2D distributed dense inverse: the recursive Schur
+// complement on a block-cyclic matrix, in place.
+//
+// The recursion splits the GLOBAL INDEX RANGE at the block boundary
+// nearest the midpoint -- not the rank range as the retired 1D
+// RecursiveSchurInverse did -- so every rank owns part of every sub-block
+// at every depth, and the ownership gating (inI/inJ, dummy operands,
+// zero-width rank ranges) a 1D scheme needs has no analogue here.
+//
+//   I = [c0,m)   J = [m,c1)      (block-aligned, m the mid block boundary)
+//   1. recurse I  :  A11 -> A11inv                        (in place)
+//   2. Bt = A11inv . A12                                  (scratch, I x J)
+//   3. Ct = A21 . A11inv                                  (scratch, J x I)
+//   4. A22 -= A21 . Bt          == S                      (in place)
+//   5. recurse J  :  S -> Sinv                            (in place)
+//   6. Tt = Sinv . Ct                                     (scratch, J x I)
+//   7. Ut = Bt . Sinv                                     (scratch, I x J)
+//   8. A11 += Ut . Ct           == X11                    (in place)
+//   9. A12 = -Ut ,  A21 = -Tt                             (window copies)
+//
+// SCRATCH SHARING.  Four full-size block-cyclic scratch matrices (Bt, Ct,
+// Tt, Ut) serve the ENTIRE tree, used through windows.  This is safe at
+// every depth because of a window-disjointness invariant:
+//
+//   * every temporary of a node has its row range in one half of the
+//     node's window and its column range in the other (I x J or J x I);
+//   * everything any DESCENDANT touches -- its A windows and its own
+//     temporaries -- has BOTH ranges inside a single half (I x I during
+//     step 1, J x J during step 5).
+//
+// Hence a descendant window and a live ancestor temporary always differ in
+// at least one dimension by disjoint ranges.  Only Bt and Ct are live
+// across the step-5 recursion (Tt, Ut are written after it), and both are
+// covered by the invariant.
+//
+// LEAF.  A leaf is a single diagonal block, and block (b,b) of a
+// block-cyclic layout lives ENTIRELY on rank (b%Pr, b%Pc).  The leaf
+// inversion is therefore purely local -- pack the strided block dense,
+// GridBLASInverse, unpack -- with NO communication and no assembly.
+// Successive leaves cycle over ranks, so leaf work is naturally spread.
+//
+// COMMUNICATION.  Every transfer in the whole inversion is a
+// SendToRecvFrom inside BlockCyclicSumma's rings: pure point-to-point, no
+// collectives on the critical path, deterministic summation order (so
+// repeated inversions are bitwise identical).  ReportTelemetry() is the
+// one optional exception: it performs reductions, and is only ever called
+// explicitly by a caller who wants the numbers.
+//
+// NUMERICS.  No pivoting: every A11 and every Schur complement met on the
+// way down must be non-singular.  The growth telemetry stands in for
+// pivoting.
+///////////////////////////////////////////////////////////////////////////////
+
+class BlockCyclicSchurInverse
+{
+public:
+  BlockCyclicSumma        SUMMA;
+  GridBLASInverse         INV;
+
+  // Telemetry: accumulated LOCALLY, no comms unless ReportTelemetry().
+  double                  telLeafMaxInv;
+  uint64_t                nLeaf;
+  // BIG LEAVES.  Below span s blocks a sub-block lives on <= s of the Pr
+  // process rows / s of the Pc columns; when s is small relative to the grid
+  // the SUMMA rings run on a few ranks while the rest block in their next
+  // SendToRecvFrom.  Instead: gather the (s*nb)^2 sub-block to one rank,
+  // invert locally (inverseLU), scatter back.  Fires only while
+  // span < min(Pr,Pc) -- when the whole grid participates the rings are not
+  // degenerate and gathering would only concentrate memory (and at the top
+  // of the tree would gather the whole matrix).  Larger s concentrates more
+  // memory on the root; smaller s loses ring parallelism.
+  int                     leafSpan = 9;
+  uint64_t                nBigLeaf = 0;  int64_t maxBigW = 0;
+  double                  tBigGather = 0, tBigInv = 0, tBigScatter = 0;
+  uint64_t                nNode;
+  double                  tLeaf;
+  double                  tGemm;      // wall in Multiply calls (comms+gemm)
+  double                  tCopy;
+
+  BlockCyclicSchurInverse()
+  {
+    telLeafMaxInv = 0.0;
+    nLeaf = nNode = 0;
+    tLeaf = tGemm = tCopy = 0.0;
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Window copy-scale:  Dst[i0:i1, j0:j1] = alpha * Src[same window].
+  // Both share one layout, so the local bands coincide; pure local kernel.
+  ///////////////////////////////////////////////////////////////////////////
+  void WindowCopyScale(ComplexD alpha,
+                       BlockCyclicMatrix &Src, BlockCyclicMatrix &Dst,
+                       int64_t i0, int64_t i1, int64_t j0, int64_t j1)
+  {
+    BlockCyclicLayout &L = Dst.layout;
+    GRID_ASSERT( Src.layout.N==L.N && Src.layout.nb==L.nb );
+    GRID_ASSERT( Src.layout.Pr==L.Pr && Src.layout.Pc==L.Pc );
+    int64_t li0,li1, lj0,lj1;
+    L.RowRange(i0,i1, li0,li1);
+    L.ColRange(j0,j1, lj0,lj1);
+    int64_t m = li1-li0, n = lj1-lj0;
+    if ( !(m && n) ) return;
+    ComplexD *src = Src.LocalWindow(li0,lj0);
+    ComplexD *dst = Dst.LocalWindow(li0,lj0);
+    int64_t ldS = Src.layout.mloc;
+    int64_t ldD = L.mloc;
+    tCopy -= usecond();
+    accelerator_for(idx, (uint64_t)(m*n), 1, {
+      int64_t jj = idx / m;
+      int64_t ii = idx - jj*m;
+      dst[ii + jj*ldD] = alpha*src[ii + jj*ldS];
+    });
+    tCopy += usecond();
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Leaf: single diagonal block (b,b), entirely on rank (b%Pr, b%Pc).
+  // Local pack -> dense inverse -> unpack; every other rank does nothing
+  // and needs no synchronisation: the next SUMMA's rings pair them up.
+  ///////////////////////////////////////////////////////////////////////////
+  void Leaf(BlockCyclicMatrix &A, int64_t b)
+  {
+    GRID_TRACE("SchurLeaf");
+    BlockCyclicLayout &L = A.layout;
+    nLeaf++;
+    if ( (int)(b % L.Pr) != L.prow ) return;
+    if ( (int)(b % L.Pc) != L.pcol ) return;
+
+    tLeaf -= usecond();
+    int64_t g0 = b*L.nb;
+    int64_t g1 = std::min(L.N, g0+L.nb);
+    int64_t w  = g1-g0;
+    int64_t lr0,lr1, lc0,lc1;
+    L.RowRange(g0,g1, lr0,lr1);
+    L.ColRange(g0,g1, lc0,lc1);
+    GRID_ASSERT( lr1-lr0 == w );
+    GRID_ASSERT( lc1-lc0 == w );
+
+    // Pack the strided block dense (inverseBatched assumes lda == w).
+    deviceVector<ComplexD> dense((uint64_t)w*w);
+    {
+      ComplexD *src = A.LocalWindow(lr0,lc0);
+      ComplexD *dst = &dense[0];
+      int64_t   ld  = L.mloc;
+      accelerator_for(idx, (uint64_t)(w*w), 1, {
+        int64_t jj = idx / w;
+        int64_t ii = idx - jj*w;
+        dst[ii + jj*w] = src[ii + jj*ld];
+      });
+    }
+    {
+      deviceVector<ComplexD*> bp(1);
+      std::vector<ComplexD*>  ptr(1);
+      ptr[0] = &dense[0];
+      acceleratorCopyToDevice(&ptr[0], &bp[0], sizeof(ComplexD*));
+      INV.inverseBatched(w, bp);
+    }
+    {
+      ComplexD *src = &dense[0];
+      ComplexD *dst = A.LocalWindow(lr0,lc0);
+      int64_t   ld  = L.mloc;
+      accelerator_for(idx, (uint64_t)(w*w), 1, {
+        int64_t jj = idx / w;
+        int64_t ii = idx - jj*w;
+        dst[ii + jj*ld] = src[ii + jj*w];
+      });
+    }
+    // Growth telemetry, local only.
+    {
+      std::vector<ComplexD> h((uint64_t)w*w);
+      acceleratorCopyFromDevice(&dense[0], &h[0], h.size()*sizeof(ComplexD));
+      double mx = 0.0;
+      for(auto &z : h){
+        double re=z.real(), im=z.imag();
+        mx = std::max(mx, re*re+im*im);
+      }
+      telLeafMaxInv = std::max(telLeafMaxInv, std::sqrt(mx));
+    }
+    tLeaf += usecond();
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // BIG LEAF on block range [b0,b1): gather to root = owner of block (b0,b0),
+  // invert there, scatter back.  Rank q's piece of the sub-block is the
+  // contiguous local window RowRange(c0,c1) x ColRange(c0,c1); its local row
+  // ii maps to global row (brq0 + (ii/nb)*Pr)*nb + ii%nb where brq0 is q's
+  // first block row >= b0 (closed form: no tables).  Transport is pairwise
+  // SendToRecvFrom with root (symmetric byte count: the reverse direction
+  // carries a same-size dummy -- a leaf-local cost, accepted for simplicity).
+  ///////////////////////////////////////////////////////////////////////////
+  static int64_t FirstBlock(int64_t b0, int p, int Pg){ int64_t r = ((b0 % Pg) <= p) ? b0 - (b0 % Pg) + p : b0 - (b0 % Pg) + Pg + p; return r; }
+  void BigLeaf(BlockCyclicMatrix &A, int64_t b0, int64_t b1)
+  {
+    GRID_TRACE("SchurBigLeaf");
+    BlockCyclicLayout &L = A.layout;
+    GridBase *grid = A.grid;
+    const int Pr=L.Pr, Pc=L.Pc, nb=(int)L.nb;
+    GRID_ASSERT( L.me == L.prow*Pc + L.pcol );          // rank convention shared with the SUMMA rings
+    const int64_t c0 = b0*L.nb, c1 = std::min(L.N, b1*L.nb), W = c1-c0;
+    const int root = (int)((b0%Pr)*Pc + (b0%Pc));
+    const int me   = L.me;
+    nBigLeaf++; maxBigW = std::max(maxBigW, W);
+
+    // my piece
+    int64_t lr0,lr1,lc0,lc1; L.RowRange(c0,c1,lr0,lr1); L.ColRange(c0,c1,lc0,lc1);
+    const int64_t mq = lr1-lr0, nq = lc1-lc0;
+
+    deviceVector<ComplexD> dense;            // root only: W x W column major
+    deviceVector<ComplexD> piece, dummy;     // piece: my mq x nq contiguous; dummy: reverse-direction filler
+    if ( me == root ) dense.resize((uint64_t)W*W);
+
+    auto pack_piece = [&](ComplexD *dst, int64_t m, int64_t n, int64_t r0, int64_t cc0){
+      ComplexD *src = A.LocalWindow(r0,cc0); const int64_t ld = L.mloc;
+      accelerator_for(idx,(uint64_t)(m*n),1,{ int64_t jj=idx/m, ii=idx-jj*m; dst[ii+jj*m] = src[ii+jj*ld]; });
+    };
+    auto unpack_piece = [&](ComplexD *src, int64_t m, int64_t n, int64_t r0, int64_t cc0){
+      ComplexD *dst = A.LocalWindow(r0,cc0); const int64_t ld = L.mloc;
+      accelerator_for(idx,(uint64_t)(m*n),1,{ int64_t jj=idx/m, ii=idx-jj*m; dst[ii+jj*ld] = src[ii+jj*m]; });
+    };
+    // root: piece of rank q  <->  dense, via the closed-form block map
+    auto root_place = [&](ComplexD *pc, int64_t m, int64_t n, int q, int to_dense){
+      const int pq=q/Pc, cq=q%Pc;
+      const int64_t brq0=FirstBlock(b0,pq,Pr), bcq0=FirstBlock(b0,cq,Pc);
+      ComplexD *dn = &dense[0]; const int64_t WW=W, NB=nb, PR=Pr, PC=Pc, C0=c0;
+      accelerator_for(idx,(uint64_t)(m*n),1,{
+        int64_t jj=idx/m, ii=idx-jj*m;
+        int64_t gr = (brq0 + (ii/NB)*PR)*NB + ii%NB - C0;
+        int64_t gc = (bcq0 + (jj/NB)*PC)*NB + jj%NB - C0;
+        if ( to_dense ) dn[gr + gc*WW] = pc[ii+jj*m]; else pc[ii+jj*m] = dn[gr + gc*WW];
+      });
+    };
+    auto piece_dims = [&](int q, int64_t &m, int64_t &n){
+      const int pq=q/Pc, cq=q%Pc;
+      m = BlockCyclicLayout::NumLocal(c1,L.nb,pq,Pr) - BlockCyclicLayout::NumLocal(c0,L.nb,pq,Pr);
+      n = BlockCyclicLayout::NumLocal(c1,L.nb,cq,Pc) - BlockCyclicLayout::NumLocal(c0,L.nb,cq,Pc);
+    };
+
+    // ---- gather ----
+    tBigGather -= usecond();
+    if ( mq*nq ) { piece.resize((uint64_t)mq*nq); dummy.resize((uint64_t)mq*nq); pack_piece(&piece[0],mq,nq,lr0,lc0); accelerator_barrier(); }
+    if ( me == root ) {
+      deviceVector<ComplexD> stage;
+      for(int q=0;q<Pr*Pc;q++){
+        int64_t m,n; piece_dims(q,m,n); if ( !(m*n) ) continue;
+        if ( q == root ) { root_place(&piece[0],m,n,q,1); continue; }
+        if ( stage.size() < (uint64_t)(m*n) ) stage.resize((uint64_t)m*n);
+        deviceVector<ComplexD> junk((uint64_t)m*n);
+        grid->SendToRecvFrom((void *)&junk[0], q, (void *)&stage[0], q, (uint64_t)m*n*sizeof(ComplexD));
+        root_place(&stage[0],m,n,q,1);
+      }
+      accelerator_barrier();
+    } else if ( mq*nq ) {
+      grid->SendToRecvFrom((void *)&piece[0], root, (void *)&dummy[0], root, (uint64_t)mq*nq*sizeof(ComplexD));
+    }
+    tBigGather += usecond();
+
+    // ---- invert on root: blocked getrf_64 + identity getrs_64 ----
+    tBigInv -= usecond();
+    if ( me == root ) INV.inverseLU(W, &dense[0]);
+    tBigInv += usecond();
+
+    // ---- scatter ----
+    tBigScatter -= usecond();
+    if ( me == root ) {
+      deviceVector<ComplexD> stage;
+      for(int q=0;q<Pr*Pc;q++){
+        int64_t m,n; piece_dims(q,m,n); if ( !(m*n) ) continue;
+        if ( q == root ) { root_place(&piece[0],m,n,q,0); accelerator_barrier(); continue; }
+        if ( stage.size() < (uint64_t)(m*n) ) stage.resize((uint64_t)m*n);
+        deviceVector<ComplexD> junk((uint64_t)m*n);
+        root_place(&stage[0],m,n,q,0); accelerator_barrier();
+        grid->SendToRecvFrom((void *)&stage[0], q, (void *)&junk[0], q, (uint64_t)m*n*sizeof(ComplexD));
+      }
+    } else if ( mq*nq ) {
+      grid->SendToRecvFrom((void *)&dummy[0], root, (void *)&piece[0], root, (uint64_t)mq*nq*sizeof(ComplexD));
+    }
+    if ( mq*nq ) { unpack_piece(&piece[0],mq,nq,lr0,lc0); accelerator_barrier(); }
+    tBigScatter += usecond();
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // The recursion, on global BLOCK range [b0,b1).  SPMD: every rank calls
+  // with identical arguments; there is no ownership gating to get wrong.
+  ///////////////////////////////////////////////////////////////////////////
+  void SchurNode(BlockCyclicMatrix &A,
+                 BlockCyclicMatrix &Bt, BlockCyclicMatrix &Ct,
+                 BlockCyclicMatrix &Tt, BlockCyclicMatrix &Ut,
+                 int64_t b0, int64_t b1)
+  {
+    BlockCyclicLayout &L = A.layout;
+    int64_t span = b1-b0;
+    GRID_ASSERT( span >= 1 );
+    if ( span == 1 ) { Leaf(A, b0); return; }
+    if ( span <= leafSpan && span < std::min(L.Pr,L.Pc) ) { BigLeaf(A, b0, b1); return; }
+    GRID_TRACE("SchurNode");
+    nNode++;
+
+    int64_t bm = b0 + span/2;
+    int64_t c0 = b0*L.nb;
+    int64_t m  = bm*L.nb;
+    int64_t c1 = std::min(L.N, b1*L.nb);
+    ComplexD one (1.0,0.0), mone(-1.0,0.0), zero(0.0,0.0);
+
+    // 1. A11 -> A11inv
+    SchurNode(A,Bt,Ct,Tt,Ut, b0,bm);
+
+    tGemm -= usecond();
+    // 2. Bt[I,J] = A11inv . A12
+    SUMMA.Multiply(one,  A,  A,  zero, Bt, c0,m,  m,c1,  c0,m );
+    // 3. Ct[J,I] = A21 . A11inv
+    SUMMA.Multiply(one,  A,  A,  zero, Ct, m,c1,  c0,m,  c0,m );
+    // 4. A22 -= A21 . Bt   (the Schur complement, in place)
+    SUMMA.Multiply(mone, A,  Bt, one,  A,  m,c1,  m,c1,  c0,m );
+    tGemm += usecond();
+
+    // 5. S -> Sinv          (Bt, Ct live across this call: see invariant)
+    SchurNode(A,Bt,Ct,Tt,Ut, bm,b1);
+
+    tGemm -= usecond();
+    // 6. Tt[J,I] = Sinv . Ct
+    SUMMA.Multiply(one,  A,  Ct, zero, Tt, m,c1,  c0,m,  m,c1 );
+    // 7. Ut[I,J] = Bt . Sinv
+    SUMMA.Multiply(one,  Bt, A,  zero, Ut, c0,m,  m,c1,  m,c1 );
+    // 8. A11 += Ut . Ct
+    SUMMA.Multiply(one,  Ut, Ct, one,  A,  c0,m,  c0,m,  m,c1 );
+    tGemm += usecond();
+
+    // 9. Off-diagonal signs
+    { GRID_TRACE("SchurCopy");
+    WindowCopyScale(mone, Ut, A, c0,m,  m,c1);
+    WindowCopyScale(mone, Tt, A, m,c1,  c0,m);
+    }
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // PUBLIC ENTRY.  In-place inverse of the whole matrix.  Scratch (4x the
+  // matrix footprint) is allocated here and released on return.
+  ///////////////////////////////////////////////////////////////////////////
+  void Invert(BlockCyclicMatrix &A)
+  {
+    BlockCyclicLayout &L = A.layout;
+    GRID_ASSERT( L.N >= 1 );
+    int64_t nblocks = (L.N + L.nb - 1)/L.nb;
+
+    BlockCyclicMatrix Bt(A.grid, L.N, L.nb, L.Pr, L.Pc);
+    BlockCyclicMatrix Ct(A.grid, L.N, L.nb, L.Pr, L.Pc);
+    BlockCyclicMatrix Tt(A.grid, L.N, L.nb, L.Pr, L.Pc);
+    BlockCyclicMatrix Ut(A.grid, L.N, L.nb, L.Pr, L.Pc);
+
+    telLeafMaxInv = 0.0;
+    nLeaf = nNode = 0;
+    tLeaf = tGemm = tCopy = 0.0;
+
+    SchurNode(A, Bt,Ct,Tt,Ut, 0, nblocks);
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Optional, and the ONLY place any reduction happens: call it if you
+  // want the numbers, never from Invert.
+  ///////////////////////////////////////////////////////////////////////////
+  void ReportTelemetry(GridBase *grid)
+  {
+    RealD mx = telLeafMaxInv;
+    grid->GlobalMax(mx);
+    std::cout << GridLogMessage << "BlockCyclicSchurInverse:"
+              << "  nodes " << nNode << "  leaves " << nLeaf
+              << "  max|leafinv| " << mx
+              << "  (boss secs: gemm+comms " << tGemm/1.0e6
+              << "  leaf " << tLeaf/1.0e6
+              << "  copy " << tCopy/1.0e6 << ")"
+              << std::endl;
+    if ( nBigLeaf ) {
+      RealD ti = tBigInv/1.0e6; grid->GlobalMax(ti);     // inverse runs on the root of each leaf: report the max over ranks
+      std::cout << GridLogMessage << "BlockCyclicSchurInverse: BIG LEAVES (span " << leafSpan
+                << ", inverseLU): " << nBigLeaf
+                << " leaves, max W " << maxBigW
+                << "  boss secs: gather " << tBigGather/1.0e6 << " scatter " << tBigScatter/1.0e6
+                << "  inverse (max over ranks) " << ti << std::endl;
+    }
+    // SUMMA breakdown: boss-rank seconds, plus the ring/gemm time spread over
+    // ranks (min/max) -- skew shows as max >> min.
+    RealD ring = (SUMMA.tRingA+SUMMA.tRingB)/1.0e6;
+    RealD rmin = -ring, rmax = ring; grid->GlobalMax(rmin); grid->GlobalMax(rmax); rmin = -rmin;   // no GlobalMin: max of negation
+    RealD gmin = -SUMMA.tGemm/1.0e6, gmax = SUMMA.tGemm/1.0e6; grid->GlobalMax(gmin); grid->GlobalMax(gmax); gmin = -gmin;
+    double gb  = SUMMA.bytesRing/1.0e9;
+    std::cout << GridLogMessage << "BlockCyclicSumma:"
+              << " multiplies " << SUMMA.nMultiply << " gemms " << SUMMA.nGemm << " ring msgs " << SUMMA.nRingMsg
+              << " | boss secs: alloc " << SUMMA.tAlloc/1.0e6
+              << " pack " << SUMMA.tPack/1.0e6
+              << " ringA " << SUMMA.tRingA/1.0e6 << " ringB " << SUMMA.tRingB/1.0e6
+              << " gemm " << SUMMA.tGemm/1.0e6
+              << " | ring min/max over ranks " << rmin << "/" << rmax
+              << " gemm min/max " << gmin << "/" << gmax
+              << " | ring bytes/rank " << gb << " GB -> " << (ring>0 ? gb/ring : 0.0) << " GB/s/rank (boss)"
+              << std::endl;
+    // Ring time decomposed by message size (boss rank; every rank sends the
+    // same sequence of sizes).  Time is wall time inside SendToRecvFrom, so it
+    // includes waiting for the partner -- a bucket whose GB/s is far below the
+    // probe's for the same size is wait, not wire.
+    std::cout << GridLogMessage << "BlockCyclicSumma ring histogram (boss):  size-bucket  msgs  GB  xfer-secs  GB/s  %time" << std::endl;
+    std::streamsize oldprec = std::cout.precision();
+    for(int b=0;b<SUMMA.NHIST;b++){
+      if ( !SUMMA.histN[b] ) continue;
+      double sec = SUMMA.histUs[b]/1.0e6, g = SUMMA.histBytes[b]/1.0e9;
+      double lo = (double)(1ull<<b);
+      char sz[32]; if (lo>=1048576) snprintf(sz,32,"%6.1f MB",lo/1048576.); else if (lo>=1024) snprintf(sz,32,"%6.1f KB",lo/1024.); else snprintf(sz,32,"%6.0f B ",lo);
+      std::cout << GridLogMessage << "   >=" << sz
+                << std::setw(8) << SUMMA.histN[b]
+                << std::setw(10) << std::setprecision(3) << g
+                << std::setw(9) << std::setprecision(3) << sec
+                << std::setw(9) << std::setprecision(3) << (sec>0 ? g/sec : 0.0)
+                << std::setw(8) << std::setprecision(3) << (ring>0 ? 100.0*sec/ring : 0.0);
+      std::cout << std::endl;
+    }
+    std::cout.precision(oldprec);
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/BlockCyclicSumma.h

@@ -0,0 +1,349 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/BlockCyclicSumma.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/multigrid/BlockCyclic.h>
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// Stage 2 of the 2D distributed dense inverse: the windowed SUMMA product
+//
+//   C[i0:i1, j0:j1]  <-  beta C[i0:i1, j0:j1]
+//                      + alpha A[i0:i1, k0:k1] . B[k0:k1, j0:j1]
+//
+// on block-cyclic matrices sharing one BlockCyclicLayout.  All ranges are
+// BLOCK ALIGNED (multiples of nb, or N itself at the top end): stage 3's
+// recursion splits on block boundaries, so nothing else is ever needed, and
+// alignment makes every local window a contiguous band of local storage
+// (BlockCyclic.h RangeToLocal, Test_blockcyclic T7).
+//
+// Transport is PURE POINT-TO-POINT: SendToRecvFrom on explicit world ranks
+// computed from the row-major rank convention.  No collectives of any kind
+// -- no sub-communicators, no broadcast, no allgather -- by design: the
+// collective pathologies measured on this machine (MPI_Allgatherv at
+// ~0.18 MB/s with skewed counts, mpir_request.h:508 aborts, allreduce at
+// 39% of the P2P rate) motivated this implementation.  SendToRecvFrom is
+// the most exercised device-buffer path in Grid and the only one never
+// implicated.
+//
+// Algorithm: round-based ring allgather SUMMA.  The k range is processed in
+// rounds of Pc consecutive global blocks.  Within a round
+//
+//   * process column c owns at most one A panel  (blocks s with s%Pc == c);
+//     the Pc panels circulate around each process-ROW ring in Pc-1 steps;
+//   * process row r owns up to ceil(Pc/Pr) B panels (blocks s%Pr == r);
+//     they circulate around each process-COLUMN ring in Pr-1 steps;
+//   * every rank then accumulates  Cloc += alpha * Apanel_s . Bpanel_s
+//     for each block s of the round, in ascending s: a fixed summation
+//     order, so REPEATED RUNS ARE BITWISE IDENTICAL (no reduction, no
+//     order ambiguity -- the property the P2P doctrine buys).
+//
+// Ring chunks are PADDED to a fixed size (full nb panels, fixed
+// panels-per-origin): SendToRecvFrom carries one byte count for both
+// directions, so symmetric transfers eliminate every variable-size edge
+// case at a worst-case ~1/Pc extra volume on ragged rounds.  Padding is
+// never read: GEMMs address only the leading nb_s x width of each slot.
+//
+// Per-rank received volume: (k-extent) * (mloc_i + nloc_j) elements --
+// the N^2 (1/Pr + 1/Pc) SUMMA optimum, ~sqrt(P)/2 below the 1D scheme.
+///////////////////////////////////////////////////////////////////////////////
+
+class BlockCyclicMatrix
+{
+public:
+  GridBase              *grid;      // borrowed, never owned
+  BlockCyclicLayout      layout;
+  deviceVector<ComplexD> data;      // column major, ld = layout.mloc
+
+  BlockCyclicMatrix(GridBase *g, int64_t N, int64_t nb, int Pr, int Pc)
+    : grid(g),
+      layout(N, nb, Pr, Pc, g->ThisRank())
+  {
+    GRID_ASSERT( Pr*Pc == g->ProcessorCount() );
+    uint64_t sz = (uint64_t)layout.mloc*layout.nloc;
+    data.resize( sz ? sz : 1 );
+  }
+
+  ComplexD *LocalWindow(int64_t li, int64_t lj)
+  {
+    return &data[0] + li + lj*layout.mloc;
+  }
+
+  /////////////////////////////////////////////////////////////////////////
+  // TEST-SCALE import/export of a replicated global matrix (host, O(N^2)
+  // loops, one collective in Export).  For unit tests and the stage-3
+  // oracle only; production data enters through the direct block-cyclic
+  // import, never through these.
+  /////////////////////////////////////////////////////////////////////////
+  void ImportGlobal(const std::vector<ComplexD> &G)
+  {
+    int64_t N = layout.N;
+    GRID_ASSERT( (int64_t)G.size() == N*N );
+    std::vector<ComplexD> h((uint64_t)layout.mloc*layout.nloc, ComplexD(0.0,0.0));
+    for(int64_t j=0;j<N;j++){
+      for(int64_t i=0;i<N;i++){
+        if ( layout.Owns(i,j) ) h[layout.LocalOffset(i,j)] = G[i + j*N];
+      }
+    }
+    if ( h.size() )
+      acceleratorCopyToDevice(&h[0], &data[0], h.size()*sizeof(ComplexD));
+  }
+  void ExportGlobal(std::vector<ComplexD> &G)
+  {
+    int64_t N = layout.N;
+    G.assign((uint64_t)N*N, ComplexD(0.0,0.0));
+    std::vector<ComplexD> h((uint64_t)layout.mloc*layout.nloc);
+    if ( h.size() )
+      acceleratorCopyFromDevice(&data[0], &h[0], h.size()*sizeof(ComplexD));
+    for(int64_t j=0;j<N;j++){
+      for(int64_t i=0;i<N;i++){
+        if ( layout.Owns(i,j) ) G[i + j*N] = h[layout.LocalOffset(i,j)];
+      }
+    }
+    if ( N ) grid->GlobalSumVector((ComplexD *)&G[0], (int)(N*N)); // zero-fill: exact
+  }
+};
+
+class BlockCyclicSumma
+{
+public:
+  GridBLAS BLAS;
+
+  // Per-rank telemetry (boss-rank seconds when printed; no reductions here).
+  // Every Multiply is: buffer alloc, pack panels, ring A along the process
+  // row, ring B along the process column, then the local GEMMs.  The rings
+  // are synchronous SendToRecvFrom, so tRing is time the GPU is idle unless
+  // a future version overlaps them with the GEMMs.
+  // PERSISTENT ring buffers: allocated once (grow-only) and reused by every
+  // Multiply, so the device addresses handed to MPI never change.  Fresh
+  // per-call buffers rotated through the caching allocator's blocks, and a
+  // GPU-direct RDMA registration cache keyed on address then re-registers
+  // per message: measured 1.26 GB/s/rank on 13 MB ring messages (production,
+  // GRID_ALLOC_NCACHE_LARGE=64) against 62 s total for the same inverse when
+  // hipMalloc returned a stable address.
+  deviceVector<ComplexD> Abuf;
+  deviceVector<ComplexD> Bbuf;
+  double   tAlloc=0, tPack=0, tRingA=0, tRingB=0, tGemm=0;
+  uint64_t bytesRing=0, nRingMsg=0, nMultiply=0, nGemm=0;
+  // Per-message-size histogram (bucket = floor(log2 bytes)): count, bytes,
+  // microseconds -- decomposes the ring time by packet size so a low average
+  // GB/s can be attributed (many small latency-bound messages vs slow large
+  // ones vs partner-wait).  Time is wall time inside SendToRecvFrom, so a
+  // bucket far below the wire rate for its size is wait, not wire.
+  static const int NHIST=48;
+  uint64_t histN[NHIST]={0}, histBytes[NHIST]={0}; double histUs[NHIST]={0};
+  void HistAdd(uint64_t bytes, double us){ int b=0; while((bytes>>b)>1) b++; histN[b]++; histBytes[b]+=bytes; histUs[b]+=us; }
+  void ResetTelemetry(void){ tAlloc=tPack=tRingA=tRingB=tGemm=0; bytesRing=nRingMsg=nMultiply=nGemm=0; for(int b=0;b<NHIST;b++){histN[b]=histBytes[b]=0; histUs[b]=0;} }
+
+  static int Overlap(int64_t a0,int64_t a1,int64_t b0,int64_t b1)
+  { return (a0 < b1) && (b0 < a1); }
+
+  void Multiply(ComplexD alpha,
+                BlockCyclicMatrix &A,
+                BlockCyclicMatrix &B,
+                ComplexD beta,
+                BlockCyclicMatrix &C,
+                int64_t i0, int64_t i1,
+                int64_t j0, int64_t j1,
+                int64_t k0, int64_t k1)
+  {
+    BlockCyclicLayout &L = C.layout;
+    GridBase *grid = C.grid;
+    const int64_t N  = L.N;
+    const int64_t nb = L.nb;
+    const int     Pr = L.Pr,   Pc   = L.Pc;
+    const int     prow = L.prow, pcol = L.pcol;
+
+    ///////////////////////////////////////////////////////////////////////
+    // Conformability: one layout, one communicator, aligned ranges.
+    ///////////////////////////////////////////////////////////////////////
+    auto same = [&](BlockCyclicLayout &X){
+      GRID_ASSERT( X.N==N ); GRID_ASSERT( X.nb==nb );
+      GRID_ASSERT( X.Pr==Pr ); GRID_ASSERT( X.Pc==Pc );
+      GRID_ASSERT( X.me==L.me );
+    };
+    same(A.layout); same(B.layout);
+    GRID_ASSERT( A.grid==grid ); GRID_ASSERT( B.grid==grid );
+    auto aligned = [&](int64_t g0,int64_t g1){
+      GRID_ASSERT( 0<=g0 ); GRID_ASSERT( g0<=g1 ); GRID_ASSERT( g1<=N );
+      GRID_ASSERT( g0%nb==0 ); GRID_ASSERT( (g1%nb==0)||(g1==N) );
+    };
+    aligned(i0,i1); aligned(j0,j1); aligned(k0,k1);
+    GRID_ASSERT( k1 > k0 );                    // pure scaling not supported here
+    // In-place windows are legal only if the written window is disjoint
+    // from anything read (stage 3 uses this; make violation loud).
+    if ( &C==&A ) GRID_ASSERT( !Overlap(j0,j1,k0,k1) );
+    if ( &C==&B ) GRID_ASSERT( !Overlap(i0,i1,k0,k1) );
+
+    ///////////////////////////////////////////////////////////////////////
+    // My local bands of the three windows (contiguous: T7).
+    ///////////////////////////////////////////////////////////////////////
+    int64_t li0,li1, lj0,lj1;
+    L.RowRange(i0,i1, li0,li1);
+    L.ColRange(j0,j1, lj0,lj1);
+    const int64_t mloc_i = li1-li0;            // my rows of the i window
+    const int64_t nloc_j = lj1-lj0;            // my cols of the j window
+
+    const int64_t kb0 = k0/nb;
+    const int64_t kb1 = (k1+nb-1)/nb;          // block-aligned or ==N: exact
+    const int64_t S   = (Pc + Pr - 1)/Pr;      // max B panels per origin row
+
+    ///////////////////////////////////////////////////////////////////////
+    // Round buffers, padded to fixed slot sizes (see header comment).
+    //   A: Pc slots of mloc_i x nb   (slot c = panel of the block owned by c)
+    //   B: Pr slots of S x (nb x nloc_j)
+    ///////////////////////////////////////////////////////////////////////
+    const uint64_t slotA  = (uint64_t)mloc_i*nb;
+    const uint64_t slotB1 = (uint64_t)nb*nloc_j;         // one panel
+    const uint64_t slotB  = (uint64_t)S*slotB1;
+    nMultiply++;
+    tAlloc -= usecond();
+    if ( Abuf.size() < std::max<uint64_t>(slotA*Pc,1) ) Abuf.resize( std::max<uint64_t>(slotA*Pc,1) );
+    if ( Bbuf.size() < std::max<uint64_t>(slotB*Pr,1) ) Bbuf.resize( std::max<uint64_t>(slotB*Pr,1) );
+    tAlloc += usecond();
+
+    deviceVector<ComplexD *> ap(1), bp(1), cp(1);
+    std::vector<ComplexD *>  ptr(1);
+
+    int firstblock = 1;
+    for(int64_t r0=kb0; r0<kb1; r0+=Pc){       // rounds of Pc k-blocks
+      int64_t r1 = std::min(kb1, r0+Pc);
+
+      /////////////////////////////////////////////////////////////////////
+      // Pack MY panels of this round into my origin slots.
+      /////////////////////////////////////////////////////////////////////
+      tPack -= usecond();
+      { GRID_TRACE("SummaPack");
+      for(int64_t s=r0; s<r1; s++){
+        int64_t nb_s = L.BlockSize(s);
+        if ( (int)(s%Pc) == pcol && mloc_i ){  // my A panel: block-column s
+          int64_t lc0, lc1;
+          L.ColRange(s*nb, std::min(N,(s+1)*nb), lc0, lc1);
+          GRID_ASSERT( lc1-lc0 == nb_s );
+          ComplexD *src = A.LocalWindow(li0, lc0);
+          ComplexD *dst = &Abuf[0] + slotA*pcol;
+          int64_t   ld  = A.layout.mloc;
+          int64_t   m   = mloc_i;
+          accelerator_for(idx, (uint64_t)(m*nb_s), 1, {
+            int64_t jj = idx / m;
+            int64_t ii = idx - jj*m;
+            dst[ii + jj*m] = src[ii + jj*ld];
+          });
+        }
+        if ( (int)(s%Pr) == prow && nloc_j ){  // my B panel: block-row s
+          int64_t lr0, lr1;
+          L.RowRange(s*nb, std::min(N,(s+1)*nb), lr0, lr1);
+          GRID_ASSERT( lr1-lr0 == nb_s );
+          int64_t idxs = (s - r0 - ((prow - r0%Pr + Pr) % Pr)) / Pr; // my panel # in round
+          GRID_ASSERT( idxs >= 0 ); GRID_ASSERT( idxs < S );
+          ComplexD *src = B.LocalWindow(lr0, lj0);
+          ComplexD *dst = &Bbuf[0] + slotB*prow + slotB1*idxs;
+          int64_t   ld  = B.layout.mloc;
+          int64_t   nn  = nloc_j;
+          accelerator_for(idx, (uint64_t)(nb_s*nn), 1, {
+            int64_t jj = idx / nb_s;
+            int64_t ii = idx - jj*nb_s;
+            dst[ii + jj*nb_s] = src[ii + jj*ld];
+          });
+        }
+      }
+      accelerator_barrier();
+      }
+      tPack += usecond();
+
+      /////////////////////////////////////////////////////////////////////
+      // Ring allgather along my process ROW: Pc-1 symmetric steps.  At
+      // step t send the slot of origin (pcol-t+1), receive origin (pcol-t).
+      /////////////////////////////////////////////////////////////////////
+      if ( Pc > 1 && slotA ){
+        GRID_TRACE("SummaRingA");
+        tRingA -= usecond();
+        int dest = prow*Pc + (pcol+1)%Pc;
+        int src  = prow*Pc + (pcol-1+Pc)%Pc;
+        for(int t=1;t<Pc;t++){
+          int cs = (pcol - t + 1 + Pc*Pc) % Pc;
+          int cr = (pcol - t     + Pc*Pc) % Pc;
+          double tm = usecond();
+          grid->SendToRecvFrom((void *)(&Abuf[0]+slotA*cs), dest,
+                               (void *)(&Abuf[0]+slotA*cr), src,
+                               slotA*sizeof(ComplexD));
+          HistAdd(slotA*sizeof(ComplexD), usecond()-tm);
+          bytesRing += slotA*sizeof(ComplexD); nRingMsg++;
+        }
+        tRingA += usecond();
+      }
+      /////////////////////////////////////////////////////////////////////
+      // Ring allgather along my process COLUMN: Pr-1 symmetric steps.
+      /////////////////////////////////////////////////////////////////////
+      if ( Pr > 1 && slotB ){
+        GRID_TRACE("SummaRingB");
+        tRingB -= usecond();
+        int dest = ((prow+1)%Pr)*Pc + pcol;
+        int src  = ((prow-1+Pr)%Pr)*Pc + pcol;
+        for(int t=1;t<Pr;t++){
+          int rs = (prow - t + 1 + Pr*Pr) % Pr;
+          int rr = (prow - t     + Pr*Pr) % Pr;
+          double tm = usecond();
+          grid->SendToRecvFrom((void *)(&Bbuf[0]+slotB*rs), dest,
+                               (void *)(&Bbuf[0]+slotB*rr), src,
+                               slotB*sizeof(ComplexD));
+          HistAdd(slotB*sizeof(ComplexD), usecond()-tm);
+          bytesRing += slotB*sizeof(ComplexD); nRingMsg++;
+        }
+        tRingB += usecond();
+      }
+
+      /////////////////////////////////////////////////////////////////////
+      // Local update, ascending s: fixed order, bitwise-reproducible.
+      /////////////////////////////////////////////////////////////////////
+      tGemm -= usecond();
+      { GRID_TRACE("SummaGEMM");
+      for(int64_t s=r0; s<r1; s++){
+        int64_t nb_s = L.BlockSize(s);
+        if ( !(mloc_i && nloc_j && nb_s) ) { firstblock = 0; continue; }
+        int cA = (int)(s%Pc);
+        int rB = (int)(s%Pr);
+        int64_t idxs = (s - r0 - ((rB - r0%Pr + Pr) % Pr)) / Pr;
+        ComplexD beta_use = firstblock ? beta : ComplexD(1.0,0.0);
+        firstblock = 0;
+
+        ptr[0] = &Abuf[0] + slotA*cA;
+        acceleratorCopyToDevice(&ptr[0], &ap[0], sizeof(ComplexD *));
+        ptr[0] = &Bbuf[0] + slotB*rB + slotB1*idxs;
+        acceleratorCopyToDevice(&ptr[0], &bp[0], sizeof(ComplexD *));
+        ptr[0] = C.LocalWindow(li0, lj0);
+        acceleratorCopyToDevice(&ptr[0], &cp[0], sizeof(ComplexD *));
+
+        BLAS.gemmBatched(GridBLAS_OP_N, GridBLAS_OP_N,
+                         (int)mloc_i, (int)nloc_j, (int)nb_s,
+                         alpha,    ap, (int)mloc_i,
+                                   bp, (int)nb_s,
+                         beta_use, cp, (int)C.layout.mloc);
+        BLAS.synchronise();
+        nGemm++;
+      }
+      }
+      tGemm += usecond();
+    }
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/CoarsenedMatrix.h

@@ -309,7 +309,7 @@ public:
     if ((out.size()!=ndir)&&(out.size()!=ndir+1)) { 
       std::cout <<"MdirAll out size "<< out.size()<<std::endl;
       std::cout <<"MdirAll ndir "<< ndir<<std::endl;
-      assert(0);
+      GRID_ASSERT(0);
     }
     for(int p=0;p<ndir;p++){
       MdirCalc(in,out[p],p);
@@ -373,7 +373,7 @@ public:
     conformable(in.Grid(), _cbgrid);    // verifies half grid
     conformable(in.Grid(), out.Grid()); // drops the cb check
 
-    assert(in.Checkerboard() == Even);
+    GRID_ASSERT(in.Checkerboard() == Even);
     out.Checkerboard() = Odd;
 
     DhopInternal(StencilEven, Aodd, in, out, dag);
@@ -383,7 +383,7 @@ public:
     conformable(in.Grid(), _cbgrid);    // verifies half grid
     conformable(in.Grid(), out.Grid()); // drops the cb check
 
-    assert(in.Checkerboard() == Odd);
+    GRID_ASSERT(in.Checkerboard() == Odd);
     out.Checkerboard() = Even;
 
     DhopInternal(StencilOdd, Aeven, in, out, dag);
@@ -391,7 +391,7 @@ public:
 
   void MooeeInternal(const CoarseVector &in, CoarseVector &out, int dag, int inv) {
     out.Checkerboard() = in.Checkerboard();
-    assert(in.Checkerboard() == Odd || in.Checkerboard() == Even);
+    GRID_ASSERT(in.Checkerboard() == Odd || in.Checkerboard() == Even);
 
     CoarseMatrix *Aself = nullptr;
     if(in.Grid()->_isCheckerBoarded) {
@@ -406,7 +406,7 @@ public:
       Aself = (inv) ? &AselfInv : &A[geom.npoint-1];
       DselfInternal(Stencil, *Aself, in, out, dag);
     }
-    assert(Aself != nullptr);
+    GRID_ASSERT(Aself != nullptr);
   }
 
   void DselfInternal(CartesianStencil<siteVector,siteVector,DefaultImplParams> &st, CoarseMatrix &a,
@@ -697,7 +697,7 @@ public:
     evenmask = where(mod(bcb,2)==(Integer)0,one,zero);
     oddmask  = one-evenmask;
 
-    assert(self_stencil!=-1);
+    GRID_ASSERT(self_stencil!=-1);
 
     for(int i=0;i<nbasis;i++){
 

Grid/algorithms/multigrid/DenseCoarseMatrix.h

@@ -0,0 +1,854 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/algorithms/multigrid/DenseCoarseMatrix.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/blas/BatchedBlas.h>
+#include <Grid/algorithms/blas/BatchedInverse.h>
+#include <Grid/algorithms/multigrid/BlockCyclicSchurInverse.h>
+#include <Grid/algorithms/multigrid/BlockCyclicRedistribute.h>
+
+#include <unordered_map>
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////////////////////
+// DenseCoarseMatrix: a coarsened operator treated as a DENSE matrix -- explicit,
+// row-distributed A^{-1} of a GeneralCoarsenedMatrix.
+//
+//  - Stencil -> dense DIRECT IMPORT.  The coarse operator IS the dense matrix
+//    unrolled: Dense[(s,a),(s+shift_p,b)] += A[p][s]_{a,b}.  Rows of my sites are
+//    assembled from purely LOCAL _A[p] data: no operator applies, no comms.
+//    ACCUMULATE (+=) because on short axes distinct shifts wrap to the same
+//    neighbour.  An IMPORT CERTIFICATE compares the dense apply against Op.M on a
+//    NON-CONSTANT vector (a constant one cannot see a shift-sign error).
+//
+//  - Inversion is END-TO-END fp64 through the 2D block-cyclic recursive Schur
+//    complement (BlockCyclicSchurInverse): fp64 rank-major import ->
+//    RowsToCyclic -> in-place recursion (pure point-to-point SUMMA rings and
+//    local leaves; bitwise reproducible) -> CyclicToRows -> ONE terminal
+//    rounding into the fp32 apply slab.  Distributed at every N and P.
+//
+//  - Split-K apply through GridBLAS.gemmBatched with EXPLICIT leading dimensions
+//    (arXiv:2409.03904 fig 11): the tiny-output/huge-K GEMM Y = slab^T X becomes
+//    SPLITK chunk-GEMMs by pointer offset into the resident slab (lda = N),
+//    partials reduced in one accelerator_for.  The source vector is assembled by
+//    a cartesian ring ALLGATHER of device buffers (pure P2P; ~8x fewer bytes
+//    than a padded allreduce, and no collective size cliffs).  Platform-agnostic:
+//    deviceVector + GridBLAS run the SAME code on HIP/CUDA/SYCL and CPU(Eigen).
+//
+// VERIFY ||A Ainv x - x||/||x|| certifies the DEVICE slab + split-K path at the
+// end of Import, since the single-RHS apply routes through the same core.
+//
+// Tensor-depth agnostic: site scalar objects treated as contiguous ComplexD
+// (iScalar wrappers add no data), so any MG level's coarse operator imports.
+//////////////////////////////////////////////////////////////////////////////////////
+//
+// Depends on a coarse operator only to extract its matrix elements; thereafter
+// it is given a coarse vector and applies the inverse. Import() is a template
+// member so any of the coarse classes will do, and the type of a
+// DenseCoarseMatrix does not record which one built it.
+//
+template<class CComplex,int nbasis>
+class DenseCoarseMatrix : public LinearFunction<Lattice<iVector<CComplex,nbasis> > > {
+public:
+  typedef iVector<CComplex,nbasis >           siteVector;
+  typedef Lattice<siteVector>                 CoarseVector;
+  typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
+  typedef CoarseVector                        Field;
+  using LinearFunction<Field>::operator();
+  typedef typename Field::vector_object vobj;
+  typedef typename vobj::scalar_object  sobj;
+  typedef typename CoarseMatrix::vector_object Mvobj;
+  typedef typename Mvobj::scalar_object        Msobj;
+
+  GridBase *grid;
+  int      nd;
+  int64_t  N;                       // dense rank = gSites * nbasis
+  int      lsites;                  // my local sites
+  int64_t  nrows;                   // my rows = lsites * nbasis
+  std::vector<Coordinate> myLcoor;  // local coordinate of my site ss
+  std::vector<int64_t>    myGsite;  // global lex site index of my site ss
+  std::vector<ComplexF>   slab;     // nrows x N row-major: A during setup, rows of A^{-1} after
+
+  static const int MRHS_MAX = 32;
+  static const int SPLITK   = 32;   // requested split-K chunk count, snapped DOWN to a divisor of N
+
+  // Apply machinery: resident slab + persistent buffers + AOT split-K pointers.
+  GridBLAS BLAS;
+  deviceVector<ComplexF>  dSlab;
+  deviceVector<ComplexF>  dX;       // N x MRHS_MAX
+  deviceVector<ComplexF>  dY;       // nrows x MRHS_MAX
+  deviceVector<ComplexF>  dG;       // N x MRHS_MAX lex-major staging for the allgather
+  deviceVector<int>       dLex2Rank;// lex index of a process coordinate -> its rank (allgather block order -> row-block order)
+  deviceVector<int64_t>   dRm2G;    // rank-major index (rank*nrows + ss*nbasis + b) -> global column (gsite*nbasis + b) of x / the slab
+  int                     myLex;
+  deviceVector<ComplexF>  dPartial; // NK x (nrows x MRHS_MAX)
+  deviceVector<ComplexF*> aptrs;    // slab K-chunk pointers   (lda = N)
+  deviceVector<ComplexF*> xptrs;    // X    K-chunk pointers   (ldb = N)
+  deviceVector<ComplexF*> cptrs;    // partial buffers         (ldc = nrows)
+  std::vector<ComplexF>   hX;
+  std::vector<ComplexF>   hY;
+  int NK;                           // split-K chunk count (divides N)
+
+  DenseCoarseMatrix(GridBase *g)
+    : grid(g)
+  {
+    GRID_ASSERT( sizeof(sobj)  == nbasis*sizeof(ComplexD) );
+    GRID_ASSERT( sizeof(Msobj) == nbasis*nbasis*sizeof(ComplexD) );
+    nd     = grid->_ndimension;
+    N      = grid->gSites() * nbasis;
+    lsites = grid->lSites();
+    nrows  = (int64_t)lsites * nbasis;
+
+    std::cout << GridLogMessage << "DenseCoarseMatrix: N = " << N
+              << " (" << grid->gSites() << " sites x " << nbasis << ")"
+              << "  rows/rank = " << nrows
+              << "  slab = " << (double)nrows*N*sizeof(ComplexF)/1024./1024. << " MB/rank"
+              << std::endl;
+
+    ////////////////////////////////////////////////////////////////////
+    // Enumerate my sites: local coords and global lexicographic indices
+    ////////////////////////////////////////////////////////////////////
+    Coordinate ldims = grid->LocalDimensions();
+    Coordinate gdims = grid->GlobalDimensions();
+    myLcoor.resize(lsites);
+    myGsite.resize(lsites);
+    for(int ss=0; ss<lsites; ss++){
+      Coordinate lcoor(nd);
+      Lexicographic::CoorFromIndex(lcoor, ss, ldims);
+      Coordinate gcoor(nd);
+      for(int d=0; d<nd; d++) gcoor[d] = grid->_lstart[d] + lcoor[d];
+      int64_t gsite;
+      Lexicographic::IndexFromCoor(gcoor, gsite, gdims);
+      myLcoor[ss] = lcoor;
+      myGsite[ss] = gsite;
+    }
+
+    slab.resize((uint64_t)nrows * N);
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // The only place a coarse operator is needed: pull its elements, invert,
+  // and make the slab resident. Any class exposing Geometry() and
+  // ExtractMatrix(p,A) will do -- single RHS or either multiRHS.
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void Import(CoarseOp &Op)
+  {
+    double t0 = usecond();
+    {
+      ImportDense(Op);        // slab <- my rows of A   (LOCAL, no comms)
+      ImportCertificate(Op);  // dense apply == Op.M, before inversion
+      InvertDense(Op);        // slab <- my rows of A^{-1}
+      double t1 = usecond();
+      std::cout << GridLogMessage << "DenseCoarseMatrix: import+invert took "
+                << (t1-t0)/1.0e6 << " s" << std::endl;
+    }
+
+    ////////////////////////////////////////////////////////////////////
+    // Device residency + persistent apply buffers + AOT split-K pointers
+    ////////////////////////////////////////////////////////////////////
+    {
+      uint64_t sbytes = (uint64_t)nrows * N * sizeof(ComplexF);
+      dSlab.resize((uint64_t)nrows*N);
+      acceleratorCopyToDevice(&slab[0],&dSlab[0],sbytes);
+
+      // Split-K chunk count, snapped DOWN to a divisor of N.
+      NK = 1;
+      for(int j=1;j<=SPLITK;j++) if ( (N % j) == 0 ) NK = j;
+      int64_t Kc = N / NK;
+
+      dX.resize((uint64_t)N*MRHS_MAX);
+      dY.resize((uint64_t)nrows*MRHS_MAX);
+      dPartial.resize((uint64_t)NK*nrows*MRHS_MAX);
+      hX.resize((uint64_t)N*MRHS_MAX);
+      hY.resize((uint64_t)nrows*MRHS_MAX);
+
+      aptrs.resize(NK); xptrs.resize(NK); cptrs.resize(NK);
+      std::vector<ComplexF*> h(NK);
+      for(int j=0;j<NK;j++) h[j] = &dSlab[0]    + (uint64_t)j*Kc;              // K-offset, lda=N
+      acceleratorCopyToDevice(&h[0],&aptrs[0],NK*sizeof(ComplexF*));
+      for(int j=0;j<NK;j++) h[j] = &dX[0]       + (uint64_t)j*Kc;              // K-offset, ldb=N
+      acceleratorCopyToDevice(&h[0],&xptrs[0],NK*sizeof(ComplexF*));
+      for(int j=0;j<NK;j++) h[j] = &dPartial[0] + (uint64_t)j*nrows*MRHS_MAX;  // compact, ldc=nrows
+      acceleratorCopyToDevice(&h[0],&cptrs[0],NK*sizeof(ComplexF*));
+
+      {
+        dG.resize((uint64_t)N*MRHS_MAX);
+        // allgather delivers blocks in lexicographic-coordinate order; the row
+        // blocks of x are in RANK order.  Same table as BuildRankMajorMap.
+        int P = grid->ProcessorCount();
+        std::vector<int> l2r(P);
+        for(int lp=0; lp<P; lp++){ Coordinate pc(nd); Lexicographic::CoorFromIndex(pc, lp, grid->_processors); l2r[lp] = grid->RankFromProcessorCoor(pc); }
+        dLex2Rank.resize(P);
+        acceleratorCopyToDevice(&l2r[0], &dLex2Rank[0], P*sizeof(int));
+        myLex = CartesianLexIndex(grid);
+        GRID_ASSERT( l2r[myLex] == grid->ThisRank() );
+        // x and the slab columns are in GLOBAL-SITE order (hX[myGsite*nbasis+b]);
+        // the gathered blocks are in RANK-MAJOR order (rank*nrows + ss*nbasis + b).
+        // The two coincide only on one rank, so scatter through the inverse map.
+        std::vector<int64_t> g2rm; BuildRankMajorMap(g2rm);
+        std::vector<int64_t> rm2g(N); for(int64_t g=0; g<N; g++) rm2g[g2rm[g]] = g;
+        for(int ss=0; ss<lsites; ss++) GRID_ASSERT( rm2g[(int64_t)grid->ThisRank()*nrows + (int64_t)ss*nbasis] == myGsite[ss]*nbasis );
+        dRm2G.resize(N);
+        acceleratorCopyToDevice(&rm2g[0], &dRm2G[0], N*sizeof(int64_t));
+      }
+      std::cout << GridLogMessage << "DenseCoarseMatrix: slab resident on device ("
+                << sbytes/1024./1024. << " MB/rank), split-K NK=" << NK << " (Kc=" << Kc
+                << "); DEVICE cartesian ring ALLGATHER (P2P)" << std::endl;
+    }
+
+    ////////////////////////////////////////////////////////////////////
+    // VERIFY: || A (Ainv x) - x || / ||x|| through the DEVICE split-K core.
+    ////////////////////////////////////////////////////////////////////
+    {
+      Field x(grid); Field y(grid); Field z(grid);
+      x = ComplexD(1.0,0.0);
+      double ta = usecond();
+      (*this)(x, y);
+      double tb = usecond();
+      ApplyOracle(Op, y, z);
+      z = z - x;
+      RealD rel = std::sqrt(norm2(z)/norm2(x));
+      std::cout << GridLogMessage << "DenseCoarseMatrix: VERIFY ||A Ainv x - x||/||x|| = "
+                << rel << "   (one apply took " << (tb-ta)/1000.0 << " ms)" << std::endl;
+      GRID_ASSERT(rel < 1.0e-2);
+    }
+    std::cout << GridLogMessage << "DenseCoarseMatrix: setup complete, total "
+              << (usecond()-t0)/1.0e6 << " s" << std::endl;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Apply the source operator to a D dimensional field, whichever kind it is.
+  //
+  // A multiRHS operator lives on the D+1 grid, so drive it with several right
+  // hand sides at once: slice r carries (r+1)*in, and linearity says the
+  // results must scale likewise. One apply, and unlike a single rhs check it
+  // also catches rhs mixing. Cheap check, not a production path.
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void ApplyOracle(CoarseOp &Op,const Field &in, Field &out)
+  {
+    if ( Op.Grid() == grid ) { Op.M(in,out); return; }
+
+    GridBase *mgrid = Op.Grid();
+    GRID_ASSERT(mgrid->_ndimension == nd+1);
+    int nr = mgrid->_fdimensions[0];
+
+    Field min(mgrid), mout(mgrid);
+    for(int r=0;r<nr;r++){
+      Field scaled(grid);
+      scaled = ComplexD(r+1.0,0.0)*in;
+      InsertSliceFast(scaled,min,r,0);
+    }
+
+    Op.M(min,mout);
+
+    ExtractSliceFast(out,mout,0,0);
+    for(int r=1;r<nr;r++){
+      Field sr(grid),d(grid);
+      ExtractSliceFast(sr,mout,r,0);
+      d = sr - ComplexD(r+1.0,0.0)*out;
+      RealD rel = std::sqrt(norm2(d)/norm2(sr));
+      if ( rel >= 1.0e-6 ) {
+        std::cout << GridLogMessage << "DenseCoarseMatrix: oracle rhs "<<r
+                  <<" inconsistent with rhs 0, rel "<<rel<<std::endl;
+      }
+      GRID_ASSERT( rel < 1.0e-6 );
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 1. Direct stencil -> dense import of MY ROWS of A (no comms):
+  //      Dense[(s,a),(wrap(s+shift_p),b)] += A[p][s]_{a,b}
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void ImportDense(CoarseOp &Op)
+  {
+    double t = -usecond();
+    Coordinate gdims = grid->GlobalDimensions();
+
+    uint64_t nelem = (uint64_t)nrows * N;
+    thread_for(i, nelem, { slab[i] = ComplexF(0.0,0.0); });
+
+    for(int p=0; p<Op.Geometry().npoint; p++){
+      Coordinate shift = Op.Geometry().shifts[p];
+      // _A[p] is PADDED after ExchangeCoarseLinks (end of CoarsenOperator):
+      // extract the unpadded field before peeking with unpadded coordinates
+      // (exactly as MultiGeneralCoarsenedMatrix::CopyMatrix does).
+      CoarseMatrix Aun(grid);  Op.ExtractMatrix(p,Aun);
+      autoView(Av, Aun, CpuRead);
+      thread_for(ss, lsites, {
+        Coordinate ncoor(nd);
+        for(int d=0; d<nd; d++){
+          int64_t g = grid->_lstart[d] + myLcoor[ss][d] + shift[d];
+          ncoor[d] = (int)((g % gdims[d] + gdims[d]) % gdims[d]);
+        }
+        int64_t nsite;
+        Lexicographic::IndexFromCoor(ncoor, nsite, gdims);
+        Msobj m;
+        peekLocalSite(m, Av, myLcoor[ss]);
+        ComplexD *md = (ComplexD *)&m;
+        // The operator contracts out(s,b) = sum_a A[p](s)(a,b) in(nbr,a)
+        // (GeneralCoarsenedMatrix.h Mult kernel): the stored site matrix
+        // acts TRANSPOSED, so element (a,b) lands at dense row (s,b),
+        // column (nbr,a).
+        for(int b=0; b<nbasis; b++){
+          ComplexF *row = &slab[(uint64_t)(ss*nbasis+b)*N + nsite*nbasis];
+          for(int a=0; a<nbasis; a++)
+            row[a] += ComplexF(md[a*nbasis+b]);   // += : wrapped shifts may collide
+        }
+      });
+    }
+    t += usecond();
+
+    // Structural diagnostic: identically-zero rows of my slab (a healthy
+    // coarse operator has none; dead rows mean a rank-deficient import
+    // or operator and the inverse will be NaN).
+    int64_t zrows = 0;
+    for(int64_t r=0; r<nrows; r++)
+    {
+      double mx = 0.0;
+      const ComplexF *row = &slab[(uint64_t)r*N];
+      for(int64_t j=0; j<N; j++)
+      {
+        mx = std::max(mx, (double)abs(row[j]));
+      }
+      if ( mx < 1.0e-30 ) zrows++;
+    }
+    RealD gz = (RealD)zrows;
+    grid->GlobalSumVector(&gz, 1);
+
+    std::cout << GridLogMessage << "DenseCoarseMatrix: stencil->dense import took "
+              << t/1.0e6 << " s  (" << Op.Geometry().npoint << " points, local, no comms)"
+              << "  zero rows " << (int64_t)gz << "/" << N << std::endl;
+
+    // Debug: coordinate pattern of live sites (mechanism fingerprint)
+    if ( (int64_t)gz > 0 )
+    {
+      int shown = 0;
+      for(int ss=0; ss<lsites && shown<24; ss++)
+      {
+        double mx = 0.0;
+        const ComplexF *row = &slab[(uint64_t)(ss*nbasis)*N];
+        for(int64_t j=0; j<N; j++)
+        {
+          mx = std::max(mx, (double)abs(row[j]));
+        }
+        if ( mx > 1.0e-30 )
+        {
+          std::cout << GridLogMessage << "DenseCoarseMatrix: LIVE site ss=" << ss
+                    << " lcoor " << myLcoor[ss] << std::endl;
+          shown++;
+        }
+      }
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 2. IMPORT CERTIFICATE: dense rows vs Op.M on a NON-CONSTANT vector.
+  //    (Constant x has x[s+d]==x[s-d]: blind to a shift-sign error.)
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void ImportCertificate(CoarseOp &Op)
+  {
+    Field x(grid); Field Ax(grid); Field Dx(grid);
+    for(int ss=0; ss<lsites; ss++){
+      sobj s;
+      for(int b=0; b<nbasis; b++){
+        double ph = 0.37*(double)(myGsite[ss]*nbasis+b);
+        ((ComplexD *)&s)[b] = ComplexD(std::cos(ph),std::sin(0.61*ph));
+      }
+      pokeLocalSite(s, x, myLcoor[ss]);
+    }
+    // gather full x (zero-fill + exact GlobalSum), dense rows on host
+    std::vector<ComplexD> xh((uint64_t)N, ComplexD(0.0,0.0));
+    for(int ss=0; ss<lsites; ss++){
+      sobj s;
+      peekLocalSite(s, x, myLcoor[ss]);
+      for(int b=0; b<nbasis; b++) xh[ myGsite[ss]*nbasis + b ] = ((ComplexD *)&s)[b];
+    }
+    grid->GlobalSumVector(&xh[0], (int)N);
+    std::vector<ComplexD> yh(nrows);
+    thread_for(r, nrows, {
+      ComplexD acc(0.0,0.0);
+      const ComplexF *row = &slab[(uint64_t)r * N];
+      for(int64_t j=0; j<N; j++) acc += ComplexD(row[j]) * xh[j];
+      yh[r] = acc;
+    });
+    for(int ss=0; ss<lsites; ss++){
+      sobj s;
+      for(int b=0; b<nbasis; b++) ((ComplexD *)&s)[b] = yh[ss*nbasis+b];
+      pokeLocalSite(s, Dx, myLcoor[ss]);
+    }
+    ApplyOracle(Op, x, Ax);
+    Field d(grid); d = Dx - Ax;
+    RealD rel = std::sqrt(norm2(d)/norm2(Ax));
+    std::cout << GridLogMessage << "DenseCoarseMatrix: IMPORT CERTIFICATE ||Dense x - A x||/||A x|| = "
+              << rel << std::endl;
+    if ( rel >= 1.0e-3 ) {
+      std::cout << GridLogMessage << "DenseCoarseMatrix: IMPORT CERTIFICATE FAILED. If O(1), the "
+                << "stencil shift-sign convention of the coarse operator has changed: "
+                << "the import in ImportDense/ImportDenseFP64 must change with it"
+                << std::endl;
+    }
+    GRID_ASSERT(rel < 1.0e-3);
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 3a. Global column -> rank-major column map, computed LOCALLY.
+  //    Rank-major ordering: rank q's rows/columns are the contiguous
+  //    block [q*nrows, (q+1)*nrows), ordered by q's local site index
+  //    (uniform local volumes make ownership arithmetic exact).
+  //    MPI_Cart_rank is queried ONCE PER RANK (serial, P calls) into a
+  //    lex-processor table; the per-site sweep is then pure arithmetic.
+  ////////////////////////////////////////////////////////////////////
+  void BuildRankMajorMap(std::vector<int64_t> &g2rm)
+  {
+    int P = grid->ProcessorCount();
+    Coordinate pdims = grid->_processors;
+    Coordinate gdims = grid->GlobalDimensions();
+    Coordinate ldims = grid->LocalDimensions();
+
+    std::vector<int> lexp2rank(P);
+    for(int lp=0; lp<P; lp++)
+    {
+      Coordinate pcoor(nd);
+      Lexicographic::CoorFromIndex(pcoor, lp, pdims);
+      lexp2rank[lp] = grid->RankFromProcessorCoor(pcoor);
+    }
+
+    int64_t gsites = grid->gSites();
+    g2rm.resize(N);
+    thread_for(gsite, gsites, {
+      Coordinate gcoor(nd);
+      Coordinate pcoor(nd);
+      Coordinate lcoor(nd);
+      Lexicographic::CoorFromIndex(gcoor, gsite, gdims);
+      for(int d=0; d<nd; d++)
+      {
+        pcoor[d] = gcoor[d]/ldims[d];
+        lcoor[d] = gcoor[d]-pcoor[d]*ldims[d];
+      }
+      int64_t lexp;
+      int64_t lsite;
+      Lexicographic::IndexFromCoor(pcoor, lexp,  pdims);
+      Lexicographic::IndexFromCoor(lcoor, lsite, ldims);
+      int64_t base = (int64_t)lexp2rank[lexp]*nrows + lsite*nbasis;
+      for(int b=0; b<nbasis; b++)
+      {
+        g2rm[(uint64_t)gsite*nbasis + b] = base + b;
+      }
+    });
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 3b. Direct stencil -> fp64 rank-major import of MY ROWS of A (the
+  //    end-to-end fp64 path: the stencil source IS ComplexD; nothing is
+  //    rounded through fp32 on the way into the inversion).  Same
+  //    loop/sign/accumulate/transposed-contraction discipline as
+  //    ImportDense; output is column-major rows x N with columns in
+  //    rank-major order (g2rm).
+  //    ALWAYS-ON CERTIFICATE: the fp64 import, rounded, must agree with
+  //    the fp32 slab entry at the corresponding global column, over the
+  //    WHOLE of my rows (few ulp: wrapped-shift collisions accumulate in
+  //    different precision order).  NaN-proof: non-finite entries are
+  //    counted explicitly since max() silently masks NaN.
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void ImportDenseFP64(CoarseOp &Op, BlockRows &S, std::vector<int64_t> &g2rm)
+  {
+    Coordinate gdims = grid->GlobalDimensions();
+
+    std::vector<ComplexD> h((uint64_t)nrows*N, ComplexD(0.0,0.0));
+    for(int p=0; p<Op.Geometry().npoint; p++)
+    {
+      Coordinate shift = Op.Geometry().shifts[p];
+      CoarseMatrix Aun(grid);  Op.ExtractMatrix(p,Aun);
+      autoView(Av, Aun, CpuRead);
+      thread_for(ss, lsites, {
+        Coordinate ncoor(nd);
+        for(int d=0; d<nd; d++)
+        {
+          int64_t g = grid->_lstart[d] + myLcoor[ss][d] + shift[d];
+          ncoor[d] = (int)((g % gdims[d] + gdims[d]) % gdims[d]);
+        }
+        int64_t nsite;
+        Lexicographic::IndexFromCoor(ncoor, nsite, gdims);
+        Msobj m;
+        peekLocalSite(m, Av, myLcoor[ss]);
+        ComplexD *md = (ComplexD *)&m;
+        // Transposed contraction as ImportDense: (a,b) lands at
+        // row (s,b), column (nbr,a); column index in rank-major order.
+        for(int a=0; a<nbasis; a++)
+        {
+          int64_t jj = g2rm[ nsite*nbasis + a ];
+          for(int b=0; b<nbasis; b++)
+          {
+            h[(uint64_t)(ss*nbasis+b) + (uint64_t)jj*nrows] += md[a*nbasis+b];
+          }
+        }
+      });
+    }
+
+    // Certificate vs the fp32 slab (slab holds A at this point)
+    double  mx   = 0.0;
+    int64_t nbad = 0;
+    for(int64_t i=0; i<nrows; i++)
+    {
+      for(int64_t gcol=0; gcol<N; gcol++)
+      {
+        ComplexD d64 = h[(uint64_t)(i + g2rm[gcol]*nrows)];
+        ComplexF f32 = slab[(uint64_t)i*N + gcol];
+        double dev = abs(ComplexD(f32) - d64);
+        if ( !std::isfinite(dev) ) nbad++;
+        else mx = std::max(mx, dev);
+      }
+    }
+    RealD gmx  = mx;
+    RealD gbad = (RealD)nbad;
+    grid->GlobalMax(gmx);
+    grid->GlobalSumVector(&gbad, 1);
+    std::cout << GridLogMessage << "DenseCoarseMatrix: fp64 import certificate "
+              << "max|A64 - A32| = " << gmx
+              << "  non-finite entries " << (int64_t)gbad << std::endl;
+    GRID_ASSERT( gbad == 0 );
+    GRID_ASSERT( gmx < 1.0e-5 );
+
+    S.Resize(nrows, N);
+    acceleratorCopyToDevice(&h[0], &S.data[0], (uint64_t)nrows*N*sizeof(ComplexD));
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 3c. The inverse: distributed recursive Schur, END-TO-END fp64.
+  //    stencil (ComplexD) -> fp64 rank-major import -> fp64 recursion ->
+  //    ONE terminal rounding into the fp32 apply slab.  Everything
+  //    downstream (device residency, split-K apply, VERIFY) is fp32.
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void InvertDense(CoarseOp &Op)
+  {
+    double t1 = usecond();
+    int P  = grid->ProcessorCount();
+    int me = grid->ThisRank();
+
+    // Uniform local volumes => contiguous uniform ownership
+    std::vector<int64_t> rowStart(P+1);
+    for(int r=0; r<=P; r++)
+    {
+      rowStart[r] = (int64_t)r*nrows;
+    }
+    GRID_ASSERT( rowStart[P] == N );
+
+    std::vector<int64_t> g2rm;
+    BuildRankMajorMap(g2rm);
+
+    // Self-certifying map: my own global rows land at my rank-major slots
+    for(int ss=0; ss<lsites; ss++)
+    {
+      for(int a=0; a<nbasis; a++)
+      {
+        GRID_ASSERT( g2rm[ myGsite[ss]*nbasis + a ] == (int64_t)me*nrows + ss*nbasis + a );
+      }
+    }
+
+    BlockRows S;
+    ImportDenseFP64(Op, S, g2rm);
+
+    ////////////////////////////////////////////////////////////////
+    // The 2D block-cyclic recursion (BlockCyclicSchurInverse).
+    // Everything in it -- redistribution, SUMMA rings, leaves -- is
+    // point-to-point SendToRecvFrom; no collectives at all.  Block size
+    // nb = rows-per-rank makes the redistribution edges maximally
+    // regular.
+    ////////////////////////////////////////////////////////////////
+    double t2, t3;
+    {
+      int Pr,Pc;
+      BlockCyclicLayout::ChooseProcessGrid(P, Pr, Pc);
+      int64_t nb = nrows;
+      GRID_ASSERT( nb >= 1 );
+      std::cout << GridLogMessage << "DenseCoarseMatrix: 2D SCHUR invert, process grid "
+                << Pr << " x " << Pc << "  nb " << nb
+                << "  (pure P2P: redistribute + SUMMA rings + local leaves)" << std::endl;
+      BlockCyclicMatrix A2(grid, N, nb, Pr, Pc);
+      BlockCyclicSchurInverse RSI2;
+      t2 = usecond();
+      BlockCyclicRedistribute::RowsToCyclic(grid, rowStart, &S.data[0], nrows, A2);
+      RSI2.Invert(A2);
+      BlockCyclicRedistribute::CyclicToRows(grid, rowStart, A2, &S.data[0], nrows);
+      t3 = usecond();
+      RSI2.ReportTelemetry(grid);
+    }
+
+    // The single terminal rounding: fp64 inverse -> fp32 apply slab
+    // (row-major, global columns)
+    {
+      std::vector<ComplexD> h((uint64_t)nrows*N);
+      acceleratorCopyFromDevice(&S.data[0], &h[0], (uint64_t)nrows*N*sizeof(ComplexD));
+      thread_for(gcol, N, {
+        int64_t jj = g2rm[gcol];
+        for(int64_t i=0; i<nrows; i++)
+        {
+          slab[(uint64_t)i*N + gcol] = ComplexF(h[(uint64_t)(i + jj*nrows)]);
+        }
+      });
+    }
+    double t4 = usecond();
+    std::cout << GridLogMessage << "DenseCoarseMatrix: SCHUR fp64 distributed invert took "
+              << (t4-t1)/1.0e6 << " s (recursion " << (t3-t2)/1.0e6 << " s)" << std::endl;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // CORE apply on packed data: hX[N x nr] (zero-filled, my sites only)
+  // -> ring allgather -> split-K GEMM against the resident slab ->
+  // reduce partials -> hY[nrows x nr] (column major).  Platform-agnostic:
+  // deviceVector + GridBLAS (Eigen fallback on CPU builds).
+  // tprof (optional): per-phase microseconds {allgather, H2D, gemm+reduce,
+  // D2H}, printed by the caller on GridLogPerformance.
+  ////////////////////////////////////////////////////////////////////
+  void SlabApplyPacked(int nr, double *tprof)
+  {
+    GRID_TRACE("DenseSlabApply");
+    GRID_ASSERT(nr <= MRHS_MAX);
+    uint64_t nY = (uint64_t)nrows * nr;
+    int64_t  Kc = N / NK;
+    double t1 = usecond();
+    double t2, t3;
+    {
+      // ALLGATHER: x is not a reduction -- every rank owns the rows of x at
+      // global columns myGsite[ss]*nbasis+b (scattered by site coordinate, NOT
+      // a contiguous block) and needs all of it.  Only MY rows go host->device
+      // (nrows x nr, ~15 KB at nr=1), packed rank-major [r][ss*nbasis+b],
+      // gathered along the process grid, then scattered through rm2g into the
+      // column-major dX (ld = N, global-site columns) the split-K GEMM reads.
+      const uint64_t chunk = (uint64_t)nrows*nr;              // my block: [r][ss*nbasis+b]
+      { GRID_TRACE("DenseH2D");
+        std::vector<ComplexF> hG(chunk);
+        for(int r=0;r<nr;r++)
+          for(int ss=0; ss<lsites; ss++)
+            memcpy(&hG[(uint64_t)r*nrows + (uint64_t)ss*nbasis], &hX[(uint64_t)r*N + (uint64_t)myGsite[ss]*nbasis], nbasis*sizeof(ComplexF));
+        acceleratorCopyToDevice(&hG[0], &dG[(uint64_t)myLex*chunk], chunk*sizeof(ComplexF));   // my slot is my LEX index
+      }
+      t2 = usecond();
+      { GRID_TRACE("DenseAllgather");
+        CartesianRingAllGather(grid, (ComplexF *)&dG[0], chunk);
+        // scatter lex block L=[r][i] -> dX[r*N + rm2g[rank(L)*nrows + i]]
+        ComplexF *g = &dG[0]; ComplexF *x = &dX[0]; int *l2r = &dLex2Rank[0]; int64_t *rm2g = &dRm2G[0];
+        const int64_t nrw = nrows; const int64_t NN = N; const int nrr = nr;
+        accelerator_for(idx, (uint64_t)N*nr, 1, {
+          int64_t r = idx / NN;  int64_t gi = idx - r*NN;
+          int64_t L = gi / nrw;  int64_t i  = gi - L*nrw;
+          x[r*NN + rm2g[(int64_t)l2r[L]*nrw + i]] = g[L*(nrw*nrr) + r*nrw + i];
+        });
+      }
+      t3 = usecond();
+    }
+    // Y = op(slab,T) . X : row-major slab (nrows x N) == col-major A^T
+    // (N x nrows, lda=N) => transpose gives the nrows x N operator.
+    // Split-K: NK chunk-GEMMs by pointer offset (AOT lists), then reduce.
+    ComplexF one (1.0,0.0);
+    ComplexF zero(0.0,0.0);
+    { GRID_TRACE("DenseSplitKGEMM");
+      BLAS.gemmBatched(GridBLAS_OP_T, GridBLAS_OP_N,
+                       (int)nrows, nr, (int)Kc,
+                       one,  aptrs, (int)N,
+                             xptrs, (int)N,
+                       zero, cptrs, (int)nrows);
+      BLAS.synchronise();
+      ComplexF *pp = &dPartial[0];
+      ComplexF *py = &dY[0];
+      uint64_t stride = (uint64_t)nrows*MRHS_MAX;
+      int nk = NK;
+      accelerator_for(i, nY, 1, {
+        ComplexF acc(0.0,0.0);
+        for(int j=0;j<nk;j++) acc += pp[(uint64_t)j*stride + i];
+        py[i] = acc;
+      });
+    }
+    double t4 = usecond();
+    { GRID_TRACE("DenseD2H");
+      acceleratorCopyFromDevice(&dY[0],&hY[0],nY*sizeof(ComplexF));
+    }
+    double t5 = usecond();
+    if (tprof) {
+      tprof[0] = t3-t2;                        // allgather
+      tprof[1] = t2-t1;                        // H2D
+      tprof[2] = t4-t3;                        // gemm+reduce
+      tprof[3] = t5-t4;                        // D2H
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Single-RHS apply (also the VERIFY path => certifies device slab).
+  ////////////////////////////////////////////////////////////////////
+  virtual void operator()(const Field &src, Field &psi)
+  {
+    GRID_TRACE("DenseApply1");
+    uint64_t nX = (uint64_t)N;
+    { GRID_TRACE("DensePack");
+      thread_for(i, nX, { hX[i]=ComplexF(0.0,0.0); });
+      for(int ss=0; ss<lsites; ss++){
+        sobj s;
+        peekLocalSite(s, src, myLcoor[ss]);
+        for(int b=0; b<nbasis; b++)
+          hX[ myGsite[ss]*nbasis + b ] = ComplexF(((ComplexD *)&s)[b]);
+      }
+    }
+    SlabApplyPacked(1, nullptr);
+    { GRID_TRACE("DenseUnpack");
+      for(int ss=0; ss<lsites; ss++){
+        sobj s;
+        for(int b=0; b<nbasis; b++)
+          ((ComplexD *)&s)[b] = ComplexD(hY[ss*nbasis + b]);
+        pokeLocalSite(s, psi, myLcoor[ss]);
+      }
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Defect of an applied inverse. Was a DENSE_CC_CHECK block inside
+  // operator(), but that is virtual and cannot take an operator, so the
+  // caller now asks for it explicitly.
+  ////////////////////////////////////////////////////////////////////
+  template<class CoarseOp>
+  void CheckApply(CoarseOp &Op,const Field &src,const Field &psi)
+  {
+    Field tmp(grid);
+    Op.M(psi, tmp);
+    tmp = tmp - src;
+    std::cout << GridLogMessage << "DenseCoarseMatrix: apply defect ||A x - b||/||b|| = "
+              << std::sqrt(norm2(tmp)/norm2(src)) << std::endl;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Batched (vector-of-fields) apply.
+  ////////////////////////////////////////////////////////////////////
+  void ApplyBatch(std::vector<Field> &src, std::vector<Field> &psi)
+  {
+    int nr = src.size();
+    GRID_ASSERT(nr <= MRHS_MAX);
+    double t0 = usecond();
+    uint64_t nX = (uint64_t)N*nr;
+    { GRID_TRACE("DensePack");
+      thread_for(i, nX, { hX[i]=ComplexF(0.0,0.0); });
+      for(int rr=0; rr<nr; rr++){
+        for(int ss=0; ss<lsites; ss++){
+          sobj s;
+          peekLocalSite(s, src[rr], myLcoor[ss]);
+          for(int b=0; b<nbasis; b++)
+            hX[ (uint64_t)rr*N + myGsite[ss]*nbasis + b ] = ComplexF(((ComplexD *)&s)[b]);
+        }
+      }
+    }
+    SlabApplyPacked(nr, nullptr);
+    { GRID_TRACE("DenseUnpack");
+      for(int rr=0; rr<nr; rr++){
+        for(int ss=0; ss<lsites; ss++){
+          sobj s;
+          for(int b=0; b<nbasis; b++)
+            ((ComplexD *)&s)[b] = ComplexD(hY[(uint64_t)rr*nrows + (ss*nbasis+b)]);
+          pokeLocalSite(s, psi[rr], myLcoor[ss]);
+        }
+      }
+    }
+    double t1 = usecond();
+    std::cout << GridLogMessage << "DenseCoarseMatrix: batched apply " << nr << " rhs took "
+              << (t1-t0)/1000.0 << " ms  (" << (t1-t0)/1000.0/nr << " ms/rhs)" << std::endl;
+  }
+
+  template<class CoarseOp>
+  void CheckApplyBatch(CoarseOp &Op,std::vector<Field> &src,std::vector<Field> &psi,int nr)
+  {
+    Field tmp(grid);
+    for(int rr=0; rr<nr; rr++){
+      Op.M(psi[rr], tmp);
+      tmp = tmp - src[rr];
+      std::cout << GridLogMessage << "DenseCoarseMatrix: batch defect["<<rr<<"] = "
+                << std::sqrt(norm2(tmp)/norm2(src[rr])) << std::endl;
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // 6D mrhs apply: operates DIRECTLY on the packed 6D field (rhs = dim 0).
+  ////////////////////////////////////////////////////////////////////
+  void ApplyBatch6D(const Field &in6, Field &out6, int nr)
+  {
+    GRID_ASSERT(nr <= MRHS_MAX);
+    GRID_ASSERT(in6.Grid()->_ndimension == nd+1);   // {rhs, s, x,y,z,t}
+    double t0 = usecond();
+    Field &in = const_cast<Field &>(in6);
+    uint64_t nX = (uint64_t)N * nr;
+    thread_for(i, nX, { hX[i]=ComplexF(0.0,0.0); });
+    { GRID_TRACE("DensePack");
+      autoView(iv, in, CpuRead);
+      Coordinate c6(nd+1);
+      for(int ss=0; ss<lsites; ss++){
+        for(int d=0; d<nd; d++) c6[d+1] = myLcoor[ss][d];
+        for(int rr=0; rr<nr; rr++){
+          c6[0] = rr;
+          sobj s;
+          peekLocalSite(s, iv, c6);
+          for(int b=0; b<nbasis; b++)
+            hX[(uint64_t)rr*N + myGsite[ss]*nbasis + b] = ComplexF(((ComplexD *)&s)[b]);
+        }
+      }
+    }
+    double t1 = usecond();
+    double tprof[4];
+    SlabApplyPacked(nr, tprof);
+    double t5 = usecond();
+    { GRID_TRACE("DenseUnpack");
+      autoView(ov, out6, CpuWrite);
+      Coordinate c6(nd+1);
+      for(int ss=0; ss<lsites; ss++){
+        for(int d=0; d<nd; d++) c6[d+1] = myLcoor[ss][d];
+        for(int rr=0; rr<nr; rr++){
+          c6[0] = rr;
+          sobj s;
+          for(int b=0; b<nbasis; b++)
+            ((ComplexD *)&s)[b] = ComplexD(hY[(uint64_t)rr*nrows + (ss*nbasis+b)]);  // Y col-major
+          pokeLocalSite(s, ov, c6);
+        }
+      }
+    }
+    double t6 = usecond();
+    std::cout << GridLogMessage << "DenseCoarseMatrix: apply6D " << nr << " rhs took "
+              << (t6-t0)/1000.0 << " ms" << std::endl;
+    std::cout << GridLogPerformance << "DenseCoarseMatrix: apply6D profile:"
+              << " pack "        << (t1-t0)/1000.0
+              << "  allgather "  << tprof[0]/1000.0
+              << "  H2D "        << tprof[1]/1000.0
+              << "  gemm+reduce "<< tprof[2]/1000.0
+              << "  D2H "        << tprof[3]/1000.0
+              << "  unpack "     << (t6-t5)/1000.0
+              << "  ms" << std::endl;
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/GeneralCoarsenedMatrix.h

@@ -31,6 +31,7 @@ Author: Peter Boyle <pboyle@bnl.gov>
 
 #include <Grid/lattice/PaddedCell.h>
 #include <Grid/stencil/GeneralLocalStencil.h>
+#include <Grid/algorithms/deflation/MultiRHSBlockProject.h>
 
 NAMESPACE_BEGIN(Grid);
 
@@ -66,9 +67,24 @@ public:
   std::vector<CoarseMatrix> _Adag;
   std::vector<CoarseVector> MultTemporaries;
 
+  int64_t MultCalls;
+  double  MultFlopsAccum;
+  double  MultUsecAccum;
+
   ///////////////////////
   // Interface
   ///////////////////////
+  //////////////////////////////////////////////////////////////////////////
+  // Bilingual accessors: everything a consumer needs to read the operator
+  // without knowing which of the three coarse classes it holds. The D
+  // dimensional grid the elements live on, the geometry they are indexed by,
+  // and one unpadded point at a time (a whole npoint vector is too much
+  // memory at production nbasis).
+  //////////////////////////////////////////////////////////////////////////
+  GridCartesian * CoarseGridD(void)             { return _CoarseGrid; };
+  NonLocalStencilGeometry & Geometry(void)      { return geom; };
+  void ExtractMatrix(int p,CoarseMatrix &A)     { A = Cell.Extract(_A[p]); };
+
   GridBase      * Grid(void)           { return _CoarseGrid; };   // this is all the linalg routines need to know
   GridBase      * FineGrid(void)       { return _FineGrid; };   // this is all the linalg routines need to know
   GridCartesian * CoarseGrid(void)     { return _CoarseGrid; };   // this is all the linalg routines need to know
@@ -99,24 +115,25 @@ public:
 	}
       }
     }
-    assert(nfound==geom.npoint);
+    GRID_ASSERT(nfound==geom.npoint);
     ExchangeCoarseLinks();
   }
   */
   
-  GeneralCoarsenedMatrix(NonLocalStencilGeometry &_geom,GridBase *FineGrid, GridCartesian * CoarseGrid)
+  GeneralCoarsenedMatrix(NonLocalStencilGeometry &_geom,GridBase *FineGrid, GridCartesian * CoarseGrid,int _herm=1)
     : geom(_geom),
       _FineGrid(FineGrid),
       _CoarseGrid(CoarseGrid),
-      hermitian(1),
+      hermitian(_herm),
       Cell(_geom.Depth(),_CoarseGrid),
-      Stencil(Cell.grids.back(),geom.shifts)
+      Stencil(Cell.grids.back(),geom.shifts),
+      MultCalls(0), MultFlopsAccum(0.0), MultUsecAccum(0.0)
   {
     {
       int npoint = _geom.npoint;
     }
     _A.resize(geom.npoint,CoarseGrid);
-    //    _Adag.resize(geom.npoint,CoarseGrid);
+    if ( !hermitian ) _Adag.resize(geom.npoint,CoarseGrid);
   }
   void M (const CoarseVector &in, CoarseVector &out)
   {
@@ -124,10 +141,10 @@ public:
   }
   void Mdag (const CoarseVector &in, CoarseVector &out)
   {
-    assert(hermitian);
-    Mult(_A,in,out);
-    //    if ( hermitian ) M(in,out);
-    //    else Mult(_Adag,in,out);
+    if(hermitian)
+      Mult(_A,in,out);
+    else
+      Mult(_Adag,in,out);
   }
   void Mult (std::vector<CoarseMatrix> &A,const CoarseVector &in, CoarseVector &out)
   {
@@ -227,29 +244,28 @@ public:
     text+=usecond();
     ttot+=usecond();
     
-    std::cout << GridLogPerformance<<"Coarse 1rhs Mult Aviews "<<tviews<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult exch "<<texch<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult mult "<<tmult<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<" of which mult2  "<<tmult2<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult ext  "<<text<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult temps "<<ttemps<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult copy  "<<tcopy<<" us"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Mult tot  "<<ttot<<" us"<<std::endl;
-    //    std::cout << GridLogPerformance<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Kernel flops "<< flops<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Kernel flop/s "<< flops/tmult<<" mflop/s"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse Kernel bytes/s "<< bytes/tmult<<" MB/s"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse overall flops/s "<< flops/ttot<<" mflop/s"<<std::endl;
-    std::cout << GridLogPerformance<<"Coarse total bytes   "<< bytes/1e6<<" MB"<<std::endl;
+    MultCalls++;
+    MultFlopsAccum += flops;
+    MultUsecAccum  += ttot;
+    std::cout << GridLogPerformance
+	      << "Coarse Mult call " << MultCalls
+	      << "  tot " << ttot << " us"
+	      << "  kernel " << tmult << " us"
+	      << "  kernel " << flops/tmult*1e-3 << " GFlop/s"
+	      << "  overall " << MultFlopsAccum/MultUsecAccum*1e-3 << " GFlop/s (cumul)"
+	      << "  bw " << bytes/tmult*1e-3 << " GB/s"
+	      << std::endl;
 
   };
   
   void PopulateAdag(void)
   {
+#if 0
+    // Serial global peek/poke reference implementation
     for(int64_t bidx=0;bidx<CoarseGrid()->gSites() ;bidx++){
       Coordinate bcoor;
       CoarseGrid()->GlobalIndexToGlobalCoor(bidx,bcoor);
-      
+
       for(int p=0;p<geom.npoint;p++){
 	Coordinate scoor = bcoor;
 	for(int mu=0;mu<bcoor.size();mu++){
@@ -262,6 +278,36 @@ public:
 	pokeSite(adj(link),_Adag[pp],bcoor);
       }
     }
+#else
+    // Parallel: _Adag[pp](x) = adj( _A[p](x + s_pp) ),  pp = Reverse(p), s_pp = -s_p.
+    // The neighbour fetch reuses the same padded-cell + stencil machinery as Mult,
+    // reading one matrix element per coalesced access so no whole site matrix
+    // (230KB at nbasis=60) ever lands on a GPU thread stack (HIP limit 128KB).
+    // Halo sites compute garbage neighbours; Cell.Extract discards them.
+    // Must run on the unpadded _A, i.e. before ExchangeCoarseLinks.
+    const int Nsimd = CComplex::Nsimd();
+    for(int p=0;p<geom.npoint;p++){
+      int pp = geom.Reverse(p);
+      CoarseMatrix Apad = Cell.ExchangePeriodic(_A[p]);
+      CoarseMatrix Dpad(Apad.Grid());
+      int64_t osites = Apad.Grid()->oSites();
+      {
+	autoView( Apad_v   , Apad, AcceleratorRead);
+	autoView( Dpad_v   , Dpad, AcceleratorWriteDiscard);
+	autoView( Stencil_v, Stencil, AcceleratorRead);
+	accelerator_for(sj, osites*nbasis, Nsimd, {
+	    int32_t ss = sj/nbasis;
+	    int32_t j  = sj%nbasis;
+	    auto SE = Stencil_v.GetEntry(pp,ss);
+	    for(int i=0;i<nbasis;i++){
+	      auto z = coalescedReadGeneralPermute(Apad_v[SE->_offset](i,j),SE->_permute,Nd);
+	      coalescedWrite(Dpad_v[ss](j,i),conjugate(z));
+	    }
+	});
+      }
+      _Adag[pp] = Cell.Extract(Dpad);
+    }
+#endif
   }
   /////////////////////////////////////////////////////////////
   // 
@@ -417,10 +463,19 @@ public:
 	int osites=CoarseGrid()->oSites();
 	autoView( A_v  , _A[k], AcceleratorWrite);
 	autoView( FT_v  , FT[k], AcceleratorRead);
-	accelerator_for(sss, osites, 1, {
+	accelerator_for(sss, osites, nbasis, {
+#ifdef GRID_SIMT
+	    int j = acceleratorSIMTlane(nbasis);
+	    A_v[sss](i,j) = FT_v[sss](j);
+#else
+	    // CPU build: acceleratorSIMTlane()==0 -- an un-looped SIMT tensor
+	    // index writes ONLY j=0 and silently drops the other nbasis-1
+	    // columns (caught by Test_schur_dense_coarse import certificate,
+	    // 2026-08-14).  Loop explicitly.
 	    for(int j=0;j<nbasis;j++){
 	      A_v[sss](i,j) = FT_v[sss](j);
 	    }
+#endif
         });
       }
       tinv+=usecond();
@@ -428,8 +483,8 @@ public:
 
     // Only needed if nonhermitian
     if ( ! hermitian ) {
-      //      std::cout << GridLogMessage<<"PopulateAdag  "<<std::endl;
-      //      PopulateAdag();
+      std::cout << GridLogMessage<<"PopulateAdag  "<<std::endl;
+      PopulateAdag();
     }
 
     // Need to write something to populate Adag from A
@@ -441,8 +496,20 @@ public:
     std::cout << GridLogMessage<<"CoarsenOperator inv    "<<tinv<<" us"<<std::endl;
   }
 #else
+  //////////////////////////////////////////////////////////////////////
+  // Galerkin projection of matrix
+  //////////////////////////////////////////////////////////////////////
   void CoarsenOperator(LinearOperatorBase<Lattice<Fobj> > &linop,
 		       Aggregation<Fobj,CComplex,nbasis> & Subspace)
+  {
+    CoarsenOperator(linop,Subspace,Subspace);
+  }
+  //////////////////////////////////////////////////////////////////////
+  // Petrov - Galerkin projection of matrix
+  //////////////////////////////////////////////////////////////////////
+  void CoarsenOperator(LinearOperatorBase<Lattice<Fobj> > &linop,
+		       Aggregation<Fobj,CComplex,nbasis> & U,
+		       Aggregation<Fobj,CComplex,nbasis> & V)
   {
     std::cout << GridLogMessage<< "GeneralCoarsenMatrix "<< std::endl;
     GridBase *grid = FineGrid();
@@ -458,11 +525,9 @@ public:
     // Orthogonalise the subblocks over the basis
     /////////////////////////////////////////////////////////////
     CoarseScalar InnerProd(CoarseGrid()); 
-    blockOrthogonalise(InnerProd,Subspace.subspace);
+    blockOrthogonalise(InnerProd,V.subspace);
+    blockOrthogonalise(InnerProd,U.subspace);
 
-    //    for(int s=0;s<Subspace.subspace.size();s++){
-      //      std::cout << " subspace norm "<<norm2(Subspace.subspace[s])<<std::endl;
-    //    }
     const int npoint = geom.npoint;
       
     Coordinate clatt = CoarseGrid()->GlobalDimensions();
@@ -507,13 +572,9 @@ public:
     // Now compute the matrix elements of linop between the orthonormal
     // set of vectors.
     ///////////////////////////////////////////////////////////////////////
-    FineField phaV(grid); // Phased block basis vector
-    FineField MphaV(grid);// Matrix applied
     std::vector<FineComplexField> phaF(npoint,grid);
     std::vector<CoarseComplexField> pha(npoint,CoarseGrid());
-    
-    CoarseVector coarseInner(CoarseGrid());
-    
+
     typedef typename CComplex::scalar_type SComplex;
     FineComplexField one(grid); one=SComplex(1.0);
     FineComplexField zz(grid); zz = Zero();
@@ -532,37 +593,52 @@ public:
       pha[p]  =exp(pha[p]*ci);
 
       blockZAXPY(phaF[p],pha[p],one,zz);
-      
+
     }
     tphase+=usecond();
-    
-    std::vector<CoarseVector> ComputeProj(npoint,CoarseGrid());
-    std::vector<CoarseVector>          FT(npoint,CoarseGrid());
+
+    // Import basis into BLAS layout once; blockProject then reads it once per
+    // basis vector rather than once per (i,p) as in scalar blockProject.
+    // Process all npoint in a single batch.
+    MultiRHSBlockProject<FineField> Projector;
+    Projector.Allocate(nbasis, grid, CoarseGrid());
+    Projector.ImportBasis(U.subspace);
+
+    std::vector<FineField>    phaV_batch(npoint, grid);
+    std::vector<FineField>    MphaV_batch(npoint, grid);
+    std::vector<CoarseVector> proj_batch(npoint, CoarseGrid());
+    std::vector<CoarseVector> ComputeProj(npoint, CoarseGrid());
+    std::vector<CoarseVector>          FT(npoint, CoarseGrid());
+
+    // Pre-allocate BLAS_F and BLAS_C to avoid repeated hipMalloc/hipFree of
+    // ~5.6 GB per blockProject call, which hangs on ROCm for large allocations.
+    Projector.BLAS_F.resize(Projector.fine_vol * Projector.words * npoint);
+    Projector.BLAS_C.resize(Projector.coarse_vol * nbasis * npoint);
+
     for(int i=0;i<nbasis;i++){// Loop over basis vectors
+      accelerator_barrier(); // ensure prior iteration's async writes are retired
       std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i<<"/"<<nbasis<< std::endl;
-      for(int p=0;p<npoint;p++){ // Loop over momenta in npoint
-	tphaseBZ-=usecond();
-	phaV = phaF[p]*Subspace.subspace[i];
-	tphaseBZ+=usecond();
 
-	/////////////////////////////////////////////////////////////////////
-	// Multiple phased subspace vector by matrix and project to subspace
-	// Remove local bulk phase to leave relative phases
-	/////////////////////////////////////////////////////////////////////
-	tmat-=usecond();
-	linop.Op(phaV,MphaV);
-	tmat+=usecond();
-	//	std::cout << i << " " <<p << " MphaV "<<norm2(MphaV)<<" "<<norm2(phaV)<<std::endl;
+      tphaseBZ-=usecond();
+      for(int p=0;p<npoint;p++)
+	phaV_batch[p] = phaF[p] * V.subspace[i];
+      tphaseBZ+=usecond();
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i<<" phaseBZ done"<< std::endl;
 
-	tproj-=usecond();
-	blockProject(coarseInner,MphaV,Subspace.subspace);
-	coarseInner = conjugate(pha[p]) * coarseInner;
+      tmat-=usecond();
+      for(int p=0;p<npoint;p++)
+	linop.Op(phaV_batch[p], MphaV_batch[p]);
+      tmat+=usecond();
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i<<" mat done"<< std::endl;
 
-	ComputeProj[p] = coarseInner;
-	tproj+=usecond();
-	//	std::cout << i << " " <<p << " ComputeProj "<<norm2(ComputeProj[p])<<std::endl;
-
-      }
+      // One batched GEMM reads BLAS_V once for all npoint vectors.
+      tproj-=usecond();
+      Projector.blockProject(MphaV_batch, proj_batch);
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i<<" blockProject done"<< std::endl;
+      for(int p=0;p<npoint;p++)
+	ComputeProj[p] = conjugate(pha[p]) * proj_batch[p];
+      tproj+=usecond();
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i<<" proj done"<< std::endl;
 
       tinv-=usecond();
       for(int k=0;k<npoint;k++){
@@ -570,14 +646,23 @@ public:
 	for(int l=0;l<npoint;l++){
 	  FT[k]= FT[k]+ invMkl(l,k)*ComputeProj[l];
 	}
-      
+
 	int osites=CoarseGrid()->oSites();
 	autoView( A_v  , _A[k], AcceleratorWrite);
 	autoView( FT_v  , FT[k], AcceleratorRead);
-	accelerator_for(sss, osites, 1, {
+	accelerator_for(sss, osites, nbasis, {
+#ifdef GRID_SIMT
+	    int j = acceleratorSIMTlane(nbasis);
+	    A_v[sss](i,j) = FT_v[sss](j);
+#else
+	    // CPU build: acceleratorSIMTlane()==0 -- an un-looped SIMT tensor
+	    // index writes ONLY j=0 and silently drops the other nbasis-1
+	    // columns (caught by Test_schur_dense_coarse import certificate,
+	    // 2026-08-14).  Loop explicitly.
 	    for(int j=0;j<nbasis;j++){
 	      A_v[sss](i,j) = FT_v[sss](j);
 	    }
+#endif
         });
       }
       tinv+=usecond();
@@ -585,13 +670,13 @@ public:
 
     // Only needed if nonhermitian
     if ( ! hermitian ) {
-      //      std::cout << GridLogMessage<<"PopulateAdag  "<<std::endl;
-      //      PopulateAdag();
+      std::cout << GridLogMessage<<"PopulateAdag  "<<std::endl;
+      PopulateAdag();
     }
 
-    for(int p=0;p<geom.npoint;p++){
-      std::cout << " _A["<<p<<"] "<<norm2(_A[p])<<std::endl;
-    }
+    //    for(int p=0;p<geom.npoint;p++){
+    //      std::cout << " _A["<<p<<"] "<<norm2(_A[p])<<std::endl;
+    //    }
 
     // Need to write something to populate Adag from A
     ExchangeCoarseLinks();
@@ -606,10 +691,10 @@ public:
   void ExchangeCoarseLinks(void){
     for(int p=0;p<geom.npoint;p++){
       _A[p] = Cell.ExchangePeriodic(_A[p]);
-      //      _Adag[p]= Cell.ExchangePeriodic(_Adag[p]);
+      if ( !hermitian ) _Adag[p]= Cell.ExchangePeriodic(_Adag[p]);
     }
   }
-  virtual  void Mdiag    (const Field &in, Field &out){ assert(0);};
+  virtual  void Mdiag    (const Field &in, Field &out){ GRID_ASSERT(0);};
   virtual  void Mdir     (const Field &in, Field &out,int dir, int disp){assert(0);};
   virtual  void MdirAll  (const Field &in, std::vector<Field> &out){assert(0);};
 };

Grid/algorithms/multigrid/GeneralCoarsenedMatrixMultiRHS.h

@@ -77,15 +77,24 @@ public:
   GridBase      * Grid(void)           { return _CoarseGridMulti; };   // this is all the linalg routines need to know
   GridCartesian * CoarseGrid(void)     { return _CoarseGridMulti; };   // this is all the linalg routines need to know
 
-  // Can be used to do I/O on the operator matrices externally
-  void SetMatrix (int p,CoarseMatrix & A)
+  //////////////////////////////////////////////////////////////////////////
+  // Bilingual accessors, matching GeneralCoarsenedMatrix. Grid() here is the
+  // D+1 multiRHS grid and this class never holds the D dimensional one, so
+  // ExtractMatrix writes into whatever grid the caller's lattice is on.
+  //////////////////////////////////////////////////////////////////////////
+  NonLocalStencilGeometry & Geometry(void)      { return geom_srhs; };
+  void ExtractMatrix(int p,CoarseMatrix &A)     { BLAStoGrid(A,BLAS_A[p]); };
+
+  // I/O on the operator matrices, via the BLAS layout array. The parameter is
+  // a vector over the geometry points; the body indexes A[p].
+  void SetMatrix (int p,std::vector<CoarseMatrix> & A)
   {
-    assert(A.size()==geom_srhs.npoint);
+    GRID_ASSERT(A.size()==geom_srhs.npoint);
     GridtoBLAS(A[p],BLAS_A[p]);
   }
-  void GetMatrix (int p,CoarseMatrix & A)
+  void GetMatrix (int p,std::vector<CoarseMatrix> & A)
   {
-    assert(A.size()==geom_srhs.npoint);
+    GRID_ASSERT(A.size()==geom_srhs.npoint);
     BLAStoGrid(A[p],BLAS_A[p]);
   }
   void CopyMatrix (GeneralCoarseOp &_Op)
@@ -178,14 +187,14 @@ public:
 	for(int32_t point = 0 ; point < geom.npoint; point++){
 	  int i=s*orhs*geom.npoint+point;
  	  int32_t nbr = Stencil._entries[i]._offset*CComplex::Nsimd(); // oSite -> lSite
-	  assert(nbr<BLAS_B.size());
+	  GRID_ASSERT(nbr<BLAS_B.size());
 	  ComplexD * ptr = (ComplexD *)&BLAS_B[nbr];
 	  acceleratorPut(BLAS_BP[point][j],ptr); // neighbour indexing in ghost zone volume
 	}
 	j++;
       }
     }
-    assert(j==unpadded_sites);
+    GRID_ASSERT(j==unpadded_sites);
   }
   template<class vobj> void GridtoBLAS(const Lattice<vobj> &from,deviceVector<typename vobj::scalar_object> &to)
   {
@@ -194,7 +203,7 @@ public:
   typedef typename vobj::vector_type vector_type;
 
   GridBase *Fg = from.Grid();
-  assert(!Fg->_isCheckerBoarded);
+  GRID_ASSERT(!Fg->_isCheckerBoarded);
   int nd = Fg->_ndimension;
 
   to.resize(Fg->lSites());
@@ -241,10 +250,10 @@ public:
   typedef typename vobj::vector_type vector_type;
 
   GridBase *Tg = grid.Grid();
-  assert(!Tg->_isCheckerBoarded);
+  GRID_ASSERT(!Tg->_isCheckerBoarded);
   int nd = Tg->_ndimension;
   
-  assert(in.size()==Tg->lSites());
+  GRID_ASSERT(in.size()==Tg->lSites());
 
   Coordinate LocalLatt = Tg->LocalDimensions();
   size_t nsite = 1;
@@ -669,7 +678,7 @@ Grid : Message : 328.193436 s : CoarsenOperator mat    122213270 us
     const int Nsimd = CComplex::Nsimd();
 
     int64_t nrhs  =pin.Grid()->GlobalDimensions()[0];
-    assert(nrhs>=1);
+    GRID_ASSERT(nrhs>=1);
 
     RealD flops,bytes;
     int64_t osites=in.Grid()->oSites(); // unpadded
@@ -721,7 +730,7 @@ Grid : Message : 328.193436 s : CoarsenOperator mat    122213270 us
     //    std::cout << GridLogMessage<<"Coarse overall flops/s "<< flops/t_tot<<" mflop/s"<<std::endl;
     //    std::cout << GridLogMessage<<"Coarse total bytes   "<< bytes/1e6<<" MB"<<std::endl;
   };
-  virtual  void Mdiag    (const Field &in, Field &out){ assert(0);};
+  virtual  void Mdiag    (const Field &in, Field &out){ GRID_ASSERT(0);};
   virtual  void Mdir     (const Field &in, Field &out,int dir, int disp){assert(0);};
   virtual  void MdirAll  (const Field &in, std::vector<Field> &out){assert(0);};
 };

Grid/algorithms/multigrid/GeneralCoarsenedMatrixMultiRHSV2.h

@@ -0,0 +1,870 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/GeneralCoarsenedMatrixMultiRHS.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+
+NAMESPACE_BEGIN(Grid);
+
+
+// Fine Object == (per site) type of fine field
+// nbasis      == number of deflation vectors
+template<class Fobj,class CComplex,int nbasis>
+class MultiGeneralCoarsenedOperatorV2 : public SparseMatrixBase<Lattice<iVector<CComplex,nbasis > > >  {
+public:
+  typedef typename CComplex::scalar_object SComplex;
+  typedef GeneralCoarsenedMatrix<Fobj,CComplex,nbasis> GeneralCoarseOp;
+  typedef MultiGeneralCoarsenedOperatorV2<Fobj,CComplex,nbasis> MultiGeneralCoarseOp;
+
+  typedef iVector<CComplex,nbasis >           siteVector;
+  typedef iMatrix<CComplex,nbasis >           siteMatrix;
+  typedef iVector<SComplex,nbasis >           calcVector;
+  typedef iMatrix<SComplex,nbasis >           calcMatrix;
+  typedef Lattice<iScalar<CComplex> >         CoarseComplexField;
+  typedef Lattice<siteVector>                 CoarseVector;
+  typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
+  typedef iMatrix<CComplex,nbasis >  Cobj;
+  typedef iVector<CComplex,nbasis >  Cvec;
+  typedef Lattice< CComplex >   CoarseScalar; // used for inner products on fine field
+  typedef Lattice<Fobj >        FineField;
+  typedef CoarseVector Field;
+
+  // Block operations on the fine vectors carry the fine layout, which need
+  // not be the coarse one
+  typedef decltype(innerProduct(Fobj(),Fobj())) FineInner;
+  typedef Lattice<FineInner>                    FineComplexField;
+  typedef Lattice<FineInner>                    BlockComplexField;
+
+  ////////////////////
+  // Data members
+  //
+  // Nrhs independent: the D dimensional coarse grid, the geometry, the padded
+  // cell that supplies the stencil grid, the stencil, and the matrix elements.
+  //
+  // Nrhs dependent: the D+1 grid, its padded cell, and the BLAS B/C buffers
+  // with their pointer tables. Owned by SetNRHS().
+  ////////////////////
+  GridCartesian *       _CoarseGrid;        // D dimensional
+  NonLocalStencilGeometry geom;
+  NonLocalStencilGeometry geom_srhs;
+  PaddedCell CellD;                         // D dimensional, supplies stencil grid
+  GeneralLocalStencil Stencil;              // D dimensional
+
+  int                   _Nrhs;
+  GridCartesian *       _CoarseGridMulti;   // D+1 dimensional, SetNRHS
+  PaddedCell *          CellMulti;          // D+1 dimensional, SetNRHS
+
+  deviceVector<calcVector> BLAS_B;
+  deviceVector<calcVector> BLAS_C;
+  std::vector<deviceVector<calcMatrix> > BLAS_A;
+
+  std::vector<deviceVector<ComplexD *> > BLAS_AP;
+  std::vector<deviceVector<ComplexD *> > BLAS_BP;
+  deviceVector<ComplexD *>               BLAS_CP;
+
+  ///////////////////////
+  // Interface
+  ///////////////////////
+  GridBase      * Grid(void)           { CheckGridSet(); return _CoarseGridMulti; };
+  GridCartesian * CoarseGrid(void)     { CheckGridSet(); return _CoarseGridMulti; };
+  GridCartesian * CoarseGridD(void)    { return _CoarseGrid; };        // lower dimensional grid
+  int             Nrhs(void)           { CheckGridSet(); return _Nrhs; };
+
+  void CheckGridSet(void)
+  {
+    if ( _CoarseGridMulti == nullptr ) {
+      std::cout << GridLogError
+		<< "MultiGeneralCoarsenedOperatorV2: the multiRHS grid has not been set."
+		<< std::endl;
+      std::cout << GridLogError
+		<< "  Call SetGrid(CoarseGridMulti) with the D+1 dimensional grid your"
+		<< std::endl;
+      std::cout << GridLogError
+		<< "  coarse vectors live on, before Grid(), Nrhs() or M()."
+		<< std::endl;
+      GRID_ASSERT(_CoarseGridMulti != nullptr);
+    }
+  }
+
+  //////////////////////////////////////////////////////////////////////////
+  // Bilingual accessors, matching GeneralCoarsenedMatrix. Note Grid() is the
+  // D+1 multiRHS grid here, so a consumer wanting the space the elements live
+  // on must ask for CoarseGridD().
+  //////////////////////////////////////////////////////////////////////////
+  NonLocalStencilGeometry & Geometry(void)      { return geom_srhs; };
+  void ExtractMatrix(int p,CoarseMatrix &A)     { BLAStoGrid(A,BLAS_A[p]); };
+
+  // I/O on the operator matrices, via the BLAS layout array. The parameter is
+  // a vector over the geometry points; the body indexes A[p].
+  void SetMatrix (int p,std::vector<CoarseMatrix> & A)
+  {
+    GRID_ASSERT(A.size()==geom_srhs.npoint);
+    GridtoBLAS(A[p],BLAS_A[p]);
+  }
+  void GetMatrix (int p,std::vector<CoarseMatrix> & A)
+  {
+    GRID_ASSERT(A.size()==geom_srhs.npoint);
+    BLAStoGrid(A[p],BLAS_A[p]);
+  }
+  void CopyMatrix (GeneralCoarseOp &_Op)
+  {
+    for(int p=0;p<geom.npoint;p++){
+      auto Aup = _Op.Cell.Extract(_Op._A[p]);
+      //Unpadded
+      GridtoBLAS(Aup,BLAS_A[p]);
+    }
+  }
+  /*
+  void CheckMatrix (GeneralCoarseOp &_Op)
+  {
+    std::cout <<"************* Checking the little direc operator mRHS"<<std::endl;
+    for(int p=0;p<geom.npoint;p++){
+      //Unpadded
+      auto Aup = _Op.Cell.Extract(_Op._A[p]);
+      auto Ack = Aup;
+      BLAStoGrid(Ack,BLAS_A[p]);
+      std::cout << p<<" Ack "<<norm2(Ack)<<std::endl;
+      std::cout << p<<" Aup "<<norm2(Aup)<<std::endl;
+    }
+    std::cout <<"************* "<<std::endl;
+  }
+  */
+  
+  ///////////////////////////////////////////////////////////////////////////
+  // Constructor takes the D dimensional coarse grid. Everything built here
+  // is independent of Nrhs, in particular the matrix elements, which must
+  // survive a change of Nrhs untouched.
+  ///////////////////////////////////////////////////////////////////////////
+  MultiGeneralCoarsenedOperatorV2(NonLocalStencilGeometry &_geom,GridCartesian *CoarseGrid) :
+    _CoarseGrid(CoarseGrid),
+    geom_srhs(_geom),
+    geom(CoarseGrid,_geom.hops,_geom.skip),
+    CellD(geom.Depth(),CoarseGrid),
+    Stencil(CellD.grids.back(),geom.shifts), // D dimensional padded cell stencil
+    _Nrhs(-1),
+    _CoarseGridMulti(nullptr),
+    CellMulti(nullptr)
+  {
+    int32_t unpadded_sites = _CoarseGrid->lSites();
+
+    /////////////////////////////////////////////////
+    // Matrix elements and their pointer table
+    /////////////////////////////////////////////////
+    BLAS_A.resize(geom.npoint);
+    BLAS_AP.resize(geom.npoint);
+    for(int p=0;p<geom.npoint;p++){
+      BLAS_A[p].resize (unpadded_sites); // no ghost zone, npoint elements
+      BLAS_AP[p].resize(unpadded_sites);
+    }
+
+    // Site identity mapping for A
+    for(int p=0;p<geom.npoint;p++){
+      for(int ss=0;ss<unpadded_sites;ss++){
+	ComplexD *ptr = (ComplexD *)&BLAS_A[p][ss];
+	acceleratorPut(BLAS_AP[p][ss],ptr);
+      }
+    }
+  }
+
+  virtual ~MultiGeneralCoarsenedOperatorV2()
+  {
+    ReleaseGrid();
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Free everything SetGrid allocated. The D+1 grid is borrowed from the
+  // caller and is never deleted here. Safe to call repeatedly and before
+  // the destructor.
+  ///////////////////////////////////////////////////////////////////////////
+  void ReleaseGrid(void)
+  {
+    if ( CellMulti != nullptr ) { delete CellMulti; CellMulti = nullptr; }
+
+    _CoarseGridMulti = nullptr;  // borrowed, not owned
+    _Nrhs            = -1;
+
+    BLAS_B.resize(0);
+    BLAS_C.resize(0);
+    for(int p=0;p<BLAS_BP.size();p++){
+      BLAS_BP[p].resize(0);
+    }
+    BLAS_BP.resize(0);
+    BLAS_CP.resize(0);
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Everything that depends on Nrhs. Idempotent; lazy called on demand.
+  //
+  // The stencil lives on the D dimensional padded grid. PaddedCell only pads
+  // a dimension when it is distributed, and the rhs direction never is, so
+  // the D+1 padded grid is exactly Nrhs copies of the D dimensional padded
+  // grid with rhs innermost. The neighbour offset therefore carries an Nrhs
+  // factor, in the same way the Nsimd factor is carried.
+  ///////////////////////////////////////////////////////////////////////////
+  void SetGrid(GridCartesian *CoarseGridMulti)
+  {
+    GRID_ASSERT(CoarseGridMulti != nullptr);
+
+    if ( CoarseGridMulti == _CoarseGridMulti ) return; // idempotent on identity
+
+    ReleaseGrid();
+
+    /////////////////////////////////////////////////
+    // The D+1 grid is supplied and owned by the caller. Two operators over
+    // the same coarse space must share one grid object or their fields will
+    // not conform, so this is never manufactured internally.
+    /////////////////////////////////////////////////
+    int nd = _CoarseGrid->_ndimension;
+
+    GRID_ASSERT(CoarseGridMulti->_ndimension == nd+1);
+    GRID_ASSERT(CoarseGridMulti->_processors[0] == 1);   // rhs is not distributed
+    for(int d=0;d<nd;d++){
+      GRID_ASSERT(CoarseGridMulti->_fdimensions[d+1] == _CoarseGrid->_fdimensions[d]);
+      GRID_ASSERT(CoarseGridMulti->_processors [d+1] == _CoarseGrid->_processors [d]);
+      GRID_ASSERT(CoarseGridMulti->_simd_layout[d+1] == _CoarseGrid->_simd_layout[d]);
+    }
+
+    _CoarseGridMulti = CoarseGridMulti;
+    _Nrhs            = CoarseGridMulti->_fdimensions[0];
+    GRID_ASSERT(_Nrhs>=1);
+
+    int nrhs = _Nrhs;
+
+    CellMulti = new PaddedCell(geom.Depth(),_CoarseGridMulti);
+
+    int32_t padded_sites   = CellD.grids.back()->lSites(); // D dimensional
+    int32_t unpadded_sites = _CoarseGrid->lSites();        // D dimensional
+
+    // The neighbour offset multiplication by nrhs is exact only if the D+1
+    // padded volume is nrhs copies of the D dimensional one. Check it.
+    GRID_ASSERT(CellMulti->grids.back()->lSites() == nrhs*padded_sites);
+    GRID_ASSERT(_CoarseGridMulti->lSites()        == nrhs*unpadded_sites);
+
+    /////////////////////////////////////////////////
+    // Device data vector storage
+    /////////////////////////////////////////////////
+    BLAS_B.resize(nrhs *padded_sites);   // includes ghost zone
+    BLAS_C.resize(nrhs *unpadded_sites); // no ghost zone
+    BLAS_BP.resize(geom.npoint);
+    for(int p=0;p<geom.npoint;p++){
+      BLAS_BP[p].resize(unpadded_sites);
+    }
+    BLAS_CP.resize(unpadded_sites);
+
+    // Site identity mapping for C
+    for(int ss=0;ss<unpadded_sites;ss++){
+      ComplexD *ptr = (ComplexD *)&BLAS_C[ss*nrhs];
+      acceleratorPut(BLAS_CP[ss],ptr);
+    }
+
+    // Neighbour table is more complicated
+    int32_t j=0; // Interior point counter (unpadded)
+    for(int32_t s=0;s<padded_sites;s++){ // D volume, padded
+      int ghost_zone=0;
+      for(int32_t point = 0 ; point < geom.npoint; point++){
+	int i=s*geom.npoint+point;
+	if( Stencil._entries[i]._wrap ) { // stencil is indexed by the oSite of the D dim grid
+	  ghost_zone=1; // If general stencil wrapped in any direction, wrap=1
+	}
+      }
+
+      if( ghost_zone==0) {
+	for(int32_t point = 0 ; point < geom.npoint; point++){
+	  int i=s*geom.npoint+point;
+ 	  int32_t nbr = Stencil._entries[i]._offset*CComplex::Nsimd(); // oSite -> lSite, D dim
+	  nbr = nbr*nrhs;                                              // D -> D+1, rhs innermost
+	  GRID_ASSERT(nbr<BLAS_B.size());
+	  ComplexD * ptr = (ComplexD *)&BLAS_B[nbr];
+	  acceleratorPut(BLAS_BP[point][j],ptr); // neighbour indexing in ghost zone volume
+	}
+	j++;
+      }
+    }
+    GRID_ASSERT(j==unpadded_sites);
+  }
+  template<class vobj> void GridtoBLAS(const Lattice<vobj> &from,deviceVector<typename vobj::scalar_object> &to)
+  {
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  GridBase *Fg = from.Grid();
+  GRID_ASSERT(!Fg->_isCheckerBoarded);
+  int nd = Fg->_ndimension;
+
+  to.resize(Fg->lSites());
+
+  Coordinate LocalLatt = Fg->LocalDimensions();
+  size_t nsite = 1;
+  for(int i=0;i<nd;i++) nsite *= LocalLatt[i];
+
+  ////////////////////////////////////////////////////////////////////////////////////////////////
+  // do the index calc on the GPU
+  ////////////////////////////////////////////////////////////////////////////////////////////////
+  Coordinate f_ostride = Fg->_ostride;
+  Coordinate f_istride = Fg->_istride;
+  Coordinate f_rdimensions = Fg->_rdimensions;
+
+  autoView(from_v,from,AcceleratorRead);
+  auto to_v = &to[0];
+
+  const int words=sizeof(vobj)/sizeof(vector_type);
+  accelerator_for(idx,nsite,1,{
+      
+      Coordinate from_coor, base;
+      Lexicographic::CoorFromIndex(base,idx,LocalLatt);
+      for(int i=0;i<nd;i++){
+	from_coor[i] = base[i];
+      }
+      int from_oidx = 0; for(int d=0;d<nd;d++) from_oidx+=f_ostride[d]*(from_coor[d]%f_rdimensions[d]);
+      int from_lane = 0; for(int d=0;d<nd;d++) from_lane+=f_istride[d]*(from_coor[d]/f_rdimensions[d]);
+
+      const vector_type* from = (const vector_type *)&from_v[from_oidx];
+      scalar_type* to = (scalar_type *)&to_v[idx];
+      
+      scalar_type stmp;
+      for(int w=0;w<words;w++){
+	stmp = getlane(from[w], from_lane);
+	to[w] = stmp;
+      }
+    });
+  }    
+  template<class vobj> void BLAStoGrid(Lattice<vobj> &grid,deviceVector<typename vobj::scalar_object> &in)
+  {
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  GridBase *Tg = grid.Grid();
+  GRID_ASSERT(!Tg->_isCheckerBoarded);
+  int nd = Tg->_ndimension;
+  
+  GRID_ASSERT(in.size()==Tg->lSites());
+
+  Coordinate LocalLatt = Tg->LocalDimensions();
+  size_t nsite = 1;
+  for(int i=0;i<nd;i++) nsite *= LocalLatt[i];
+
+  ////////////////////////////////////////////////////////////////////////////////////////////////
+  // do the index calc on the GPU
+  ////////////////////////////////////////////////////////////////////////////////////////////////
+  Coordinate t_ostride = Tg->_ostride;
+  Coordinate t_istride = Tg->_istride;
+  Coordinate t_rdimensions = Tg->_rdimensions;
+
+  autoView(to_v,grid,AcceleratorWrite);
+  auto from_v = &in[0];
+
+  const int words=sizeof(vobj)/sizeof(vector_type);
+  accelerator_for(idx,nsite,1,{
+      
+      Coordinate to_coor, base;
+      Lexicographic::CoorFromIndex(base,idx,LocalLatt);
+      for(int i=0;i<nd;i++){
+	to_coor[i] = base[i];
+      }
+      int to_oidx = 0; for(int d=0;d<nd;d++) to_oidx+=t_ostride[d]*(to_coor[d]%t_rdimensions[d]);
+      int to_lane = 0; for(int d=0;d<nd;d++) to_lane+=t_istride[d]*(to_coor[d]/t_rdimensions[d]);
+
+      vector_type* to = (vector_type *)&to_v[to_oidx];
+      scalar_type* from = (scalar_type *)&from_v[idx];
+      
+      scalar_type stmp;
+      for(int w=0;w<words;w++){
+	stmp=from[w];
+	putlane(to[w], stmp, to_lane);
+      }
+    });
+  }
+  ///////////////////////////////////////////////////////////////////////////
+  // Shared by both CoarsenOperator variants
+  //
+  //     conj(pha[block]) proj[k (which mom)][j (basis vec cpt)][block]
+  //       =  \sum_{l in ball}  e^{i q_k . delta_l} < phi_{block,j} | MdagM | phi_{(block+delta_l),i} >
+  //       =  \sum_{l in ball} e^{iqk.delta_l} A_ji^{b.b+l}
+  //       = M_{kl} A_ji^{b.b+l}
+  //
+  //     Where q_k = delta_k . (2*M_PI/global_nb[mu])
+  //     Then A{ji}^{b,b+l} = M^{-1}_{lm} ComputeProj_{m,b,i,j}
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenFourierMatrix(GridBase *CoarseGrid,Eigen::MatrixXcd &invMkl)
+  {
+    const int npoint = geom_srhs.npoint;
+    Coordinate clatt = CoarseGrid->GlobalDimensions();
+    int Nd = CoarseGrid->Nd();
+
+    Eigen::MatrixXcd Mkl = Eigen::MatrixXcd::Zero(npoint,npoint);
+    ComplexD ci(0.0,1.0);
+    for(int k=0;k<npoint;k++){ // Loop over momenta
+      for(int l=0;l<npoint;l++){ // Loop over nbr relative
+	ComplexD phase(0.0,0.0);
+	for(int mu=0;mu<Nd;mu++){
+	  RealD TwoPiL =  M_PI * 2.0/ clatt[mu];
+	  phase=phase+TwoPiL*geom_srhs.shifts[k][mu]*geom_srhs.shifts[l][mu];
+	}
+	phase=exp(phase*ci);
+	Mkl(k,l) = phase;
+      }
+    }
+    invMkl = Mkl.inverse();
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // blockOrthogonalise and blockZAXPY are block operations on the fine
+  // vectors, using a coarse shaped field only as an index set. They need a
+  // grid carrying the fine SIMD layout, which the coarse space no longer
+  // does. Constructed local to the caller so it cannot be mistaken for the
+  // coarse grid.
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenBlockGridLayout(GridBase *grid,GridBase *CoarseGrid,
+			      Coordinate &latt,Coordinate &simd,Coordinate &mpi)
+  {
+    int nd = CoarseGrid->_ndimension;
+    latt.resize(nd); simd.resize(nd); mpi.resize(nd);
+    for(int d=0;d<nd;d++){
+      latt[d] = CoarseGrid->_fdimensions[d];
+      simd[d] = grid->_simd_layout[d];
+      mpi [d] = CoarseGrid->_processors[d];
+    }
+  }
+
+  // D+1 coarse grid holding the batch, rhs innermost and unvectorised
+  void CoarsenBatchGridLayout(GridBase *CoarseGrid,int batch,
+			      Coordinate &latt,Coordinate &simd,Coordinate &mpi)
+  {
+    latt.resize(1,batch); simd.resize(1,1); mpi.resize(1,1);
+    latt[0]=batch; simd[0]=1; mpi[0]=1;
+    for(int d=0;d<CoarseGrid->_ndimension;d++){
+      latt.push_back(CoarseGrid->_fdimensions[d]);
+      simd.push_back(CoarseGrid->_simd_layout[d]);
+      mpi .push_back(CoarseGrid->_processors[d]);
+    }
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // The Fourier inverse needs the phase in the coarse layout and the basis
+  // phasing needs it in the fine layout; each is built from its own
+  // coordinates rather than transferred.
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenPhases(GridBase *grid,GridBase *CoarseGrid,GridCartesian *BlockGrid,
+		     std::vector<CoarseComplexField> &pha,
+		     std::vector<FineComplexField> &phaF)
+  {
+    const int npoint = geom_srhs.npoint;
+    Coordinate clatt = CoarseGrid->GlobalDimensions();
+    int Nd = CoarseGrid->Nd();
+    ComplexD ci(0.0,1.0);
+
+    typedef typename CComplex::scalar_type SComplex;
+    FineComplexField one(grid); one=SComplex(1.0);
+    FineComplexField zz(grid);  zz = Zero();
+    BlockComplexField pha_blk (BlockGrid);
+    BlockComplexField blk_coor(BlockGrid);
+
+    for(int p=0;p<npoint;p++){ // Loop over momenta in npoint
+      CoarseComplexField coor(CoarseGrid);
+      pha[p] =Zero();
+      pha_blk=Zero();
+      for(int mu=0;mu<Nd;mu++){
+	RealD TwoPiL =  M_PI * 2.0/ clatt[mu];
+	LatticeCoordinate(coor,mu);
+	pha[p]  = pha[p]  + (TwoPiL * geom_srhs.shifts[p][mu]) * coor;
+	LatticeCoordinate(blk_coor,mu);
+	pha_blk = pha_blk + (TwoPiL * geom_srhs.shifts[p][mu]) * blk_coor;
+      }
+      pha[p] =exp(pha[p] *ci);
+      pha_blk=exp(pha_blk*ci);
+
+      blockZAXPY(phaF[p],pha_blk,one,zz);
+    }
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Remove the bulk phase from the batch of coarse projections and
+  // accumulate the Fourier inverse into A. Both variants reach here with
+  // TmpProj in the same batch coarse order, so this is shared verbatim.
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenAccumulate(int p,int i0,int nbv,int batch,
+			 Eigen::MatrixXcd &invMkl,
+			 std::vector<CoarseComplexField> &pha,
+			 CoarseComplexField &phaB,
+			 CoarseVector &TmpProj,
+			 std::vector<CoarseMatrix> &_A,
+			 GridBase *CoarseGrid)
+  {
+    typedef typename CComplex::scalar_type SComplex;
+    const int npoint = geom_srhs.npoint;
+
+    for(int b=0;b<batch;b++) InsertSliceFast(pha[p],phaB,b,0);
+    TmpProj = conjugate(phaB)*TmpProj;
+
+    int osites=CoarseGrid->oSites();
+    for(int k=0;k<npoint;k++){
+      SComplex sc(invMkl(p,k).real(),invMkl(p,k).imag());
+      CComplex coef(sc);
+      autoView( A_v  , _A[k], AcceleratorWrite);
+      autoView( TP_v , TmpProj, AcceleratorRead);
+      accelerator_for(sss, osites, 1, {
+	  for(int b=0;b<nbv;b++){
+	    for(int j=0;j<nbasis;j++){
+	      A_v[sss](i0+b,j) = A_v[sss](i0+b,j) + coef*TP_v[b+batch*sss](j);
+	    }
+	  }
+      });
+    }
+  }
+
+  void CoarsenReport(RealD tphase,RealD tphaseBZ,RealD tslice,
+		     RealD tmat,RealD tproj,RealD tinv)
+  {
+    std::cout << GridLogMessage<<"CoarsenOperator phase  "<<tphase<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"CoarsenOperator phaseBZ "<<tphaseBZ<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"CoarsenOperator slice  "<<tslice <<" us"<<std::endl;
+    std::cout << GridLogMessage<<"CoarsenOperator mat    "<<tmat <<" us"<<std::endl;
+    std::cout << GridLogMessage<<"CoarsenOperator proj   "<<tproj<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"CoarsenOperator inv    "<<tinv<<" us"<<std::endl;
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Coarsen a NATIVELY multiRHS fine operator.
+  //
+  // linop acts on the D+1 dimensional fine grid FineGridMulti, with the batch
+  // of phased basis vectors carried in the rhs direction. A single RHS
+  // operator can be promoted with MrhsPromotedOperator, but that pays an
+  // ExtractSlice/InsertSlice pair per rhs; prefer the single RHS variant
+  // below in that case.
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenOperator(LinearOperatorBase<Lattice<Fobj> > &linop,
+		       GridCartesian *FineGridMulti,
+		       std::vector<FineField> &Subspace,
+		       GridBase *CoarseGrid)
+  {
+    RealD tproj=0.0, tmat=0.0, tphase=0.0, tphaseBZ=0.0, tslice=0.0, tinv=0.0;
+
+    std::cout << GridLogMessage<< "GeneralCoarsenMatrixMrhs (multiRHS fine operator)"<< std::endl;
+
+    GRID_ASSERT(Subspace.size()==nbasis);
+    GridBase *grid = Subspace[0].Grid();
+
+    GRID_ASSERT(FineGridMulti->_ndimension == grid->_ndimension+1);
+    GRID_ASSERT(FineGridMulti->_processors[0] == 1);
+    for(int d=0;d<grid->_ndimension;d++){
+      GRID_ASSERT(FineGridMulti->_fdimensions[d+1] == grid->_fdimensions[d]);
+      GRID_ASSERT(FineGridMulti->_processors [d+1] == grid->_processors [d]);
+    }
+    int batch = FineGridMulti->_fdimensions[0];
+
+    Coordinate blatt,bsimd,bmpi;
+    CoarsenBlockGridLayout(grid,CoarseGrid,blatt,bsimd,bmpi);
+    GridCartesian BlockGrid(blatt,bsimd,bmpi);
+
+    BlockComplexField InnerProd(&BlockGrid);
+    blockOrthogonalise(InnerProd,Subspace);
+
+    MultiRHSBlockProject<Lattice<Fobj> >    Projector;
+    Projector.Allocate(nbasis,grid,CoarseGrid);
+    Projector.ImportBasis(Subspace);
+
+    const int npoint = geom_srhs.npoint;
+
+    Eigen::MatrixXcd invMkl;
+    CoarsenFourierMatrix(CoarseGrid,invMkl);
+
+    FineField phaV(grid);
+    std::vector<FineComplexField>   phaF(npoint,grid);
+    std::vector<CoarseComplexField> pha (npoint,CoarseGrid);
+
+    tphase=-usecond();
+    CoarsenPhases(grid,CoarseGrid,&BlockGrid,pha,phaF);
+    tphase+=usecond();
+
+    std::vector<CoarseMatrix> _A;
+    _A.resize(npoint,CoarseGrid);
+    for(int k=0;k<npoint;k++) _A[k] = Zero();
+
+    Coordinate cmlatt,cmsimd,cmmpi;
+    CoarsenBatchGridLayout(CoarseGrid,batch,cmlatt,cmsimd,cmmpi);
+    GridCartesian CoarseBatchGrid(cmlatt,cmsimd,cmmpi);
+
+    CoarseVector       TmpProj(&CoarseBatchGrid);
+    CoarseComplexField phaB(&CoarseBatchGrid);
+
+    FineField hi_in (FineGridMulti);
+    FineField hi_out(FineGridMulti);
+    FineField zzF(grid); zzF = Zero();
+
+    for(int i0=0;i0<nbasis;i0+=batch){ // Loop over batches of basis vectors
+
+      int nbv = MIN(batch,nbasis-i0);
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i0<<"/"<<nbasis<< std::endl;
+
+      for(int p=0;p<npoint;p++){ // Loop over momenta
+
+	// One phase, applied to the whole batch. Tail slices are zeroed so
+	// the operator never sees undefined data.
+	for(int b=0;b<nbv;b++){
+	  tphaseBZ-=usecond();
+	  phaV = phaF[p]*Subspace[i0+b];
+	  tphaseBZ+=usecond();
+	  tslice-=usecond();
+	  InsertSliceFast(phaV,hi_in,b,0);
+	  tslice+=usecond();
+	}
+	tslice-=usecond();
+	for(int b=nbv;b<batch;b++){
+	  InsertSliceFast(zzF,hi_in,b,0);
+	}
+	tslice+=usecond();
+
+	tmat-=usecond();
+	linop.Op(hi_in,hi_out);
+	tmat+=usecond();
+
+	tproj-=usecond();
+	Projector.blockProject(hi_out,TmpProj);
+	tproj+=usecond();
+
+	tinv-=usecond();
+	CoarsenAccumulate(p,i0,nbv,batch,invMkl,pha,phaB,TmpProj,_A,CoarseGrid);
+	tinv+=usecond();
+      }
+    }
+
+    for(int p=0;p<npoint;p++){
+      GridtoBLAS(_A[p],BLAS_A[p]);
+    }
+    CoarsenReport(tphase,tphaseBZ,tslice,tmat,tproj,tinv);
+  }
+
+  ///////////////////////////////////////////////////////////////////////////
+  // Coarsen a SINGLE RHS fine operator.
+  //
+  // No multiRHS packing: the operator is applied once per phased basis
+  // vector and the batch is assembled on the coarse side by the mixed
+  // blockProject, which takes a vector of fine fields and writes the batch
+  // coarse field the accumulate expects. Only nbv applications per momentum,
+  // so a batch that does not divide nbasis wastes nothing, and the live fine
+  // storage is batch fields rather than two D+1 fields of extent batch.
+  ///////////////////////////////////////////////////////////////////////////
+  void CoarsenOperator(LinearOperatorBase<Lattice<Fobj> > &linop,
+		       std::vector<FineField> &Subspace,
+		       GridBase *CoarseGrid,
+		       int batch)
+  {
+    RealD tproj=0.0, tmat=0.0, tphase=0.0, tphaseBZ=0.0, tslice=0.0, tinv=0.0;
+
+    std::cout << GridLogMessage<< "GeneralCoarsenMatrixMrhs (single RHS fine operator)"<< std::endl;
+
+    GRID_ASSERT(Subspace.size()==nbasis);
+    GRID_ASSERT(batch>=1);
+    GridBase *grid = Subspace[0].Grid();
+
+    Coordinate blatt,bsimd,bmpi;
+    CoarsenBlockGridLayout(grid,CoarseGrid,blatt,bsimd,bmpi);
+    GridCartesian BlockGrid(blatt,bsimd,bmpi);
+
+    BlockComplexField InnerProd(&BlockGrid);
+    blockOrthogonalise(InnerProd,Subspace);
+
+    MultiRHSBlockProject<Lattice<Fobj> >    Projector;
+    Projector.Allocate(nbasis,grid,CoarseGrid);
+    Projector.ImportBasis(Subspace);
+
+    const int npoint = geom_srhs.npoint;
+
+    Eigen::MatrixXcd invMkl;
+    CoarsenFourierMatrix(CoarseGrid,invMkl);
+
+    FineField phaV(grid);
+    std::vector<FineComplexField>   phaF(npoint,grid);
+    std::vector<CoarseComplexField> pha (npoint,CoarseGrid);
+
+    tphase=-usecond();
+    CoarsenPhases(grid,CoarseGrid,&BlockGrid,pha,phaF);
+    tphase+=usecond();
+
+    std::vector<CoarseMatrix> _A;
+    _A.resize(npoint,CoarseGrid);
+    for(int k=0;k<npoint;k++) _A[k] = Zero();
+
+    Coordinate cmlatt,cmsimd,cmmpi;
+    CoarsenBatchGridLayout(CoarseGrid,batch,cmlatt,cmsimd,cmmpi);
+    GridCartesian CoarseBatchGrid(cmlatt,cmsimd,cmmpi);
+
+    CoarseVector       TmpProj(&CoarseBatchGrid);
+    CoarseComplexField phaB(&CoarseBatchGrid);
+
+    std::vector<FineField> MphaV(batch,grid);
+
+    for(int i0=0;i0<nbasis;i0+=batch){ // Loop over batches of basis vectors
+
+      int nbv = MIN(batch,nbasis-i0);
+      std::cout << GridLogMessage<< "CoarsenMatrixColoured vec "<<i0<<"/"<<nbasis<< std::endl;
+
+      for(int p=0;p<npoint;p++){ // Loop over momenta
+
+	for(int b=0;b<nbv;b++){
+	  tphaseBZ-=usecond();
+	  phaV = phaF[p]*Subspace[i0+b];
+	  tphaseBZ+=usecond();
+	  tmat-=usecond();
+	  linop.Op(phaV,MphaV[b]);
+	  tmat+=usecond();
+	}
+	// The accumulate reads only the first nbv slices, but the projector
+	// sees the whole vector, so the tail must not be undefined.
+	for(int b=nbv;b<batch;b++) MphaV[b] = Zero();
+
+	tproj-=usecond();
+	Projector.blockProject(MphaV,TmpProj);
+	tproj+=usecond();
+
+	tinv-=usecond();
+	CoarsenAccumulate(p,i0,nbv,batch,invMkl,pha,phaB,TmpProj,_A,CoarseGrid);
+	tinv+=usecond();
+      }
+    }
+
+    for(int p=0;p<npoint;p++){
+      GridtoBLAS(_A[p],BLAS_A[p]);
+    }
+    CoarsenReport(tphase,tphaseBZ,tslice,tmat,tproj,tinv);
+  }
+  void Mdag(const CoarseVector &in, CoarseVector &out)
+  {
+    this->M(in,out);
+  }
+  void M (const CoarseVector &in, CoarseVector &out)
+  {
+    //    std::cout << GridLogMessage << "New Mrhs coarse"<<std::endl;
+    conformable(CoarseGrid(),in.Grid());
+    conformable(in.Grid(),out.Grid());
+    out.Checkerboard() = in.Checkerboard();
+
+    RealD t_tot;
+    RealD t_exch;
+    RealD t_GtoB;
+    RealD t_BtoG;
+    RealD t_mult;
+
+    CheckGridSet();
+    if ( in.Grid() != _CoarseGridMulti ) {
+      std::cout << GridLogError
+		<< "MultiGeneralCoarsenedOperatorV2::M called with a field on a"
+		<< std::endl;
+      std::cout << GridLogError
+		<< "  different grid object from the one given to SetGrid(). Two"
+		<< std::endl;
+      std::cout << GridLogError
+		<< "  grids of identical shape do not conform; share one object."
+		<< std::endl;
+      GRID_ASSERT(in.Grid() == _CoarseGridMulti);
+    }
+
+    GRID_TRACE("CoarseV2Mult");
+    t_tot=-usecond();
+    CoarseVector tin=in;
+    t_exch=-usecond();
+    // lambda scope so the roctx range covers exactly the exchange; the
+    // PaddedCellFwd/BwdMPI markers inside it then nest properly.
+    CoarseVector pin = [&](){ GRID_TRACE("CoarseV2Exchange");
+                              return CellMulti->ExchangePeriodic(tin); }(); //padded input
+    t_exch+=usecond();
+
+    CoarseVector pout(pin.Grid());
+
+    int npoint = geom.npoint;
+    typedef calcMatrix* Aview;
+    typedef LatticeView<Cvec> Vview;
+      
+    const int Nsimd = CComplex::Nsimd();
+
+    int64_t nrhs  =pin.Grid()->GlobalDimensions()[0];
+    GRID_ASSERT(nrhs>=1);
+
+    RealD flops,bytes;
+    int64_t osites=in.Grid()->oSites(); // unpadded
+    int64_t unpadded_vol = CoarseGrid()->lSites()/nrhs;
+    
+    flops = 1.0* npoint * nbasis * nbasis * 8.0 * osites * CComplex::Nsimd();
+    bytes = 1.0*osites*sizeof(siteMatrix)*npoint/pin.Grid()->GlobalDimensions()[0]
+          + 2.0*osites*sizeof(siteVector)*npoint;
+    
+
+    t_GtoB=-usecond();
+    { GRID_TRACE("CoarseV2GridToBLAS");
+      GridtoBLAS(pin,BLAS_B);
+    }
+    t_GtoB+=usecond();
+
+    GridBLAS BLAS;
+
+    t_mult=-usecond();
+    { GRID_TRACE("CoarseV2StencilGEMM");
+    for(int p=0;p<geom.npoint;p++){
+      RealD c = 1.0;
+      if (p==0) c = 0.0;
+      ComplexD beta(c);
+
+      BLAS.gemmBatched(nbasis,nrhs,nbasis,
+		       ComplexD(1.0),
+		       BLAS_AP[p], 
+		       BLAS_BP[p], 
+		       ComplexD(c), 
+		       BLAS_CP);
+    }
+    BLAS.synchronise();
+    }
+    t_mult+=usecond();
+
+    t_BtoG=-usecond();
+    { GRID_TRACE("CoarseV2BLASToGrid");
+      BLAStoGrid(out,BLAS_C);
+    }
+    t_BtoG+=usecond();
+    t_tot+=usecond();
+    /*
+    std::cout << GridLogMessage << "New Mrhs coarse DONE "<<std::endl;
+    std::cout << GridLogMessage<<"Coarse Mult exch "<<t_exch<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"Coarse Mult mult "<<t_mult<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"Coarse Mult GtoB  "<<t_GtoB<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"Coarse Mult BtoG  "<<t_BtoG<<" us"<<std::endl;
+    std::cout << GridLogMessage<<"Coarse Mult tot  "<<t_tot<<" us"<<std::endl;
+    */
+    //    std::cout << GridLogMessage<<std::endl;
+    //    std::cout << GridLogMessage<<"Coarse Kernel flops "<< flops<<std::endl;
+    //    std::cout << GridLogMessage<<"Coarse Kernel flop/s "<< flops/t_mult<<" mflop/s"<<std::endl;
+    //    std::cout << GridLogMessage<<"Coarse Kernel bytes/s "<< bytes/t_mult/1000<<" GB/s"<<std::endl;
+    //    std::cout << GridLogMessage<<"Coarse overall flops/s "<< flops/t_tot<<" mflop/s"<<std::endl;
+    //    std::cout << GridLogMessage<<"Coarse total bytes   "<< bytes/1e6<<" MB"<<std::endl;
+  };
+  virtual  void Mdiag    (const Field &in, Field &out){ GRID_ASSERT(0);};
+  virtual  void Mdir     (const Field &in, Field &out,int dir, int disp){assert(0);};
+  virtual  void MdirAll  (const Field &in, std::vector<Field> &out){assert(0);};
+};
+  
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/Geometry.h

@@ -67,8 +67,8 @@ public:
   }
 
   int point(int dir, int disp) {
-    assert(disp == -1 || disp == 0 || disp == 1);
-    assert(base+0 <= dir && dir < base+4);
+    GRID_ASSERT(disp == -1 || disp == 0 || disp == 1);
+    GRID_ASSERT(base+0 <= dir && dir < base+4);
 
     // directions faster index = new indexing
     // 4d (base = 0):
@@ -131,7 +131,7 @@ public:
 	return p;
       }
     }
-    assert(0);
+    GRID_ASSERT(0);
     return -1;
   }
   void BuildShifts(void)

Grid/algorithms/multigrid/MrhsMultiGrid.h

@@ -0,0 +1,256 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/MrhsMultiGrid.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/deflation/MultiRHSBlockProject.h>
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////
+// mrhs LinearFunction interface: vector-of-fields in, vector out.
+// The outer level carries mrhs as std::vector<Field>; below it mrhs
+// is PACKED into a single D+1 field (rhs = dim 0) and the coarse
+// classes are plain LinearFunctions on that.
+//////////////////////////////////////////////////////////////////////
+template<class Field>
+class MrhsLinearFunction {
+public:
+  virtual void operator()(std::vector<Field> &in, std::vector<Field> &out) = 0;
+};
+
+//////////////////////////////////////////////////////////////////////
+// Single-polynomial mrhs PGCR: one step length and one set of
+// orthogonalisation coefficients SHARED across the right-hand sides
+// (vnorm2/vinnerProduct sum over rhs).  For DWF the dense spectrum
+// makes sharing near-free and the summed reductions amortise; see the
+// mrhs-HDCG paper (arXiv:2409.03904).  Orthogonalisation is classical
+// Gram-Schmidt, batched per rhs (rankInnerProductMulti/axpyMulti) with
+// ONE GlobalSumVector for all coefficients.
+//
+// OnStep is a programmatic per-step hook for algorithmic studies
+// (e.g. driving the GCR coefficient recorder); it is not a consumer
+// parameter.
+//////////////////////////////////////////////////////////////////////
+template<class Field>
+class MrhsPGCRNonHermitian {
+public:
+  RealD Tolerance; Integer MaxIterations; int mmax,nstep,steps,level;
+  int ZeroGuess = 0; int FirstCycle = 0;
+  std::string name = "Level 1";
+  LinearOperatorBase<Field> &Linop;
+  MrhsLinearFunction<Field> &Preconditioner;
+  std::function<void(int)> OnStep;      // called with the outer step count after every step
+  void Level(int lv){ name = "Level " + std::to_string(lv); level=lv; }
+  void Name(std::string n){ name = n; }
+  void SetZeroGuess(int z){ ZeroGuess=z; }
+  MrhsPGCRNonHermitian(RealD tol,Integer maxit,LinearOperatorBase<Field> &_Linop,MrhsLinearFunction<Field> &Prec,int _mmax,int _nstep)
+    : Tolerance(tol),MaxIterations(maxit),Linop(_Linop),Preconditioner(Prec),mmax(_mmax),nstep(_nstep){ level=1; }
+  static RealD vnorm2(std::vector<Field> &x){ RealD s=0; for(auto &f:x) s+=norm2(f); return s; }
+  static ComplexD vinnerProduct(std::vector<Field> &x,std::vector<Field> &y){ ComplexD s(0); for(int r=0;r<(int)x.size();r++) s+=innerProduct(x[r],y[r]); return s; }
+  static void vaxpy(std::vector<Field> &z,ComplexD a,std::vector<Field> &x,std::vector<Field> &y){ for(int r=0;r<(int)z.size();r++) axpy(z[r],a,x[r],y[r]); }
+  void vOp(std::vector<Field> &in,std::vector<Field> &out){ GRID_TRACE("MrhsPGCR::vOp"); for(int r=0;r<(int)in.size();r++) Linop.Op(in[r],out[r]); }
+  void operator()(std::vector<Field> &src,std::vector<Field> &psi){
+    RealD cp,ssq,rsq; int nrhs=src.size(); GridBase *grid=src[0].Grid();
+    ssq=vnorm2(src); rsq=Tolerance*Tolerance*ssq;
+    std::vector<Field> r(nrhs,grid);
+    GridStopWatch T; T.Start(); steps=0; FirstCycle=1;
+    for(int k=0;k<MaxIterations;k++){
+      cp=GCRnStep(src,psi,rsq);
+      std::cout<<GridLogMessage<<std::string(level,'\t')<<" "<<name<<" MrhsPGCR("<<mmax<<","<<nstep<<") "<<steps<<" steps cp = "<<cp<<" target "<<rsq<<std::endl;
+      if(cp<rsq){
+        T.Stop(); vOp(psi,r); for(int rr=0;rr<nrhs;rr++) axpy(r[rr],-1.0,src[rr],r[rr]);
+        RealD tr=vnorm2(r);
+        std::cout<<GridLogMessage<<std::string(level,'\t')<<" "<<name<<" MrhsPGCR: Converged on iteration "<<steps
+                 <<" computed residual "<<std::sqrt(cp/ssq)<<" true residual "<<std::sqrt(tr/ssq)<<" target "<<Tolerance<<std::endl;
+        std::cout<<GridLogMessage<<std::string(level,'\t')<<" "<<name<<" MrhsPGCR Time elapsed: Total "<<T.Elapsed()<<std::endl;
+        return;
+      }
+    }
+    std::cout<<GridLogMessage<<"MrhsPGCR: did not converge"<<std::endl;
+  }
+  RealD GCRnStep(std::vector<Field> &src,std::vector<Field> &psi,RealD rsq){
+    RealD cp; ComplexD a,rq; int nrhs=src.size(); GridBase *grid=src[0].Grid();
+    std::vector<Field> r(nrhs,grid),Az(nrhs,grid);   // Az: restart residual scratch only
+    std::vector< std::vector<Field> > q(mmax,std::vector<Field>(nrhs,grid));
+    std::vector< std::vector<Field> > p(mmax,std::vector<Field>(nrhs,grid));
+    std::vector<RealD> qq(mmax);
+    if (ZeroGuess && FirstCycle) { for(int rr=0;rr<nrhs;rr++){ psi[rr]=Zero(); r[rr]=src[rr]; } }
+    else                         { vOp(psi,Az); for(int rr=0;rr<nrhs;rr++) r[rr]=src[rr]-Az[rr]; }
+    FirstCycle=0;
+    // p[0]=Prec(r), q[0]=A p[0], produced directly in the history slots (no copies)
+    Preconditioner(r,p[0]); vOp(p[0],q[0]); qq[0]=vnorm2(q[0]); cp=vnorm2(r);
+    for(int k=0;k<nstep;k++){
+      steps++; int kp=k+1, peri_k=k%mmax, peri_kp=kp%mmax;
+      if ( OnStep ) OnStep(steps);
+      rq=vinnerProduct(q[peri_k],r); a=rq/qq[peri_k];
+      vaxpy(psi,a,p[peri_k],psi); vaxpy(r,-a,q[peri_k],r); cp=vnorm2(r);
+      std::cout<<GridLogMessage<<std::string(level,'\t')<<" "<<name<<" MrhsPGCR step["<<steps<<"]  resid "<<cp<<" target "<<rsq<<std::endl;
+      if((k==nstep-1)||(cp<rsq)) return cp;
+      // New direction straight into its history slot: p=Prec(r), q=A p.
+      Preconditioner(r,p[peri_kp]);
+      vOp(p[peri_kp],q[peri_kp]);
+      int northog=((kp)>(mmax-1))?(mmax-1):(kp);
+      {
+        GRID_TRACE("MrhsPGCR orthog");
+        // Classical Gram-Schmidt: all coefficients against the UN-updated new q
+        // (independent, batchable), then apply.  Complex coefficient: the
+        // operator is non-Hermitian, real(<q_j,Aq>) alone left q's non-orthogonal.
+        // Batched per rhs (one fused kernel + one reduction each), the shared
+        // coefficient summed over rhs on the host, ONE GlobalSumVector.
+        std::vector<ComplexD> bcoef(northog,ComplexD(0.0)), part;
+        for(int rr=0;rr<nrhs;rr++){
+          std::vector<const Field*> qwin(northog);
+          for(int back=0;back<northog;back++){ int peri_back=(k-back)%mmax; GRID_ASSERT((k-back)>=0); qwin[back]=&q[peri_back][rr]; }
+          rankInnerProductMulti(part,qwin,q[peri_kp][rr]);
+          for(int back=0;back<northog;back++) bcoef[back]+=part[back];
+        }
+        if(northog) grid->GlobalSumVector(&bcoef[0],northog);
+        for(int back=0;back<northog;back++){ int peri_back=(k-back)%mmax; bcoef[back]=-bcoef[back]/qq[peri_back]; }
+        for(int rr=0;rr<nrhs;rr++){
+          std::vector<const Field*> qwin(northog), pwin(northog);
+          for(int back=0;back<northog;back++){ int peri_back=(k-back)%mmax; qwin[back]=&q[peri_back][rr]; pwin[back]=&p[peri_back][rr]; }
+          axpyMulti(p[peri_kp][rr],bcoef,pwin);
+          axpyMulti(q[peri_kp][rr],bcoef,qwin);
+        }
+      }
+      qq[peri_kp]=vnorm2(q[peri_kp]);
+    }
+    GRID_ASSERT(0); return cp;
+  }
+};
+
+//////////////////////////////////////////////////////////////////////
+// Dense L3 solve on the packed D+1 coarse-coarse field.  Adapts any
+// class exposing ApplyBatch6D (DenseCoarseMatrix) to a LinearFunction.
+//////////////////////////////////////////////////////////////////////
+template<class DenseType, class CoarseCoarseField>
+class MrhsDenseCCSolve : public LinearFunction<CoarseCoarseField> {
+public:
+  DenseType &_Dense;
+  int _nrhs;
+  MrhsDenseCCSolve(DenseType &D, int nrhs) : _Dense(D), _nrhs(nrhs) {}
+  using LinearFunction<CoarseCoarseField>::operator();
+  virtual void operator()(const CoarseCoarseField &in, CoarseCoarseField &out){
+    _Dense.ApplyBatch6D(in, out, _nrhs);
+  }
+};
+
+//////////////////////////////////////////////////////////////////////
+// L2->L3 half V-cycle on the D+1 coarse field: coarse-coarse correction
+// through the mixed blockProject, then post-smooth.
+//////////////////////////////////////////////////////////////////////
+template<class CoarseField, class CoarseCoarseField>
+class MrhsCoarseThreeLevelPrec : public LinearFunction<CoarseField> {
+public:
+  LinearOperatorBase<CoarseField>          &_CoarseOp;
+  LinearFunction<CoarseField>              &_CoarseSmoother;
+  MultiRHSBlockProject<CoarseField>        &_Projector;
+  LinearFunction<CoarseCoarseField>        &_CoarseCoarseSolve;
+  GridBase *_Coarse5d, *_CoarseCoarse5d, *_CoarseCoarseMrhs;
+  int _nrhs;
+  MrhsCoarseThreeLevelPrec(LinearOperatorBase<CoarseField> &CoarseOp,
+                           LinearFunction<CoarseField> &CoarseSmoother,
+                           MultiRHSBlockProject<CoarseField> &Projector,
+                           LinearFunction<CoarseCoarseField> &CoarseCoarseSolve,
+                           GridBase *Coarse5d, GridBase *CoarseCoarse5d, GridBase *CoarseCoarseMrhs, int nrhs)
+    : _CoarseOp(CoarseOp), _CoarseSmoother(CoarseSmoother), _Projector(Projector),
+      _CoarseCoarseSolve(CoarseCoarseSolve),
+      _Coarse5d(Coarse5d), _CoarseCoarse5d(CoarseCoarse5d), _CoarseCoarseMrhs(CoarseCoarseMrhs), _nrhs(nrhs) {}
+  using LinearFunction<CoarseField>::operator();
+  virtual void operator()(const CoarseField &in, CoarseField &out) {
+    CoarseField vec1(in.Grid());
+    CoarseField vec2(in.Grid());
+    out = in;
+    _CoarseOp.Op(out,vec1);  sub(vec1,in,vec1);
+
+    // restrict, through the mixed blockProject: D+1 coarse in, D+1 cc out
+    CoarseCoarseField CCsrc(_CoarseCoarseMrhs);
+    CoarseCoarseField CCsol(_CoarseCoarseMrhs);
+    _Projector.blockProject(vec1,CCsrc);
+
+    _CoarseCoarseSolve(CCsrc,CCsol);
+
+    _Projector.blockPromote(vec1,CCsol);
+    add(out,out,vec1);
+
+    _CoarseOp.Op(out,vec1);  sub(vec1,in,vec1);
+    _CoarseSmoother(vec1,vec2);
+    add(out,out,vec2);
+  }
+};
+
+//////////////////////////////////////////////////////////////////////
+// L1->L2 mrhs V-cycle.  The whole V-cycle is preconditioner: its fine
+// residuals and the smoother may run with sloppy halos; the caller
+// (the outer Krylov) gets the exact operator back on exit.  SetSloppy
+// is wired by the composer to PVdagMLinearOperator::SloppyComms (a
+// no-op by default), replacing the file-scope global the example used.
+//////////////////////////////////////////////////////////////////////
+template<class FineField, class MrhsCoarseVector, class FineSmoother>
+class MrhsTwoLevelMG : public MrhsLinearFunction<FineField> {
+public:
+  typedef MrhsCoarseVector CoarseVector;
+  LinearOperatorBase<FineField>   &_FineOperator;
+  FineSmoother                    &_PostSmoother;
+  MultiRHSBlockProject<FineField> &_Projector;
+  LinearFunction<CoarseVector>    &_CoarseSolve;
+  GridBase *_CoarseGrid, *_CoarseGridMrhs;
+  std::function<void(int)> SetSloppy = [](int){};
+  int SloppyComms = 0;                 // value passed to SetSloppy on entry
+  MrhsTwoLevelMG(LinearOperatorBase<FineField> &FineOp, FineSmoother &Post,
+                 MultiRHSBlockProject<FineField> &Projector, LinearFunction<CoarseVector> &CoarseSolve,
+                 GridBase *CoarseGrid, GridBase *CoarseGridMrhs)
+    : _FineOperator(FineOp),_PostSmoother(Post),_Projector(Projector),_CoarseSolve(CoarseSolve),
+      _CoarseGrid(CoarseGrid),_CoarseGridMrhs(CoarseGridMrhs){}
+  virtual void operator()(std::vector<FineField> &in, std::vector<FineField> &out){
+    GRID_TRACE("MGVcycle");
+    SetSloppy(SloppyComms);
+    int nrhs=in.size(); GridBase *fgrid=in[0].Grid();
+    std::vector<FineField> vec1(nrhs,fgrid),vec2(nrhs,fgrid);
+    for(int r=0;r<nrhs;r++) out[r]=in[r];
+    { GRID_TRACE("MGFineResidual");
+      for(int r=0;r<nrhs;r++){ _FineOperator.Op(out[r],vec1[r]); sub(vec1[r],in[r],vec1[r]); }
+    }
+    // fine vector -> D+1 coarse, via the mixed blockProject
+    CoarseVector CsrcMrhs(_CoarseGridMrhs), CsolMrhs(_CoarseGridMrhs);
+    { GRID_TRACE("MGProject");
+      _Projector.blockProject(vec1,CsrcMrhs);
+    }
+    CsolMrhs=Zero();
+    { GRID_TRACE("MGCoarseSolve");
+      _CoarseSolve(CsrcMrhs,CsolMrhs);
+    }
+    { GRID_TRACE("MGPromote");
+      _Projector.blockPromote(vec1,CsolMrhs);
+      for(int r=0;r<nrhs;r++) add(out[r],out[r],vec1[r]);
+    }
+    { GRID_TRACE("MGFineResidual2");
+      for(int r=0;r<nrhs;r++){ _FineOperator.Op(out[r],vec1[r]); sub(vec1[r],in[r],vec1[r]); }
+    }
+    { GRID_TRACE("MGPostSmooth");
+      for(int r=0;r<nrhs;r++){
+        _PostSmoother(vec1[r],vec2[r]); add(out[r],out[r],vec2[r]);
+      }
+    }
+    SetSloppy(0);
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/MrhsPromotedOperator.h

@@ -0,0 +1,161 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/algorithms/multigrid/MrhsPromotedOperator.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////
+// Present a D dimensional operator as a D+1 dimensional operator with Nrhs in
+// dimension 0.  Field type is unchanged; only the Grid differs, so this is a
+// LinearOperatorBase and callers need no template parameter: a native mrhs
+// operator derives from the same base and substitutes without a call site
+// change.
+//
+// Slices in and out around the wrapped operator.  No arithmetic beyond the
+// wrapped call, but one ExtractSliceFast/InsertSliceFast pair per right hand
+// side: data motion, not work.
+//
+// AdjOp is carried so that A^dag may be coarsened as a separate coarse
+// operator when needed, rather than doubling coarse storage.
+///////////////////////////////////////////////////////////////////////////////
+template<class Field>
+class MrhsPromotedOperator : public LinearOperatorBase<Field>
+{
+private:
+
+  LinearOperatorBase<Field> &_LinOp;
+  GridBase                  *_LowGrid;
+  int                        _Nrhs;
+
+public:
+
+  MrhsPromotedOperator(LinearOperatorBase<Field> &LinOp,GridBase *LowGrid,int Nrhs)
+    : _LinOp(LinOp), _LowGrid(LowGrid), _Nrhs(Nrhs)
+  {
+    GRID_ASSERT(_Nrhs>=1);
+  }
+
+  GridBase *LowGrid(void) { return _LowGrid; }
+  int       Nrhs(void)    { return _Nrhs;    }
+
+  // Reset on each call; retrieve and accumulate in the caller
+  RealD tslice;
+  RealD top;
+
+  void OpDiag (const Field &in, Field &out)
+  {
+    SliceLoop(in,out,[&](Field &i,Field &o){ _LinOp.OpDiag(i,o); });
+  }
+
+  void Op     (const Field &in, Field &out)
+  {
+    SliceLoop(in,out,[&](Field &i,Field &o){ _LinOp.Op(i,o); });
+  }
+
+  void AdjOp  (const Field &in, Field &out)
+  {
+    SliceLoop(in,out,[&](Field &i,Field &o){ _LinOp.AdjOp(i,o); });
+  }
+
+  void HermOp (const Field &in, Field &out)
+  {
+    SliceLoop(in,out,[&](Field &i,Field &o){ _LinOp.HermOp(i,o); });
+  }
+
+  void OpDir  (const Field &in, Field &out,int dir,int disp)
+  {
+    SliceLoop(in,out,[&](Field &i,Field &o){ _LinOp.OpDir(i,o,dir,disp); });
+  }
+
+  //////////////////////////////////////////////////////////////////
+  // Norms of the D+1 field are the sums over slices
+  //////////////////////////////////////////////////////////////////
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2)
+  {
+    Conformable(in,out);
+    Field lo_in (_LowGrid);
+    Field lo_out(_LowGrid);
+    n1=0.0;
+    n2=0.0;
+    for(int r=0;r<_Nrhs;r++){
+      RealD r1,r2;
+      ExtractSliceFast(lo_in,in,r,0);
+      _LinOp.HermOpAndNorm(lo_in,lo_out,r1,r2);
+      InsertSliceFast(lo_out,out,r,0);
+      n1=n1+r1;
+      n2=n2+r2;
+    }
+  }
+
+  void OpDirAll(const Field &in, std::vector<Field> &out)
+  {
+    int npoint = out.size();
+    Field lo_in(_LowGrid);
+    std::vector<Field> lo_out(npoint,_LowGrid);
+    for(int r=0;r<_Nrhs;r++){
+      ExtractSliceFast(lo_in,in,r,0);
+      _LinOp.OpDirAll(lo_in,lo_out);
+      for(int p=0;p<npoint;p++){
+        InsertSliceFast(lo_out[p],out[p],r,0);
+      }
+    }
+  }
+
+private:
+
+  void Conformable(const Field &in,const Field &out)
+  {
+    conformable(in.Grid(),out.Grid());
+    GRID_ASSERT(in.Grid()->_ndimension == _LowGrid->_ndimension+1);
+    GRID_ASSERT(in.Grid()->_fdimensions[0] == _Nrhs);
+  }
+
+  template<class Kernel>
+  void SliceLoop(const Field &in,Field &out,Kernel K)
+  {
+    Conformable(in,out);
+    Field lo_in (_LowGrid);
+    Field lo_out(_LowGrid);
+    tslice=0.0;
+    top=0.0;
+    for(int r=0;r<_Nrhs;r++){
+      tslice-=usecond();
+      ExtractSliceFast(lo_in,in,r,0);
+      tslice+=usecond();
+      top-=usecond();
+      K(lo_in,lo_out);
+      top+=usecond();
+      tslice-=usecond();
+      InsertSliceFast(lo_out,out,r,0);
+      tslice+=usecond();
+    }
+  }
+
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/MultiGrid.h

@@ -32,3 +32,11 @@ Author: Peter Boyle <pboyle@bnl.gov>
 #include <Grid/algorithms/multigrid/CoarsenedMatrix.h>
 #include <Grid/algorithms/multigrid/GeneralCoarsenedMatrix.h>
 #include <Grid/algorithms/multigrid/GeneralCoarsenedMatrixMultiRHS.h>
+#include <Grid/algorithms/multigrid/GeneralCoarsenedMatrixMultiRHSV2.h>
+#include <Grid/algorithms/multigrid/MrhsPromotedOperator.h>
+#include <Grid/algorithms/multigrid/Smoothers.h>
+#include <Grid/algorithms/multigrid/PVdagMMultiGridParams.h>
+// PVdagMOperators.h / MrhsMultiGrid.h / PVdagMMultiGrid.h /
+// DenseCoarseMatrix.h are NOT in this umbrella: consumers of the PVdagM
+// chain include PVdagMMultiGrid.h explicitly (it pulls the dense stack
+// and BLAS).

Grid/algorithms/multigrid/PVdagMMultiGrid.h

@@ -0,0 +1,536 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/PVdagMMultiGrid.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/algorithms/multigrid/DenseCoarseMatrix.h>
+#include <Grid/algorithms/multigrid/PVdagMOperators.h>
+#include <Grid/algorithms/multigrid/MrhsMultiGrid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////////////////////
+// The three-level mrhs PVdagM multigrid, as objects:
+//
+//   MGCoarseGrids              the derived coarse grids (owned here, since
+//                              conformability is pointer identity and someone
+//                              must hold them; declare it BEFORE anything that
+//                              borrows from it)
+//   PVdagMMultiGridCoarsening  ALL the coarsening information for one gauge
+//                              configuration: the raw near-null basis, the
+//                              transfer operators it induces, the Galerkin
+//                              coarse operator at each level, and the dense
+//                              bottom inverse.  Everything here is a function
+//                              of the gauge field -- the state HMC must
+//                              rebuild or maintain as U evolves -- and
+//                              everything is solver-independent.
+//   PVdagMMultiGridSolver      the solve chain composed on a borrowed
+//                              coarsening: smoothers, coarse Krylov, V-cycle,
+//                              outer mrhs PGCR.
+//
+// Scope discipline: grids outlive the coarsening outlives the solver.
+//////////////////////////////////////////////////////////////////////////////////////
+
+//////////////////////////////////////////////////////////////////////
+// Subspace I/O: bare-vector scidac records.  Loaded vectors are RAW --
+// deliberately NOT re-orthogonalised: CoarsenOperator block
+// orthonormalises in place, and projecting a block-orthonormal vector
+// onto its own block-orthonormalised aggregation gives e_k with the
+// near-null content silently gone.  GramGuard below catches that.
+//////////////////////////////////////////////////////////////////////
+template <class Field>
+void saveSubspace(std::vector<Field> &subspace, std::string const fname){
+#ifdef HAVE_LIME
+  Grid::emptyUserRecord record;
+  Grid::ScidacWriter SW(subspace[0].Grid()->IsBoss());
+  SW.open(fname);
+  for (int k = 0; k < (int)subspace.size(); k++) {
+    SW.writeScidacFieldRecord(subspace[k], record);
+  }
+  SW.close();
+#endif
+}
+template <class Field>
+void loadSubspace(std::vector<Field> &subspace, std::string const fname){
+#ifdef HAVE_LIME
+  Grid::emptyUserRecord record;
+  Grid::ScidacReader SR;
+  SR.open(fname);
+  for (int k = 0; k < (int)subspace.size(); k++) {
+    SR.readScidacFieldRecord(subspace[k], record);
+  }
+  SR.close();
+#endif
+}
+
+//////////////////////////////////////////////////////////////////////
+// ||<v|v> - I||_F over a set of coarse vectors.  Small means the raw
+// near-null content survived the projection.
+//////////////////////////////////////////////////////////////////////
+template<class CoarseField>
+RealD GramDefect(std::vector<CoarseField> &v)
+{
+  RealD s2=0.0;
+  for(int i=0;i<(int)v.size();i++){
+    for(int j=0;j<(int)v.size();j++){
+      ComplexD sij=TensorRemove(innerProduct(v[i],v[j]));
+      ComplexD d=sij-(i==j?ComplexD(1.0):ComplexD(0.0));
+      s2+=real(d)*real(d)+imag(d)*imag(d);
+    }
+  }
+  return std::sqrt(s2);
+}
+
+// On a leak every image collapses to the block unit e_k, the Gram becomes
+// N*I, and the defect lands at (N-1)*sqrt(nbasis) -- orders above the ~0.2
+// of a content-preserving projection.  Trip well below that so a mis-set
+// threshold costs a log line rather than the run.
+template<class CoarseField>
+void GramGuard(const std::string &name,std::vector<CoarseField> &v,GridBase *grid)
+{
+  RealD defect = GramDefect(v);
+  RealD N      = (RealD)grid->gSites();
+  RealD leak   = (N-1.0)*std::sqrt((RealD)v.size());
+  RealD trip   = std::sqrt(N);
+  std::cout << GridLogMessage << "GUARD: ||<"<<name<<"|"<<name<<"> - I||_F = " << defect
+            << "   (e_k leak would be " << leak << ", trip at " << trip << ")" << std::endl;
+  GRID_ASSERT( defect < trip );
+}
+
+//////////////////////////////////////////////////////////////////////
+// The derived grids of the three-level chain, owned in one place.
+// The coarse space is UNVECTORISED (sComplex scalar, simd {1,..,1});
+// the 5D/6D grids are built directly so the SIMD layout is ours.
+// rhs/batch is dim 0 of the 6D grids, undistributed -- no divisibility
+// constraint on nrhs.
+//////////////////////////////////////////////////////////////////////
+class MGCoarseGrids {
+public:
+  GridCartesian *FGrid;                     // borrowed
+  int Ls;
+  int batch;
+  Coordinate clatt;                         // 4d coarse lattice
+  Coordinate cclatt;                        // 4d coarse-coarse lattice
+  Coordinate c5simd, c5mpi;                 // 5D coarse simd/mpi
+  Coordinate cmsimd, cmmpi;                 // 6D coarse simd/mpi
+  // owned:
+  GridCartesian *Coarse5d;
+  GridCartesian *CoarseBatch;               // 6D at the coarsening batch
+  GridCartesian *CoarseCoarse5d;
+  GridCartesian *CoarseCoarseBatch;
+
+  MGCoarseGrids(GridCartesian *_FGrid, const MGSetupParams &P)
+    : FGrid(_FGrid)
+  {
+    Coordinate fdims = FGrid->FullDimensions();      // {Ls, x,y,z,t}
+    Coordinate fmpi  = FGrid->_processors;
+    GRID_ASSERT( fdims.size() == 5 );
+    Ls    = fdims[0];
+    batch = P.CoarsenBatch;
+
+    clatt.resize(4); cclatt.resize(4);
+    for(int d=0;d<4;d++){
+      GRID_ASSERT( fdims[d+1] % P.Block[d] == 0 );
+      clatt[d] = fdims[d+1] / P.Block[d];
+    }
+    for(int d=0;d<4;d++){
+      GRID_ASSERT( clatt[d] % P.Block2[d] == 0 );
+      cclatt[d] = clatt[d] / P.Block2[d];
+    }
+    std::cout << GridLogMessage << "MGCoarseGrids: Block  " << P.Block  << "  coarse lattice        " << clatt  << std::endl;
+    std::cout << GridLogMessage << "MGCoarseGrids: Block2 " << P.Block2 << "  coarse-coarse lattice " << cclatt << std::endl;
+
+    Coordinate c5latt({1,clatt[0],clatt[1],clatt[2],clatt[3]});
+    c5simd = Coordinate({1,1,1,1,1});
+    c5mpi  = Coordinate({1,fmpi[1],fmpi[2],fmpi[3],fmpi[4]});
+    Coarse5d = new GridCartesian(c5latt,c5simd,c5mpi);
+
+    cmsimd = Coordinate({1,1,1,1,1,1});
+    cmmpi  = Coordinate({1,1,fmpi[1],fmpi[2],fmpi[3],fmpi[4]});
+    Coordinate cblatt({batch,1,clatt[0],clatt[1],clatt[2],clatt[3]});
+    CoarseBatch = new GridCartesian(cblatt,cmsimd,cmmpi);
+
+    Coordinate cc5latt({1,cclatt[0],cclatt[1],cclatt[2],cclatt[3]});
+    CoarseCoarse5d = new GridCartesian(cc5latt,c5simd,c5mpi);
+
+    Coordinate ccblatt({batch,1,cclatt[0],cclatt[1],cclatt[2],cclatt[3]});
+    CoarseCoarseBatch = new GridCartesian(ccblatt,cmsimd,cmmpi);
+  }
+  ~MGCoarseGrids()
+  {
+    delete CoarseCoarseBatch;
+    delete CoarseCoarse5d;
+    delete CoarseBatch;
+    delete Coarse5d;
+  }
+};
+
+//////////////////////////////////////////////////////////////////////
+// ALL the coarsening information for one gauge configuration.
+//
+// Owns: the RAW near-null bases (fine and coarse -- retained so the
+// coarse operators can be REBUILT on a changed gauge field with a
+// fixed basis, the cheap HMC maintenance step; DiscardBasis() frees
+// them for valence use), the two coarsened operators, the two block
+// projectors, the dense bottom inverse (raw owning pointer: null in
+// the constructor, allocated by BuildDenseBottom, deleted here), and
+// the transient per-Nrhs solve grids (created by SetNrhs, which owns
+// the ReleaseGrid -> delete -> new -> SetGrid ordering in ONE place).
+//
+// Borrows: the grid bundle.  Declare MGCoarseGrids first.
+//////////////////////////////////////////////////////////////////////
+template<class Fobj,class CComplex,int nbasis>
+class PVdagMMultiGridCoarsening {
+public:
+  typedef Lattice<Fobj>                                                 FineField;
+  typedef MultiGeneralCoarsenedOperatorV2<Fobj,CComplex,nbasis>         CoarseOperator;
+  typedef typename CoarseOperator::CoarseVector                         CoarseVector;
+  typedef typename CoarseVector::vector_object                          CoarseSiteObj;
+  typedef iScalar<CComplex>                                             CComplex2;  // coarsening deepens the nest by one iScalar
+  typedef MultiGeneralCoarsenedOperatorV2<CoarseSiteObj,CComplex2,nbasis> CoarseCoarseOperator;
+  typedef typename CoarseCoarseOperator::CoarseVector                   CoarseCoarseVector;
+  typedef DenseCoarseMatrix<CComplex2,nbasis>                           DenseBottom;
+
+  MGCoarseGrids                     &Grids;      // borrowed
+  MGSetupParams                      Params;
+  NextToNearestStencilGeometry5D     geom;
+  NextToNearestStencilGeometry5D     geom2;
+  CoarseOperator                     CoarseOpPV;
+  CoarseCoarseOperator               CoarseOpL2;
+  MultiRHSBlockProject<FineField>    MrhsProjector;
+  MultiRHSBlockProject<CoarseVector> MrhsProjectorL2;
+  DenseBottom                       *DenseCC;
+  std::vector<FineField>             rawNull;    // RAW fine near-null basis
+  std::vector<CoarseVector>          rawPsi;     // RAW coarse near-null basis
+  GridCartesian                     *CMrhs;      // transient solve grids, owned
+  GridCartesian                     *CCMrhs;
+  int                                nrhs;
+
+  PVdagMMultiGridCoarsening(MGCoarseGrids &_Grids, const MGSetupParams &P)
+    : Grids(_Grids),
+      Params(P),
+      geom (_Grids.Coarse5d),
+      geom2(_Grids.CoarseCoarse5d),
+      CoarseOpPV(geom ,_Grids.Coarse5d),
+      CoarseOpL2(geom2,_Grids.CoarseCoarse5d),
+      DenseCC(nullptr),
+      CMrhs(nullptr),
+      CCMrhs(nullptr),
+      nrhs(-1)
+  {};
+
+  ~PVdagMMultiGridCoarsening()
+  {
+    if ( DenseCC ) delete DenseCC;
+    CoarseOpPV.ReleaseGrid();     // borrowers let go before their grids die
+    CoarseOpL2.ReleaseGrid();
+    if ( CMrhs  ) delete CMrhs;
+    if ( CCMrhs ) delete CCMrhs;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // The RAW fine basis: load from Params.SubspaceFile if it exists,
+  // else GCR inverse iteration from noise (and save if a file name
+  // was given).  The Aggregation is scaffolding for CreateSubspaceGCR
+  // only -- that runs entirely on the fine grid, so the coarse grid
+  // it holds is never dereferenced.
+  ////////////////////////////////////////////////////////////////////
+  void GetSubspace(GridParallelRNG &RNG, LinearOperatorBase<FineField> &FineOp)
+  {
+    uint64_t file_exists=0;
+    if ( Params.SubspaceFile.length() ) {
+      if ( Grids.FGrid->IsBoss() ){ std::ifstream f(Params.SubspaceFile); file_exists=f.good()?1:0; }
+      Grids.FGrid->GlobalSum(file_exists);
+    }
+    rawNull.clear();
+    rawNull.reserve(nbasis);
+    for(int k=0;k<nbasis;k++) rawNull.push_back(FineField(Grids.FGrid));
+    if ( file_exists ){
+      std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: loading subspace "
+                << Params.SubspaceFile << " (kept RAW)" << std::endl;
+      loadSubspace(rawNull, Params.SubspaceFile);
+    } else {
+      std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: GCR subspace generation" << std::endl;
+      Aggregation<Fobj,CComplex,nbasis> Agg(Grids.Coarse5d,Grids.FGrid,0);
+      Agg.CreateSubspaceGCR(RNG,FineOp,nbasis);
+      for(int k=0;k<nbasis;k++)
+	rawNull[k]=Agg.subspace[k];
+      if ( Params.SubspaceFile.length() )
+	saveSubspace(rawNull, Params.SubspaceFile);
+    }
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Both coarsenings from the retained RAW basis.  Calling this again
+  // on a CHANGED fine operator is the fixed-basis rebuild: for the
+  // PVdagM coarsening the coarse operator with a fixed basis is exact
+  // in the changed links.  CoarsenOperator block-orthonormalises its
+  // input in place, so working copies are taken and the RAW vectors
+  // survive for the next rebuild.
+  //
+  // The fine operator must be the PVdagM wrapper (it needs
+  // SloppyComms: the coarsening builds the PRECONDITIONER, so its
+  // halos follow the FineSloppyComms policy, restored EXACT on exit).
+  ////////////////////////////////////////////////////////////////////
+  template<class PVdagMOp>
+  void Coarsen(PVdagMOp &FineOp)
+  {
+    GRID_ASSERT( rawNull.size() == nbasis );
+
+    // L1: working copy of the raw fine basis, orthonormalised in place
+    std::vector<FineField> sub(rawNull);
+
+    CoarseOpPV.SetGrid(Grids.CoarseBatch);
+
+    std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: L1 CoarsenOperator, batch "
+              << Grids.batch << std::endl;
+
+    FineOp.SloppyComms(Params.FineSloppyComms);
+    CoarseOpPV.CoarsenOperator(FineOp,sub,Grids.Coarse5d,Grids.batch);
+    FineOp.SloppyComms(0);
+
+    // Transfer operators from the orthonormalised basis; then the
+    // Galerkin images of the RAW basis define the L2 null space.
+    MrhsProjector.Allocate(nbasis,Grids.FGrid,Grids.Coarse5d);
+    MrhsProjector.ImportBasis(sub);
+    sub.clear(); sub.shrink_to_fit();
+
+    std::vector<CoarseVector> psi(nbasis,Grids.Coarse5d);
+    MrhsProjector.blockProject(rawNull,psi);
+    GramGuard("psi_coarse",psi,Grids.Coarse5d);
+
+    rawPsi.clear();
+    rawPsi.reserve(nbasis);
+    for(int k=0;k<nbasis;k++) rawPsi.push_back(psi[k]);
+
+    // L2: the V2 L1 operator is natively multiRHS at the batch grid
+    CoarseOpL2.SetGrid(Grids.CoarseCoarseBatch);
+    NonHermitianLinearOperator<CoarseOperator,CoarseVector> LinOpCoarse(CoarseOpPV);
+    std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: L2 CoarsenOperator, batch "
+              << Grids.batch << std::endl;
+    CoarseOpL2.CoarsenOperator(LinOpCoarse,Grids.CoarseBatch,psi,Grids.CoarseCoarse5d);
+
+    MrhsProjectorL2.Allocate(nbasis,Grids.Coarse5d,Grids.CoarseCoarse5d);
+    MrhsProjectorL2.ImportBasis(psi);              // now block orthonormal
+    {
+      std::vector<CoarseCoarseVector> psi_cc(nbasis,Grids.CoarseCoarse5d);
+      MrhsProjectorL2.blockProject(rawPsi,psi_cc); // RAW vectors in
+      GramGuard("psi_cc",psi_cc,Grids.CoarseCoarse5d);
+    }
+    nrhs = -1;                                     // both operators left at the batch grids
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Dense bottom: import the L2 matrix elements, invert (2D
+  // block-cyclic, fp64), then the check Import cannot run itself --
+  // the mrhs operator applies only on a D+1 grid, so drive it at
+  // Nrhs 1 through a slice: ||A Ainv x - x||/||x||.
+  ////////////////////////////////////////////////////////////////////
+  void BuildDenseBottom(void)
+  {
+    if ( DenseCC ) delete DenseCC;
+    std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: L3 dense bottom import" << std::endl;
+    DenseCC = new DenseBottom(Grids.CoarseCoarse5d);
+    DenseCC->Import(CoarseOpL2);
+
+    Coordinate cc1latt({1,1,Grids.cclatt[0],Grids.cclatt[1],Grids.cclatt[2],Grids.cclatt[3]});
+    GridCartesian *CoarseCoarseOne = new GridCartesian(cc1latt,Grids.cmsimd,Grids.cmmpi);
+    CoarseOpL2.SetGrid(CoarseCoarseOne);
+
+    CoarseCoarseVector x(Grids.CoarseCoarse5d),y(Grids.CoarseCoarse5d),z(Grids.CoarseCoarse5d);
+    GridParallelRNG dRNG(Grids.CoarseCoarse5d); dRNG.SeedFixedIntegers(std::vector<int>({11,12,13,14}));
+    random(dRNG,x);
+
+    (*DenseCC)(x,y);                    // y = Ainv x
+
+    CoarseCoarseVector y1(CoarseCoarseOne),z1(CoarseCoarseOne);
+    InsertSliceFast(y,y1,0,0);
+    CoarseOpL2.M(y1,z1);                // z = A y
+    ExtractSliceFast(z,z1,0,0);
+
+    z = z - x;
+    RealD rel = std::sqrt(norm2(z)/norm2(x));
+    std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: L3 dense ||A Ainv x - x||/||x|| = "
+              << rel << std::endl;
+    GRID_ASSERT( rel < 1.0e-2 );
+
+    CoarseOpL2.ReleaseGrid();           // let go before the grid dies
+    delete CoarseCoarseOne;
+    nrhs = -1;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Retarget both operators to a solve Nrhs.  The matrix elements are
+  // Nrhs independent and survive; only the D+1 grids and BLAS buffers
+  // change.  The ordering discipline (borrowers release BEFORE their
+  // grids are destroyed) lives here and nowhere else.
+  ////////////////////////////////////////////////////////////////////
+  void SetNrhs(int nr)
+  {
+    if ( nr == nrhs ) return;
+    CoarseOpPV.ReleaseGrid();
+    CoarseOpL2.ReleaseGrid();
+    if ( CMrhs  ) { delete CMrhs;  CMrhs =nullptr; }
+    if ( CCMrhs ) { delete CCMrhs; CCMrhs=nullptr; }
+    Coordinate cml ({nr,1,Grids.clatt[0], Grids.clatt[1], Grids.clatt[2], Grids.clatt[3]});
+    Coordinate ccml({nr,1,Grids.cclatt[0],Grids.cclatt[1],Grids.cclatt[2],Grids.cclatt[3]});
+    CMrhs  = new GridCartesian(cml, Grids.cmsimd,Grids.cmmpi);
+    CCMrhs = new GridCartesian(ccml,Grids.cmsimd,Grids.cmmpi);
+    CoarseOpPV.SetGrid(CMrhs);
+    CoarseOpL2.SetGrid(CCMrhs);
+    nrhs = nr;
+    std::cout << GridLogMessage << "PVdagMMultiGridCoarsening: operators at Nrhs " << nr << std::endl;
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // Free the retained RAW bases (valence use: coarsening is final for
+  // this configuration and the fine basis is ~GB-scale host memory).
+  ////////////////////////////////////////////////////////////////////
+  void DiscardBasis(void)
+  {
+    rawNull.clear(); rawNull.shrink_to_fit();
+    rawPsi.clear();  rawPsi.shrink_to_fit();
+  }
+};
+
+//////////////////////////////////////////////////////////////////////
+// The solve chain on a borrowed coarsening: adaptive PGCR smoothers
+// at both levels (the one correct route for this non-Hermitian
+// chain), dense bottom, mrhs V-cycle, outer shared-coefficient mrhs
+// PGCR.  Construction retargets the coarsening to this Nrhs.
+//
+// Halo policy: the whole V-cycle is preconditioner and runs with
+// sloppy halos when FineSloppyComms is set; the outer Krylov is EXACT
+// (its applications define what "converged" means), asserted by the
+// FINAL true-residual report in Solve.
+//////////////////////////////////////////////////////////////////////
+template<class Matrix,class Coarsening>
+class PVdagMMultiGridSolver {
+public:
+  typedef typename Coarsening::FineField            FineField;
+  typedef typename Coarsening::CoarseOperator       CoarseOperator;
+  typedef typename Coarsening::CoarseVector         CoarseVector;
+  typedef typename Coarsening::CoarseCoarseOperator CoarseCoarseOperator;
+  typedef typename Coarsening::CoarseCoarseVector   CoarseCoarseVector;
+  typedef typename Coarsening::DenseBottom          DenseBottom;
+  typedef PrecGeneralisedConjugateResidualNonHermitian<FineField>    FineSmoother_t;
+  typedef PrecGeneralisedConjugateResidualNonHermitian<CoarseVector> CoarseKrylov_t;
+
+  Coarsening            &C;
+  PVdagMMultiGridParams  Params;
+  int                    nrhs;
+  int                    _regrid;    // FIRST member-like init: SetNrhs before members capture grids
+
+  PVdagMLinearOperator<Matrix,FineField>        PVdagM;
+  ShiftedPVdagMLinearOperator<Matrix,FineField> ShiftedPVdagM;
+  TrivialPrecon<FineField>                      simple_fine;
+  TrivialPrecon<CoarseVector>                   simpleC;
+  NonHermitianLinearOperator<CoarseOperator,CoarseVector> LinOpC;
+  ShiftedLinearOperator<CoarseVector>           ShiftedC;
+  MrhsDenseCCSolve<DenseBottom,CoarseCoarseVector> ccSolve;
+  CoarseKrylov_t                                CoarseSmootherGCR;
+  MrhsCoarseThreeLevelPrec<CoarseVector,CoarseCoarseVector> L2to3Precon;
+  CoarseKrylov_t                                L2PGCR;
+  FineSmoother_t                                SmootherGCR;
+  MrhsTwoLevelMG<FineField,CoarseVector,FineSmoother_t> ThreeLevelPrecon;
+  MrhsPGCRNonHermitian<FineField>               L1PGCR;
+
+  PVdagMMultiGridSolver(Matrix &Ddwf, Matrix &Dpv,
+                        Coarsening &_C, const PVdagMMultiGridParams &P, int nr)
+    : C(_C), Params(P), nrhs(nr),
+      _regrid((C.SetNrhs(nr),0)),              // members below capture C.CMrhs/C.CCMrhs
+      PVdagM(Ddwf,Dpv),
+      ShiftedPVdagM(P.FineSmoother.Shift,Ddwf,Dpv),
+      LinOpC(C.CoarseOpPV),
+      ShiftedC(P.CoarseSmoother.Shift,LinOpC),
+      ccSolve(*C.DenseCC,nr),
+      // Coarse smoother: one preconditioned restart of nstep GCR steps.
+      CoarseSmootherGCR(0.01,1,ShiftedC,simpleC,P.CoarseSmoother.Mmax,P.CoarseSmoother.Nstep),
+      L2to3Precon(LinOpC,CoarseSmootherGCR,C.MrhsProjectorL2,ccSolve,
+                  C.Grids.Coarse5d,C.Grids.CoarseCoarse5d,C.CCMrhs,nr),
+      // Coarse Krylov: Order/16 restarts of 16 steps.
+      L2PGCR(P.CoarseSolver.Tol,P.CoarseSolver.Order/16,LinOpC,L2to3Precon,P.CoarseSolver.Mmax,16),
+      // Fine smoother: one restart of nstep GCR steps, tolerance 0 = fixed work.
+      SmootherGCR(0.0,1,ShiftedPVdagM,simple_fine,P.FineSmoother.Mmax,P.FineSmoother.Nstep),
+      ThreeLevelPrecon(PVdagM,SmootherGCR,C.MrhsProjector,L2PGCR,C.Grids.Coarse5d,C.CMrhs),
+      L1PGCR(P.Outer.Tol,P.Outer.MaxIterations,PVdagM,ThreeLevelPrecon,P.Outer.Mmax,P.Outer.Nstep)
+  {
+    GRID_ASSERT( C.DenseCC != nullptr );
+    
+    CoarseSmootherGCR.Level(2);
+    CoarseSmootherGCR.Name("Csmoother");
+    CoarseSmootherGCR.SetZeroGuess(1);
+
+    L2PGCR.Level(2);
+    L2PGCR.Name("Couter");
+    L2PGCR.SetZeroGuess(1);
+    
+    SmootherGCR.Level(1);
+    SmootherGCR.Name("Fsmoother");
+    SmootherGCR.SetZeroGuess(1);
+
+    L1PGCR.Level(1);
+    L1PGCR.Name("Fouter");
+    L1PGCR.SetZeroGuess(1);
+    
+    // The V-cycle is preconditioner: sloppy halos inside, exact restored on exit.
+    ThreeLevelPrecon.SetSloppy    = [this](int s){ PVdagM.SloppyComms(s); };
+    ThreeLevelPrecon.SloppyComms  = Params.Setup.FineSloppyComms;
+    PVdagM.SloppyComms(0);
+    std::cout << GridLogMessage << "PVdagMMultiGridSolver: Nrhs " << nr
+              << ", fine halo policy: preconditioner+coarsening "
+              << (Params.Setup.FineSloppyComms ? "SLOPPY (fp32 wire)" : "exact")
+              << ", outer Krylov EXACT" << std::endl;
+  }
+
+  void Solve(std::vector<FineField> &src, std::vector<FineField> &sol)
+  {
+    GRID_ASSERT( (int)src.size() == nrhs );
+    // Start the solve from a clean device: evict setup-era Lattice copies
+    // and release the allocation caches' held blocks, so the LRU cap
+    // applies to the solve's own working set (OOM policy, Frontier).
+    MemoryManager::EvictAll();
+    MemoryManager::DropCache();
+
+    GridStopWatch w; w.Start();
+    L1PGCR(src,sol);
+    w.Stop();
+    std::cout << GridLogMessage << "PVdagMMultiGridSolver: Nrhs "<<nrhs<<" total " << w.Elapsed()
+              << "  (per RHS: " << w.useconds()/1.0e6/nrhs << " s)" << std::endl;
+
+    // The outer operator is exact by policy; assert the state rather than
+    // trust it -- a preconditioner that failed to restore would surface here.
+    PVdagM.SloppyComms(0);
+    {
+      FineField Ax(src[0].Grid());
+      RealD worst=0.0;
+      for(int r=0;r<nrhs;r++){
+	PVdagM.Op(sol[r],Ax);
+	Ax=Ax-src[r];
+        RealD rn=std::sqrt(norm2(Ax)/norm2(src[r]));
+        std::cout << GridLogMessage << "FINAL Nrhs "<<nrhs<<": rhs["<<r<<"] true residual = " << rn << std::endl;
+        worst=std::max(worst,rn);
+      }
+      std::cout << GridLogMessage << "FINAL Nrhs "<<nrhs<<": worst-case residual = " << worst
+                << "   (exact-halo verification)" << std::endl;
+    }
+  }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/PVdagMMultiGridParams.h

@@ -0,0 +1,155 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/PVdagMMultiGridParams.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////////////////////
+// Serialisable parameters for the mrhs PVdagM multigrid solver chain.
+//
+// Named for the multigrid FORM it configures, as WilsonMGParams is for the
+// Wilson MG tests; an MrhsHDCGParams will compose the same generic MG*
+// sub-structs for the Hermitian chain.  Read from XML (or JSON -- the
+// serialisation macros give both) via ReadPVdagMMultiGridParams below and
+// printed at startup by the macro's operator<<, so every log identifies its
+// own run.  The constructor defaults ARE a tuned operating point, not
+// arbitrary: an unconfigured run reproduces it, so changing them changes what
+// an unconfigured run does.  Smoother mmax == nstep (full GCR history).
+//
+// There are NO environment-variable controls anywhere in this subsystem:
+// parameters live here, library-internal constants are hard defaults in their
+// classes, verbosity is the --log channels.
+// Research instruments (GCR coefficient recording/replay, Chebyshev and
+// stationary smoother variants, power iteration) are deliberately NOT part
+// of this interface: they remain programmatic, for algorithmic studies,
+// not consumer knobs.
+//////////////////////////////////////////////////////////////////////////////////////
+
+// The smoother is the adaptive shifted PGCR -- the one correct route for
+// this non-Hermitian chain (stationary replay and Chebyshev were explored
+// and did not win here; Chebyshev remains the HERMITIAN chain's smoother).
+struct MGSmootherParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MGSmootherParams,
+                                  RealD, Shift,
+                                  int,   Nstep,
+                                  int,   Mmax);
+  MGSmootherParams(RealD shift=0.1, int nstep=6, int mmax=6)
+    : Shift(shift), Nstep(nstep), Mmax(mmax) {};
+};
+
+struct MGCoarseSolverParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MGCoarseSolverParams,
+                                  RealD, Tol,
+                                  int,   Order,
+                                  int,   Mmax);
+  MGCoarseSolverParams() : Tol(0.05), Order(200), Mmax(16) {};
+};
+
+struct MGOuterParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MGOuterParams,
+                                  RealD, Tol,
+                                  int,   MaxIterations,
+                                  int,   Mmax,
+                                  int,   Nstep);
+  MGOuterParams() : Tol(1.0e-8), MaxIterations(1000), Mmax(4), Nstep(8) {};
+};
+
+struct MGDenseParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MGDenseParams,
+                                  int, LeafSpan);   // big-leaf span in blocks; see BlockCyclicSchurInverse
+  MGDenseParams() : LeafSpan(9) {};
+};
+
+struct MGSetupParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MGSetupParams,
+                                  std::vector<int>, Block,           // fine -> coarse blocking
+                                  std::vector<int>, Block2,          // coarse -> coarse-coarse blocking
+                                  int,              CoarsenBatch,
+                                  std::string,      SubspaceFile,    // scidac; empty = create from noise, no I/O
+                                  int,              FineSloppyComms);// fp32 wire INSIDE the preconditioner only
+  MGSetupParams()
+    : Block({2,2,3,3}), Block2({4,4,2,4}), CoarsenBatch(9),
+      SubspaceFile(""), FineSloppyComms(1) {};
+};
+
+struct PVdagMMultiGridParams : Serializable {
+  GRID_SERIALIZABLE_CLASS_MEMBERS(PVdagMMultiGridParams,
+                                  MGSetupParams,        Setup,
+                                  MGSmootherParams,     FineSmoother,
+                                  MGSmootherParams,     CoarseSmoother,
+                                  MGCoarseSolverParams, CoarseSolver,
+                                  MGOuterParams,        Outer,
+                                  MGDenseParams,        Dense);
+  PVdagMMultiGridParams()
+    : FineSmoother  (0.1, 6, 6),
+      CoarseSmoother(0.1, 2, 2) {};
+};
+
+inline void CheckValidity(const PVdagMMultiGridParams &P)
+{
+  GRID_ASSERT( P.Setup.Block.size()  == 4 );
+  GRID_ASSERT( P.Setup.Block2.size() == 4 );
+  GRID_ASSERT( P.Setup.CoarsenBatch  >= 1 );
+  GRID_ASSERT( P.FineSmoother.Nstep   > 0 );
+  GRID_ASSERT( P.CoarseSmoother.Nstep > 0 );
+  GRID_ASSERT( P.CoarseSolver.Tol     > 0.0 );
+  GRID_ASSERT( P.Outer.Tol            > 0.0 );
+  GRID_ASSERT( P.Dense.LeafSpan      >= 1 );
+}
+
+//////////////////////////////////////////////////////////////////////////////////////
+// Read from an XML file; on a MISSING file the boss writes <file>.templ from
+// the defaults and returns false so the caller can Grid_finalize() and exit
+// -- the produced template is the documentation of every parameter.  An
+// empty filename means "defaults": no I/O, returns true.  Always prints the
+// effective parameters, so the log describes the run.
+//
+// Drivers fetch the filename themselves, e.g.
+//   std::string pfile("");
+//   if( GridCmdOptionExists(argv,argv+argc,"--multigrid-params") )
+//     pfile = GridCmdOptionPayload(argv,argv+argc,"--multigrid-params");
+//   if( !ReadPVdagMMultiGridParams(Params, pfile) ) { Grid_finalize(); return 0; }
+//////////////////////////////////////////////////////////////////////////////////////
+inline bool ReadPVdagMMultiGridParams(PVdagMMultiGridParams &P, const std::string &xmlfile)
+{
+  if ( xmlfile.length() ) {
+    bool good;
+    { std::ifstream f(xmlfile); good = f.good(); }
+    if ( !good ) {
+      std::cout << GridLogMessage << "PVdagMMultiGridParams: " << xmlfile << " does not exist" << std::endl;
+      if ( GlobalSharedMemory::WorldRank == 0 ) {
+        XmlWriter WR(xmlfile+".templ");
+        write(WR, "PVdagMMultiGridParams", P);
+        std::cout << GridLogMessage << "PVdagMMultiGridParams: template written to "
+                  << xmlfile << ".templ" << std::endl;
+      }
+      return false;
+    }
+    XmlReader RD(xmlfile);
+    read(RD, "PVdagMMultiGridParams", P);
+  }
+  CheckValidity(P);
+  std::cout << GridLogMessage << "PVdagMMultiGridParams ("
+            << (xmlfile.length() ? xmlfile : std::string("defaults")) << "):" << std::endl;
+  std::cout << P << std::endl;
+  return true;
+}
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/PVdagMOperators.h

@@ -0,0 +1,88 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/PVdagMOperators.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////
+// A = PV^dag M (non-Hermitian).  The operator the PVdagM multigrid
+// coarsens and solves.
+//
+// SloppyComms delegates to BOTH underlying fermion operators, so one
+// call controls every fine halo in the chain -- including any shifted
+// wrapper built on the same two matrices.  Halo-precision POLICY:
+// reduced-precision (fp32 wire) halos belong in the PRECONDITIONER --
+// the smoother, the V-cycle's own residuals, the coarsening -- and
+// NEVER in the outer Krylov, whose applications define what
+// "converged" means.  The operators default to EXACT; the V-cycle
+// turns sloppiness on for its own scope only (MrhsTwoLevelMG).
+//////////////////////////////////////////////////////////////////////
+template<class Matrix,class Field>
+class PVdagMLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat; Matrix &_PV;
+public:
+  PVdagMLinearOperator(Matrix &Mat,Matrix &PV): _Mat(Mat),_PV(PV) {};
+  void SloppyComms(int sloppy) { _Mat.SloppyComms(sloppy); _PV.SloppyComms(sloppy); }
+  void OpDiag (const Field &in, Field &out) { GRID_ASSERT(0); }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) { GRID_ASSERT(0); }
+  void OpDirAll  (const Field &in, std::vector<Field> &out){ GRID_ASSERT(0); };
+  void Op     (const Field &in, Field &out){ Field tmp(in.Grid()); _Mat.M(in,tmp); _PV.Mdag(tmp,out); }
+  void AdjOp  (const Field &in, Field &out){ Field tmp(in.Grid()); _PV.M(in,tmp); _Mat.Mdag(tmp,out); }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){ HermOp(in,out); ComplexD d=innerProduct(in,out); n1=real(d); n2=norm2(out); }
+  void HermOp(const Field &in, Field &out){ Field tmp(in.Grid()); Op(in,tmp); AdjOp(tmp,out); }
+};
+
+//////////////////////////////////////////////////////////////////////
+// A + shift, built directly on the two matrices (fine level: avoids
+// an extra fine-field pass through a generic shifted wrapper).
+//////////////////////////////////////////////////////////////////////
+template<class Matrix,class Field>
+class ShiftedPVdagMLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat; Matrix &_PV;
+public:
+  RealD shift;
+  ShiftedPVdagMLinearOperator(RealD _shift,Matrix &Mat,Matrix &PV): shift(_shift),_Mat(Mat),_PV(PV){};
+  void OpDiag (const Field &in, Field &out) { GRID_ASSERT(0); }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) { GRID_ASSERT(0); }
+  void OpDirAll  (const Field &in, std::vector<Field> &out){ GRID_ASSERT(0); };
+  void Op     (const Field &in, Field &out){ Field tmp(in.Grid()); _Mat.M(in,tmp); _PV.Mdag(tmp,out); out = out + shift*in; }
+  void AdjOp  (const Field &in, Field &out){ Field tmp(in.Grid()); _PV.M(in,tmp); _Mat.Mdag(tmp,out); out = out + shift*in; }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){ GRID_ASSERT(0); }
+  void HermOp(const Field &in, Field &out){ Field tmp(in.Grid()); Op(in,tmp); AdjOp(tmp,out); }
+};
+
+//////////////////////////////////////////////////////////////////////
+// Op + shift for any operator (coarse levels).
+//////////////////////////////////////////////////////////////////////
+template<class Field>
+class ShiftedLinearOperator : public LinearOperatorBase<Field> {
+  LinearOperatorBase<Field> &_Op; RealD shift;
+public:
+  ShiftedLinearOperator(RealD _shift, LinearOperatorBase<Field> &Op) : _Op(Op), shift(_shift) {}
+  void OpDiag  (const Field &in, Field &out) { GRID_ASSERT(0); }
+  void OpDir   (const Field &in, Field &out,int dir,int disp) { GRID_ASSERT(0); }
+  void OpDirAll (const Field &in, std::vector<Field> &out) { GRID_ASSERT(0); }
+  void Op      (const Field &in, Field &out) { _Op.Op(in,out);    out = out + shift*in; }
+  void AdjOp   (const Field &in, Field &out) { _Op.AdjOp(in,out); out = out + shift*in; }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){ GRID_ASSERT(0); }
+  void HermOp  (const Field &in, Field &out) { Field tmp(in.Grid()); Op(in,tmp); AdjOp(tmp,out); }
+};
+
+NAMESPACE_END(Grid);

Grid/algorithms/multigrid/Smoothers.h

@@ -0,0 +1,292 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/algorithms/multigrid/Smoothers.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+#include <Grid/algorithms/iterative/GCRCoefficients.h>
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////////////////////////////
+// Multigrid smoothers as LinearFunction<Field> objects.
+//
+// Gathered here from ~15 copies in tests/solver and tests/debug (HDCR /
+// HDCG era), plus the fixed-polynomial smoothers of 2026:
+//
+//   ChebyshevSmoother          Chebyshev approx to 1/x on [lo,hi], applied
+//                              through HermOp (the original HDCG smoother on
+//                              a Hermitian shifted operator).
+//   ChebyshevNonHermitianSmoother  same polynomial applied through Op(), for
+//                              a non-Hermitian (near-normal, real-spectrum)
+//                              smoother operator such as shifted PVdagM.
+//   ChebyshevInverter          one Chebyshev-corrected step with residual print.
+//   MirsSmoother               shifted-MdagM CG, HDCG arXiv:1402.2585.
+//   GCRReplaySmoother          replays a GCR's recorded step lengths a_k and
+//                              orthogonalisation coefficients b_kj with NO
+//                              inner products: one matvec per step, zero
+//                              reductions.  The "PreconditionerMirsPoly"
+//                              idea of 1402.2585 p.13 applied to GCR.
+//
+// Recording: PrecGeneralisedConjugateResidualNonHermitian::SetCoefficientRecorder
+// accumulates per-step means over calls into a GCRCoefficients; construct a
+// GCRReplaySmoother from it.
+//////////////////////////////////////////////////////////////////////////////
+
+inline RealD InverseApproximation(RealD x){ return 1.0/x; }
+
+//////////////////////////////////////////////////////////////////////////////
+// HermOp-based Chebyshev smoother.  The second template parameter and the
+// 5-argument constructor exist only so the historical call sites
+//   ChebyshevSmoother<LatticeFermion,DomainWallFermionD> S(lo,hi,ord,HermOp,Ddwf);
+// compile unchanged; the Matrix argument was never used.
+//////////////////////////////////////////////////////////////////////////////
+template<class Field,class Matrix=void> class ChebyshevSmoother : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  typedef LinearOperatorBase<Field> FineOperator;
+  FineOperator   & _SmootherOperator;
+  Chebyshev<Field> Cheby;
+  ChebyshevSmoother(RealD _lo,RealD _hi,int _ord, FineOperator &SmootherOperator) :
+    _SmootherOperator(SmootherOperator),
+    Cheby(_lo,_hi,_ord,InverseApproximation)
+  {
+    std::cout << GridLogMessage<<" Chebyshev smoother order "<<_ord<<" ["<<_lo<<","<<_hi<<"]"<<std::endl;
+  };
+  template<class M>
+  ChebyshevSmoother(RealD _lo,RealD _hi,int _ord, FineOperator &SmootherOperator, M &) :
+    ChebyshevSmoother(_lo,_hi,_ord,SmootherOperator) {};
+  void operator() (const Field &in, Field &out)
+  {
+    Cheby(_SmootherOperator,in,out);
+  }
+};
+
+//////////////////////////////////////////////////////////////////////////////
+// Op()-based Chebyshev smoother: x = S(A) r with S the Chebyshev fit to 1/x
+// on [lo,hi].  Same three-term recurrence as Chebyshev<Field>::operator()
+// but through Op, for the non-Hermitian smoother operators of the PVdagM
+// multigrid (real coefficients / near-normal, as the recorded GCR
+// coefficients show).  `order` matvecs, no reductions.
+//////////////////////////////////////////////////////////////////////////////
+template<class Field> class ChebyshevNonHermitianSmoother : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  LinearOperatorBase<Field> &Linop;
+  RealD lo, hi; int order;
+  std::vector<RealD> Coeffs;
+  int Verbose = 0;
+  std::string name = "cheb";
+  ChebyshevNonHermitianSmoother(RealD _lo,RealD _hi,int _order,LinearOperatorBase<Field> &Op)
+    : Linop(Op), lo(_lo), hi(_hi), order(_order)
+  {
+    GRID_ASSERT(order>=2);
+    Coeffs.resize(order);
+    for(int j=0;j<order;j++){
+      RealD s=0;
+      for(int k=0;k<order;k++){
+        RealD y=std::cos(M_PI*(k+0.5)/order);
+        RealD x=0.5*(y*(hi-lo)+(hi+lo));
+        s=s+InverseApproximation(x)*std::cos( j*M_PI*(k+0.5)/order );
+      }
+      Coeffs[j] = s * 2.0/order;
+    }
+    std::cout << GridLogMessage<<" ChebyshevNonHermitian smoother order "<<order<<" ["<<lo<<","<<hi<<"]"<<std::endl;
+  }
+  void operator() (const Field &in, Field &out)
+  {
+    GRID_TRACE("ChebyshevSmoother");
+    GridBase *grid=in.Grid();
+    Field T0(grid); T0 = in;
+    Field T1(grid), T2(grid), y(grid);
+    Field *Tnm=&T0, *Tn=&T1, *Tnp=&T2;
+    RealD xscale = 2.0/(hi-lo);
+    RealD mscale = -(hi+lo)/(hi-lo);
+    Linop.Op(T0,y);
+    axpby(T1,xscale,mscale,y,in);
+    axpby(out,0.5*Coeffs[0],Coeffs[1],T0,T1);
+    for(int n=2;n<order;n++){
+      Linop.Op(*Tn,y);
+      axpby(y,xscale,mscale,y,(*Tn));
+      axpby(*Tnp,2.0,-1.0,y,(*Tnm));
+      if ( Coeffs[n] != 0.0 ) axpy(out,Coeffs[n],*Tnp,out);
+      Field *swizzle=Tnm; Tnm=Tn; Tn=Tnp; Tnp=swizzle;
+    }
+    if ( Verbose ) { Linop.Op(out,y); y = y - in; std::cout << GridLogMessage << " " << name << " cheb |r|/|r0| = " << std::sqrt(norm2(y)/norm2(in)) << std::endl; }
+  }
+};
+
+template<class Field> class ChebyshevInverter : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  typedef LinearOperatorBase<Field> FineOperator;
+  FineOperator   & _Operator;
+  Chebyshev<Field> Cheby;
+  ChebyshevInverter(RealD _lo,RealD _hi,int _ord, FineOperator &Operator) :
+    _Operator(Operator),
+    Cheby(_lo,_hi,_ord,InverseApproximation)
+  {
+    std::cout << GridLogMessage<<" Chebyshev Inverter order "<<_ord<<" ["<<_lo<<","<<_hi<<"]"<<std::endl;
+  };
+  void operator() (const Field &in, Field &out)
+  {
+    Field r(in.Grid());
+    Field AinvR(in.Grid());
+    _Operator.HermOp(out,r);
+    r = in - r; // b - A x
+    Cheby(_Operator,r,AinvR); // A^{-1} ( b - A x ) ~ A^{-1} b - x
+    out = out + AinvR;
+    _Operator.HermOp(out,r);
+    r = in - r; // b - A x
+    RealD rr = norm2(r);
+    RealD ss = norm2(in);
+    std::cout << GridLogMessage << "ChebshevInverse resid " <<::sqrt(rr/ss)<<std::endl;
+  }
+};
+
+//////////////////////////////////////////////////////////////////////////////
+// MIRS: CG on the infra-red shifted MdagM (HDCG, arXiv:1402.2585).
+//////////////////////////////////////////////////////////////////////////////
+template<class Field,class Matrix> class MirsSmoother : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  typedef LinearOperatorBase<Field>                            FineOperator;
+  Matrix         & SmootherMatrix;
+  FineOperator   & SmootherOperator;
+  RealD tol;
+  RealD shift;
+  int   maxit;
+  MirsSmoother(RealD _shift,RealD _tol,int _maxit,FineOperator &_SmootherOperator,Matrix &_SmootherMatrix) :
+    shift(_shift),tol(_tol),maxit(_maxit),
+    SmootherOperator(_SmootherOperator),
+    SmootherMatrix(_SmootherMatrix)
+  {};
+  void operator() (const Field &in, Field &out)
+  {
+    ZeroGuesser<Field> Guess;
+    ConjugateGradient<Field>  CG(tol,maxit,false);
+    Field src(in.Grid());
+    ShiftedMdagMLinearOperator<SparseMatrixBase<Field>,Field> MdagMOp(SmootherMatrix,shift);
+    SmootherOperator.AdjOp(in,src);
+    Guess(src,out);
+    CG(MdagMOp,src,out);
+  }
+};
+
+//////////////////////////////////////////////////////////////////////////////
+// Replay of a recorded GCR with a trivial preconditioner:
+//   p_0 = r_0 ;  x_{k+1} = x_k + a_k p_k ;  r_{k+1} = r_k - a_k A p_k ;
+//   p_{k+1} = r_{k+1} + sum_{j<northog(k)} b_kj p_{k-j}
+// One matvec per step, no reductions, history of mmax p vectors held
+// persistently (allocated on first use, per grid).
+//////////////////////////////////////////////////////////////////////////////
+template<class Field> class GCRReplaySmoother : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  LinearOperatorBase<Field> &Linop;
+  int mmax, nstep;
+  std::vector<ComplexD>               a;
+  std::vector<std::vector<ComplexD> > b;
+  GridBase *hist_grid = nullptr;
+  std::vector<Field> p;
+  int Verbose = 0;                       // 1: print |r_m|/|r_0| per call (one extra reduction)
+  std::string name = "replay";
+  GCRReplaySmoother(LinearOperatorBase<Field> &Op, const GCRCoefficients &c) : Linop(Op)
+  {
+    mmax  = c.mmax;  GRID_ASSERT(mmax>=1);
+    nstep = c.Steps(); GRID_ASSERT(nstep>=1);
+    a.resize(nstep); b.resize(nstep);
+    // No std::abs / std::isfinite on ComplexD: it is thrust::complex on HIP
+    // builds and neither overload exists.  Work on the real and imaginary
+    // parts explicitly.
+    auto finite = [](ComplexD z){ RealD x=real(z), y=imag(z); return (x==x) && (y==y) && (x-x==0.0) && (y-y==0.0); };
+    auto cabs   = [](ComplexD z){ RealD x=real(z), y=imag(z); return std::sqrt(x*x+y*y); };
+    RealD amax=0.0, bmax=0.0;
+    for(int k=0;k<nstep;k++){
+      a[k] = c.A(k);
+      GRID_ASSERT( finite(a[k]) );
+      amax = std::max(amax, cabs(a[k]));
+      b[k].resize(c.NB(k));
+      GRID_ASSERT( c.NB(k) <= mmax-1 );
+      for(int j=0;j<c.NB(k);j++){
+        b[k][j] = c.B(k,j);
+        GRID_ASSERT( finite(b[k][j]) );
+        bmax = std::max(bmax, cabs(b[k][j]));
+      }
+    }
+    std::cout << GridLogMessage << " GCRReplaySmoother: max|a| " << amax << " max|b| " << bmax << std::endl;
+    std::cout << GridLogMessage << " GCRReplaySmoother: " << nstep << " steps, mmax " << mmax
+              << ", from " << c.Calls() << " recorded calls" << std::endl;
+  }
+  void operator() (const Field &src, Field &psi)
+  {
+    GRID_TRACE("GCRReplaySmoother");
+    GridBase *grid = src.Grid();
+    if ( hist_grid != grid ) {
+      p.clear(); p.reserve(mmax);
+      for(int i=0;i<mmax;i++) p.emplace_back(grid);
+      hist_grid = grid;
+    }
+    Field r(grid), q(grid);
+    r = src;
+    psi = Zero();
+    p[0] = r;
+    RealD r0 = Verbose ? norm2(src) : 0.0;
+    for(int k=0;k<nstep;k++){
+      int kp=k+1, peri_k=k%mmax, peri_kp=kp%mmax;
+      Linop.Op(p[peri_k],q);                 // q_k = A p_k
+      axpy(psi,  a[k], p[peri_k], psi);
+      if ( k==nstep-1 ) {
+        if ( Verbose ) { axpy(r,-a[k],q,r); std::cout << GridLogMessage << " " << name << " replay |r|/|r0| = " << std::sqrt(norm2(r)/r0) << std::endl; }
+        break;
+      }
+      axpy(r,   -a[k], q,         r);
+      p[peri_kp] = r;
+      for(int j=0;j<(int)b[k].size();j++){
+        int peri_back=(k-j)%mmax;
+        axpy(p[peri_kp], b[k][j], p[peri_back], p[peri_kp]);
+      }
+    }
+  }
+};
+
+//////////////////////////////////////////////////////////////////////////////
+// A LinearFunction that forwards to a replaceable target: lets a V-cycle be
+// built once around a smoother slot whose implementation is swapped at run
+// time (record with the adaptive GCR, then replay the polynomial).
+//////////////////////////////////////////////////////////////////////////////
+template<class Field> class SwitchableSmoother : public LinearFunction<Field>
+{
+public:
+  using LinearFunction<Field>::operator();
+  LinearFunction<Field> *current;
+  std::string label;
+  SwitchableSmoother(LinearFunction<Field> &initial, const std::string &l="initial") : current(&initial), label(l) {}
+  void Set(LinearFunction<Field> &f, const std::string &l)
+  {
+    current = &f; label = l;
+    std::cout << GridLogMessage << " SwitchableSmoother -> " << l << std::endl;
+  }
+  void operator() (const Field &in, Field &out) { (*current)(in,out); }
+};
+
+NAMESPACE_END(Grid);

Grid/allocator/AlignedAllocator.h

@@ -57,7 +57,7 @@ public:
     if ( (_Tp*)ptr == (_Tp *) NULL ) {
       printf("Grid CPU Allocator got NULL for %lu bytes\n",(unsigned long) bytes );
     }
-    assert( ( (_Tp*)ptr != (_Tp *)NULL ) );
+    GRID_ASSERT( ( (_Tp*)ptr != (_Tp *)NULL ) );
     return ptr;
   }
 
@@ -69,7 +69,7 @@ public:
   }
 
   // FIXME: hack for the copy constructor: it must be avoided to avoid single thread loop
-  void construct(pointer __p, const _Tp& __val) { assert(0);};
+  void construct(pointer __p, const _Tp& __val) { };
   void construct(pointer __p) { };
   void destroy(pointer __p) { };
 };
@@ -106,7 +106,7 @@ public:
     if ( (_Tp*)ptr == (_Tp *) NULL ) {
       printf("Grid Shared Allocator got NULL for %lu bytes\n",(unsigned long) bytes );
     }
-    assert( ( (_Tp*)ptr != (_Tp *)NULL ) );
+    GRID_ASSERT( ( (_Tp*)ptr != (_Tp *)NULL ) );
     return ptr;
   }
 
@@ -154,7 +154,7 @@ public:
     if ( (_Tp*)ptr == (_Tp *) NULL ) {
       printf("Grid Device Allocator got NULL for %lu bytes\n",(unsigned long) bytes );
     }
-    assert( ( (_Tp*)ptr != (_Tp *)NULL ) );
+    GRID_ASSERT( ( (_Tp*)ptr != (_Tp *)NULL ) );
     return ptr;
   }
 
@@ -175,10 +175,11 @@ template<typename _Tp>  inline bool operator!=(const devAllocator<_Tp>&, const d
 // Template typedefs
 ////////////////////////////////////////////////////////////////////////////////
 template<class T> using hostVector          = std::vector<T,alignedAllocator<T> >;           // Needs autoview
-template<class T> using Vector              = std::vector<T,uvmAllocator<T> >;               // 
+template<class T> using Vector              = std::vector<T,uvmAllocator<T> >;               // Really want to deprecate
 template<class T> using uvmVector           = std::vector<T,uvmAllocator<T> >;               // auto migrating page
 template<class T> using deviceVector        = std::vector<T,devAllocator<T> >;               // device vector
 
+/*
 template<class T> class vecView
 {
  protected:
@@ -214,6 +215,7 @@ template<class T> vecView<T> VectorView(Vector<T> &vec,ViewMode _mode)
 #define autoVecView(v_v,v,mode)					\
   auto v_v = VectorView(v,mode);				\
   ViewCloser<decltype(v_v)> _autoView##v_v(v_v);
+*/
 
 NAMESPACE_END(Grid);
 

Grid/allocator/MemoryManager.cc

@@ -63,12 +63,10 @@ void MemoryManager::PrintBytes(void)
   std::cout << " MemoryManager : "<<(total_device>>20)<<" accelerator Mbytes "<<std::endl;
   std::cout << " MemoryManager : "<<(total_host>>20)  <<" cpu         Mbytes "<<std::endl;
   uint64_t cacheBytes;
-  cacheBytes = CacheBytes[Cpu];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" cpu cache Mbytes "<<std::endl;
-  cacheBytes = CacheBytes[Acc];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" acc cache Mbytes "<<std::endl;
-  cacheBytes = CacheBytes[Shared];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" shared cache Mbytes "<<std::endl;
+  cacheBytes = HostCacheBytes();
+  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" cpu alloc cache Mbytes "<<std::endl;
+  cacheBytes = DeviceCacheBytes();
+  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" acc alloc cache Mbytes "<<std::endl;
   
 #ifdef GRID_CUDA
   cuda_mem();
@@ -86,6 +84,31 @@ MemoryManager::AllocationCacheEntry MemoryManager::Entries[MemoryManager::Nalloc
 int MemoryManager::Victim[MemoryManager::NallocType];
 int MemoryManager::Ncache[MemoryManager::NallocType] = { 2, 0, 8, 8, 0, 16, 8, 0, 16 };
 uint64_t MemoryManager::CacheBytes[MemoryManager::NallocType];
+uint64_t MemoryManager::DeviceAllocCalls;
+uint64_t MemoryManager::DeviceFreeCalls;
+uint64_t MemoryManager::DeviceAllocBytes;
+uint64_t MemoryManager::DeviceFreeBytes;
+uint64_t MemoryManager::DeviceCacheHits;
+
+void MemoryManager::PrintAllocCounts(void)
+{
+  std::cout << GridLogMemory << "MemoryManager: device allocator calls: acceleratorAllocDevice "
+	    << DeviceAllocCalls <<" ("<< DeviceAllocBytes <<" bytes), acceleratorFreeDevice "
+	    << DeviceFreeCalls  <<" ("<< DeviceFreeBytes  <<" bytes), served from ring cache "
+	    << DeviceCacheHits  << std::endl;
+  std::cout << GridLogMemory << "MemoryManager: view traffic: HostToDevice "
+	    << HostToDeviceXfer <<" transfers ("<< HostToDeviceBytes <<" bytes), DeviceToHost "
+	    << DeviceToHostXfer <<" transfers ("<< DeviceToHostBytes <<" bytes), evictions "
+	    << DeviceEvictions  << std::endl;
+}
+void MemoryManager::Snapshot(const std::string &where)
+{
+  if ( !GridLogMemory.isActive() ) return;
+  std::cout << GridLogMemory << "---------------- memory snapshot: "<< where <<" ----------------"<<std::endl;
+  PrintAllocCounts();
+  PrintBytes();
+  acceleratorMem();
+}
 //////////////////////////////////////////////////////////////////////
 // Actual allocation and deallocation utils
 //////////////////////////////////////////////////////////////////////
@@ -95,6 +118,11 @@ void *MemoryManager::AcceleratorAllocate(size_t bytes)
   void *ptr = (void *) Lookup(bytes,Acc);
   if ( ptr == (void *) NULL ) {
     ptr = (void *) acceleratorAllocDevice(bytes);
+    DeviceAllocCalls++;  DeviceAllocBytes+=bytes;
+    std::cout << GridLogMemory << "MemoryManager: acceleratorAllocDevice size "<< bytes
+	      <<" AccPtr "<< std::hex << (uint64_t)ptr << std::dec << std::endl;
+  } else {
+    DeviceCacheHits++;
   }
 #ifdef GRID_MM_VERBOSE
   std::cout <<"AcceleratorAllocate "<<std::endl;
@@ -105,9 +133,13 @@ void *MemoryManager::AcceleratorAllocate(size_t bytes)
 void  MemoryManager::AcceleratorFree    (void *ptr,size_t bytes)
 {
   total_device-=bytes;
-  void *__freeme = Insert(ptr,bytes,Acc);
+  size_t freed_bytes=0;
+  void *__freeme = Insert(ptr,bytes,Acc,&freed_bytes);
   if ( __freeme ) {
     acceleratorFreeDevice(__freeme);
+    DeviceFreeCalls++;  DeviceFreeBytes+=freed_bytes;
+    std::cout << GridLogMemory << "MemoryManager: acceleratorFreeDevice size "<< freed_bytes
+	      <<" AccPtr "<< std::hex << (uint64_t)__freeme << std::dec << std::endl;
   }
 #ifdef GRID_MM_VERBOSE
   std::cout <<"AcceleratorFree "<<std::endl;
@@ -274,8 +306,7 @@ void MemoryManager::InitMessage(void) {
 #endif
 
 }
-
-void *MemoryManager::Insert(void *ptr,size_t bytes,int type) 
+void *MemoryManager::Insert(void *ptr,size_t bytes,int type,size_t *freed) 
 {
 #ifdef ALLOCATION_CACHE
   int cache;
@@ -283,19 +314,59 @@ void *MemoryManager::Insert(void *ptr,size_t bytes,int type)
   else if (bytes >= GRID_ALLOC_HUGE_LIMIT) cache = type + 1;
   else                                     cache = type;
 
-  return Insert(ptr,bytes,Entries[cache],Ncache[cache],Victim[cache],CacheBytes[cache]);  
+  return Insert(ptr,bytes,Entries[cache],Ncache[cache],Victim[cache],CacheBytes[cache],freed);  
 #else
   return ptr;
 #endif
 }
+void MemoryManager::DropCache(void)
+{
+  // Release every block held in the allocation caches (the "recent
+  // allocations" pools: Small/Large/Huge x Cpu/Acc/Shared).  Blocks in these
+  // pools are freed from the caller's point of view but still occupy memory;
+  // after a large setup phase (SUMMA scratch, coarsening temporaries) they can
+  // hold gigabytes of device memory that hipMalloc then cannot get.  Each pool
+  // is freed with the call that pairs with its allocator.
+  for(int sz=0;sz<3;sz++) {
+    int t;
+    t = Acc+sz;
+    for(int e=0;e<Ncache[t];e++) {
+      if( Entries[t][e].valid ) {
+	acceleratorFreeDevice(Entries[t][e].address);
+	Entries[t][e].valid = 0; Entries[t][e].bytes = 0; Entries[t][e].address = NULL;
+      }
+    }
+    CacheBytes[t]=0; Victim[t]=0;
+    t = Shared+sz;
+    for(int e=0;e<Ncache[t];e++) {
+      if( Entries[t][e].valid ) {
+	acceleratorFreeShared(Entries[t][e].address);
+	Entries[t][e].valid = 0; Entries[t][e].bytes = 0; Entries[t][e].address = NULL;
+      }
+    }
+    CacheBytes[t]=0; Victim[t]=0;
+    t = Cpu+sz;
+    for(int e=0;e<Ncache[t];e++) {
+      if( Entries[t][e].valid ) {
+#ifdef GRID_UVM
+	acceleratorFreeShared(Entries[t][e].address);
+#else
+	acceleratorFreeCpu(Entries[t][e].address);
+#endif
+	Entries[t][e].valid = 0; Entries[t][e].bytes = 0; Entries[t][e].address = NULL;
+      }
+    }
+    CacheBytes[t]=0; Victim[t]=0;
+  }
+}
 
-void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim, uint64_t &cacheBytes) 
+void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim, uint64_t &cacheBytes,size_t *freed) 
 {
 #ifdef GRID_OMP
-  assert(omp_in_parallel()==0);
+  GRID_ASSERT(omp_in_parallel()==0);
 #endif 
 
-  if (ncache == 0) return ptr;
+  if (ncache == 0) { if(freed) *freed = bytes; return ptr; }   // uncached: the incoming block is the one freed
 
   void * ret = NULL;
   int v = -1;
@@ -314,6 +385,7 @@ void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries
 
   if ( entries[v].valid ) {
     ret = entries[v].address;
+    if ( freed ) *freed = entries[v].bytes;   // the DISPLACED block is what actually gets freed
     cacheBytes -= entries[v].bytes;
     entries[v].valid = 0;
     entries[v].address = NULL;
@@ -345,7 +417,7 @@ void *MemoryManager::Lookup(size_t bytes,int type)
 void *MemoryManager::Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache,uint64_t & cacheBytes) 
 {
 #ifdef GRID_OMP
-  assert(omp_in_parallel()==0);
+  GRID_ASSERT(omp_in_parallel()==0);
 #endif 
   for(int e=0;e<ncache;e++){
     if ( entries[e].valid && ( entries[e].bytes == bytes ) ) {

Grid/allocator/MemoryManager.h

@@ -108,12 +108,12 @@ private:
   /////////////////////////////////////////////////
   // Free pool
   /////////////////////////////////////////////////
-  static void *Insert(void *ptr,size_t bytes,int type) ;
+  static void *Insert(void *ptr,size_t bytes,int type,size_t *freed=nullptr) ;
   static void *Lookup(size_t bytes,int type) ;
-  static void *Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim,uint64_t &cbytes) ;
+  static void *Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim,uint64_t &cbytes,size_t *freed=nullptr) ;
   static void *Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache,uint64_t &cbytes) ;
 
- public:
+public:  
   static void PrintBytes(void);
   static void Audit(std::string s);
   static void Init(void);
@@ -137,7 +137,19 @@ private:
   static uint64_t     DeviceToHostXfer;
   static uint64_t     DeviceEvictions;
   static uint64_t     DeviceDestroy;
-  
+  // Calls that actually reach the runtime, as distinct from the Evict/Clone traffic above:
+  // a free only reaches acceleratorFreeDevice when displaced from the allocation ring cache,
+  // and an allocate only reaches acceleratorAllocDevice on a ring miss.
+  static uint64_t     DeviceAllocCalls;
+  static uint64_t     DeviceFreeCalls;
+  static uint64_t     DeviceAllocBytes;
+  static uint64_t     DeviceFreeBytes;
+  static uint64_t     DeviceCacheHits;
+  static void         PrintAllocCounts(void);
+  // Labelled snapshot: allocator counts + footprint + device free/total.  Silent unless
+  // --log Memory is on, so it can be left in hot code.
+  static void         Snapshot(const std::string &where);
+
   static uint64_t     DeviceCacheBytes();
   static uint64_t     HostCacheBytes();
 
@@ -215,6 +227,8 @@ private:
   static void NotifyDeletion(void * CpuPtr);
   static void Print(void);
   static void PrintAll(void);
+  static void EvictAll(void);
+  static void DropCache(void);
   static void PrintState( void* CpuPtr);
   static int   isOpen   (void* CpuPtr);
   static void  ViewClose(void* CpuPtr,ViewMode mode);

Grid/allocator/MemoryManagerCache.cc

@@ -9,6 +9,7 @@ static char print_buffer [ MAXLINE ];
 #define mprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogMemory << print_buffer << std::endl;
 #define dprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogDebug  << print_buffer << std::endl;
 //#define dprintf(...) 
+//#define mprintf(...) 
 
 ////////////////////////////////////////////////////////////
 // For caching copies of data on device
@@ -49,12 +50,12 @@ int   MemoryManager::EntryPresent(uint64_t CpuPtr)
 {
   if(AccViewTable.empty()) return 0;
 
-  auto count = AccViewTable.count(CpuPtr);  assert((count==0)||(count==1));
+  auto count = AccViewTable.count(CpuPtr);  GRID_ASSERT((count==0)||(count==1));
   return count;
 }
 void  MemoryManager::EntryCreate(uint64_t CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint)
 {
-  assert(!EntryPresent(CpuPtr));
+  GRID_ASSERT(!EntryPresent(CpuPtr));
   AcceleratorViewEntry AccCache;
   AccCache.CpuPtr = CpuPtr;
   AccCache.AccPtr = (uint64_t)NULL;
@@ -68,9 +69,9 @@ void  MemoryManager::EntryCreate(uint64_t CpuPtr,size_t bytes,ViewMode mode,View
 }
 MemoryManager::AccViewTableIterator MemoryManager::EntryLookup(uint64_t CpuPtr)
 {
-  assert(EntryPresent(CpuPtr));
+  GRID_ASSERT(EntryPresent(CpuPtr));
   auto AccCacheIterator = AccViewTable.find(CpuPtr);
-  assert(AccCacheIterator!=AccViewTable.end());
+  GRID_ASSERT(AccCacheIterator!=AccViewTable.end());
   return AccCacheIterator;
 }
 void MemoryManager::EntryErase(uint64_t CpuPtr)
@@ -78,9 +79,28 @@ void MemoryManager::EntryErase(uint64_t CpuPtr)
   auto AccCache = EntryLookup(CpuPtr);
   AccViewTable.erase(CpuPtr);
 }
+/////////////////////////////////////////////////////////////////////////////////
+// LRU membership invariant:
+//
+//    LRU_valid == 1   <=>   AccPtr != NULL  &&  accLock == 0  &&  cpuLock == 0
+//
+// i.e. the LRU queue contains exactly the device-resident, completely unlocked
+// entries -- the evictable set.  Membership is maintained EAGERLY at the lock
+// 0<->1 edges, O(1) via the stored LRU_entry iterator:
+//
+//    AcceleratorViewOpen   lock 0->1 : LRUremove   (gated on LRU_valid)
+//    AcceleratorViewClose  accLock->0: LRUinsert   (AccPtr necessarily exists)
+//    CpuViewOpen           lock 0->1 : LRUremove   (gated on LRU_valid)
+//    CpuViewClose          cpuLock->0: LRUinsert   (iff AccPtr exists)
+//    Evict/AccDiscard                : LRUremove   (frees the device copy)
+//
+// Consequences: victims taken from LRU.back() are evictable by construction;
+// Evict() on a locked entry is an invariant violation (asserted), and the
+// eviction loops (EvictVictims/EvictAll) cannot spin.
+/////////////////////////////////////////////////////////////////////////////////
 void  MemoryManager::LRUinsert(AcceleratorViewEntry &AccCache)
 {
-  assert(AccCache.LRU_valid==0);
+  GRID_ASSERT(AccCache.LRU_valid==0);
   if (AccCache.transient) { 
     LRU.push_back(AccCache.CpuPtr);
     AccCache.LRU_entry = --LRU.end();
@@ -93,7 +113,7 @@ void  MemoryManager::LRUinsert(AcceleratorViewEntry &AccCache)
 }
 void  MemoryManager::LRUremove(AcceleratorViewEntry &AccCache)
 {
-  assert(AccCache.LRU_valid==1);
+  GRID_ASSERT(AccCache.LRU_valid==1);
   LRU.erase(AccCache.LRU_entry);
   AccCache.LRU_valid = 0;
   DeviceLRUBytes-=AccCache.bytes;
@@ -107,19 +127,19 @@ void MemoryManager::AccDiscard(AcceleratorViewEntry &AccCache)
   // Remove from Accelerator, remove entry, without flush
   // Cannot be locked. If allocated Must be in LRU pool.
   ///////////////////////////////////////////////////////////
-  assert(AccCache.state!=Empty);
+  GRID_ASSERT(AccCache.state!=Empty);
   
-  dprintf("MemoryManager: Discard(%lx) %lx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
-  assert(AccCache.accLock==0);
-  assert(AccCache.cpuLock==0);
-  assert(AccCache.CpuPtr!=(uint64_t)NULL);
+  dprintf("MemoryManager: Discard(%lx) %lx",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
+  GRID_ASSERT(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.cpuLock==0);
+  GRID_ASSERT(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.AccPtr) {
     AcceleratorFree((void *)AccCache.AccPtr,AccCache.bytes);
     DeviceDestroy++;
     DeviceBytes   -=AccCache.bytes;
     LRUremove(AccCache);
     AccCache.AccPtr=(uint64_t) NULL;
-    dprintf("MemoryManager: Free(%lx) LRU %ld Total %ld\n",(uint64_t)AccCache.AccPtr,DeviceLRUBytes,DeviceBytes);  
+    dprintf("MemoryManager: Free(%lx) LRU %ld Total %ld",(uint64_t)AccCache.AccPtr,DeviceLRUBytes,DeviceBytes);  
   }
   uint64_t CpuPtr = AccCache.CpuPtr;
   EntryErase(CpuPtr);
@@ -129,21 +149,21 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
 {
   ///////////////////////////////////////////////////////////////////////////
   // Make CPU consistent, remove from Accelerator, remove from LRU, LEAVE CPU only entry
-  // Cannot be acclocked. If allocated must be in LRU pool.
+  // Cannot be locked. If allocated must be in LRU pool.
   //
-  // Nov 2022... Felix issue: Allocating two CpuPtrs, can have an entry in LRU-q with CPUlock.
-  //                          and require to evict the AccPtr copy. Eviction was a mistake in CpuViewOpen
-  //                          but there is a weakness where CpuLock entries are attempted for erase
-  //                          Take these OUT LRU queue when CPU locked?
-  //                          Cannot take out the table as cpuLock data is important.
+  // (Historical: a Nov 2022 incident (two CpuPtrs; eviction called from
+  //  CpuViewOpen -- since excised) could present a cpuLocked entry here, and
+  //  silent-return guards were added.  The LRU membership invariant (see
+  //  LRUinsert) now excludes ALL locked entries from the queue eagerly at the
+  //  lock edges, so a locked victim is an invariant violation: asserted.)
   ///////////////////////////////////////////////////////////////////////////
-  assert(AccCache.state!=Empty);
+  GRID_ASSERT(AccCache.state!=Empty);
   
-  mprintf("MemoryManager: Evict CpuPtr %lx AccPtr %lx cpuLock %ld accLock %ld\n",
+  mprintf("MemoryManager: Evict CpuPtr %lx AccPtr %lx cpuLock %ld accLock %ld",
 	  (uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr,
 	  (uint64_t)AccCache.cpuLock,(uint64_t)AccCache.accLock); 
-  if (AccCache.accLock!=0) return;
-  if (AccCache.cpuLock!=0) return;
+  GRID_ASSERT(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.cpuLock==0);
   if(AccCache.state==AccDirty) {
     Flush(AccCache);
   }
@@ -153,7 +173,7 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
     AccCache.AccPtr=(uint64_t)NULL;
     AccCache.state=CpuDirty; // CPU primary now
     DeviceBytes   -=AccCache.bytes;
-    dprintf("MemoryManager: Free(AccPtr %lx) footprint now %ld \n",(uint64_t)AccCache.AccPtr,DeviceBytes);  
+    dprintf("MemoryManager: Free(AccPtr %lx) footprint now %ld ",(uint64_t)AccCache.AccPtr,DeviceBytes);  
   }
   //  uint64_t CpuPtr = AccCache.CpuPtr;
   DeviceEvictions++;
@@ -161,28 +181,28 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
 }
 void MemoryManager::Flush(AcceleratorViewEntry &AccCache)
 {
-  assert(AccCache.state==AccDirty);
-  assert(AccCache.cpuLock==0);
-  assert(AccCache.accLock==0);
-  assert(AccCache.AccPtr!=(uint64_t)NULL);
-  assert(AccCache.CpuPtr!=(uint64_t)NULL);
+  GRID_ASSERT(AccCache.state==AccDirty);
+  GRID_ASSERT(AccCache.cpuLock==0);
+  GRID_ASSERT(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.AccPtr!=(uint64_t)NULL);
+  GRID_ASSERT(AccCache.CpuPtr!=(uint64_t)NULL);
   acceleratorCopyFromDevice((void *)AccCache.AccPtr,(void *)AccCache.CpuPtr,AccCache.bytes);
-  mprintf("MemoryManager: acceleratorCopyFromDevice Flush size %ld AccPtr %lx -> CpuPtr %lx\n",(uint64_t)AccCache.bytes,(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
+  mprintf("MemoryManager: acceleratorCopyFromDevice Flush size %ld AccPtr %lx -> CpuPtr %lx",(uint64_t)AccCache.bytes,(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
   DeviceToHostBytes+=AccCache.bytes;
   DeviceToHostXfer++;
   AccCache.state=Consistent;
 }
 void MemoryManager::Clone(AcceleratorViewEntry &AccCache)
 {
-  assert(AccCache.state==CpuDirty);
-  assert(AccCache.cpuLock==0);
-  assert(AccCache.accLock==0);
-  assert(AccCache.CpuPtr!=(uint64_t)NULL);
+  GRID_ASSERT(AccCache.state==CpuDirty);
+  GRID_ASSERT(AccCache.cpuLock==0);
+  GRID_ASSERT(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.AccPtr==(uint64_t)NULL){
     AccCache.AccPtr=(uint64_t)AcceleratorAllocate(AccCache.bytes);
     DeviceBytes+=AccCache.bytes;
   }
-  mprintf("MemoryManager: acceleratorCopyToDevice   Clone size %ld AccPtr %lx <- CpuPtr %lx\n",
+  mprintf("MemoryManager: acceleratorCopyToDevice   Clone size %ld AccPtr %lx <- CpuPtr %lx",
 	  (uint64_t)AccCache.bytes,
 	  (uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
   acceleratorCopyToDevice((void *)AccCache.CpuPtr,(void *)AccCache.AccPtr,AccCache.bytes);
@@ -193,10 +213,10 @@ void MemoryManager::Clone(AcceleratorViewEntry &AccCache)
 
 void MemoryManager::CpuDiscard(AcceleratorViewEntry &AccCache)
 {
-  assert(AccCache.state!=Empty);
-  assert(AccCache.cpuLock==0);
-  assert(AccCache.accLock==0);
-  assert(AccCache.CpuPtr!=(uint64_t)NULL);
+  GRID_ASSERT(AccCache.state!=Empty);
+  GRID_ASSERT(AccCache.cpuLock==0);
+  GRID_ASSERT(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.AccPtr==(uint64_t)NULL){
     AccCache.AccPtr=(uint64_t)AcceleratorAllocate(AccCache.bytes);
     DeviceBytes+=AccCache.bytes;
@@ -210,33 +230,49 @@ void MemoryManager::CpuDiscard(AcceleratorViewEntry &AccCache)
 void MemoryManager::ViewClose(void* Ptr,ViewMode mode)
 {
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
-    dprintf("AcceleratorViewClose %lx\n",(uint64_t)Ptr);
+    dprintf("AcceleratorViewClose %lx",(uint64_t)Ptr);
     AcceleratorViewClose((uint64_t)Ptr);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     CpuViewClose((uint64_t)Ptr);
   } else { 
-    assert(0);
+    GRID_ASSERT(0);
   }
 }
 void *MemoryManager::ViewOpen(void* _CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint)
 {
   uint64_t CpuPtr = (uint64_t)_CpuPtr;
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
-    dprintf("AcceleratorViewOpen %lx\n",(uint64_t)CpuPtr);
+    dprintf("AcceleratorViewOpen %lx",(uint64_t)CpuPtr);
     return (void *) AcceleratorViewOpen(CpuPtr,bytes,mode,hint);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     return (void *)CpuViewOpen(CpuPtr,bytes,mode,hint);
   } else { 
-    assert(0);
+    GRID_ASSERT(0);
     return NULL;
   }
 }
 void  MemoryManager::EvictVictims(uint64_t bytes)
 {
-  assert(bytes<DeviceMaxBytes);
+  if(bytes>=DeviceMaxBytes) {
+    printf("EvictVictims bytes %ld DeviceMaxBytes %ld\n",bytes,DeviceMaxBytes);
+  }
+  GRID_ASSERT(bytes<DeviceMaxBytes);
   while(bytes+DeviceLRUBytes > DeviceMaxBytes){
     if ( DeviceLRUBytes > 0){
-      assert(LRU.size()>0);
+      GRID_ASSERT(LRU.size()>0);
+      uint64_t victim = LRU.back(); // From the LRU
+      auto AccCacheIterator = EntryLookup(victim);
+      auto & AccCache = AccCacheIterator->second;
+      Evict(AccCache);
+    } else {
+      return;
+    }
+  }
+}
+void  MemoryManager::EvictAll(void)
+{
+  while(LRU.size()>0){
+    if ( DeviceLRUBytes > 0){
       uint64_t victim = LRU.back(); // From the LRU
       auto AccCacheIterator = EntryLookup(victim);
       auto & AccCache = AccCacheIterator->second;
@@ -260,19 +296,19 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
   if (!AccCache.AccPtr) {
     EvictVictims(bytes); 
   } 
-  assert((mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard));
+  GRID_ASSERT((mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard));
 
-  assert(AccCache.cpuLock==0);  // Programming error
+  GRID_ASSERT(AccCache.cpuLock==0);  // Programming error
 
   if(AccCache.state!=Empty) {
-    dprintf("ViewOpen found entry %lx %lx : sizes %ld %ld accLock %ld\n",
+    dprintf("ViewOpen found entry %lx %lx : sizes %ld %ld accLock %ld",
 		    (uint64_t)AccCache.CpuPtr,
 		    (uint64_t)CpuPtr,
 		    (uint64_t)AccCache.bytes,
 	            (uint64_t)bytes,
 		    (uint64_t)AccCache.accLock);
-    assert(AccCache.CpuPtr == CpuPtr);
-    assert(AccCache.bytes  ==bytes);
+    GRID_ASSERT(AccCache.CpuPtr == CpuPtr);
+    GRID_ASSERT(AccCache.bytes  ==bytes);
   }
 /*
  *  State transitions and actions
@@ -289,7 +325,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
  *  AccWrite AccDirty   AccDirty       -        - 
  */
   if(AccCache.state==Empty) {
-    assert(AccCache.LRU_valid==0);
+    GRID_ASSERT(AccCache.LRU_valid==0);
     AccCache.CpuPtr = CpuPtr;
     AccCache.AccPtr = (uint64_t)NULL;
     AccCache.bytes  = bytes;
@@ -305,7 +341,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // Empty + AccRead => Consistent
     }
     AccCache.accLock= 1;
-    dprintf("Copied Empty entry into device accLock= %d\n",AccCache.accLock);
+    dprintf("Copied Empty entry into device accLock= %d",AccCache.accLock);
   } else if(AccCache.state==CpuDirty ){
     if(mode==AcceleratorWriteDiscard) {
       CpuDiscard(AccCache);
@@ -318,30 +354,30 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // CpuDirty + AccRead => Consistent
     }
     AccCache.accLock++;
-    dprintf("CpuDirty entry into device ++accLock= %d\n",AccCache.accLock);
+    dprintf("CpuDirty entry into device ++accLock= %d",AccCache.accLock);
   } else if(AccCache.state==Consistent) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty;   // Consistent + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = Consistent; // Consistent + AccRead => Consistent
     AccCache.accLock++;
-    dprintf("Consistent entry into device ++accLock= %d\n",AccCache.accLock);
+    dprintf("Consistent entry into device ++accLock= %d",AccCache.accLock);
   } else if(AccCache.state==AccDirty) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty; // AccDirty + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = AccDirty; // AccDirty + AccRead => AccDirty
     AccCache.accLock++;
-    dprintf("AccDirty entry ++accLock= %d\n",AccCache.accLock);
+    dprintf("AccDirty entry ++accLock= %d",AccCache.accLock);
   } else {
-    assert(0);
+    GRID_ASSERT(0);
   }
 
-  assert(AccCache.accLock>0);
+  GRID_ASSERT(AccCache.accLock>0);
   // If view is opened on device must remove from LRU
   if(AccCache.LRU_valid==1){
     // must possibly remove from LRU as now locked on GPU
-    dprintf("AccCache entry removed from LRU \n");
+    dprintf("AccCache entry removed from LRU ");
     LRUremove(AccCache);
   }
 
@@ -358,16 +394,16 @@ void MemoryManager::AcceleratorViewClose(uint64_t CpuPtr)
   auto AccCacheIterator = EntryLookup(CpuPtr);
   auto & AccCache = AccCacheIterator->second;
 
-  assert(AccCache.cpuLock==0);
-  assert(AccCache.accLock>0);
+  GRID_ASSERT(AccCache.cpuLock==0);
+  GRID_ASSERT(AccCache.accLock>0);
 
   AccCache.accLock--;
   // Move to LRU queue if not locked and close on device
   if(AccCache.accLock==0) {
-    dprintf("AccleratorViewClose %lx AccLock decremented to %ld move to LRU queue\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld move to LRU queue",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
     LRUinsert(AccCache);
   } else {
-    dprintf("AccleratorViewClose %lx AccLock decremented to %ld\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
   }
 }
 void MemoryManager::CpuViewClose(uint64_t CpuPtr)
@@ -375,10 +411,17 @@ void MemoryManager::CpuViewClose(uint64_t CpuPtr)
   auto AccCacheIterator = EntryLookup(CpuPtr);
   auto & AccCache = AccCacheIterator->second;
 
-  assert(AccCache.cpuLock>0);
-  assert(AccCache.accLock==0);
+  GRID_ASSERT(AccCache.cpuLock>0);
+  GRID_ASSERT(AccCache.accLock==0);
 
   AccCache.cpuLock--;
+  // Return to LRU queue when fully unlocked -- mirrors AcceleratorViewClose.
+  // Asymmetry vs the Acc side: a device copy need not exist for a host view;
+  // only device-resident entries belong in the (evictable) LRU queue.
+  if( (AccCache.cpuLock==0) && (AccCache.AccPtr!=(uint64_t)NULL) ) {
+    dprintf("CpuViewClose %lx cpuLock decremented to zero, move to LRU queue",(uint64_t)CpuPtr);
+    LRUinsert(AccCache);
+  }
 }
 /*
  *  Action  State   StateNext         Flush    Clone
@@ -409,12 +452,12 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V
   //    EvictVictims(bytes);
   //  }
 
-  assert((mode==CpuRead)||(mode==CpuWrite));
-  assert(AccCache.accLock==0);  // Programming error
+  GRID_ASSERT((mode==CpuRead)||(mode==CpuWrite));
+  GRID_ASSERT(AccCache.accLock==0);  // Programming error
 
   if(AccCache.state!=Empty) {
-    assert(AccCache.CpuPtr == CpuPtr);
-    assert(AccCache.bytes==bytes);
+    GRID_ASSERT(AccCache.CpuPtr == CpuPtr);
+    GRID_ASSERT(AccCache.bytes==bytes);
   }
 
   if(AccCache.state==Empty) {
@@ -429,20 +472,28 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V
     AccCache.state = CpuDirty; // CpuDirty +CpuRead/CpuWrite => CpuDirty
     AccCache.cpuLock++;
   } else if(AccCache.state==Consistent) {
-    assert(AccCache.AccPtr != (uint64_t)NULL);
+    GRID_ASSERT(AccCache.AccPtr != (uint64_t)NULL);
     if(mode==CpuWrite)
       AccCache.state = CpuDirty;   // Consistent +CpuWrite => CpuDirty
     else 
       AccCache.state = Consistent; // Consistent +CpuRead  => Consistent
     AccCache.cpuLock++;
   } else if(AccCache.state==AccDirty) {
-    assert(AccCache.AccPtr != (uint64_t)NULL);
+    GRID_ASSERT(AccCache.AccPtr != (uint64_t)NULL);
     Flush(AccCache);
     if(mode==CpuWrite) AccCache.state = CpuDirty;   // AccDirty +CpuWrite => CpuDirty, Flush
     else            AccCache.state = Consistent; // AccDirty +CpuRead  => Consistent, Flush
     AccCache.cpuLock++;
   } else {
-    assert(0); // should be unreachable
+    GRID_ASSERT(0); // should be unreachable
+  }
+
+  GRID_ASSERT(AccCache.cpuLock>0);
+  // If view is opened on host must remove from LRU -- mirrors AcceleratorViewOpen.
+  // LRU_valid==1 here implies this is the 0->1 lock edge of a device-resident entry.
+  if(AccCache.LRU_valid==1){
+    dprintf("CpuViewOpen: entry removed from LRU ");
+    LRUremove(AccCache);
   }
 
   AccCache.transient= transient? EvictNext : 0;
@@ -524,12 +575,12 @@ void MemoryManager::Audit(std::string s)
   std::cout << " Memory Manager::Audit() from "<<s<<std::endl;
   for(auto it=LRU.begin();it!=LRU.end();it++){
     uint64_t cpuPtr = *it;
-    assert(EntryPresent(cpuPtr));
+    GRID_ASSERT(EntryPresent(cpuPtr));
     auto AccCacheIterator = EntryLookup(cpuPtr);
     auto & AccCache = AccCacheIterator->second;
     LruBytes2+=AccCache.bytes;
-    assert(AccCache.LRU_valid==1);
-    assert(AccCache.LRU_entry==it);
+    GRID_ASSERT(AccCache.LRU_valid==1);
+    GRID_ASSERT(AccCache.LRU_entry==it);
   }
   std::cout << " Memory Manager::Audit() LRU queue matches table entries "<<std::endl;
 
@@ -548,7 +599,7 @@ void MemoryManager::Audit(std::string s)
     if( AccCache.LRU_valid ) LruCnt++;
     
     if ( AccCache.cpuLock || AccCache.accLock ) {
-      assert(AccCache.LRU_valid==0);
+      GRID_ASSERT(AccCache.LRU_valid==0);
 
       std::cout << GridLogError << s<< "\n\t 0x"<<std::hex<<AccCache.CpuPtr<<std::dec
 		<< "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
@@ -557,16 +608,16 @@ void MemoryManager::Audit(std::string s)
 		<< "\t LRUvalid " << AccCache.LRU_valid<<std::endl;
     }
 
-    assert( AccCache.cpuLock== 0 ) ;
-    assert( AccCache.accLock== 0 ) ;
+    GRID_ASSERT( AccCache.cpuLock== 0 ) ;
+    GRID_ASSERT( AccCache.accLock== 0 ) ;
   }
   std::cout << " Memory Manager::Audit() no locked table entries "<<std::endl;
-  assert(LruBytes1==LruBytes2);
-  assert(LruBytes1==DeviceLRUBytes);
+  GRID_ASSERT(LruBytes1==LruBytes2);
+  GRID_ASSERT(LruBytes1==DeviceLRUBytes);
   std::cout << " Memory Manager::Audit() evictable bytes matches sum over table "<<std::endl;
-  assert(AccBytes==DeviceBytes);
+  GRID_ASSERT(AccBytes==DeviceBytes);
   std::cout << " Memory Manager::Audit() device bytes matches sum over table "<<std::endl;
-  assert(LruCnt == LRU.size());
+  GRID_ASSERT(LruCnt == LRU.size());
   std::cout << " Memory Manager::Audit() LRU entry count matches "<<std::endl;
 
 }

Grid/allocator/MemoryManagerShared.cc

@@ -24,6 +24,7 @@ void  MemoryManager::PrintState(void* CpuPtr)
 std::cout << GridLogMessage << "Host<->Device memory movement not currently managed by Grid." << std::endl;
 };
 void  MemoryManager::Print(void){};
+void  MemoryManager::EvictAll(void){};   // nothing to evict in unified memory
 void  MemoryManager::PrintAll(void){};
 void  MemoryManager::NotifyDeletion(void *ptr){};
 

Grid/allocator/MemoryStats.cc

@@ -10,16 +10,16 @@ void check_huge_pages(void *Buf,uint64_t BYTES)
 {
 #ifdef __linux__
   int fd = open("/proc/self/pagemap", O_RDONLY);
-  assert(fd >= 0);
+  GRID_ASSERT(fd >= 0);
   const int page_size = 4096;
   uint64_t virt_pfn = (uint64_t)Buf / page_size;
   off_t offset = sizeof(uint64_t) * virt_pfn;
   uint64_t npages = (BYTES + page_size-1) / page_size;
   std::vector<uint64_t> pagedata(npages);
   uint64_t ret = lseek(fd, offset, SEEK_SET);
-  assert(ret == offset);
+  GRID_ASSERT(ret == offset);
   ret = ::read(fd, &pagedata[0], sizeof(uint64_t)*npages);
-  assert(ret == sizeof(uint64_t) * npages);
+  GRID_ASSERT(ret == sizeof(uint64_t) * npages);
   int nhugepages = npages / 512;
   int n4ktotal, nnothuge;
   n4ktotal = 0;

Grid/cartesian/Cartesian_base.h

@@ -165,7 +165,7 @@ public:
     //
     if ( _simd_layout[dimension] > 2 ) { 
       for(int d=0;d<_ndimension;d++){
-	if ( d != dimension ) assert ( (_simd_layout[d]==1)  );
+	if ( d != dimension ) GRID_ASSERT ( (_simd_layout[d]==1)  );
       }
       permute_type = RotateBit; // How to specify distance; this is not just direction.
       return permute_type;
@@ -187,7 +187,7 @@ public:
   inline int64_t gSites(void) const { return (int64_t)_isites*(int64_t)_osites*(int64_t)_Nprocessors; }; 
   inline int Nd    (void) const { return _ndimension;};
 
-  inline const Coordinate LocalStarts(void)             { return _lstart;    };
+  inline const Coordinate &LocalStarts(void)            { return _lstart;    };
   inline const Coordinate &FullDimensions(void)         { return _fdimensions;};
   inline const Coordinate &GlobalDimensions(void)       { return _gdimensions;};
   inline const Coordinate &LocalDimensions(void)        { return _ldimensions;};
@@ -216,11 +216,11 @@ public:
   // Global addressing
   ////////////////////////////////////////////////////////////////
   void GlobalIndexToGlobalCoor(int64_t gidx,Coordinate &gcoor){
-    assert(gidx< gSites());
+    GRID_ASSERT(gidx< gSites());
     Lexicographic::CoorFromIndex(gcoor,gidx,_gdimensions);
   }
   void LocalIndexToLocalCoor(int lidx,Coordinate &lcoor){
-    assert(lidx<lSites());
+    GRID_ASSERT(lidx<lSites());
     Lexicographic::CoorFromIndex(lcoor,lidx,_ldimensions);
   }
   void GlobalCoorToGlobalIndex(const Coordinate & gcoor,int64_t & gidx){

Grid/cartesian/Cartesian_full.h

@@ -128,10 +128,10 @@ public:
         // Use a reduced simd grid
         _ldimensions[d] = _gdimensions[d] / _processors[d]; //local dimensions
         //std::cout << _ldimensions[d] << "  " << _gdimensions[d] << "  " << _processors[d] << std::endl;
-        assert(_ldimensions[d] * _processors[d] == _gdimensions[d]);
+        GRID_ASSERT(_ldimensions[d] * _processors[d] == _gdimensions[d]);
 
         _rdimensions[d] = _ldimensions[d] / _simd_layout[d]; //overdecomposition
-        assert(_rdimensions[d] * _simd_layout[d] == _ldimensions[d]);
+        GRID_ASSERT(_rdimensions[d] * _simd_layout[d] == _ldimensions[d]);
 
         _lstart[d] = _processor_coor[d] * _ldimensions[d];
         _lend[d] = _processor_coor[d] * _ldimensions[d] + _ldimensions[d] - 1;

Grid/cartesian/Cartesian_red_black.h

@@ -67,7 +67,7 @@ public:
   }
   virtual int CheckerBoard(const Coordinate &site){
     int linear=0;
-    assert(site.size()==_ndimension);
+    GRID_ASSERT(site.size()==_ndimension);
     for(int d=0;d<_ndimension;d++){ 
       if(_checker_dim_mask[d])
 	linear=linear+site[d];
@@ -160,11 +160,11 @@ public:
 
       _isCheckerBoarded = true;
     _checker_dim = checker_dim;
-    assert(checker_dim_mask[checker_dim] == 1);
+    GRID_ASSERT(checker_dim_mask[checker_dim] == 1);
     _ndimension = dimensions.size();
-    assert(checker_dim_mask.size() == _ndimension);
-    assert(processor_grid.size() == _ndimension);
-    assert(simd_layout.size() == _ndimension);
+    GRID_ASSERT(checker_dim_mask.size() == _ndimension);
+    GRID_ASSERT(processor_grid.size() == _ndimension);
+    GRID_ASSERT(simd_layout.size() == _ndimension);
 
     _fdimensions.resize(_ndimension);
     _gdimensions.resize(_ndimension);
@@ -190,20 +190,20 @@ public:
 
         if (d == _checker_dim)
 	  {
-	    assert((_gdimensions[d] & 0x1) == 0);
+	    GRID_ASSERT((_gdimensions[d] & 0x1) == 0);
 	    _gdimensions[d] = _gdimensions[d] / 2; // Remove a checkerboard
 	    _gsites /= 2;
 	  }
         _ldimensions[d] = _gdimensions[d] / _processors[d];
-        assert(_ldimensions[d] * _processors[d] == _gdimensions[d]);
+        GRID_ASSERT(_ldimensions[d] * _processors[d] == _gdimensions[d]);
         _lstart[d] = _processor_coor[d] * _ldimensions[d];
         _lend[d] = _processor_coor[d] * _ldimensions[d] + _ldimensions[d] - 1;
 
         // Use a reduced simd grid
         _simd_layout[d] = simd_layout[d];
         _rdimensions[d] = _ldimensions[d] / _simd_layout[d]; // this is not checking if this is integer
-        assert(_rdimensions[d] * _simd_layout[d] == _ldimensions[d]);
-        assert(_rdimensions[d] > 0);
+        GRID_ASSERT(_rdimensions[d] * _simd_layout[d] == _ldimensions[d]);
+        GRID_ASSERT(_rdimensions[d] > 0);
 
         // all elements of a simd vector must have same checkerboard.
         // If Ls vectorised, this must still be the case; e.g. dwf rb5d

Grid/communicator/Communicator_base.cc

@@ -34,6 +34,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 NAMESPACE_BEGIN(Grid);
 
 bool Stencil_force_mpi = true;
+bool Stencil_force_barrier = false;
 
 ///////////////////////////////////////////////////////////////
 // Info that is setup once and indept of cartesian layout

Grid/communicator/Communicator_base.h

@@ -33,9 +33,12 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 ///////////////////////////////////
 #include <Grid/communicator/SharedMemory.h>
 
+#define NVLINK_GET
+
 NAMESPACE_BEGIN(Grid);
 
 extern bool Stencil_force_mpi ;
+extern bool Stencil_force_barrier ;
 
 class CartesianCommunicator : public SharedMemory {
 
@@ -106,7 +109,7 @@ public:
   // very VERY rarely (Log, serial RNG) we need world without a grid
   ////////////////////////////////////////////////////////////////////////////////
   static int  RankWorld(void) ;
-  static void BroadcastWorld(int root,void* data, int bytes);
+  static void BroadcastWorld(int root,void* data, uint64_t bytes);
   static void BarrierWorld(void);
   
   ////////////////////////////////////////////////////////////
@@ -147,7 +150,8 @@ public:
 			    sizeof(obj),d*100+p);
 
       }
-      CommsComplete(list);
+      if (!list.empty()) // avoid triggering GRID_ASSERT in comms == none
+	CommsComplete(list);
       for(int p=1;p<_processors[d];p++){
 	accum = accum + column[p];
       }
@@ -172,32 +176,38 @@ public:
 			   int dest,
 			   void *recv,
 			   int from,
-			   int bytes,int dir);
+			   uint64_t bytes,int dir);
   
   void SendToRecvFrom(void *xmit,
 		      int xmit_to_rank,
 		      void *recv,
 		      int recv_from_rank,
-		      int bytes);
+		      uint64_t bytes);
   
+  int IsOffNode(int rank);
   double StencilSendToRecvFrom(void *xmit,
 			       int xmit_to_rank,int do_xmit,
 			       void *recv,
 			       int recv_from_rank,int do_recv,
-			       int bytes,int dir);
+			       uint64_t bytes,int dir);
 
   double StencilSendToRecvFromPrepare(std::vector<CommsRequest_t> &list,
 				      void *xmit,
 				      int xmit_to_rank,int do_xmit,
 				      void *recv,
 				      int recv_from_rank,int do_recv,
-				      int xbytes,int rbytes,int dir);
+				      uint64_t xbytes,uint64_t rbytes,int dir);
+
+  // Could do a PollHtoD and have a CommsMerge dependence
+  void StencilSendToRecvFromPollDtoH (std::vector<CommsRequest_t> &list);
+  void StencilSendToRecvFromPollIRecv(std::vector<CommsRequest_t> &list);
+
   double StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
-				    void *xmit,
+				    void *xmit,void *xmit_comp,
 				    int xmit_to_rank,int do_xmit,
-				    void *recv,
+				    void *recv,void *recv_comp,
 				    int recv_from_rank,int do_recv,
-				    int xbytes,int rbytes,int dir);
+				    uint64_t xbytes,uint64_t rbytes,int dir);
   
   
   void StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall,int i);
@@ -211,24 +221,51 @@ public:
   ////////////////////////////////////////////////////////////
   // Broadcast a buffer and composite larger
   ////////////////////////////////////////////////////////////
-  void Broadcast(int root,void* data, int bytes);
+  void Broadcast(int root,void* data, uint64_t bytes);
 
   ////////////////////////////////////////////////////////////
   // All2All down one dimension
   ////////////////////////////////////////////////////////////
   template<class T> void AllToAll(int dim,std::vector<T> &in, std::vector<T> &out){
-    assert(dim>=0);
-    assert(dim<_ndimension);
-    assert(in.size()==out.size());
+    GRID_ASSERT(dim>=0);
+    GRID_ASSERT(dim<_ndimension);
+    GRID_ASSERT(in.size()==out.size());
     int numnode = _processors[dim];
     uint64_t bytes=sizeof(T);
     uint64_t words=in.size()/numnode;
-    assert(numnode * words == in.size());
-    assert(words < (1ULL<<31));
+    GRID_ASSERT(numnode * words == in.size());
+    GRID_ASSERT(words < (1ULL<<31));
     AllToAll(dim,(void *)&in[0],(void *)&out[0],words,bytes);
   }
   void AllToAll(int dim  ,void *in,void *out,uint64_t words,uint64_t bytes);
   void AllToAll(void  *in,void *out,uint64_t words         ,uint64_t bytes);
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Variable count all to all. Counts and displacements are in units of
+  // "bytes" sized words and are indexed by rank within this communicator.
+  // For exchanges that are a permutation but do not divide evenly between
+  // ranks; AllToAll above is the uniform count special case.
+  ////////////////////////////////////////////////////////////////////////////
+  void AllToAllV(void *in ,const std::vector<int> &sendcounts,const std::vector<int> &senddispls,
+                 void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                 uint64_t bytes);
+
+  ////////////////////////////////////////////////////////////////////////////
+  // Gather to all. Every rank contributes sendcount words from "in"; every
+  // rank receives the concatenation of all contributions in rank order.
+  // Counts and displacements are in units of "bytes" sized words and are
+  // indexed by rank within this communicator; recvcounts[ThisRank()] must
+  // equal sendcount.  AllGather below is the uniform count special case.
+  //
+  // This is the primitive form of the zero-fill + GlobalSumVector idiom used
+  // wherever each element of the result has exactly one contributing rank.
+  // That idiom moves the payload twice and reduces over zeros; this moves it
+  // once and performs no arithmetic.
+  ////////////////////////////////////////////////////////////////////////////
+  void AllGatherV(void *in ,int sendcount,
+                  void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                  uint64_t bytes);
+  void AllGather (void *in ,void *out,uint64_t words,uint64_t bytes);
   
   template<class obj> void Broadcast(int root,obj &data)
   {

Grid/communicator/Communicator_mpi3.cc

@@ -28,9 +28,17 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/GridCore.h>
 #include <Grid/communicator/SharedMemory.h>
 
+void GridAbort(void) { MPI_Abort(MPI_COMM_WORLD,SIGABRT); }
+extern void * Grid_backtrace_buffer[_NBACKTRACE];
+
 NAMESPACE_BEGIN(Grid);
 
+
 Grid_MPI_Comm       CartesianCommunicator::communicator_world;
+#ifdef GRID_CHECKSUM_COMMS
+uint64_t checksum_index = 1;
+#endif
+
 
 ////////////////////////////////////////////
 // First initialise of comms system
@@ -55,11 +63,11 @@ void CartesianCommunicator::Init(int *argc, char ***argv)
 #endif
     //If only 1 comms thread we require any threading mode other than SINGLE, but for multiple comms threads we need MULTIPLE
     if( (nCommThreads == 1) && (provided == MPI_THREAD_SINGLE) ) {
-      assert(0);
+      GRID_ASSERT(0);
     }
 
     if( (nCommThreads > 1) && (provided != MPI_THREAD_MULTIPLE) ) {
-      assert(0);
+      GRID_ASSERT(0);
     }
   }
 
@@ -80,20 +88,20 @@ void CartesianCommunicator::Init(int *argc, char ***argv)
 void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest)
 {
   int ierr=MPI_Cart_shift(communicator,dim,shift,&source,&dest);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 int CartesianCommunicator::RankFromProcessorCoor(Coordinate &coor)
 {
   int rank;
   int ierr=MPI_Cart_rank  (communicator, &coor[0], &rank);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
   return rank;
 }
 void  CartesianCommunicator::ProcessorCoorFromRank(int rank, Coordinate &coor)
 {
   coor.resize(_ndimension);
   int ierr=MPI_Cart_coords  (communicator, rank, _ndimension,&coor[0]);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -120,8 +128,8 @@ CartesianCommunicator::CartesianCommunicator(const Coordinate &processors)
 //////////////////////////////////
 CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const CartesianCommunicator &parent,int &srank)
 {
-  _ndimension = processors.size();  assert(_ndimension>=1);
-  int parent_ndimension = parent._ndimension; assert(_ndimension >= parent._ndimension);
+  _ndimension = processors.size();  GRID_ASSERT(_ndimension>=1);
+  int parent_ndimension = parent._ndimension; GRID_ASSERT(_ndimension >= parent._ndimension);
   Coordinate parent_processor_coor(_ndimension,0);
   Coordinate parent_processors    (_ndimension,1);
   Coordinate shm_processors       (_ndimension,1);
@@ -145,7 +153,7 @@ CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const
     childsize *= processors[d];
   }
   int Nchild = Nparent/childsize;
-  assert (childsize * Nchild == Nparent);
+  GRID_ASSERT (childsize * Nchild == Nparent);
 
   Coordinate ccoor(_ndimension); // coor within subcommunicator
   Coordinate scoor(_ndimension); // coor of split within parent
@@ -171,12 +179,12 @@ CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const
     // Split the communicator
     ////////////////////////////////////////////////////////////////
     int ierr= MPI_Comm_split(parent.communicator,srank,crank,&comm_split);
-    assert(ierr==0);
+    GRID_ASSERT(ierr==0);
 
   } else {
     srank = 0;
     int ierr = MPI_Comm_dup (parent.communicator,&comm_split);
-    assert(ierr==0);
+    GRID_ASSERT(ierr==0);
   }
 
   //////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -201,7 +209,7 @@ CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const
     }
   }
   for(int d=0;d<processors.size();d++){
-    assert(_processor_coor[d] == ccoor[d] );
+    GRID_ASSERT(_processor_coor[d] == ccoor[d] );
   }
 }
 
@@ -243,7 +251,7 @@ void CartesianCommunicator::InitFromMPICommunicator(const Coordinate &processors
   for(int i=0;i<_ndimension*2;i++){
     MPI_Comm_dup(communicator,&communicator_halo[i]);
   }
-  assert(Size==_Nprocessors);
+  GRID_ASSERT(Size==_Nprocessors);
 }
 
 CartesianCommunicator::~CartesianCommunicator()
@@ -259,62 +267,74 @@ CartesianCommunicator::~CartesianCommunicator()
 }
 #ifdef USE_GRID_REDUCTION
 void CartesianCommunicator::GlobalSum(float &f){
+  FlightRecorder::StepLog("GlobalSumP2P");
   CartesianCommunicator::GlobalSumP2P(f);
 }
 void CartesianCommunicator::GlobalSum(double &d)
 {
+  FlightRecorder::StepLog("GlobalSumP2P");
   CartesianCommunicator::GlobalSumP2P(d);
 }
 #else
 void CartesianCommunicator::GlobalSum(float &f){
+  FlightRecorder::StepLog("AllReduce float");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&f,1,MPI_FLOAT,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalSum(double &d)
 {
+  FlightRecorder::StepLog("AllReduce double");
   int ierr = MPI_Allreduce(MPI_IN_PLACE,&d,1,MPI_DOUBLE,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 #endif
 void CartesianCommunicator::GlobalSum(uint32_t &u){
+  FlightRecorder::StepLog("AllReduce uint32_t");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT32_T,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalSum(uint64_t &u){
+  FlightRecorder::StepLog("AllReduce uint64_t");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalSumVector(uint64_t* u,int N){
+  FlightRecorder::StepLog("AllReduceVector");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,u,N,MPI_UINT64_T,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalXOR(uint32_t &u){
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT32_T,MPI_BXOR,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalXOR(uint64_t &u){
+  FlightRecorder::StepLog("GlobalXOR");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_BXOR,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalMax(float &f)
 {
+  FlightRecorder::StepLog("GlobalMax");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&f,1,MPI_FLOAT,MPI_MAX,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalMax(double &d)
 {
+  FlightRecorder::StepLog("GlobalMax");
   int ierr = MPI_Allreduce(MPI_IN_PLACE,&d,1,MPI_DOUBLE,MPI_MAX,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalSumVector(float *f,int N)
 {
+  FlightRecorder::StepLog("GlobalSumVector(float *)");
   int ierr=MPI_Allreduce(MPI_IN_PLACE,f,N,MPI_FLOAT,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void CartesianCommunicator::GlobalSumVector(double *d,int N)
 {
+  FlightRecorder::StepLog("GlobalSumVector(double *)");
   int ierr = MPI_Allreduce(MPI_IN_PLACE,d,N,MPI_DOUBLE,MPI_SUM,communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 
 void CartesianCommunicator::SendToRecvFromBegin(std::vector<MpiCommsRequest_t> &list,
@@ -322,24 +342,23 @@ void CartesianCommunicator::SendToRecvFromBegin(std::vector<MpiCommsRequest_t> &
 						int dest,
 						void *recv,
 						int from,
-						int bytes,int dir)
+						uint64_t bytes,int dir)
 {
   MPI_Request xrq;
   MPI_Request rrq;
 
-  assert(dest != _processor);
-  assert(from != _processor);
-
+  GRID_ASSERT(dest != _processor);
+  GRID_ASSERT(from != _processor);
   int tag;
 
   tag= dir+from*32;
-  int ierr=MPI_Irecv(recv, bytes, MPI_CHAR,from,tag,communicator,&rrq);
-  assert(ierr==0);
+  int ierr=MPI_Irecv(recv,(int)( bytes/sizeof(int32_t)), MPI_INT32_T,from,tag,communicator,&rrq);
+  GRID_ASSERT(ierr==0);
   list.push_back(rrq);
   
   tag= dir+_processor*32;
-  ierr =MPI_Isend(xmit, bytes, MPI_CHAR,dest,tag,communicator,&xrq);
-  assert(ierr==0);
+  ierr =MPI_Isend(xmit,(int)(bytes/sizeof(int32_t)), MPI_INT32_T,dest,tag,communicator,&xrq);
+  GRID_ASSERT(ierr==0);
   list.push_back(xrq);
 }
 void CartesianCommunicator::CommsComplete(std::vector<MpiCommsRequest_t> &list)
@@ -350,7 +369,7 @@ void CartesianCommunicator::CommsComplete(std::vector<MpiCommsRequest_t> &list)
 
   std::vector<MPI_Status> status(nreq);
   int ierr = MPI_Waitall(nreq,&list[0],&status[0]);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
   list.resize(0);
 }
 
@@ -359,61 +378,63 @@ void CartesianCommunicator::SendToRecvFrom(void *xmit,
 					   int dest,
 					   void *recv,
 					   int from,
-					   int bytes)
+					   uint64_t bytes)
 {
   std::vector<MpiCommsRequest_t> reqs(0);
-  unsigned long  xcrc = crc32(0L, Z_NULL, 0);
-  unsigned long  rcrc = crc32(0L, Z_NULL, 0);
 
   int myrank = _processor;
   int ierr;
 
   // Enforce no UVM in comms, device or host OK
-  assert(acceleratorIsCommunicable(xmit));
-  assert(acceleratorIsCommunicable(recv));
+  GRID_ASSERT(acceleratorIsCommunicable(xmit));
+  GRID_ASSERT(acceleratorIsCommunicable(recv));
 
   // Give the CPU to MPI immediately; can use threads to overlap optionally
   //  printf("proc %d SendToRecvFrom %d bytes Sendrecv \n",_processor,bytes);
-  ierr=MPI_Sendrecv(xmit,bytes,MPI_CHAR,dest,myrank,
-		    recv,bytes,MPI_CHAR,from, from,
+  ierr=MPI_Sendrecv(xmit,(int)(bytes/sizeof(int32_t)),MPI_INT32_T,dest,myrank,
+		    recv,(int)(bytes/sizeof(int32_t)),MPI_INT32_T,from, from,
 		    communicator,MPI_STATUS_IGNORE);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 
-  //  xcrc = crc32(xcrc,(unsigned char *)xmit,bytes);
-  //  rcrc = crc32(rcrc,(unsigned char *)recv,bytes);
-  //  printf("proc %d SendToRecvFrom %d bytes xcrc %lx rcrc %lx\n",_processor,bytes,xcrc,rcrc); fflush
 }
 // Basic Halo comms primitive
 double CartesianCommunicator::StencilSendToRecvFrom( void *xmit,
 						     int dest, int dox,
 						     void *recv,
 						     int from, int dor,
-						     int bytes,int dir)
+						     uint64_t bytes,int dir)
 {
   std::vector<CommsRequest_t> list;
   double offbytes = StencilSendToRecvFromPrepare(list,xmit,dest,dox,recv,from,dor,bytes,bytes,dir);
-  offbytes       += StencilSendToRecvFromBegin(list,xmit,dest,dox,recv,from,dor,bytes,bytes,dir);
+  offbytes       += StencilSendToRecvFromBegin(list,xmit,xmit,dest,dox,recv,recv,from,dor,bytes,bytes,dir);
   StencilSendToRecvFromComplete(list,dir);
   return offbytes;
 }
-
+int CartesianCommunicator::IsOffNode(int rank)
+{
+  int grank = ShmRanks[rank];
+  if ( grank == MPI_UNDEFINED ) return true;
+  else return false;
+}
 
 #ifdef ACCELERATOR_AWARE_MPI
+void CartesianCommunicator::StencilSendToRecvFromPollIRecv(std::vector<CommsRequest_t> &list) {};
+void CartesianCommunicator::StencilSendToRecvFromPollDtoH(std::vector<CommsRequest_t> &list) {};
 double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequest_t> &list,
 							   void *xmit,
 							   int dest,int dox,
 							   void *recv,
 							   int from,int dor,
-							   int xbytes,int rbytes,int dir)
+							   uint64_t xbytes,uint64_t rbytes,int dir)
 {
   return 0.0; // Do nothing -- no preparation required
 }
 double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
-							 void *xmit,
+							 void *xmit,void *xmit_comp,
 							 int dest,int dox,
-							 void *recv,
+							 void *recv,void *recv_comp,
 							 int from,int dor,
-							 int xbytes,int rbytes,int dir)
+							 uint64_t xbytes,uint64_t rbytes,int dir)
 {
   int ncomm  =communicator_halo.size();
   int commdir=dir%ncomm;
@@ -426,33 +447,44 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
   int gfrom = ShmRanks[from];
   int gme   = ShmRanks[_processor];
 
-  assert(dest != _processor);
-  assert(from != _processor);
-  assert(gme  == ShmRank);
+  GRID_ASSERT(dest != _processor);
+  GRID_ASSERT(from != _processor);
+  GRID_ASSERT(gme  == ShmRank);
   double off_node_bytes=0.0;
   int tag;
   
   if ( dor ) {
     if ( (gfrom ==MPI_UNDEFINED) || Stencil_force_mpi ) {
       tag= dir+from*32;
-      ierr=MPI_Irecv(recv, rbytes, MPI_CHAR,from,tag,communicator_halo[commdir],&rrq);
-      assert(ierr==0);
+      //      std::cout << " StencilSendToRecvFrom "<<dir<<" MPI_Irecv "<<std::hex<<recv<<std::dec<<std::endl;
+      ierr=MPI_Irecv(recv_comp,(int)(rbytes/sizeof(int32_t)), MPI_INT32_T,from,tag,communicator_halo[commdir],&rrq);
+      GRID_ASSERT(ierr==0);
       list.push_back(rrq);
       off_node_bytes+=rbytes;
     }
+#ifdef NVLINK_GET
+    else { 
+      void *shm = (void *) this->ShmBufferTranslate(from,xmit);
+      GRID_ASSERT(shm!=NULL);
+      //      std::cout << " StencilSendToRecvFrom "<<dir<<" CopyDeviceToDevice recv "<<std::hex<<recv<<" remote "<<shm <<std::dec<<std::endl;
+      acceleratorCopyDeviceToDeviceAsynch(shm,recv,rbytes);
+    }
+#endif
   }
-  
+  // This is a NVLINK PUT  
   if (dox) {
     if ( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) {
       tag= dir+_processor*32;
-      ierr =MPI_Isend(xmit, xbytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
-      assert(ierr==0);
+      ierr =MPI_Isend(xmit_comp,(int)(xbytes/sizeof(int32_t)), MPI_INT32_T,dest,tag,communicator_halo[commdir],&xrq);
+      GRID_ASSERT(ierr==0);
       list.push_back(xrq);
       off_node_bytes+=xbytes;
     } else {
+#ifndef NVLINK_GET
       void *shm = (void *) this->ShmBufferTranslate(dest,recv);
-      assert(shm!=NULL);
+      GRID_ASSERT(shm!=NULL);
       acceleratorCopyDeviceToDeviceAsynch(xmit,shm,xbytes);
+#endif
     }
   }
   return off_node_bytes;
@@ -461,13 +493,13 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
 void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &list,int dir)
 {
   int nreq=list.size();
-
+  /*finishes Get/Put*/
   acceleratorCopySynchronise();
 
   if (nreq==0) return;
   std::vector<MPI_Status> status(nreq);
   int ierr = MPI_Waitall(nreq,&list[0],&status[0]);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
   list.resize(0);
   this->StencilBarrier(); 
 }
@@ -510,7 +542,7 @@ double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequ
 							   int dest,int dox,
 							   void *recv,
 							   int from,int dor,
-							   int xbytes,int rbytes,int dir)
+							   uint64_t xbytes,uint64_t rbytes,int dir)
 {
 /*
  * Bring sequence from Stencil.h down to lower level.
@@ -527,9 +559,9 @@ double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequ
   int gfrom = ShmRanks[from];
   int gme   = ShmRanks[_processor];
 
-  assert(dest != _processor);
-  assert(from != _processor);
-  assert(gme  == ShmRank);
+  GRID_ASSERT(dest != _processor);
+  GRID_ASSERT(from != _processor);
+  GRID_ASSERT(gme  == ShmRank);
   double off_node_bytes=0.0;
   int tag;
 
@@ -542,18 +574,24 @@ double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequ
    * - post device - host send buffer transfer asynch
    */
   
+#ifdef GRID_CHECKSUM_COMMS
+  rbytes += 8;
+  xbytes += 8;
+#endif
+
   if ( dor ) {
     if ( (gfrom ==MPI_UNDEFINED) || Stencil_force_mpi ) {
       tag= dir+from*32;
       host_recv = this->HostBufferMalloc(rbytes);
-      ierr=MPI_Irecv(host_recv, rbytes, MPI_CHAR,from,tag,communicator_halo[commdir],&rrq);
-      assert(ierr==0);
+      ierr=MPI_Irecv(host_recv,(int)(rbytes/sizeof(int32_t)), MPI_INT32_T,from,tag,communicator_halo[commdir],&rrq);
+      GRID_ASSERT(ierr==0);
       CommsRequest_t srq;
       srq.PacketType = InterNodeRecv;
       srq.bytes      = rbytes;
       srq.req        = rrq;
       srq.host_buf   = host_recv;
       srq.device_buf = recv;
+      srq.tag        = tag;
       list.push_back(srq);
       off_node_bytes+=rbytes;
     }
@@ -561,60 +599,125 @@ double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequ
   
   if (dox) {
     if ( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) {
-#undef DEVICE_TO_HOST_CONCURRENT // pipeline
-#ifdef DEVICE_TO_HOST_CONCURRENT
+
       tag= dir+_processor*32;
 
       host_xmit = this->HostBufferMalloc(xbytes);
-      acceleratorCopyFromDeviceAsynch(xmit, host_xmit,xbytes); // Make this Asynch
+      CommsRequest_t srq;
+
+#ifdef GRID_CHECKSUM_COMMS
+      uint64_t xbytes_data = xbytes - 8;
+      srq.ev = acceleratorCopyFromDeviceAsynch(xmit, host_xmit,xbytes_data); // Make this Asynch
+      GRID_ASSERT(xbytes % 8 == 0);
+      // flip one bit so that a zero buffer is not consistent
+      uint64_t xsum = checksum_gpu((uint64_t*)xmit, xbytes_data / 8) ^ (checksum_index + 1 + 1000 * tag); 
+      *(uint64_t*)(((char*)host_xmit) + xbytes_data) = xsum;
+#else
+      srq.ev = acceleratorCopyFromDeviceAsynch(xmit, host_xmit,xbytes); // Make this Asynch
+#endif
       
       //      ierr =MPI_Isend(host_xmit, xbytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
-      //      assert(ierr==0);
+      //      GRID_ASSERT(ierr==0);
       //      off_node_bytes+=xbytes;
 
-      CommsRequest_t srq;
       srq.PacketType = InterNodeXmit;
       srq.bytes      = xbytes;
       //      srq.req        = xrq;
       srq.host_buf   = host_xmit;
       srq.device_buf = xmit;
+      srq.tag        = tag;
+      srq.dest       = dest;
+      srq.commdir    = commdir;
       list.push_back(srq);
-#else
-      tag= dir+_processor*32;
-
-      host_xmit = this->HostBufferMalloc(xbytes);
-      const int chunks=1;
-      for(int n=0;n<chunks;n++){
-	void * host_xmitc = (void *)( (uint64_t) host_xmit + n*xbytes/chunks);
-	void * xmitc      = (void *)( (uint64_t) xmit      + n*xbytes/chunks);
-	acceleratorCopyFromDeviceAsynch(xmitc, host_xmitc,xbytes/chunks); // Make this Asynch
-      }
-      acceleratorCopySynchronise(); // Complete all pending copy transfers
-      
-      ierr =MPI_Isend(host_xmit, xbytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
-      assert(ierr==0);
-      off_node_bytes+=xbytes;
-
-      CommsRequest_t srq;
-      srq.PacketType = InterNodeXmit;
-      srq.bytes      = xbytes;
-      srq.req        = xrq;
-      srq.host_buf   = host_xmit;
-      srq.device_buf = xmit;
-      list.push_back(srq);
-#endif
     }
   }
 
   return off_node_bytes;
 }
+/*
+ * In the interest of better pipelining, poll for completion on each DtoH and 
+ * start MPI_ISend in the meantime
+ */
+void CartesianCommunicator::StencilSendToRecvFromPollIRecv(std::vector<CommsRequest_t> &list)
+{
+  int pending = 0;
+  do {
+
+    pending = 0;
+
+    for(int idx = 0; idx<list.size();idx++){
+
+      if ( list[idx].PacketType==InterNodeRecv ) {
+
+	int flag = 0;
+	MPI_Status status;
+	int ierr = MPI_Test(&list[idx].req,&flag,&status);
+	assert(ierr==0);
+
+	if ( flag ) {
+	  //	  std::cout << " PollIrecv "<<idx<<" flag "<<flag<<std::endl;
+#ifdef GRID_CHECKSUM_COMMS
+ 	  acceleratorCopyToDeviceAsynch(list[idx].host_buf,list[idx].device_buf,list[idx].bytes - 8);
+#else
+	  acceleratorCopyToDeviceAsynch(list[idx].host_buf,list[idx].device_buf,list[idx].bytes);
+#endif
+	  list[idx].PacketType=InterNodeReceiveHtoD;
+	} else {
+	  pending ++;
+	}
+      }
+    }
+    //    std::cout << " PollIrecv "<<pending<<" pending requests"<<std::endl;
+  } while ( pending );
+  
+}
+void CartesianCommunicator::StencilSendToRecvFromPollDtoH(std::vector<CommsRequest_t> &list)
+{
+  int pending = 0;
+  do {
+
+    pending = 0;
+
+    for(int idx = 0; idx<list.size();idx++){
+
+      if ( list[idx].PacketType==InterNodeXmit ) {
+
+	if ( acceleratorEventIsComplete(list[idx].ev) ) {
+
+	  void *host_xmit = list[idx].host_buf;
+	  uint64_t xbytes = list[idx].bytes;
+	  int dest        = list[idx].dest;
+	  int tag         = list[idx].tag;
+	  int commdir     = list[idx].commdir;
+	  ///////////////////
+	  // Send packet
+	  ///////////////////
+
+	  //	  std::cout << " DtoH is complete for index "<<idx<<" calling MPI_Isend "<<std::endl;
+	  
+	  MPI_Request xrq;
+	  int ierr =MPI_Isend(host_xmit, (int)(xbytes/sizeof(int32_t)), MPI_INT32_T,dest,tag,communicator_halo[commdir],&xrq);
+	  GRID_ASSERT(ierr==0);
+
+	  list[idx].req        = xrq; // Update the MPI request in the list
+
+	  list[idx].PacketType=InterNodeXmitISend;
+
+	} else {
+	  // not done, so return to polling loop
+	  pending++;
+	}
+      }
+    }
+  } while (pending);
+}  
 
 double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
-							 void *xmit,
+							 void *xmit,void *xmit_comp,
 							 int dest,int dox,
-							 void *recv,
+							 void *recv,void *recv_comp,
 							 int from,int dor,
-							 int xbytes,int rbytes,int dir)
+							 uint64_t xbytes,uint64_t rbytes,int dir)
 {
   int ncomm  =communicator_halo.size();
   int commdir=dir%ncomm;
@@ -627,9 +730,9 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
   int gfrom = ShmRanks[from];
   int gme   = ShmRanks[_processor];
 
-  assert(dest != _processor);
-  assert(from != _processor);
-  assert(gme  == ShmRank);
+  GRID_ASSERT(dest != _processor);
+  GRID_ASSERT(from != _processor);
+  GRID_ASSERT(gme  == ShmRank);
   double off_node_bytes=0.0;
   int tag;
 
@@ -644,69 +747,122 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
    * - complete all copies
    * - post MPI send asynch
    */
+#ifdef NVLINK_GET
+  if ( dor ) {
 
-  //  static int printed;
-  //  if((printed<8) && this->IsBoss() ) {
-  //    printf("dir %d doX %d doR %d Face size %ld %ld\n",dir,dox,dor,xbytes,rbytes);
-  //    printed++;
-  //  }
-  
+    if ( ! ( (gfrom ==MPI_UNDEFINED) || Stencil_force_mpi ) ) {
+      // Intranode
+      void *shm = (void *) this->ShmBufferTranslate(from,xmit);
+      GRID_ASSERT(shm!=NULL);
+
+      CommsRequest_t srq;
+
+      srq.ev = acceleratorCopyDeviceToDeviceAsynch(shm,recv,rbytes);
+
+      srq.PacketType = IntraNodeRecv;
+      srq.bytes      = xbytes;
+      //      srq.req        = xrq;
+      srq.host_buf   = NULL;
+      srq.device_buf = xmit;
+      srq.tag        = -1;
+      srq.dest       = dest;
+      srq.commdir    = dir;
+      list.push_back(srq);
+    }
+  }  
+#else
   if (dox) {
 
-    if ( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) {
-#ifdef DEVICE_TO_HOST_CONCURRENT
-      tag= dir+_processor*32;
-      // Find the send in the prepared list
-      int list_idx=-1;
-      for(int idx = 0; idx<list.size();idx++){
-
-	if ( (list[idx].device_buf==xmit)
-	   &&(list[idx].PacketType==InterNodeXmit)
-	   &&(list[idx].bytes==xbytes) ) {
-
-	  list_idx = idx;
-	  host_xmit = list[idx].host_buf;
-	}
-      }
-      assert(list_idx != -1); // found it
-      ierr =MPI_Isend(host_xmit, xbytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
-      assert(ierr==0);
-      list[list_idx].req        = xrq; // Update the MPI request in the list
-      off_node_bytes+=xbytes;
-#endif      
-    } else {
+    if ( !( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) ) {
+      // Intranode
       void *shm = (void *) this->ShmBufferTranslate(dest,recv);
-      assert(shm!=NULL);
-      acceleratorCopyDeviceToDeviceAsynch(xmit,shm,xbytes);
+      GRID_ASSERT(shm!=NULL);
+
+      CommsRequest_t srq;
+      
+      srq.ev = acceleratorCopyDeviceToDeviceAsynch(xmit,shm,xbytes);
+
+      srq.PacketType = IntraNodeXmit;
+      srq.bytes      = xbytes;
+      //      srq.req        = xrq;
+      srq.host_buf   = NULL;
+      srq.device_buf = xmit;
+      srq.tag        = -1;
+      srq.dest       = dest;
+      srq.commdir    = dir;
+      list.push_back(srq);
+      
     }
   }
+#endif
   return off_node_bytes;
 }
 void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &list,int dir)
 {
-  int nreq=list.size();
+  acceleratorCopySynchronise(); // Complete all pending copy transfers D2D
 
-  if (nreq==0) return;
-  std::vector<MPI_Status> status(nreq);
-  std::vector<MPI_Request> MpiRequests(nreq);
+  std::vector<MPI_Status> status;
+  std::vector<MPI_Request> MpiRequests;
+    
+  for(int r=0;r<list.size();r++){
+    // Must check each Send buf is clear to reuse
+    if ( list[r].PacketType == InterNodeXmitISend ) MpiRequests.push_back(list[r].req);
+    //    if ( list[r].PacketType == InterNodeRecv ) MpiRequests.push_back(list[r].req); // Already "Test" passed
+  }
 
-  for(int r=0;r<nreq;r++){
-    MpiRequests[r] = list[r].req;
+  int nreq=MpiRequests.size();
+
+  if (nreq>0) {
+    status.resize(MpiRequests.size());
+    int ierr = MPI_Waitall(MpiRequests.size(),&MpiRequests[0],&status[0]); // Sends are guaranteed in order. No harm in not completing.
+    GRID_ASSERT(ierr==0);
   }
   
-  int ierr = MPI_Waitall(nreq,&MpiRequests[0],&status[0]);
-  assert(ierr==0);
+  //  for(int r=0;r<nreq;r++){
+  //    if ( list[r].PacketType==InterNodeRecv ) {
+  //      acceleratorCopyToDeviceAsynch(list[r].host_buf,list[r].device_buf,list[r].bytes);
+  //    }
+  //  }
+#ifdef GRID_CHECKSUM_COMMS
+  for(int r=0;r<list.size();r++){
+    if ( list[r].PacketType == InterNodeReceiveHtoD ) {
+      uint64_t rbytes_data = list[r].bytes - 8;
+      uint64_t expected_cs = *(uint64_t*)(((char*)list[r].host_buf) + rbytes_data);
+      uint64_t computed_cs = checksum_gpu((uint64_t*)list[r].device_buf, rbytes_data / 8) ^ (checksum_index + 1 + 1000 * list[r].tag); //
+      if (expected_cs != computed_cs) {
+	// TODO: error message, backtrace, quit
 
-  for(int r=0;r<nreq;r++){
-    if ( list[r].PacketType==InterNodeRecv ) {
-      acceleratorCopyToDeviceAsynch(list[r].host_buf,list[r].device_buf,list[r].bytes);
+	fprintf(stderr, "GRID_CHECKSUM_COMMS error:\n");
+	fprintf(stderr, " processor = %d\n", (int)_processor);
+	for(int d=0;d<_processors.size();d++)
+	  fprintf(stderr, " processor_coord[%d] = %d\n", d, _processor_coor[d]);
+	fprintf(stderr, " hostname: %s\n", GridHostname());
+	fprintf(stderr, " expected_cs: %ld\n", expected_cs);
+	fprintf(stderr, " computed_cs: %ld\n", computed_cs);
+	fprintf(stderr, " dest: %d\n", list[r].dest);
+	fprintf(stderr, " tag: %d\n", list[r].tag);
+	fprintf(stderr, " commdir: %d\n", list[r].commdir);
+	fprintf(stderr, " bytes: %ld\n", (uint64_t)list[r].bytes);
+
+	fflush(stderr);
+
+	// backtrace
+	int symbols = backtrace(Grid_backtrace_buffer,_NBACKTRACE);
+	backtrace_symbols_fd(Grid_backtrace_buffer,symbols, 2);
+
+	exit(1);
+      }
     }
   }
+
+  checksum_index += 1;
+#endif
   
-  acceleratorCopySynchronise(); // Complete all pending copy transfers
   list.resize(0);               // Delete the list
   this->HostBufferFreeAll();    // Clean up the buffer allocs
-  this->StencilBarrier(); 
+#ifndef NVLINK_GET
+  this->StencilBarrier(); // if PUT must check our nbrs have filled our receive buffers.
+#endif   
 }
 #endif
 ////////////////////////////////////////////
@@ -715,6 +871,9 @@ void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsReque
 
 void CartesianCommunicator::StencilBarrier(void)
 {
+  FlightRecorder::StepLog("NodeBarrier");
+  // Only orders the device-to-device PUT/GET into a peer's window; not needed under --shm-mpi 1.
+  if ( Stencil_force_mpi && !Stencil_force_barrier ) return;
   MPI_Barrier  (ShmComm);
 }
 //void CartesianCommunicator::SendToRecvFromComplete(std::vector<CommsRequest_t> &list)
@@ -722,17 +881,19 @@ void CartesianCommunicator::StencilBarrier(void)
 //}
 void CartesianCommunicator::Barrier(void)
 {
+  FlightRecorder::StepLog("GridBarrier");
   int ierr = MPI_Barrier(communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
-void CartesianCommunicator::Broadcast(int root,void* data, int bytes)
+void CartesianCommunicator::Broadcast(int root,void* data,uint64_t bytes)
 {
+  FlightRecorder::StepLog("Broadcast");
   int ierr=MPI_Bcast(data,
-		     bytes,
+		     (int)bytes,
 		     MPI_BYTE,
 		     root,
 		     communicator);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 int CartesianCommunicator::RankWorld(void){
   int r;
@@ -740,23 +901,25 @@ int CartesianCommunicator::RankWorld(void){
   return r;
 }
 void CartesianCommunicator::BarrierWorld(void){
+  FlightRecorder::StepLog("BarrierWorld");
   int ierr = MPI_Barrier(communicator_world);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
-void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
+void CartesianCommunicator::BroadcastWorld(int root,void* data, uint64_t bytes)
 {
+  FlightRecorder::StepLog("BroadcastWorld");
   int ierr= MPI_Bcast(data,
-		      bytes,
+		      (int)bytes,
 		      MPI_BYTE,
 		      root,
 		      communicator_world);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 
 void CartesianCommunicator::AllToAll(int dim,void  *in,void *out,uint64_t words,uint64_t bytes)
 {
   Coordinate row(_ndimension,1);
-  assert(dim>=0 && dim<_ndimension);
+  GRID_ASSERT(dim>=0 && dim<_ndimension);
 
   //  Split the communicator
   row[dim] = _processors[dim];
@@ -767,6 +930,7 @@ void CartesianCommunicator::AllToAll(int dim,void  *in,void *out,uint64_t words,
 }
 void CartesianCommunicator::AllToAll(void  *in,void *out,uint64_t words,uint64_t bytes)
 {
+  FlightRecorder::StepLog("AllToAll");
   // MPI is a pain and uses "int" arguments
   // 64*64*64*128*16 == 500Million elements of data.
   // When 24*4 bytes multiples get 50x 10^9 >>> 2x10^9 Y2K bug.
@@ -776,12 +940,67 @@ void CartesianCommunicator::AllToAll(void  *in,void *out,uint64_t words,uint64_t
   int ibytes;
   iwords = words;
   ibytes = bytes;
-  assert(words == iwords); // safe to cast to int ?
-  assert(bytes == ibytes); // safe to cast to int ?
+  GRID_ASSERT(words == iwords); // safe to cast to int ?
+  GRID_ASSERT(bytes == ibytes); // safe to cast to int ?
   MPI_Type_contiguous(ibytes,MPI_BYTE,&object);
   MPI_Type_commit(&object);
   MPI_Alltoall(in,iwords,object,out,iwords,object,communicator);
   MPI_Type_free(&object);
 }
+void CartesianCommunicator::AllToAllV(void *in ,const std::vector<int> &sendcounts,const std::vector<int> &senddispls,
+                                      void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                                      uint64_t bytes)
+{
+  FlightRecorder::StepLog("AllToAllV");
+  GRID_ASSERT(sendcounts.size()==(size_t)_Nprocessors);
+  GRID_ASSERT(senddispls.size()==(size_t)_Nprocessors);
+  GRID_ASSERT(recvcounts.size()==(size_t)_Nprocessors);
+  GRID_ASSERT(recvdispls.size()==(size_t)_Nprocessors);
+  // MPI counts are "int"; the caller sizes the word to keep them in range
+  int ibytes = bytes;
+  GRID_ASSERT(bytes == (uint64_t)ibytes);
+  MPI_Datatype object;
+  MPI_Type_contiguous(ibytes,MPI_BYTE,&object);
+  MPI_Type_commit(&object);
+  int ierr = MPI_Alltoallv(in ,(int *)&sendcounts[0],(int *)&senddispls[0],object,
+                           out,(int *)&recvcounts[0],(int *)&recvdispls[0],object,communicator);
+  GRID_ASSERT(ierr==0);
+  MPI_Type_free(&object);
+}
+
+void CartesianCommunicator::AllGatherV(void *in ,int sendcount,
+                                       void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                                       uint64_t bytes)
+{
+  FlightRecorder::StepLog("AllGatherV");
+  GRID_ASSERT(recvcounts.size()==(size_t)_Nprocessors);
+  GRID_ASSERT(recvdispls.size()==(size_t)_Nprocessors);
+  GRID_ASSERT(recvcounts[_processor]==sendcount);
+  // MPI counts are "int"; the caller sizes the word to keep them in range
+  int ibytes = bytes;
+  GRID_ASSERT(bytes == (uint64_t)ibytes);
+  MPI_Datatype object;
+  MPI_Type_contiguous(ibytes,MPI_BYTE,&object);
+  MPI_Type_commit(&object);
+  int ierr = MPI_Allgatherv(in ,sendcount,object,
+                            out,(int *)&recvcounts[0],(int *)&recvdispls[0],object,communicator);
+  GRID_ASSERT(ierr==0);
+  MPI_Type_free(&object);
+}
+
+void CartesianCommunicator::AllGather(void *in,void *out,uint64_t words,uint64_t bytes)
+{
+  FlightRecorder::StepLog("AllGather");
+  int iwords = words;
+  int ibytes = bytes;
+  GRID_ASSERT(words == (uint64_t)iwords);
+  GRID_ASSERT(bytes == (uint64_t)ibytes);
+  MPI_Datatype object;
+  MPI_Type_contiguous(ibytes,MPI_BYTE,&object);
+  MPI_Type_commit(&object);
+  int ierr = MPI_Allgather(in,iwords,object,out,iwords,object,communicator);
+  GRID_ASSERT(ierr==0);
+  MPI_Type_free(&object);
+}
 
 NAMESPACE_END(Grid);

Grid/communicator/Communicator_none.cc

@@ -27,6 +27,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 /*  END LEGAL */
 #include <Grid/GridCore.h>
 
+void GridAbort(void) { abort(); }
+
 NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
@@ -34,6 +36,7 @@ NAMESPACE_BEGIN(Grid);
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 Grid_MPI_Comm       CartesianCommunicator::communicator_world;
 
+
 void CartesianCommunicator::Init(int *argc, char *** arv)
 {
   GlobalSharedMemory::Init(communicator_world);
@@ -54,14 +57,14 @@ CartesianCommunicator::CartesianCommunicator(const Coordinate &processors)
 {
   _shm_processors = Coordinate(processors.size(),1);
   _processors = processors;
-  _ndimension = processors.size();  assert(_ndimension>=1);
+  _ndimension = processors.size();  GRID_ASSERT(_ndimension>=1);
   _processor_coor.resize(_ndimension);
   
   // Require 1^N processor grid for fake
   _Nprocessors=1;
   _processor = 0;
   for(int d=0;d<_ndimension;d++) {
-    assert(_processors[d]==1);
+    GRID_ASSERT(_processors[d]==1);
     _processor_coor[d] = 0;
   }
   SetCommunicator(communicator_world);
@@ -87,19 +90,19 @@ void CartesianCommunicator::SendToRecvFrom(void *xmit,
 					   int dest,
 					   void *recv,
 					   int from,
-					   int bytes)
+					   uint64_t bytes)
 {
-  assert(0);
+  GRID_ASSERT(0);
 }
-void CartesianCommunicator::CommsComplete(std::vector<CommsRequest_t> &list){ assert(0);}
+void CartesianCommunicator::CommsComplete(std::vector<CommsRequest_t> &list){ GRID_ASSERT(list.size()==0);}
 void CartesianCommunicator::SendToRecvFromBegin(std::vector<CommsRequest_t> &list,
 						void *xmit,
 						int dest,
 						void *recv,
 						int from,
-						int bytes,int dir)
+						uint64_t bytes,int dir)
 {
-  assert(0);
+  GRID_ASSERT(0);
 }
 
 void CartesianCommunicator::AllToAll(int dim,void  *in,void *out,uint64_t words,uint64_t bytes)
@@ -110,11 +113,39 @@ void CartesianCommunicator::AllToAll(void  *in,void *out,uint64_t words,uint64_t
 {
   bcopy(in,out,bytes*words);
 }
+void CartesianCommunicator::AllToAllV(void *in ,const std::vector<int> &sendcounts,const std::vector<int> &senddispls,
+                                      void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                                      uint64_t bytes)
+{
+  // Single rank: the exchange degenerates to a copy of our own segment
+  GRID_ASSERT(sendcounts.size()==1);
+  GRID_ASSERT(recvcounts.size()==1);
+  GRID_ASSERT(sendcounts[0]==recvcounts[0]);
+  bcopy((char *)in +(uint64_t)senddispls[0]*bytes,
+        (char *)out+(uint64_t)recvdispls[0]*bytes,bytes*(uint64_t)sendcounts[0]);
+}
+
+void CartesianCommunicator::AllGatherV(void *in ,int sendcount,
+                                       void *out,const std::vector<int> &recvcounts,const std::vector<int> &recvdispls,
+                                       uint64_t bytes)
+{
+  // Single rank: the gather degenerates to a copy of our own contribution
+  GRID_ASSERT(recvcounts.size()==1);
+  GRID_ASSERT(recvdispls.size()==1);
+  GRID_ASSERT(recvcounts[0]==sendcount);
+  bcopy((char *)in,
+        (char *)out+(uint64_t)recvdispls[0]*bytes,bytes*(uint64_t)sendcount);
+}
+
+void CartesianCommunicator::AllGather(void *in,void *out,uint64_t words,uint64_t bytes)
+{
+  bcopy((char *)in,(char *)out,bytes*words);
+}
 
 int  CartesianCommunicator::RankWorld(void){return 0;}
 void CartesianCommunicator::Barrier(void){}
-void CartesianCommunicator::Broadcast(int root,void* data, int bytes) {}
-void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes) { }
+void CartesianCommunicator::Broadcast(int root,void* data, uint64_t bytes) {}
+void CartesianCommunicator::BroadcastWorld(int root,void* data, uint64_t bytes) { }
 void CartesianCommunicator::BarrierWorld(void) { }
 int  CartesianCommunicator::RankFromProcessorCoor(Coordinate &coor) {  return 0;}
 void CartesianCommunicator::ProcessorCoorFromRank(int rank, Coordinate &coor){  coor = _processor_coor; }
@@ -124,29 +155,33 @@ void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest
   dest=0;
 }
 
+int CartesianCommunicator::IsOffNode(int rank) { return false; }
+
 double CartesianCommunicator::StencilSendToRecvFrom( void *xmit,
 						     int xmit_to_rank,int dox,
 						     void *recv,
 						     int recv_from_rank,int dor,
-						     int bytes, int dir)
+						     uint64_t bytes, int dir)
 {
   return 2.0*bytes;
 }
+void CartesianCommunicator::StencilSendToRecvFromPollIRecv(std::vector<CommsRequest_t> &list) {};
+void CartesianCommunicator::StencilSendToRecvFromPollDtoH(std::vector<CommsRequest_t> &list) {};
 double CartesianCommunicator::StencilSendToRecvFromPrepare(std::vector<CommsRequest_t> &list,
 							   void *xmit,
 							   int xmit_to_rank,int dox,
 							   void *recv,
 							   int recv_from_rank,int dor,
-							   int xbytes,int rbytes, int dir)
+							   uint64_t xbytes,uint64_t rbytes, int dir)
 {
-  return xbytes+rbytes;
+  return 0.0;
 }
 double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
-							 void *xmit,
+							 void *xmit, void *xmit_comp,
 							 int xmit_to_rank,int dox,
-							 void *recv,
+							 void *recv, void *recv_comp,
 							 int recv_from_rank,int dor,
-							 int xbytes,int rbytes, int dir)
+							 uint64_t xbytes,uint64_t rbytes, int dir)
 {
   return xbytes+rbytes;
 }

Grid/communicator/RingAllReduce.h

@@ -0,0 +1,254 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./Grid/communicator/RingAllReduce.h
+
+    Copyright (C) 2026
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    See the full license in the file "LICENSE" in the top level distribution
+    directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+/////////////////////////////////////////////////////////////////////////////
+// Vector all-reduce on point-to-point only (SendToRecvFrom), no MPI
+// collectives: reduce-scatter ring followed by all-gather ring.  A
+// communicator-level primitive: needs only CartesianCommunicator.
+//
+//   RingAllReduce(comm, buf, n)           flat ring over all P ranks
+//   CartesianRingAllReduce(comm, buf, n, orthogDim=-1)
+//                                         ring along each processor dimension
+//                                         in turn (P_d ranks per ring).
+//                                         orthogDim in 0..Nd-1: that dimension is
+//                                         SKIPPED, so the result is the sum over
+//                                         all ranks sharing my coordinate in it
+//                                         (e.g. orthogDim=3: sum over each
+//                                         time-slice of processors separately).
+//                                         -1: all dimensions (full allreduce).
+//
+// Why: Cray MPICH device-buffer MPI_Allreduce aborts above ~8 MB (MPI_FLOAT,
+// measured 4.4 MB pass / 13.3 MB fail) and delivers 5.8 GB/s where P2P rings
+// deliver ~17 GB/s.  A ring has no size cliff -- every step is one symmetric
+// SendToRecvFrom of one chunk.
+//
+// Cost model, N bytes per rank:
+//   flat      : 2(P-1) steps, 2N(P-1)/P bytes per rank -- bandwidth-optimal,
+//               latency 2(P-1) x step  (P=288: 574 steps)
+//   cartesian : sum_d 2(P_d-1) steps, ~2N per dimension -- few steps, each on
+//               a neighbour link (3.6.4.4: 26 steps, ~8N bytes)
+// Rule of thumb: cartesian for tens of MB, flat for hundreds of MB.
+//
+// Buffer memory space: whatever SendToRecvFrom accepts on this build (device
+// under ACCELERATOR_AWARE_MPI); the reduction runs as accelerator_for on a
+// deviceVector working copy, so the caller's buffer is only ever memcpy'd.
+// Deterministic: summation order is fixed by rank and chunk index, so the
+// result is bitwise reproducible run to run (MPI_Allreduce need not be),
+// though it differs from MPI's order at rounding level.
+//
+// n elements of T; T must support + on the accelerator (RealF/RealD,
+// ComplexF/ComplexD).  sizeof(T)*chunk must be a multiple of 4 bytes
+// (SendToRecvFrom counts int32 words); true for all supported T.
+/////////////////////////////////////////////////////////////////////////////
+
+// Ring all-reduce among P ranks with given next/prev neighbours; `me` is this
+// rank's position in the ring.  work has P*c elements, scratch has c.
+template<class T>
+void RingAllReduceCore(CartesianCommunicator *comm,
+		       T *work, T *scratch, uint64_t c, int P, int me, int next, int prev)
+{
+  if ( P==1 ) return;
+  uint64_t bytes = c*sizeof(T);
+  // reduce-scatter: after P-1 steps rank me owns fully reduced chunk (me+1)%P
+  for(int s=0;s<P-1;s++){
+    int sendc = (me - s     + 2*P) % P;
+    int recvc = (me - s - 1 + 2*P) % P;
+    comm->SendToRecvFrom((void *)&work[sendc*c], next, (void *)scratch, prev, bytes);
+    T *dst = &work[recvc*c];
+    accelerator_for(i, c, 1, { dst[i] = dst[i] + scratch[i]; });
+  }
+  // all-gather: circulate the reduced chunks
+  for(int s=0;s<P-1;s++){
+    int sendc = (me - s + 1 + 2*P) % P;
+    int recvc = (me - s     + 2*P) % P;
+    comm->SendToRecvFrom((void *)&work[sendc*c], next, (void *)&work[recvc*c], prev, bytes);
+  }
+}
+
+template<class T>
+void RingAllReduce(CartesianCommunicator *comm, T *buf, uint64_t n)
+{
+  int P  = comm->ProcessorCount();
+  int me = comm->ThisRank();
+  if ( P==1 || n==0 ) return;
+  uint64_t c = (n + P - 1)/P;
+  deviceVector<T> work(c*P);
+  deviceVector<T> scratch(c);
+  T *w = &work[0];
+  accelerator_for(i, c*P, 1, { w[i] = T(0.0); });
+  acceleratorCopyDeviceToDevice((void *)buf, (void *)w, n*sizeof(T));
+  RingAllReduceCore(comm, w, &scratch[0], c, P, me, (me+1)%P, (me+P-1)%P);
+  acceleratorCopyDeviceToDevice((void *)w, (void *)buf, n*sizeof(T));
+}
+
+template<class T>
+void CartesianRingAllReduce(CartesianCommunicator *comm, T *buf, uint64_t n, int orthogDim=-1)
+{
+  if ( comm->ProcessorCount()==1 || n==0 ) return;
+  int Nd = comm->_ndimension;
+  GRID_ASSERT( orthogDim >= -1 && orthogDim < Nd );
+  for(int d=0;d<Nd;d++){
+    if ( d==orthogDim ) continue;              // leave this dimension unsummed
+    int P = comm->_processors[d];
+    if ( P==1 ) continue;
+    int me = comm->_processor_coor[d];
+    int next, prev;
+    comm->ShiftedRanks(d, 1, prev, next);   // (dim, shift, source, dest)
+    uint64_t c = (n + P - 1)/P;
+    deviceVector<T> work(c*P);
+    deviceVector<T> scratch(c);
+    T *w = &work[0];
+    accelerator_for(i, c*P, 1, { w[i] = T(0.0); });
+    acceleratorCopyDeviceToDevice((void *)buf, (void *)w, n*sizeof(T));
+    RingAllReduceCore(comm, w, &scratch[0], c, P, me, next, prev);
+    acceleratorCopyDeviceToDevice((void *)w, (void *)buf, n*sizeof(T));
+  }
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Cartesian ring ALLGATHER, point-to-point only.
+//
+//   CartesianRingAllGather(comm, buf, chunk)
+//     buf holds P*chunk elements of T.  Block index = the Grid LEXICOGRAPHIC
+//     index of the owning process coordinate (dimension 0 fastest,
+//     Lexicographic::CoorFromIndex convention), NOT the MPI rank: on entry my
+//     chunk is at buf[CartesianLexIndex(comm)*chunk]; on exit block L is the
+//     chunk of the process at coordinate CoorFromIndex(L).  Map to ranks with
+//     comm->RankFromProcessorCoor.  (Ranks and coordinates are NOT related
+//     lexicographically on Frontier -- the OptimalCommunicator relabels ranks
+//     for shared-memory locality; assuming rank order gave a wrong inverse,
+//     VERIFY 0.9965, 2026-08-26.)
+//
+// Dimension by dimension from dimension 0 (fastest) upward: each stage is a
+// ring over the P_d ranks of that line, after which the held block is the
+// concatenation over that coordinate in the lexicographic nesting.  Bytes
+// sent per rank ~N = P*chunk in total (dominated by the last stage): 8x less
+// than a zero-padded CartesianRingAllReduce.  Steps: sum_d (P_d-1).  Exact.
+//
+// Written for the dense coarse-coarse apply (every rank owns rows of A^{-1}
+// and needs the whole x): 1.86 ms with the allreduce ring at 288 ranks was
+// wire speed but moved 35 MB per rank to deliver 4.4.
+/////////////////////////////////////////////////////////////////////////////
+inline int CartesianLexIndex(CartesianCommunicator *comm)
+{
+  int idx=0, stride=1;
+  for(int d=0; d<(int)comm->_ndimension; d++){ idx += comm->_processor_coor[d]*stride; stride *= comm->_processors[d]; }
+  return idx;
+}
+
+//
+//   CartesianRingAllGather(comm, buf, chunk, dim)   dim in 0..Nd-1: gather along
+//     ONE processor dimension only.  buf holds P_dim*chunk elements; on entry my
+//     chunk is at buf[coor[dim]*chunk], on exit block c is the chunk of the rank
+//     at coordinate c along dim with all other coordinates equal to mine.  Every
+//     rank of the line holds the same result (e.g. dim=3 after a
+//     CartesianRingAllReduce(orthogDim=3): the boss of each spatial line can then
+//     write the P_t-times-longer vector, not the P-times-longer one).
+/////////////////////////////////////////////////////////////////////////////
+template<class T>
+void CartesianRingAllGather(CartesianCommunicator *comm, T *buf, uint64_t chunk, int dim=-1)
+{
+  int P  = comm->ProcessorCount();
+  int Nd = comm->_ndimension;
+  GRID_ASSERT( dim >= -1 && dim < Nd );
+  if ( dim >= 0 ) P = comm->_processors[dim];      // ranks in my line along dim
+  if ( P==1 || chunk==0 ) return;
+  int mylex = (dim<0) ? CartesianLexIndex(comm) : comm->_processor_coor[dim];
+  deviceVector<T> work((uint64_t)P*chunk);
+  // ping-pong between buf and work; the held block lives at offset `off` in `cur`
+  T *cur = buf;       uint64_t off = (uint64_t)mylex*chunk;
+  T *oth = &work[0];
+  uint64_t blk = chunk;                        // elements in the held block
+  for(int d=0; d<Nd; d++){                     // dimension 0 first: it is the fastest lex index
+    if ( dim>=0 && d!=dim ) continue;          // single-dimension gather
+    int Pd = comm->_processors[d];
+    if ( Pd==1 ) continue;
+    int med = comm->_processor_coor[d];
+    int next, prev;
+    comm->ShiftedRanks(d, 1, prev, next);      // (dim, shift, source, dest)
+    GRID_ASSERT( (blk*sizeof(T))%4 == 0 );
+    acceleratorCopyDeviceToDevice((void *)(cur+off), (void *)(oth+(uint64_t)med*blk), blk*sizeof(T));
+    for(int t=1;t<Pd;t++){
+      int sendslot = (med - t + 1 + Pd) % Pd;
+      int recvslot = (med - t     + Pd) % Pd;
+      comm->SendToRecvFrom((void *)(oth+(uint64_t)sendslot*blk), next,
+                           (void *)(oth+(uint64_t)recvslot*blk), prev, blk*sizeof(T));
+    }
+    T *tmp = cur; cur = oth; oth = tmp; off = 0;
+    blk *= Pd;
+  }
+  GRID_ASSERT( blk == (uint64_t)P*chunk );
+  if ( cur != buf ) acceleratorCopyDeviceToDevice((void *)cur, (void *)buf, blk*sizeof(T));
+}
+
+/////////////////////////////////////////////////////////////////////////////
+// Cartesian ALL-TO-ALL along one processor dimension, point-to-point only.
+//
+//   CartesianRingAllToAll(comm, sbuf, rbuf, chunk, dim)
+//     sbuf, rbuf each hold P_dim*chunk elements of T.  Block index is the
+//     process COORDINATE along dim (not the MPI rank -- ShiftedRanks resolves
+//     the OptimalCommunicator relabelling): sbuf[c*chunk] is my data destined
+//     for the rank at coordinate c along dim; on exit rbuf[c*chunk] is the
+//     block that rank sent me.  All other coordinates equal mine.
+//
+// Direct-pairwise, NOT ring-forwarding: P_dim-1 symmetric exchanges, step k
+// with the +/-k neighbours (ShiftedRanks == MPI_Cart_shift, arbitrary shift),
+// each moving one chunk with no store-and-forward.  Per-rank traffic
+// (P_dim-1)*chunk -- a factor P below the AllGather's replicated P*chunk, which
+// is the point for the pencil-FFT transpose (partition the orthogonal coords
+// rather than replicate the whole transform line on every rank).
+//
+// On a switched fabric a distance-k send is one logical hop, so direct-pairwise
+// moves the minimum; ring-forwarding would only win on a true physical ring.
+// Every step is one SendToRecvFrom of one chunk: no collective, no size cliff.
+// Send/recv-rank binding follows Cshift_mpi.h: ShiftedRanks(dim,k,xmit,recv)
+// sends to coord-k, receives from coord+k.  Steps: P_dim-1.  Exact.
+/////////////////////////////////////////////////////////////////////////////
+template<class T>
+void CartesianRingAllToAll(CartesianCommunicator *comm,
+                           T *sbuf, T *rbuf, uint64_t chunk, int dim)
+{
+  int Nd = comm->_ndimension;
+  GRID_ASSERT( dim >= 0 && dim < Nd );
+  int P  = comm->_processors[dim];
+  int me = comm->_processor_coor[dim];
+  if ( chunk==0 ) return;
+  GRID_ASSERT( (chunk*sizeof(T))%4 == 0 );      // SendToRecvFrom counts int32 words
+  uint64_t bytes = chunk*sizeof(T);
+
+  // my own block never goes on the wire
+  acceleratorCopyDeviceToDevice((void *)&sbuf[(uint64_t)me*chunk],
+                                (void *)&rbuf[(uint64_t)me*chunk], bytes);
+  if ( P==1 ) return;
+
+  for(int k=1;k<P;k++){
+    int xmit_to_rank, recv_from_rank;
+    comm->ShiftedRanks(dim, k, xmit_to_rank, recv_from_rank); // xmit=coord-k, recv=coord+k
+    uint64_t sidx = (uint64_t)((me - k + P) % P);   // block destined for the coord-k rank
+    uint64_t ridx = (uint64_t)((me + k) % P);       // block arriving from the coord+k rank
+    comm->SendToRecvFrom((void *)&sbuf[sidx*chunk], xmit_to_rank,
+                         (void *)&rbuf[ridx*chunk], recv_from_rank, bytes);
+  }
+}
+
+NAMESPACE_END(Grid);

Grid/communicator/SharedMemory.cc

@@ -58,8 +58,8 @@ int                 GlobalSharedMemory::WorldNode;
 
 void GlobalSharedMemory::SharedMemoryFree(void)
 {
-  assert(_ShmAlloc);
-  assert(_ShmAllocBytes>0);
+  GRID_ASSERT(_ShmAlloc);
+  GRID_ASSERT(_ShmAllocBytes>0);
   for(int r=0;r<WorldShmSize;r++){
     munmap(WorldShmCommBufs[r],_ShmAllocBytes);
   }
@@ -80,7 +80,7 @@ void *SharedMemory::HostBufferMalloc(size_t bytes){
     std::cout<< " Current alloc is " << (bytes/(1024*1024)) <<"MB"<<std::endl;
     std::cout<< " Current bytes is " << (host_heap_bytes/(1024*1024)) <<"MB"<<std::endl;
     std::cout<< " Current heap  is " << (host_heap_size/(1024*1024)) <<"MB"<<std::endl;
-    assert(host_heap_bytes<host_heap_size);
+    GRID_ASSERT(host_heap_bytes<host_heap_size);
   }
   return ptr;
 }
@@ -100,7 +100,7 @@ void *SharedMemory::ShmBufferMalloc(size_t bytes){
     std::cout<< " Current alloc is " << (bytes/(1024*1024)) <<"MB"<<std::endl;
     std::cout<< " Current bytes is " << (heap_bytes/(1024*1024)) <<"MB"<<std::endl;
     std::cout<< " Current heap  is " << (heap_size/(1024*1024)) <<"MB"<<std::endl;
-    assert(heap_bytes<heap_size);
+    GRID_ASSERT(heap_bytes<heap_size);
   }
   //std::cerr << "ShmBufferMalloc "<<std::hex<< ptr<<" - "<<((uint64_t)ptr+bytes)<<std::dec<<std::endl;
   return ptr;
@@ -127,13 +127,13 @@ void GlobalSharedMemory::GetShmDims(const Coordinate &WorldDims,Coordinate &ShmD
   if ( str ) {
     std::vector<int> IntShmDims;
     GridCmdOptionIntVector(std::string(str),IntShmDims);
-    assert(IntShmDims.size() == WorldDims.size());
+    GRID_ASSERT(IntShmDims.size() == WorldDims.size());
     long ShmSize = 1;
     for (int dim=0;dim<WorldDims.size();dim++) {
       ShmSize *= (ShmDims[dim] = IntShmDims[dim]);
-      assert(divides(ShmDims[dim],WorldDims[dim]));
+      GRID_ASSERT(divides(ShmDims[dim],WorldDims[dim]));
     }
-    assert(ShmSize == WorldShmSize);
+    GRID_ASSERT(ShmSize == WorldShmSize);
     return;
   }
   

Grid/communicator/SharedMemory.h

@@ -50,12 +50,30 @@ typedef MPI_Request MpiCommsRequest_t;
 #ifdef ACCELERATOR_AWARE_MPI
 typedef MPI_Request CommsRequest_t;
 #else
-enum PacketType_t { InterNodeXmit, InterNodeRecv, IntraNodeXmit, IntraNodeRecv };
+/*
+ * Enable state transitions as each packet flows.
+ */
+enum PacketType_t {
+  FaceGather,
+  InterNodeXmit,
+  InterNodeRecv,
+  IntraNodeXmit,
+  IntraNodeRecv,
+  InterNodeXmitISend,
+  InterNodeReceiveHtoD
+};
+/*
+ *Package arguments needed for various actions along packet flow
+ */
 typedef struct {
   PacketType_t PacketType;
   void *host_buf;
   void *device_buf;
+  int dest;
+  int tag;
+  int commdir;
   unsigned long bytes;
+  acceleratorEvent_t ev;
   MpiCommsRequest_t req;
 } CommsRequest_t;
 #endif
@@ -119,7 +137,7 @@ public:
   ///////////////////////////////////////////////////
   static void SharedMemoryAllocate(uint64_t bytes, int flags);
   static void SharedMemoryFree(void);
-  static void SharedMemoryCopy(void *dest,void *src,size_t bytes);
+  //  static void SharedMemoryCopy(void *dest,void *src,size_t bytes);
   static void SharedMemoryZero(void *dest,size_t bytes);
 
 };

Grid/communicator/SharedMemoryMPI.cc

@@ -43,10 +43,6 @@ Author: Christoph Lehner <christoph@lhnr.de>
 #define GRID_SYCL_LEVEL_ZERO_IPC
 #define SHM_SOCKETS
 #else
-#ifdef HAVE_NUMAIF_H
-  #warning " Using NUMAIF "
-#include <numaif.h>
-#endif 
 #endif 
 #include <syscall.h>
 #endif
@@ -71,7 +67,7 @@ public:
   {
     int errnum;
 
-    sock = socket(AF_UNIX, SOCK_DGRAM, 0);  assert(sock>0);
+    sock = socket(AF_UNIX, SOCK_DGRAM, 0);  GRID_ASSERT(sock>0);
 
     struct sockaddr_un sa_un = { 0 };
     sa_un.sun_family = AF_UNIX;
@@ -162,7 +158,7 @@ public:
 /*Construct from an MPI communicator*/
 void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
 {
-  assert(_ShmSetup==0);
+  GRID_ASSERT(_ShmSetup==0);
   WorldComm = comm;
   MPI_Comm_rank(WorldComm,&WorldRank);
   MPI_Comm_size(WorldComm,&WorldSize);
@@ -188,7 +184,7 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
 
   // WorldNodes
   WorldNodes = WorldSize/WorldShmSize;
-  assert( (WorldNodes * WorldShmSize) == WorldSize );
+  GRID_ASSERT( (WorldNodes * WorldShmSize) == WorldSize );
 
 
   // FIXME: Check all WorldShmSize are the same ?
@@ -213,7 +209,7 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
   MyGroup.resize(WorldShmSize);
   for(int rank=0;rank<WorldSize;rank++){
     if(WorldShmRanks[rank]!=MPI_UNDEFINED){
-      assert(g<WorldShmSize);
+      GRID_ASSERT(g<WorldShmSize);
       MyGroup[g++] = rank;
     }
   }
@@ -229,7 +225,7 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
   // global sum leaders over comm world
   ///////////////////////////////////////////////////////////////////
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&leaders_1hot[0],WorldSize,MPI_INT,MPI_SUM,WorldComm);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 
   ///////////////////////////////////////////////////////////////////
   // find the group leaders world rank
@@ -250,7 +246,7 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
       WorldNode=g;
     }
   }
-  assert(WorldNode!=-1);
+  GRID_ASSERT(WorldNode!=-1);
   _ShmSetup=1;
 }
 // Gray encode support 
@@ -292,7 +288,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processo
   // Assert power of two shm_size.
   ////////////////////////////////////////////////////////////////
   int log2size = Log2Size(WorldShmSize,MAXLOG2RANKSPERNODE);
-  assert(log2size != -1);
+  GRID_ASSERT(log2size != -1);
 
   ////////////////////////////////////////////////////////////////
   // Identify the hypercube coordinate of this node using hostname
@@ -313,7 +309,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processo
   // Parse ICE-XA hostname to get hypercube location
   gethostname(name,namelen);
   int nscan = sscanf(name,"r%di%dn%d",&R,&I,&N) ;
-  assert(nscan==3);
+  GRID_ASSERT(nscan==3);
 
   int nlo = N%9;
   int nhi = N/9;
@@ -337,8 +333,8 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processo
   //////////////////////////////////////////////////////////////////
   MPI_Bcast(&rootcoor, sizeof(rootcoor), MPI_BYTE, 0, WorldComm); 
   hypercoor=hypercoor-rootcoor;
-  assert(hypercoor<WorldSize);
-  assert(hypercoor>=0);
+  GRID_ASSERT(hypercoor<WorldSize);
+  GRID_ASSERT(hypercoor>=0);
 
   //////////////////////////////////////
   // Printing
@@ -386,7 +382,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processo
   for(int i=0;i<ndimension;i++){
     Nprocessors*=processors[i];
   }
-  assert(WorldSize==Nprocessors);
+  GRID_ASSERT(WorldSize==Nprocessors);
 
   ////////////////////////////////////////////////////////////////
   // Establish mapping between lexico physics coord and WorldRank
@@ -405,7 +401,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processo
   // Build the new communicator
   /////////////////////////////////////////////////////////////////
   int ierr= MPI_Comm_split(WorldComm,0,rank,&optimal_comm);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const Coordinate &processors,Grid_MPI_Comm & optimal_comm,Coordinate &SHM)
 {
@@ -435,7 +431,8 @@ void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const Coordinate &proce
   for(int i=0;i<ndimension;i++){
     Nprocessors*=processors[i];
   }
-  assert(WorldSize==Nprocessors);
+  //  std::cerr << " WorldSize "<<WorldSize << " Nprocessors "<<Nprocessors<<" "<<processors<<std::endl; 
+  GRID_ASSERT(WorldSize==Nprocessors);
 
   ////////////////////////////////////////////////////////////////
   // Establish mapping between lexico physics coord and WorldRank
@@ -451,7 +448,7 @@ void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const Coordinate &proce
   // Build the new communicator
   /////////////////////////////////////////////////////////////////
   int ierr= MPI_Comm_split(WorldComm,0,rank,&optimal_comm);
-  assert(ierr==0);
+  GRID_ASSERT(ierr==0);
 }
 ////////////////////////////////////////////////////////////////////////////////////////////
 // SHMGET
@@ -460,8 +457,8 @@ void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const Coordinate &proce
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   std::cout << Mheader "SharedMemoryAllocate "<< bytes<< " shmget implementation "<<std::endl;
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
 
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
   // allocate the shared windows for our group
@@ -522,8 +519,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   void * ShmCommBuf ; 
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
 
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
   // allocate the pointer array for shared windows for our group
@@ -542,50 +539,22 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
   // Each MPI rank should allocate our own buffer
   ///////////////////////////////////////////////////////////////////////////////////////////////////////////
 #ifndef ACCELERATOR_AWARE_MPI
-  printf("Host buffer allocate for GPU non-aware MPI\n");
-#if 0
-  HostCommBuf= acceleratorAllocHost(bytes);
-#else 
+  // printf("Host buffer allocate for GPU non-aware MPI\n");
   HostCommBuf= malloc(bytes); /// CHANGE THIS TO malloc_host
-#ifdef HAVE_NUMAIF_H
-  #warning "Moving host buffers to specific NUMA domain"
-  int numa;
-  char *numa_name=(char *)getenv("MPI_BUF_NUMA");
-  if(numa_name) {
-    unsigned long page_size = sysconf(_SC_PAGESIZE);
-    numa = atoi(numa_name);
-    unsigned long page_count = bytes/page_size;
-    std::vector<void *> pages(page_count);
-    std::vector<int>    nodes(page_count,numa);
-    std::vector<int>    status(page_count,-1);
-    for(unsigned long p=0;p<page_count;p++){
-      pages[p] =(void *) ((uint64_t) HostCommBuf + p*page_size);
-    }
-    int ret = move_pages(0,
-			 page_count,
-			 &pages[0],
-			 &nodes[0],
-			 &status[0],
-			 MPOL_MF_MOVE);
-    printf("Host buffer move to numa domain %d : move_pages returned %d\n",numa,ret);
-    if (ret) perror(" move_pages failed for reason:");
-  }
-#endif  
-  acceleratorPin(HostCommBuf,bytes);
-#endif  
-
 #endif  
   ShmCommBuf = acceleratorAllocDevice(bytes);
   if (ShmCommBuf == (void *)NULL ) {
-    std::cerr << " SharedMemoryMPI.cc acceleratorAllocDevice failed NULL pointer for " << bytes<<" bytes " << std::endl;
+    std::cerr << "SharedMemoryMPI.cc acceleratorAllocDevice failed NULL pointer for " << bytes<<" bytes " << std::endl;
     exit(EXIT_FAILURE);  
   }
   if ( WorldRank == 0 ){
-    std::cout << WorldRank << Mheader " SharedMemoryMPI.cc acceleratorAllocDevice "<< bytes 
+    std::cout << Mheader " acceleratorAllocDevice "<< bytes 
 	      << "bytes at "<< std::hex<< ShmCommBuf << " - "<<(bytes-1+(uint64_t)ShmCommBuf) <<std::dec<<" for comms buffers " <<std::endl;
   }
   SharedMemoryZero(ShmCommBuf,bytes);
-  std::cout<< "Setting up IPC"<<std::endl;
+  if ( WorldRank == 0 ){
+    std::cout<< Mheader "Setting up IPC"<<std::endl;
+  }
   ///////////////////////////////////////////////////////////////////////////////////////////////////////////
   // Loop over ranks/gpu's on our node
   ///////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -616,8 +585,6 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
       if ( err != ZE_RESULT_SUCCESS ) {
 	std::cerr << "SharedMemoryMPI.cc zeMemGetIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
 	exit(EXIT_FAILURE);
-      } else {
-	std::cout << "SharedMemoryMPI.cc zeMemGetIpcHandle succeeded for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
       }
       memcpy((void *)&handle.fd,(void *)&ihandle,sizeof(int));
       handle.pid = getpid();
@@ -662,7 +629,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 			 MPI_BYTE,
 			 r,
 			 WorldShmComm);
-      assert(ierr==0);
+      GRID_ASSERT(ierr==0);
     }
     
     ///////////////////////////////////////////////////////////////
@@ -676,12 +643,12 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 #ifdef SHM_SOCKETS
       myfd=UnixSockets::RecvFileDescriptor();
 #else
-      std::cout<<"mapping seeking remote pid/fd "
-	       <<handle.pid<<"/"
-	       <<handle.fd<<std::endl;
+      //      std::cout<<"mapping seeking remote pid/fd "
+      //	       <<handle.pid<<"/"
+      //	       <<handle.fd<<std::endl;
 
       int pidfd = syscall(SYS_pidfd_open,handle.pid,0);
-      std::cout<<"Using IpcHandle pidfd "<<pidfd<<"\n";
+      //      std::cout<<"Using IpcHandle pidfd "<<pidfd<<"\n";
       //      int myfd  = syscall(SYS_pidfd_getfd,pidfd,handle.fd,0);
       myfd  = syscall(438,pidfd,handle.fd,0);
       int err_t = errno;
@@ -691,7 +658,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 	assert(0);
       }
 #endif
-      std::cout<<"Using IpcHandle mapped remote pid "<<handle.pid <<" FD "<<handle.fd <<" to myfd "<<myfd<<"\n";
+      //      std::cout<<"Using IpcHandle mapped remote pid "<<handle.pid <<" FD "<<handle.fd <<" to myfd "<<myfd<<"\n";
       memcpy((void *)&ihandle,(void *)&handle.ze,sizeof(ihandle));
       memcpy((void *)&ihandle,(void *)&myfd,sizeof(int));
 
@@ -700,11 +667,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 	std::cerr << "SharedMemoryMPI.cc "<<zeContext<<" "<<zeDevice<<std::endl;
 	std::cerr << "SharedMemoryMPI.cc zeMemOpenIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl; 
 	exit(EXIT_FAILURE);
-      } else {
-	std::cout << "SharedMemoryMPI.cc zeMemOpenIpcHandle succeeded for rank "<<r<<std::endl;
-	std::cout << "SharedMemoryMPI.cc zeMemOpenIpcHandle pointer is "<<std::hex<<thisBuf<<std::dec<<std::endl;
       }
-      assert(thisBuf!=nullptr);
+      GRID_ASSERT(thisBuf!=nullptr);
     }
 #endif
 #ifdef GRID_CUDA
@@ -745,8 +709,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   std::cout << Mheader "SharedMemoryAllocate "<< bytes<< " MMAP implementation "<< GRID_SHM_PATH <<std::endl;
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
   // allocate the shared windows for our group
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -776,13 +740,14 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
     void *ptr = (void *) mmap(NULL, bytes, PROT_READ | PROT_WRITE, mmap_flag,fd, 0); 
     if ( ptr == (void *)MAP_FAILED ) {    
       printf("mmap %s failed\n",shm_name);
-      perror("failed mmap");      assert(0);    
+      perror("failed mmap");      GRID_ASSERT(0);    
     }
-    assert(((uint64_t)ptr&0x3F)==0);
+    GRID_ASSERT(((uint64_t)ptr&0x3F)==0);
     close(fd);
     WorldShmCommBufs[r] =ptr;
     //    std::cout << Mheader "Set WorldShmCommBufs["<<r<<"]="<<ptr<< "("<< bytes<< "bytes)"<<std::endl;
   }
+  std::cout<< Mheader " Intra-node IPC setup is complete "<<std::endl;
   _ShmAlloc=1;
   _ShmAllocBytes  = bytes;
 };
@@ -792,8 +757,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   std::cout << Mheader "SharedMemoryAllocate "<< bytes<< " MMAP anonymous implementation "<<std::endl;
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
   // allocate the shared windows for our group
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -804,7 +769,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
   // Hugetlbf and others map filesystems as mappable huge pages
   ////////////////////////////////////////////////////////////////////////////////////////////
   char shm_name [NAME_MAX];
-  assert(WorldShmSize == 1);
+  GRID_ASSERT(WorldShmSize == 1);
   for(int r=0;r<WorldShmSize;r++){
     
     int fd=-1;
@@ -818,9 +783,9 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
     void *ptr = (void *) mmap(NULL, bytes, PROT_READ | PROT_WRITE, mmap_flag,fd, 0); 
     if ( ptr == (void *)MAP_FAILED ) {    
       printf("mmap %s failed\n",shm_name);
-      perror("failed mmap");      assert(0);    
+      perror("failed mmap");      GRID_ASSERT(0);    
     }
-    assert(((uint64_t)ptr&0x3F)==0);
+    GRID_ASSERT(((uint64_t)ptr&0x3F)==0);
     close(fd);
     WorldShmCommBufs[r] =ptr;
     //    std::cout << "Set WorldShmCommBufs["<<r<<"]="<<ptr<< "("<< bytes<< "bytes)"<<std::endl;
@@ -839,8 +804,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 { 
   std::cout << Mheader "SharedMemoryAllocate "<< bytes<< " SHMOPEN implementation "<<std::endl;
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0); 
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0); 
   MPI_Barrier(WorldShmComm);
   WorldShmCommBufs.resize(WorldShmSize);
 
@@ -871,7 +836,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 	perror("failed mmap");     
 	assert(0);    
       }
-      assert(((uint64_t)ptr&0x3F)==0);
+      GRID_ASSERT(((uint64_t)ptr&0x3F)==0);
       
       WorldShmCommBufs[r] =ptr;
       close(fd);
@@ -892,8 +857,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
       if ( fd<0 ) {	perror("failed shm_open");	assert(0);      }
       
       void * ptr =  mmap(NULL,size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
-      if ( ptr == MAP_FAILED ) {       perror("failed mmap");      assert(0);    }
-      assert(((uint64_t)ptr&0x3F)==0);
+      if ( ptr == MAP_FAILED ) {       perror("failed mmap");      GRID_ASSERT(0);    }
+      GRID_ASSERT(((uint64_t)ptr&0x3F)==0);
       WorldShmCommBufs[r] =ptr;
 
       close(fd);
@@ -916,14 +881,14 @@ void GlobalSharedMemory::SharedMemoryZero(void *dest,size_t bytes)
   bzero(dest,bytes);
 #endif
 }
-void GlobalSharedMemory::SharedMemoryCopy(void *dest,void *src,size_t bytes)
-{
-#if defined(GRID_CUDA) || defined(GRID_HIP) || defined(GRID_SYCL)
-  acceleratorCopyToDevice(src,dest,bytes);
-#else   
-  bcopy(src,dest,bytes);
-#endif
-}
+//void GlobalSharedMemory::SharedMemoryCopy(void *dest,void *src,size_t bytes)
+//{
+//#if defined(GRID_CUDA) || defined(GRID_HIP) || defined(GRID_SYCL)
+//  acceleratorCopyToDevice(src,dest,bytes);
+//#else   
+//  bcopy(src,dest,bytes);
+//#endif
+//}
 ////////////////////////////////////////////////////////
 // Global shared functionality finished
 // Now move to per communicator functionality
@@ -950,7 +915,7 @@ void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
   //////////////////////////////////////////////////////////////////////
   // Map ShmRank to WorldShmRank and use the right buffer
   //////////////////////////////////////////////////////////////////////
-  assert (GlobalSharedMemory::ShmAlloc()==1);
+  GRID_ASSERT (GlobalSharedMemory::ShmAlloc()==1);
   heap_size = GlobalSharedMemory::ShmAllocBytes();
   for(int r=0;r<ShmSize;r++){
 
@@ -959,6 +924,7 @@ void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
     MPI_Allreduce(MPI_IN_PLACE,&wsr,1,MPI_UINT32_T,MPI_SUM,ShmComm);
 
     ShmCommBufs[r] = GlobalSharedMemory::WorldShmCommBufs[wsr];
+    //    std::cerr << " SetCommunicator rank "<<r<<" comm "<<ShmCommBufs[r] <<std::endl;
   }
   ShmBufferFreeAll();
 
@@ -989,7 +955,7 @@ void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
   }
 #endif
 
-  //SharedMemoryTest();
+  //  SharedMemoryTest();
 }
 //////////////////////////////////////////////////////////////////
 // On node barrier
@@ -1011,19 +977,18 @@ void SharedMemory::SharedMemoryTest(void)
        check[0]=GlobalSharedMemory::WorldNode;
        check[1]=r;
        check[2]=magic;
-       GlobalSharedMemory::SharedMemoryCopy( ShmCommBufs[r], check, 3*sizeof(uint64_t));
+       acceleratorCopyToDevice(check,ShmCommBufs[r],3*sizeof(uint64_t));
     }
   }
   ShmBarrier();
   for(uint64_t r=0;r<ShmSize;r++){
-    ShmBarrier();
-    GlobalSharedMemory::SharedMemoryCopy(check,ShmCommBufs[r], 3*sizeof(uint64_t));
-    ShmBarrier();
-    assert(check[0]==GlobalSharedMemory::WorldNode);
-    assert(check[1]==r);
-    assert(check[2]==magic);
-    ShmBarrier();
+    acceleratorCopyFromDevice(ShmCommBufs[r],check,3*sizeof(uint64_t));
+    GRID_ASSERT(check[0]==GlobalSharedMemory::WorldNode);
+    GRID_ASSERT(check[1]==r);
+    GRID_ASSERT(check[2]==magic);
   }
+  ShmBarrier();
+  std::cout << GridLogDebug << " SharedMemoryTest has passed "<<std::endl;
 }
 
 void *SharedMemory::ShmBuffer(int rank)
@@ -1038,12 +1003,14 @@ void *SharedMemory::ShmBuffer(int rank)
 void *SharedMemory::ShmBufferTranslate(int rank,void * local_p)
 {
   int gpeer = ShmRanks[rank];
-  assert(gpeer!=ShmRank); // never send to self
+  GRID_ASSERT(gpeer!=ShmRank); // never send to self
+  //  std::cout << "ShmBufferTranslate for rank " << rank<<" peer "<<gpeer<<std::endl;
   if (gpeer == MPI_UNDEFINED){
     return NULL;
   } else { 
     uint64_t offset = (uint64_t)local_p - (uint64_t)ShmCommBufs[ShmRank];
     uint64_t remote = (uint64_t)ShmCommBufs[gpeer]+offset;
+    //    std::cout << "ShmBufferTranslate : local,offset,remote "<<std::hex<<local_p<<" "<<offset<<" "<<remote<<std::dec<<std::endl;
     return (void *) remote;
   }
 }

Grid/communicator/SharedMemoryNone.cc

@@ -34,7 +34,7 @@ NAMESPACE_BEGIN(Grid);
 /*Construct from an MPI communicator*/
 void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
 {
-  assert(_ShmSetup==0);
+  GRID_ASSERT(_ShmSetup==0);
   WorldComm = 0;
   WorldRank = 0;
   WorldSize = 1;
@@ -62,8 +62,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   std::cout << header "SharedMemoryAllocate "<< bytes<< " GPU implementation "<<std::endl;
   void * ShmCommBuf ; 
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
 
   ///////////////////////////////////////////////////////////////////////////////////////////////////////////
   // Each MPI rank should allocate our own buffer
@@ -92,8 +92,8 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
   void * ShmCommBuf ; 
-  assert(_ShmSetup==1);
-  assert(_ShmAlloc==0);
+  GRID_ASSERT(_ShmSetup==1);
+  GRID_ASSERT(_ShmAlloc==0);
   int mmap_flag =0;
 #ifdef MAP_ANONYMOUS
   mmap_flag = mmap_flag| MAP_SHARED | MAP_ANONYMOUS;
@@ -122,17 +122,17 @@ void GlobalSharedMemory::SharedMemoryZero(void *dest,size_t bytes)
 {
   acceleratorMemSet(dest,0,bytes);
 }
-void GlobalSharedMemory::SharedMemoryCopy(void *dest,void *src,size_t bytes)
-{
-  acceleratorCopyToDevice(src,dest,bytes);
-}
+//void GlobalSharedMemory::SharedMemoryCopy(void *dest,void *src,size_t bytes)
+//{
+//  acceleratorCopyToDevice(src,dest,bytes);
+//}
 ////////////////////////////////////////////////////////
 // Global shared functionality finished
 // Now move to per communicator functionality
 ////////////////////////////////////////////////////////
 void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
 {
-  assert(GlobalSharedMemory::ShmAlloc()==1);
+  GRID_ASSERT(GlobalSharedMemory::ShmAlloc()==1);
   ShmRanks.resize(1);
   ShmCommBufs.resize(1);
   ShmRanks[0] = 0;

Grid/cshift/Cshift_common.h

@@ -202,7 +202,7 @@ template<class vobj> void Scatter_plane_simple (Lattice<vobj> &rhs,deviceVector<
   {
     auto buffer_p = & buffer[0];
     auto table = MapCshiftTable();
-    autoView( rhs_v, rhs, AcceleratorWrite);
+    autoView( rhs_v, rhs, AcceleratorWriteDiscard);
     accelerator_for(i,ent,vobj::Nsimd(),{
 	coalescedWrite(rhs_v[table[i].first],coalescedRead(buffer_p[table[i].second]));
     });
@@ -228,7 +228,7 @@ template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,ExtractPointerA
   if(cbmask ==0x3 ) {
     int _slice_stride = rhs.Grid()->_slice_stride[dimension];
     int _slice_block = rhs.Grid()->_slice_block[dimension];
-    autoView( rhs_v , rhs, AcceleratorWrite);
+    autoView( rhs_v , rhs, AcceleratorWriteDiscard);
     accelerator_for(nn,e1*e2,1,{
 	int n = nn%e1;
 	int b = nn/e1;
@@ -240,9 +240,9 @@ template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,ExtractPointerA
 
     // Case of SIMD split AND checker dim cannot currently be hit, except in 
     // Test_cshift_red_black code.
-    std::cout << "Scatter_plane merge assert(0); think this is buggy FIXME "<< std::endl;// think this is buggy FIXME
+    std::cout << "Scatter_plane merge GRID_ASSERT(0); think this is buggy FIXME "<< std::endl;// think this is buggy FIXME
     std::cout<<" Unthreaded warning -- buffer is not densely packed ??"<<std::endl;
-    assert(0); // This will fail if hit on GPU
+    GRID_ASSERT(0); // This will fail if hit on GPU
     autoView( rhs_v, rhs, CpuWrite);
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
@@ -302,7 +302,7 @@ template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs
   {
     auto table = MapCshiftTable();
     autoView(rhs_v , rhs, AcceleratorRead);
-    autoView(lhs_v , lhs, AcceleratorWrite);
+    autoView(lhs_v , lhs, AcceleratorWriteDiscard);
     accelerator_for(i,ent,vobj::Nsimd(),{
       coalescedWrite(lhs_v[table[i].first],coalescedRead(rhs_v[table[i].second]));
     });

Grid/cshift/Cshift_mpi.h

@@ -29,8 +29,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef _GRID_CSHIFT_MPI_H_
 #define _GRID_CSHIFT_MPI_H_
 
-
 NAMESPACE_BEGIN(Grid); 
+
+#ifdef GRID_CHECKSUM_COMMS
+extern uint64_t checksum_index;
+#endif
+
 const int Cshift_verbose=0;
 template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension,int shift)
 {
@@ -45,6 +49,20 @@ template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension
   // Map to always positive shift modulo global full dimension.
   shift = (shift+fd)%fd;
 
+  if( shift ==0 ) {
+    ret = rhs;
+    return ret;
+  }
+  //
+  // Potential easy fast cases:
+  // Shift is a multiple of the local lattice extent.
+  // Then need only to shift whole subvolumes
+  int L = rhs.Grid()->_ldimensions[dimension];
+  if ( (shift%L )==0 && !rhs.Grid()->CheckerBoarded(dimension) ) {
+    Cshift_simple(ret,rhs,dimension,shift);
+    return ret;
+  }
+  
   ret.Checkerboard() = rhs.Grid()->CheckerBoardDestination(rhs.Checkerboard(),shift,dimension);
         
   // the permute type
@@ -68,7 +86,56 @@ template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension
   if(Cshift_verbose) std::cout << GridLogPerformance << "Cshift took "<< (t1-t0)/1e3 << " ms"<<std::endl;
   return ret;
 }
-#if 1
+
+template<class vobj> void Cshift_simple(Lattice<vobj>& ret,const Lattice<vobj> &rhs,int dimension,int shift)
+{
+  GridBase *grid=rhs.Grid();
+  int comm_proc, xmit_to_rank, recv_from_rank;
+  
+  int fd              = rhs.Grid()->_fdimensions[dimension];
+  int rd              = rhs.Grid()->_rdimensions[dimension];
+  int ld              = rhs.Grid()->_ldimensions[dimension];
+  int pd              = rhs.Grid()->_processors[dimension];
+  int simd_layout     = rhs.Grid()->_simd_layout[dimension];
+  int comm_dim        = rhs.Grid()->_processors[dimension] >1 ;
+
+  comm_proc = ((shift)/ld)%pd;
+
+  grid->ShiftedRanks(dimension,comm_proc,xmit_to_rank,recv_from_rank);
+  if(comm_dim) {
+
+    int64_t bytes = sizeof(vobj) * grid->oSites();
+
+    autoView(rhs_v , rhs, AcceleratorRead);
+    autoView(ret_v , ret, AcceleratorWrite);
+    void *send_buf  = (void *)&rhs_v[0];
+    void *recv_buf  = (void *)&ret_v[0];
+
+#ifdef ACCELERATOR_AWARE_MPI
+    grid->SendToRecvFrom(send_buf,
+			 xmit_to_rank,
+			 recv_buf,
+			 recv_from_rank,
+			 bytes);
+#else
+    static hostVector<vobj> hrhs; hrhs.resize(grid->oSites());
+    static hostVector<vobj> hret; hret.resize(grid->oSites());
+
+    void *hsend_buf = (void *)&hrhs[0];
+    void *hrecv_buf = (void *)&hret[0];
+
+    acceleratorCopyFromDevice(send_buf,hsend_buf,bytes);
+
+    grid->SendToRecvFrom(hsend_buf,
+			 xmit_to_rank,
+			 hrecv_buf,
+			 recv_from_rank,
+			 bytes);
+
+    acceleratorCopyToDevice(hrecv_buf,recv_buf,bytes);
+#endif
+  }
+}
 template<class vobj> void Cshift_comms(Lattice<vobj>& ret,const Lattice<vobj> &rhs,int dimension,int shift)
 {
   int sshift[2];
@@ -117,15 +184,20 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
   int pd              = rhs.Grid()->_processors[dimension];
   int simd_layout     = rhs.Grid()->_simd_layout[dimension];
   int comm_dim        = rhs.Grid()->_processors[dimension] >1 ;
-  assert(simd_layout==1);
-  assert(comm_dim==1);
-  assert(shift>=0);
-  assert(shift<fd);
+  GRID_ASSERT(simd_layout==1);
+  GRID_ASSERT(comm_dim==1);
+  GRID_ASSERT(shift>=0);
+  GRID_ASSERT(shift<fd);
   
   int buffer_size = rhs.Grid()->_slice_nblock[dimension]*rhs.Grid()->_slice_block[dimension];
   static deviceVector<vobj> send_buf; send_buf.resize(buffer_size);
   static deviceVector<vobj> recv_buf; recv_buf.resize(buffer_size);
-    
+#ifndef ACCELERATOR_AWARE_MPI
+  int pad = (8 + sizeof(vobj) - 1) / sizeof(vobj);
+  static hostVector<vobj> hsend_buf; hsend_buf.resize(buffer_size+pad);
+  static hostVector<vobj> hrecv_buf; hrecv_buf.resize(buffer_size+pad);
+#endif
+  
   int cb= (cbmask==0x2)? Odd : Even;
   int sshift= rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
   RealD tcopy=0.0;
@@ -139,9 +211,11 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
     int comm_proc = ((x+sshift)/rd)%pd;
     
     if (comm_proc==0) {
+      FlightRecorder::StepLog("Cshift_Copy_plane");
       tcopy-=usecond();
       Copy_plane(ret,rhs,dimension,x,sx,cbmask); 
       tcopy+=usecond();
+      FlightRecorder::StepLog("Cshift_Copy_plane_complete");
     } else {
 
       int words = buffer_size;
@@ -149,26 +223,71 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
 
       int bytes = words * sizeof(vobj);
 
+      FlightRecorder::StepLog("Cshift_Gather_plane");
       tgather-=usecond();
       Gather_plane_simple (rhs,send_buf,dimension,sx,cbmask);
       tgather+=usecond();
+      FlightRecorder::StepLog("Cshift_Gather_plane_complete");
 
       //      int rank           = grid->_processor;
       int recv_from_rank;
       int xmit_to_rank;
+
       grid->ShiftedRanks(dimension,comm_proc,xmit_to_rank,recv_from_rank);
       
       tcomms-=usecond();
       grid->Barrier();
 
+      FlightRecorder::StepLog("Cshift_SendRecv");
+#ifdef ACCELERATOR_AWARE_MPI
       grid->SendToRecvFrom((void *)&send_buf[0],
 			   xmit_to_rank,
 			   (void *)&recv_buf[0],
 			   recv_from_rank,
 			   bytes);
+#else
+      // bouncy bouncy
+      acceleratorCopyFromDevice(&send_buf[0],&hsend_buf[0],bytes);
+
+#ifdef GRID_CHECKSUM_COMMS
+      GRID_ASSERT(bytes % 8 == 0);
+      checksum_index++;
+      uint64_t xsum = checksum_gpu((uint64_t*)&send_buf[0], bytes / 8) ^ (1 + checksum_index);
+      *(uint64_t*)(((char*)&hsend_buf[0]) + bytes) = xsum;
+      bytes += 8;
+#endif
+
+      grid->SendToRecvFrom((void *)&hsend_buf[0],
+			   xmit_to_rank,
+			   (void *)&hrecv_buf[0],
+			   recv_from_rank,
+			   bytes);
+
+#ifdef GRID_CHECKSUM_COMMS
+      bytes -= 8;
+      acceleratorCopyToDevice(&hrecv_buf[0],&recv_buf[0],bytes);
+      uint64_t expected_cs = *(uint64_t*)(((char*)&hrecv_buf[0]) + bytes);
+      uint64_t computed_cs = checksum_gpu((uint64_t*)&recv_buf[0], bytes / 8) ^ (1 + checksum_index);
+      std::cout << GridLogComms<< " Cshift: "
+		<<" dim"<<dimension
+		<<" shift "<<shift
+		<< " rank "<< grid->ThisRank()
+		<<" Coor "<<grid->ThisProcessorCoor()
+		<<" send "<<xsum<<" to   "<<xmit_to_rank
+		<<" recv "<<computed_cs<<" from "<<recv_from_rank
+		<<std::endl;
+      GRID_ASSERT(expected_cs == computed_cs);
+#else
+      acceleratorCopyToDevice(&hrecv_buf[0],&recv_buf[0],bytes);
+#endif
+
+#endif
+      FlightRecorder::StepLog("Cshift_SendRecv_complete");
+
       xbytes+=bytes;
       grid->Barrier();
       tcomms+=usecond();
+      FlightRecorder::StepLog("Cshift_barrier_complete");
 
       tscatter-=usecond();
       Scatter_plane_simple (ret,recv_buf,dimension,x,cbmask);
@@ -203,10 +322,10 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
   //	    << " ld "<<ld<<" pd " << pd<<" simd_layout "<<simd_layout 
   //	    << " comm_dim " << comm_dim << " cbmask " << cbmask <<std::endl;
 
-  assert(comm_dim==1);
-  assert(simd_layout==2);
-  assert(shift>=0);
-  assert(shift<fd);
+  GRID_ASSERT(comm_dim==1);
+  GRID_ASSERT(simd_layout==2);
+  GRID_ASSERT(shift>=0);
+  GRID_ASSERT(shift<fd);
 
   RealD tcopy=0.0;
   RealD tgather=0.0;
@@ -226,12 +345,24 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
   static std::vector<deviceVector<scalar_object> >  recv_buf_extract; recv_buf_extract.resize(Nsimd);
   scalar_object *  recv_buf_extract_mpi;
   scalar_object *  send_buf_extract_mpi;
- 
+
   for(int s=0;s<Nsimd;s++){
     send_buf_extract[s].resize(buffer_size);
     recv_buf_extract[s].resize(buffer_size);
   }
+#ifndef ACCELERATOR_AWARE_MPI
+#ifdef GRID_CHECKSUM_COMMS
+  buffer_size += (8 + sizeof(vobj) - 1) / sizeof(vobj);
+#endif
 
+  static hostVector<vobj> hsend_buf; hsend_buf.resize(buffer_size);
+  static hostVector<vobj> hrecv_buf; hrecv_buf.resize(buffer_size);
+
+#ifdef GRID_CHECKSUM_COMMS
+  buffer_size -= (8 + sizeof(vobj) - 1) / sizeof(vobj);
+#endif
+#endif
+  
   int bytes = buffer_size*sizeof(scalar_object);
 
   ExtractPointerArray<scalar_object>  pointers(Nsimd); // 
@@ -273,7 +404,7 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
 
       if (nbr_ic) nbr_lane|=inner_bit;
 
-      assert (sx == nbr_ox);
+      GRID_ASSERT (sx == nbr_ox);
 
       if(nbr_proc){
 	grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank); 
@@ -283,11 +414,47 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
 
 	send_buf_extract_mpi = &send_buf_extract[nbr_lane][0];
 	recv_buf_extract_mpi = &recv_buf_extract[i][0];
+#ifdef ACCELERATOR_AWARE_MPI
 	grid->SendToRecvFrom((void *)send_buf_extract_mpi,
 			     xmit_to_rank,
 			     (void *)recv_buf_extract_mpi,
 			     recv_from_rank,
 			     bytes);
+#else
+      // bouncy bouncy
+	acceleratorCopyFromDevice((void *)send_buf_extract_mpi,(void *)&hsend_buf[0],bytes);
+#ifdef GRID_CHECKSUM_COMMS
+	assert(bytes % 8 == 0);
+	checksum_index++;
+	uint64_t xsum = checksum_gpu((uint64_t*)send_buf_extract_mpi, bytes / 8) ^ (1 + checksum_index);
+	*(uint64_t*)(((char*)&hsend_buf[0]) + bytes) = xsum;
+	bytes += 8;
+#endif
+	grid->SendToRecvFrom((void *)&hsend_buf[0],
+			     xmit_to_rank,
+			     (void *)&hrecv_buf[0],
+			     recv_from_rank,
+			     bytes);
+#ifdef GRID_CHECKSUM_COMMS
+	bytes -= 8;
+	acceleratorCopyToDevice((void *)&hrecv_buf[0],(void *)recv_buf_extract_mpi,bytes);
+	uint64_t expected_cs = *(uint64_t*)(((char*)&hrecv_buf[0]) + bytes);
+	uint64_t computed_cs = checksum_gpu((uint64_t*)recv_buf_extract_mpi, bytes / 8) ^ (1 + checksum_index);
+
+	std::cout << GridLogComms<< " Cshift_comms_simd: "
+		<<" dim"<<dimension
+		<<" shift "<<shift
+		<< " rank "<< grid->ThisRank()
+		<<" Coor "<<grid->ThisProcessorCoor()
+		<<" send "<<xsum<<" to   "<<xmit_to_rank
+		<<" recv "<<computed_cs<<" from "<<recv_from_rank
+		<<std::endl;
+	assert(expected_cs == computed_cs);
+#else
+	acceleratorCopyToDevice((void *)&hrecv_buf[0],(void *)recv_buf_extract_mpi,bytes);
+#endif
+
+#endif
 
 	xbytes+=bytes;
 	grid->Barrier();
@@ -311,234 +478,6 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
     std::cout << GridLogPerformance << " Cshift BW      "<<(2.0*xbytes)/tcomms<<" MB/s "<<2*xbytes<< " Bytes "<<std::endl;
   }
 }
-#else 
-template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
-{
-  typedef typename vobj::vector_type vector_type;
-  typedef typename vobj::scalar_type scalar_type;
-
-  GridBase *grid=rhs.Grid();
-  Lattice<vobj> temp(rhs.Grid());
-
-  int fd              = rhs.Grid()->_fdimensions[dimension];
-  int rd              = rhs.Grid()->_rdimensions[dimension];
-  int pd              = rhs.Grid()->_processors[dimension];
-  int simd_layout     = rhs.Grid()->_simd_layout[dimension];
-  int comm_dim        = rhs.Grid()->_processors[dimension] >1 ;
-  assert(simd_layout==1);
-  assert(comm_dim==1);
-  assert(shift>=0);
-  assert(shift<fd);
-  RealD tcopy=0.0;
-  RealD tgather=0.0;
-  RealD tscatter=0.0;
-  RealD tcomms=0.0;
-  uint64_t xbytes=0;
-  
-  int buffer_size = rhs.Grid()->_slice_nblock[dimension]*rhs.Grid()->_slice_block[dimension];
-  static cshiftVector<vobj> send_buf_v; send_buf_v.resize(buffer_size);
-  static cshiftVector<vobj> recv_buf_v; recv_buf_v.resize(buffer_size);
-  vobj *send_buf;
-  vobj *recv_buf;
-  {
-    grid->ShmBufferFreeAll();
-    size_t bytes = buffer_size*sizeof(vobj);
-    send_buf=(vobj *)grid->ShmBufferMalloc(bytes);
-    recv_buf=(vobj *)grid->ShmBufferMalloc(bytes);
-  }
-    
-  int cb= (cbmask==0x2)? Odd : Even;
-  int sshift= rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
-
-  for(int x=0;x<rd;x++){       
-
-    int sx        =  (x+sshift)%rd;
-    int comm_proc = ((x+sshift)/rd)%pd;
-    
-    if (comm_proc==0) {
-
-      tcopy-=usecond();
-      Copy_plane(ret,rhs,dimension,x,sx,cbmask); 
-      tcopy+=usecond();
-
-    } else {
-
-      int words = buffer_size;
-      if (cbmask != 0x3) words=words>>1;
-
-      int bytes = words * sizeof(vobj);
-
-      tgather-=usecond();
-      Gather_plane_simple (rhs,send_buf_v,dimension,sx,cbmask);
-      tgather+=usecond();
-
-      //      int rank           = grid->_processor;
-      int recv_from_rank;
-      int xmit_to_rank;
-      grid->ShiftedRanks(dimension,comm_proc,xmit_to_rank,recv_from_rank);
-
-
-      tcomms-=usecond();
-      //      grid->Barrier();
-
-      acceleratorCopyDeviceToDevice((void *)&send_buf_v[0],(void *)&send_buf[0],bytes);
-      grid->SendToRecvFrom((void *)&send_buf[0],
-			   xmit_to_rank,
-			   (void *)&recv_buf[0],
-			   recv_from_rank,
-			   bytes);
-      xbytes+=bytes;
-      acceleratorCopyDeviceToDevice((void *)&recv_buf[0],(void *)&recv_buf_v[0],bytes);
-
-      //      grid->Barrier();
-      tcomms+=usecond();
-
-      tscatter-=usecond();
-      Scatter_plane_simple (ret,recv_buf_v,dimension,x,cbmask);
-      tscatter+=usecond();
-    }
-  }
-  if(Cshift_verbose){
-    std::cout << GridLogPerformance << " Cshift copy    "<<tcopy/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift gather  "<<tgather/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift scatter "<<tscatter/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift comm    "<<tcomms/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift BW      "<<(2.0*xbytes)/tcomms<<" MB/s "<<2*xbytes<< " Bytes "<<std::endl;
-  }
-}
-
-template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
-{
-  GridBase *grid=rhs.Grid();
-  const int Nsimd = grid->Nsimd();
-  typedef typename vobj::vector_type vector_type;
-  typedef typename vobj::scalar_object scalar_object;
-  typedef typename vobj::scalar_type scalar_type;
-   
-  int fd = grid->_fdimensions[dimension];
-  int rd = grid->_rdimensions[dimension];
-  int ld = grid->_ldimensions[dimension];
-  int pd = grid->_processors[dimension];
-  int simd_layout     = grid->_simd_layout[dimension];
-  int comm_dim        = grid->_processors[dimension] >1 ;
-
-  //std::cout << "Cshift_comms_simd dim "<< dimension << " fd "<<fd<<" rd "<<rd
-  //    << " ld "<<ld<<" pd " << pd<<" simd_layout "<<simd_layout 
-  //    << " comm_dim " << comm_dim << " cbmask " << cbmask <<std::endl;
-
-  assert(comm_dim==1);
-  assert(simd_layout==2);
-  assert(shift>=0);
-  assert(shift<fd);
-  RealD tcopy=0.0;
-  RealD tgather=0.0;
-  RealD tscatter=0.0;
-  RealD tcomms=0.0;
-  uint64_t xbytes=0;
-
-  int permute_type=grid->PermuteType(dimension);
-
-  ///////////////////////////////////////////////
-  // Simd direction uses an extract/merge pair
-  ///////////////////////////////////////////////
-  int buffer_size = grid->_slice_nblock[dimension]*grid->_slice_block[dimension];
-  //  int words = sizeof(vobj)/sizeof(vector_type);
-
-  static std::vector<cshiftVector<scalar_object> >  send_buf_extract; send_buf_extract.resize(Nsimd);
-  static std::vector<cshiftVector<scalar_object> >  recv_buf_extract; recv_buf_extract.resize(Nsimd);
-  scalar_object *  recv_buf_extract_mpi;
-  scalar_object *  send_buf_extract_mpi;
-  {
-    size_t bytes = sizeof(scalar_object)*buffer_size;
-    grid->ShmBufferFreeAll();
-    send_buf_extract_mpi = (scalar_object *)grid->ShmBufferMalloc(bytes);
-    recv_buf_extract_mpi = (scalar_object *)grid->ShmBufferMalloc(bytes);
-  }
-  for(int s=0;s<Nsimd;s++){
-    send_buf_extract[s].resize(buffer_size);
-    recv_buf_extract[s].resize(buffer_size);
-  }
-
-  int bytes = buffer_size*sizeof(scalar_object);
-
-  ExtractPointerArray<scalar_object>  pointers(Nsimd); // 
-  ExtractPointerArray<scalar_object> rpointers(Nsimd); // received pointers
-
-  ///////////////////////////////////////////
-  // Work out what to send where
-  ///////////////////////////////////////////
-  int cb    = (cbmask==0x2)? Odd : Even;
-  int sshift= grid->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
-
-  // loop over outer coord planes orthog to dim
-  for(int x=0;x<rd;x++){       
-
-    // FIXME call local permute copy if none are offnode.
-    for(int i=0;i<Nsimd;i++){       
-      pointers[i] = &send_buf_extract[i][0];
-    }
-    tgather-=usecond();
-    int sx   = (x+sshift)%rd;
-    Gather_plane_extract(rhs,pointers,dimension,sx,cbmask);
-    tgather+=usecond();
-
-    for(int i=0;i<Nsimd;i++){
-      
-      int inner_bit = (Nsimd>>(permute_type+1));
-      int ic= (i&inner_bit)? 1:0;
-
-      int my_coor          = rd*ic + x;
-      int nbr_coor         = my_coor+sshift;
-      int nbr_proc = ((nbr_coor)/ld) % pd;// relative shift in processors
-
-      int nbr_ic   = (nbr_coor%ld)/rd;    // inner coord of peer
-      int nbr_ox   = (nbr_coor%rd);       // outer coord of peer
-      int nbr_lane = (i&(~inner_bit));
-
-      int recv_from_rank;
-      int xmit_to_rank;
-
-      if (nbr_ic) nbr_lane|=inner_bit;
-
-      assert (sx == nbr_ox);
-
-      if(nbr_proc){
-	grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank); 
-
-	tcomms-=usecond();
-	//	grid->Barrier();
-
-	acceleratorCopyDeviceToDevice((void *)&send_buf_extract[nbr_lane][0],(void *)send_buf_extract_mpi,bytes);
-	grid->SendToRecvFrom((void *)send_buf_extract_mpi,
-			     xmit_to_rank,
-			     (void *)recv_buf_extract_mpi,
-			     recv_from_rank,
-			     bytes);
-	acceleratorCopyDeviceToDevice((void *)recv_buf_extract_mpi,(void *)&recv_buf_extract[i][0],bytes);
-	xbytes+=bytes;
-
-	//	grid->Barrier();
-	tcomms+=usecond();
-	rpointers[i] = &recv_buf_extract[i][0];
-      } else { 
-	rpointers[i] = &send_buf_extract[nbr_lane][0];
-      }
-
-    }
-    tscatter-=usecond();
-    Scatter_plane_merge(ret,rpointers,dimension,x,cbmask);
-    tscatter+=usecond();
-
-  }
-  if(Cshift_verbose){
-    std::cout << GridLogPerformance << " Cshift (s) copy    "<<tcopy/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift (s) gather  "<<tgather/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift (s) scatter "<<tscatter/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift (s) comm    "<<tcomms/1e3<<" ms"<<std::endl;
-    std::cout << GridLogPerformance << " Cshift BW      "<<(2.0*xbytes)/tcomms<<" MB/s"<<std::endl;
-  }
-}
-#endif
 
 NAMESPACE_END(Grid); 
 

Grid/json/json.hpp

@@ -21895,7 +21895,7 @@ inline void swap(nlohmann::NLOHMANN_BASIC_JSON_TPL& j1, nlohmann::NLOHMANN_BASIC
 /// @brief user-defined string literal for JSON values
 /// @sa https://json.nlohmann.me/api/basic_json/operator_literal_json/
 JSON_HEDLEY_NON_NULL(1)
-inline nlohmann::json operator "" _json(const char* s, std::size_t n)
+inline nlohmann::json operator ""_json(const char* s, std::size_t n)
 {
     return nlohmann::json::parse(s, s + n);
 }
@@ -21903,7 +21903,7 @@ inline nlohmann::json operator "" _json(const char* s, std::size_t n)
 /// @brief user-defined string literal for JSON pointer
 /// @sa https://json.nlohmann.me/api/basic_json/operator_literal_json_pointer/
 JSON_HEDLEY_NON_NULL(1)
-inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std::size_t n)
+inline nlohmann::json::json_pointer operator ""_json_pointer(const char* s, std::size_t n)
 {
     return nlohmann::json::json_pointer(std::string(s, n));
 }

Grid/lattice/Lattice_ET.h

@@ -74,7 +74,7 @@ accelerator_inline vobj predicatedWhere(const iobj &predicate,
 
   extract(iftrue, truevals);
   extract(iffalse, falsevals);
-  extract<vInteger, Integer>(TensorRemove(predicate), mask);
+  extract(TensorRemove(predicate), mask);
 
   for (int s = 0; s < Nsimd; s++) {
     if (mask[s]) falsevals[s] = truevals[s];
@@ -245,7 +245,7 @@ template <class T1,typename std::enable_if<is_lattice<T1>::value, T1>::type * =
 inline void CBFromExpression(int &cb, const T1 &lat)  // Lattice leaf
 {
   if ((cb == Odd) || (cb == Even)) {
-    assert(cb == lat.Checkerboard());
+    GRID_ASSERT(cb == lat.Checkerboard());
   }
   cb = lat.Checkerboard();
 }

Grid/lattice/Lattice_base.h

@@ -120,12 +120,12 @@ public:
     GRID_TRACE("ExpressionTemplateEval");
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
-    assert(egrid!=nullptr);
+    GRID_ASSERT(egrid!=nullptr);
     conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
     
     auto exprCopy = expr;
@@ -144,12 +144,12 @@ public:
     GRID_TRACE("ExpressionTemplateEval");
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
-    assert(egrid!=nullptr);
+    GRID_ASSERT(egrid!=nullptr);
     conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
 
     auto exprCopy = expr;
@@ -168,12 +168,12 @@ public:
     GRID_TRACE("ExpressionTemplateEval");
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
-    assert(egrid!=nullptr);
+    GRID_ASSERT(egrid!=nullptr);
     conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
@@ -191,11 +191,11 @@ public:
   Lattice(const LatticeUnaryExpression<Op,T1> & expr) {
     this->_grid = nullptr;
     GridFromExpression(this->_grid,expr);
-    assert(this->_grid!=nullptr);
+    GRID_ASSERT(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
 
     resize(this->_grid->oSites());
@@ -206,11 +206,11 @@ public:
   Lattice(const LatticeBinaryExpression<Op,T1,T2> & expr) {
     this->_grid = nullptr;
     GridFromExpression(this->_grid,expr);
-    assert(this->_grid!=nullptr);
+    GRID_ASSERT(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
 
     resize(this->_grid->oSites());
@@ -221,11 +221,11 @@ public:
   Lattice(const LatticeTrinaryExpression<Op,T1,T2,T3> & expr) {
     this->_grid = nullptr;
     GridFromExpression(this->_grid,expr);
-    assert(this->_grid!=nullptr);
+    GRID_ASSERT(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
-    assert( (cb==Odd) || (cb==Even));
+    GRID_ASSERT( (cb==Odd) || (cb==Even));
     this->checkerboard=cb;
 
     resize(this->_grid->oSites());
@@ -236,7 +236,7 @@ public:
   template<class sobj> inline Lattice<vobj> & operator = (const sobj & r){
     vobj vtmp;
     vtmp = r;
-#if 0
+#if 1
     deviceVector<vobj> vvtmp(1);
     acceleratorPut(vvtmp[0],vtmp);
     vobj *vvtmp_p = & vvtmp[0];
@@ -264,7 +264,7 @@ public:
   Lattice(GridBase *grid,ViewMode mode=AcceleratorWriteDiscard) { 
     this->_grid = grid;
     resize(this->_grid->oSites());
-    assert((((uint64_t)&this->_odata[0])&0xF) ==0);
+    GRID_ASSERT((((uint64_t)&this->_odata[0])&0xF) ==0);
     this->checkerboard=0;
     SetViewMode(mode);
   }
@@ -289,7 +289,7 @@ public:
   ///////////////////////////////////////////
   // move constructor
   ///////////////////////////////////////////
-  Lattice(Lattice && r){ 
+  Lattice(Lattice && r) noexcept { 
     this->_grid = r.Grid();
     this->_odata      = r._odata;
     this->_odata_size = r._odata_size;
@@ -330,7 +330,7 @@ public:
   ///////////////////////////////////////////
   // Move assignment possible if same type
   ///////////////////////////////////////////
-  inline Lattice<vobj> & operator = (Lattice<vobj> && r){
+  inline Lattice<vobj> & operator = (Lattice<vobj> && r) noexcept {
 
     resize(0); // deletes if appropriate
     this->_grid       = r.Grid();

Grid/lattice/Lattice_basis.h

@@ -166,9 +166,9 @@ void basisReorderInPlace(std::vector<Field> &_v,std::vector<RealD>& sort_vals, s
 {
   int vlen = idx.size();
 
-  assert(vlen>=1);
-  assert(vlen<=sort_vals.size());
-  assert(vlen<=_v.size());
+  GRID_ASSERT(vlen>=1);
+  GRID_ASSERT(vlen<=sort_vals.size());
+  GRID_ASSERT(vlen<=_v.size());
 
   for (size_t i=0;i<vlen;i++) {
 
@@ -186,7 +186,7 @@ void basisReorderInPlace(std::vector<Field> &_v,std::vector<RealD>& sort_vals, s
 	if (idx[j]==i)
 	  break;
 
-      assert(idx[i] > i);     assert(j!=idx.size());      assert(idx[j]==i);
+      GRID_ASSERT(idx[i] > i);     GRID_ASSERT(j!=idx.size());      GRID_ASSERT(idx[j]==i);
 
       swap(_v[i],_v[idx[i]]); // should use vector move constructor, no data copy
       std::swap(sort_vals[i],sort_vals[idx[i]]);
@@ -224,7 +224,7 @@ void basisSortInPlace(std::vector<Field> & _v,std::vector<RealD>& sort_vals, boo
 template<class Field>
 void basisDeflate(const std::vector<Field> &_v,const std::vector<RealD>& eval,const Field& src_orig,Field& result) {
   result = Zero();
-  assert(_v.size()==eval.size());
+  GRID_ASSERT(_v.size()==eval.size());
   int N = (int)_v.size();
   for (int i=0;i<N;i++) {
     Field& tmp = _v[i];

Grid/lattice/Lattice_comparison.h

@@ -40,16 +40,16 @@ NAMESPACE_BEGIN(Grid);
 //Query supporting logical &&, ||, 
 //////////////////////////////////////////////////////////////////////////
 
-typedef iScalar<vInteger> vPredicate ;
+template<class vobj> using vPredicate = iScalar<IntegerPredicate<vobj> > ;
 
 //////////////////////////////////////////////////////////////////////////
 // compare lattice to lattice
 //////////////////////////////////////////////////////////////////////////
 
 template<class vfunctor,class lobj,class robj>  
-inline Lattice<vPredicate> LLComparison(vfunctor op,const Lattice<lobj> &lhs,const Lattice<robj> &rhs)
+inline Lattice<vPredicate<lobj> > LLComparison(vfunctor op,const Lattice<lobj> &lhs,const Lattice<robj> &rhs)
 {
-  Lattice<vPredicate> ret(rhs.Grid());
+  Lattice<vPredicate<lobj> > ret(rhs.Grid());
   autoView( lhs_v, lhs, CpuRead);
   autoView( rhs_v, rhs, CpuRead);
   autoView( ret_v, ret, CpuWrite);
@@ -62,9 +62,9 @@ inline Lattice<vPredicate> LLComparison(vfunctor op,const Lattice<lobj> &lhs,con
 // compare lattice to scalar
 //////////////////////////////////////////////////////////////////////////
 template<class vfunctor,class lobj,class robj> 
-inline Lattice<vPredicate> LSComparison(vfunctor op,const Lattice<lobj> &lhs,const robj &rhs)
+inline Lattice<vPredicate<lobj> > LSComparison(vfunctor op,const Lattice<lobj> &lhs,const robj &rhs)
 {
-  Lattice<vPredicate> ret(lhs.Grid());
+  Lattice<vPredicate<lobj> > ret(lhs.Grid());
   autoView( lhs_v, lhs, CpuRead);
   autoView( ret_v, ret, CpuWrite);
   thread_for( ss, lhs_v.size(), {
@@ -76,9 +76,9 @@ inline Lattice<vPredicate> LSComparison(vfunctor op,const Lattice<lobj> &lhs,con
 // compare scalar to lattice
 //////////////////////////////////////////////////////////////////////////
 template<class vfunctor,class lobj,class robj> 
-inline Lattice<vPredicate> SLComparison(vfunctor op,const lobj &lhs,const Lattice<robj> &rhs)
+inline Lattice<vPredicate<robj> > SLComparison(vfunctor op,const lobj &lhs,const Lattice<robj> &rhs)
 {
-  Lattice<vPredicate> ret(rhs.Grid());
+  Lattice<vPredicate<robj> > ret(rhs.Grid());
   autoView( rhs_v, rhs, CpuRead);
   autoView( ret_v, ret, CpuWrite);
   thread_for( ss, rhs_v.size(), {
@@ -92,88 +92,151 @@ inline Lattice<vPredicate> SLComparison(vfunctor op,const lobj &lhs,const Lattic
 //////////////////////////////////////////////////////////////////////////
 // Less than
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator < (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator < (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(vlt<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator < (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator < (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(vlt<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator < (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator < (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(vlt<lobj,robj>(),lhs,rhs);
 }
   
 // Less than equal
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator <= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator <= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(vle<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator <= (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator <= (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(vle<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator <= (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator <= (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(vle<lobj,robj>(),lhs,rhs);
 }
   
 // Greater than 
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator > (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator > (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(vgt<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator > (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator > (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(vgt<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator > (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator > (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(vgt<lobj,robj>(),lhs,rhs);
 }
   
   
 // Greater than equal
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator >= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator >= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(vge<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator >= (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator >= (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(vge<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator >= (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator >= (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(vge<lobj,robj>(),lhs,rhs);
 }
    
 // equal
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator == (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator == (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(veq<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator == (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator == (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(veq<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator == (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator == (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(veq<lobj,robj>(),lhs,rhs);
 }
    
    
 // not equal
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator != (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<lobj> > operator != (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
   return LLComparison(vne<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator != (const Lattice<lobj> & lhs, const robj & rhs) {
+inline Lattice<vPredicate<lobj> > operator != (const Lattice<lobj> & lhs, const robj & rhs) {
   return LSComparison(vne<lobj,robj>(),lhs,rhs);
 }
 template<class lobj,class robj>
-inline Lattice<vPredicate> operator != (const lobj & lhs, const Lattice<robj> & rhs) {
+inline Lattice<vPredicate<robj> > operator != (const lobj & lhs, const Lattice<robj> & rhs) {
   return SLComparison(vne<lobj,robj>(),lhs,rhs);
 }
+
+//////////////////////////////////////////////////////////////////////////
+// Real-part relational comparison for COMPLEX lattices.
+//
+// Complex has no ordering, so operator<,>,<=,>= are (deliberately) undefined for
+// complex operands -- Comparison() in Lattice_comparison_utils.h is IfNotComplex --
+// and where() cannot be driven by a complex lattice directly.  It is however often
+// useful to threshold on the REAL PART of a complex (scalar/singlet) field, e.g. a
+// momentum-magnitude mask phat^2 > pc^2.  These free functions compare the real part
+// to a real threshold and return the matching IntegerPredicate, so the result feeds
+// where() exactly like the built-in relationals.
+//
+// Written with the per-lane getlane/putlane accessors (as in FFT.h, PaddedCell.h),
+// not extract/merge buffers.  One wrinkle: the predicate type IntegerPredicate<CComplex>
+// is vInteger, whose Nsimd (>= the widest real type's) exceeds the complex operand's
+// Nsimd by s = Npred/Nsimd.  where() only reads the ii=0 representative of each group,
+// at physical lane lane*s (the "s-fold skip" -- cf. extract()'s getlane(i*s)), so
+// filling the rest of the group is not functionally required; we replicate the value
+// across all s lanes anyway because some Grid code asserts the s replicas are equal
+// (and it mirrors what merge() does).  When vInteger is reworked to carry the operand
+// Nsimd (scope later) s becomes 1 and the inner loop drops out.
+//////////////////////////////////////////////////////////////////////////
+template<class scalar> class sRealLt { public:
+  accelerator_inline Integer operator()(const scalar &a, RealD b) const { return a.real() <  b ? 1 : 0; } };
+template<class scalar> class sRealLe { public:
+  accelerator_inline Integer operator()(const scalar &a, RealD b) const { return a.real() <= b ? 1 : 0; } };
+template<class scalar> class sRealGt { public:
+  accelerator_inline Integer operator()(const scalar &a, RealD b) const { return a.real() >  b ? 1 : 0; } };
+template<class scalar> class sRealGe { public:
+  accelerator_inline Integer operator()(const scalar &a, RealD b) const { return a.real() >= b ? 1 : 0; } };
+
+template<class sfunctor,class CComplex>
+inline Lattice<vPredicate<CComplex> > RealPartComparison(sfunctor op,const Lattice<CComplex> &lhs, RealD thr)
+{
+  Lattice<vPredicate<CComplex> > ret(lhs.Grid());
+  autoView( lv, lhs, AcceleratorRead);
+  autoView( rv, ret, AcceleratorWrite);
+  typedef typename CComplex::vector_type vsimd;
+  const int Nsimd = vsimd::Nsimd();
+  const int s     = IntegerPredicate<CComplex>::Nsimd() / Nsimd;   // lane-replication factor (see note)
+  accelerator_for(ss, lhs.Grid()->oSites(), Nsimd, {
+    vsimd v = TensorRemove(lv[ss]);                 // strip iScalar nest to the bare complex SIMD word
+#ifdef GRID_SIMT
+    { int lane = acceleratorSIMTlane(Nsimd);        // GPU: this thread == this operand lane
+#else
+    for(int lane=0;lane<Nsimd;lane++){              // CPU: walk the packed operand lanes
+#endif
+      Integer p = op(v.getlane(lane), thr);
+      for(int ii=0;ii<s;ii++) rv[ss]._internal.putlane(p, lane*s+ii);   // replicate across the group
+#ifdef GRID_SIMT
+    }
+#else
+    }
+#endif
+  });
+  return ret;
+}
+
+template<class CComplex> inline Lattice<vPredicate<CComplex> > RealPartLessThan    (const Lattice<CComplex> &a, RealD b){ return RealPartComparison(sRealLt<typename CComplex::vector_type::scalar_type>(),a,b); }
+template<class CComplex> inline Lattice<vPredicate<CComplex> > RealPartLessEqual    (const Lattice<CComplex> &a, RealD b){ return RealPartComparison(sRealLe<typename CComplex::vector_type::scalar_type>(),a,b); }
+template<class CComplex> inline Lattice<vPredicate<CComplex> > RealPartGreaterThan  (const Lattice<CComplex> &a, RealD b){ return RealPartComparison(sRealGt<typename CComplex::vector_type::scalar_type>(),a,b); }
+template<class CComplex> inline Lattice<vPredicate<CComplex> > RealPartGreaterEqual (const Lattice<CComplex> &a, RealD b){ return RealPartComparison(sRealGe<typename CComplex::vector_type::scalar_type>(),a,b); }
+
 NAMESPACE_END(Grid);
 #endif

Grid/lattice/Lattice_comparison_utils.h

@@ -39,47 +39,53 @@ NAMESPACE_BEGIN(Grid);
   //
   // Should guard this with isGridTensor<> enable if?
   /////////////////////////////////////////
+  //
+  // Predicate type matching the operand's vectorisation
+  //
+  template<class obj> using IntegerPredicate =
+    typename GridTypeMapper<typename GridTypeMapper<obj>::vector_type>::Integerified;
+
   //
   // Generic list of functors
   //
   template<class lobj,class robj> class veq {
   public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) == (rhs);
     }
   };
   template<class lobj,class robj> class vne {
   public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) != (rhs);
     }
   };
   template<class lobj,class robj> class vlt {
   public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) < (rhs);
     }
   };
   template<class lobj,class robj> class vle {
   public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) <= (rhs);
     }
   };
   template<class lobj,class robj> class vgt {
   public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) > (rhs);
     }
   };
   template<class lobj,class robj> class vge {
     public:
-    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator IntegerPredicate<lobj> operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) >= (rhs);
     }
@@ -133,78 +139,78 @@ NAMESPACE_BEGIN(Grid);
   // Integer and real get extra relational functions.
   //////////////////////////////////////////////////////////////////////////////////////////////////////
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    accelerator_inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const vsimd & rhs)
+    accelerator_inline IntegerPredicate<vsimd> Comparison(sfunctor sop,const vsimd & lhs, const vsimd & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
       ExtractBuffer<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
       ExtractBuffer<scalar> vrhs(vsimd::Nsimd());
       ExtractBuffer<Integer> vpred(vsimd::Nsimd());
-      vInteger ret;
+      IntegerPredicate<vsimd> ret;
       extract<vsimd,scalar>(lhs,vlhs);
       extract<vsimd,scalar>(rhs,vrhs);
       for(int s=0;s<vsimd::Nsimd();s++){
 	vpred[s] = sop(vlhs[s],vrhs[s]);
       }
-      merge<vInteger,Integer>(ret,vpred);
+      merge<IntegerPredicate<vsimd>,Integer>(ret,vpred);
       return ret;
     }
 
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    accelerator_inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const typename vsimd::scalar_type & rhs)
+    accelerator_inline IntegerPredicate<vsimd> Comparison(sfunctor sop,const vsimd & lhs, const typename vsimd::scalar_type & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
       ExtractBuffer<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
       ExtractBuffer<Integer> vpred(vsimd::Nsimd());
-      vInteger ret;
+      IntegerPredicate<vsimd> ret;
       extract<vsimd,scalar>(lhs,vlhs);
       for(int s=0;s<vsimd::Nsimd();s++){
 	vpred[s] = sop(vlhs[s],rhs);
       }
-      merge<vInteger,Integer>(ret,vpred);
+      merge<IntegerPredicate<vsimd>,Integer>(ret,vpred);
       return ret;
     }
 
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    accelerator_inline vInteger Comparison(sfunctor sop,const typename vsimd::scalar_type & lhs, const vsimd & rhs)
+    accelerator_inline IntegerPredicate<vsimd> Comparison(sfunctor sop,const typename vsimd::scalar_type & lhs, const vsimd & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
       ExtractBuffer<scalar> vrhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
       ExtractBuffer<Integer> vpred(vsimd::Nsimd());
-      vInteger ret;
+      IntegerPredicate<vsimd> ret;
       extract<vsimd,scalar>(rhs,vrhs);
       for(int s=0;s<vsimd::Nsimd();s++){
 	vpred[s] = sop(lhs,vrhs[s]);
       }
-      merge<vInteger,Integer>(ret,vpred);
+      merge<IntegerPredicate<vsimd>,Integer>(ret,vpred);
       return ret;
     }
 
 #define DECLARE_RELATIONAL_EQ(op,functor) \
   template<class vsimd,IfSimd<vsimd> = 0>\
-    accelerator_inline vInteger operator op (const vsimd & lhs, const vsimd & rhs)\
+    accelerator_inline IntegerPredicate<vsimd> operator op (const vsimd & lhs, const vsimd & rhs)\
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd,IfSimd<vsimd> = 0>\
-    accelerator_inline vInteger operator op (const vsimd & lhs, const typename vsimd::scalar_type & rhs) \
+    accelerator_inline IntegerPredicate<vsimd> operator op (const vsimd & lhs, const typename vsimd::scalar_type & rhs) \
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd,IfSimd<vsimd> = 0>\
-    accelerator_inline vInteger operator op (const typename vsimd::scalar_type & lhs, const vsimd & rhs) \
+    accelerator_inline IntegerPredicate<vsimd> operator op (const typename vsimd::scalar_type & lhs, const vsimd & rhs) \
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd>\
-    accelerator_inline vInteger operator op(const iScalar<vsimd> &lhs,const typename vsimd::scalar_type &rhs) \
+    accelerator_inline IntegerPredicate<vsimd> operator op(const iScalar<vsimd> &lhs,const typename vsimd::scalar_type &rhs) \
     {									\
       return lhs._internal op rhs;					\
     }									\
   template<class vsimd>\
-    accelerator_inline vInteger operator op(const typename vsimd::scalar_type &lhs,const iScalar<vsimd> &rhs) \
+    accelerator_inline IntegerPredicate<vsimd> operator op(const typename vsimd::scalar_type &lhs,const iScalar<vsimd> &rhs) \
     {									\
       return lhs op rhs._internal;					\
     }									\
@@ -212,7 +218,7 @@ NAMESPACE_BEGIN(Grid);
 #define DECLARE_RELATIONAL(op,functor) \
   DECLARE_RELATIONAL_EQ(op,functor)    \
   template<class vsimd>\
-    accelerator_inline vInteger operator op(const iScalar<vsimd> &lhs,const iScalar<vsimd> &rhs)\
+    accelerator_inline IntegerPredicate<vsimd> operator op(const iScalar<vsimd> &lhs,const iScalar<vsimd> &rhs)\
     {									\
       return lhs._internal op rhs._internal;				\
     }									

Grid/lattice/Lattice_conformable.h

@@ -32,8 +32,8 @@ NAMESPACE_BEGIN(Grid);
 
 template<class obj1,class obj2> void conformable(const Lattice<obj1> &lhs,const Lattice<obj2> &rhs)
 {
-  assert(lhs.Grid() == rhs.Grid());
-  assert(lhs.Checkerboard() == rhs.Checkerboard());
+  GRID_ASSERT(lhs.Grid() == rhs.Grid());
+  GRID_ASSERT(lhs.Checkerboard() == rhs.Checkerboard());
 }
 
 NAMESPACE_END(Grid);

Grid/lattice/Lattice_matrix_reduction.h

@@ -42,7 +42,7 @@ static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice
   //  Lattice<vobj> Xslice(SliceGrid);
   //  Lattice<vobj> Rslice(SliceGrid);
 
-  assert( FullGrid->_simd_layout[Orthog]==1);
+  GRID_ASSERT( FullGrid->_simd_layout[Orthog]==1);
 
   //FIXME package in a convenient iterator
   //Should loop over a plane orthogonal to direction "Orthog"
@@ -86,7 +86,7 @@ static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<
   int Nblock = X.Grid()->GlobalDimensions()[Orthog];
 
   GridBase *FullGrid  = X.Grid();
-  assert( FullGrid->_simd_layout[Orthog]==1);
+  GRID_ASSERT( FullGrid->_simd_layout[Orthog]==1);
 
   //FIXME package in a convenient iterator
   //Should loop over a plane orthogonal to direction "Orthog"
@@ -140,7 +140,7 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
   
   mat = Eigen::MatrixXcd::Zero(Nblock,Nblock);
 
-  assert( FullGrid->_simd_layout[Orthog]==1);
+  GRID_ASSERT( FullGrid->_simd_layout[Orthog]==1);
   //  int nh =  FullGrid->_ndimension;
   //  int nl = SliceGrid->_ndimension;
   //  int nl = nh-1;

Grid/lattice/Lattice_peekpoke.h

@@ -98,8 +98,8 @@ void pokeSite(const sobj &s,Lattice<vobj> &l,const Coordinate &site){
 
   int Nsimd = grid->Nsimd();
 
-  assert( l.Checkerboard()== l.Grid()->CheckerBoard(site));
-  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+  GRID_ASSERT( l.Checkerboard()== l.Grid()->CheckerBoard(site));
+  GRID_ASSERT( sizeof(sobj)*Nsimd == sizeof(vobj));
 
   int rank,odx,idx;
   // Optional to broadcast from node 0.
@@ -135,7 +135,7 @@ void peekSite(sobj &s,const Lattice<vobj> &l,const Coordinate &site){
 
   int Nsimd = grid->Nsimd();
 
-  assert( l.Checkerboard() == l.Grid()->CheckerBoard(site));
+  GRID_ASSERT( l.Checkerboard() == l.Grid()->CheckerBoard(site));
 
   int rank,odx,idx;
   grid->GlobalCoorToRankIndex(rank,odx,idx,site);
@@ -159,14 +159,14 @@ template<class vobj,class sobj>
 inline void peekLocalSite(sobj &s,const LatticeView<vobj> &l,Coordinate &site)
 {
   GridBase *grid = l.getGrid();
-  assert(l.mode==CpuRead);
+  GRID_ASSERT(l.mode==CpuRead);
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nsimd = grid->Nsimd();
 
-  //  assert( l.Checkerboard()== grid->CheckerBoard(site));
-  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+  //  GRID_ASSERT( l.Checkerboard()== grid->CheckerBoard(site));
+  GRID_ASSERT( sizeof(sobj)*Nsimd == sizeof(vobj));
 
   static const int words=sizeof(vobj)/sizeof(vector_type);
   int odx,idx;
@@ -195,15 +195,15 @@ template<class vobj,class sobj>
 inline void pokeLocalSite(const sobj &s,LatticeView<vobj> &l,Coordinate &site)
 {
   GridBase *grid=l.getGrid();
-  assert(l.mode==CpuWrite);
+  GRID_ASSERT(l.mode==CpuWrite);
 
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nsimd = grid->Nsimd();
 
-  //  assert( l.Checkerboard()== grid->CheckerBoard(site));
-  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+  //  GRID_ASSERT( l.Checkerboard()== grid->CheckerBoard(site));
+  GRID_ASSERT( sizeof(sobj)*Nsimd == sizeof(vobj));
 
   static const int words=sizeof(vobj)/sizeof(vector_type);
   int odx,idx;

Grid/lattice/Lattice_reduction.h

@@ -292,26 +292,26 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
 
   bool ok;
 #ifdef GRID_SYCL
-  uint64_t csum=0;
-  uint64_t csum2=0;
-  if ( FlightRecorder::LoggingMode != FlightRecorder::LoggingModeNone)
-  {
-    // Hack
-    // Fast integer xor checksum. Can also be used in comms now.
-    autoView(l_v,left,AcceleratorRead);
-    Integer words = left.Grid()->oSites()*sizeof(vobj)/sizeof(uint64_t);
-    uint64_t *base= (uint64_t *)&l_v[0];
-    csum=svm_xor(base,words);
-    ok = FlightRecorder::CsumLog(csum);
-    if ( !ok ) {
-      csum2=svm_xor(base,words);
-      std::cerr<< " Bad CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
-    } else {
-      //      csum2=svm_xor(base,words);
-      //      std::cerr<< " ok CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
-    }
-    assert(ok);
-  }
+  //  uint64_t csum=0;
+  //  uint64_t csum2=0;
+  //  if ( FlightRecorder::LoggingMode != FlightRecorder::LoggingModeNone)
+  //  {
+  // Hack
+  // Fast integer xor checksum. Can also be used in comms now.
+  //    autoView(l_v,left,AcceleratorRead);
+  //    Integer words = left.Grid()->oSites()*sizeof(vobj)/sizeof(uint64_t);
+  //    uint64_t *base= (uint64_t *)&l_v[0];
+  //    csum=svm_xor(base,words);
+  //    ok = FlightRecorder::CsumLog(csum);
+  //    if ( !ok ) {
+  //      csum2=svm_xor(base,words);
+  //      std::cerr<< " Bad CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
+  //    } else {
+  //      csum2=svm_xor(base,words);
+  //      std::cerr<< " ok CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
+  //    }
+  //    GRID_ASSERT(ok);
+  // }
 #endif
   FlightRecorder::StepLog("rank inner product");
   ComplexD nrm = rankInnerProduct(left,right);
@@ -322,11 +322,11 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
     ComplexD nrm2 = rankInnerProduct(left,right);
     RealD local2 = real(nrm2);
     std::cerr<< " Bad NORM " << local << " recomputed as "<<local2<<std::endl;
-    assert(ok);
+    GRID_ASSERT(ok);
   }
   FlightRecorder::StepLog("Start global sum");
-  //  grid->GlobalSumP2P(nrm);
-  grid->GlobalSum(nrm);
+  grid->GlobalSumP2P(nrm);
+  //  grid->GlobalSum(nrm);
   FlightRecorder::StepLog("Finished global sum");
   //  std::cout << " norm "<< nrm << " p2p norm "<<nrmck<<std::endl;
   FlightRecorder::ReductionLog(local,real(nrm)); 
@@ -376,40 +376,9 @@ axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Latt
       coalescedWrite(z_v[ss],tmp);
   });
   bool ok;
-#ifdef GRID_SYCL
-  uint64_t csum=0;
-  uint64_t csum2=0;
-  if ( FlightRecorder::LoggingMode != FlightRecorder::LoggingModeNone)
-  {
-    // z_v
-    {
-      Integer words = sites*sizeof(vobj)/sizeof(uint64_t);
-      uint64_t *base= (uint64_t *)&z_v[0];
-      csum=svm_xor(base,words);
-      ok = FlightRecorder::CsumLog(csum);
-      if ( !ok ) {
-	csum2=svm_xor(base,words);
-	std::cerr<< " Bad z_v CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
-      }
-      assert(ok);
-    }
-    // inner_v
-    {
-      Integer words = sites*sizeof(inner_t)/sizeof(uint64_t);
-      uint64_t *base= (uint64_t *)&inner_tmp_v[0];
-      csum=svm_xor(base,words);
-      ok = FlightRecorder::CsumLog(csum);
-      if ( !ok ) {
-	csum2=svm_xor(base,words);
-	std::cerr<< " Bad inner_tmp_v CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
-      }
-      assert(ok);
-    }
-  }
-#endif
   nrm = real(TensorRemove(sumD(inner_tmp_v,sites)));
   ok = FlightRecorder::NormLog(real(nrm));
-  assert(ok);
+  GRID_ASSERT(ok);
   RealD local = real(nrm);
   grid->GlobalSum(nrm);
   FlightRecorder::ReductionLog(local,real(nrm));
@@ -495,13 +464,13 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,
   typedef typename vobj::scalar_object sobj;
   typedef typename vobj::scalar_object::scalar_type scalar_type;
   GridBase  *grid = Data.Grid();
-  assert(grid!=NULL);
+  GRID_ASSERT(grid!=NULL);
 
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
 
-  assert(orthogdim >= 0);
-  assert(orthogdim < Nd);
+  GRID_ASSERT(orthogdim >= 0);
+  GRID_ASSERT(orthogdim < Nd);
 
   int fd=grid->_fdimensions[orthogdim];
   int ld=grid->_ldimensions[orthogdim];
@@ -588,14 +557,14 @@ static void sliceInnerProductVector( std::vector<ComplexD> & result, const Latti
   typedef typename vobj::vector_type   vector_type;
   typedef typename vobj::scalar_type   scalar_type;
   GridBase  *grid = lhs.Grid();
-  assert(grid!=NULL);
+  GRID_ASSERT(grid!=NULL);
   conformable(grid,rhs.Grid());
 
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
 
-  assert(orthogdim >= 0);
-  assert(orthogdim < Nd);
+  GRID_ASSERT(orthogdim >= 0);
+  GRID_ASSERT(orthogdim < Nd);
 
   int fd=grid->_fdimensions[orthogdim];
   int ld=grid->_ldimensions[orthogdim];
@@ -824,6 +793,188 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
   delete SliceGrid;
 }
 
+/////////////////////////////////////////////////////////////////////////////
+// Batched linear algebra for Krylov orthogonalisation (GCR history windows).
+//
+//   innerProductMulti(out,left,right):  out[j] = <left[j],right> for all j in
+//        ONE kernel (right read once), ONE device reduction / sync (the per-site
+//        partials are an iVector over the batch) and ONE GlobalSumVector.
+//   axpyMulti(z,b,x):                   z = z + sum_j b[j] x[j] in ONE pass.
+//   axpyMultiNorm(z,b,x):               same, returning global |z|^2 from the
+//        same pass (one reduction, one GlobalSum).
+//
+// The batch width is a template parameter chosen at runtime from {2,4,8,16}
+// so a short window (mmax=2 smoother) does not pay for 16 partial lanes;
+// windows longer than 16 are processed in chunks of 16, one reduction each.
+// Same code path for every Lattice<vobj>: fine fermion fields and coarse
+// multi-RHS fields (nrhs folded into the grid) alike.
+/////////////////////////////////////////////////////////////////////////////
+// The batch of views and coefficients is passed to the kernel BY VALUE as
+// lambda-captured kernel arguments (a ViewPack), so there is no per-call
+// deviceVector allocation and no synchronous host->device memcpy of pointer
+// tables: the only sync points are the reductions themselves.
+// LatticeView has no default constructor, so the pack holds raw aligned
+// storage and views are copied in bytewise (as basisRotateJ does through
+// acceleratorPut); the struct is trivially copyable as a kernel argument.
+template<class View,int B>
+struct ViewPack {
+  alignas(View) unsigned char raw[B*sizeof(View)];
+  ComplexD b[B];
+  accelerator_inline const View & v(int j) const { return reinterpret_cast<const View *>(raw)[j]; }
+  void set(int j,const View &view){ memcpy(raw+j*sizeof(View),&view,sizeof(View)); }
+};
+
+template<int B,class vobj>
+void rankInnerProductMultiChunk(ComplexD *out,int m,
+				const std::vector<const Lattice<vobj>*> &left,
+				const Lattice<vobj> &right)
+{
+  typedef decltype(innerProductD(vobj(),vobj())) inner_t;
+  typedef iVector<inner_t,B> batch_t;
+  typedef decltype(right.View(AcceleratorRead)) View;
+
+  GRID_ASSERT(m>=1 && m<=B);
+  GridBase *grid = right.Grid();
+  const uint64_t sites = grid->oSites();
+
+  std::vector<View> h_v; h_v.reserve(m);
+  ViewPack<View,B> pack;
+  for(int j=0;j<m;j++){
+    conformable(*left[j],right);
+    h_v.push_back(left[j]->View(AcceleratorRead));
+    pack.set(j,h_v[j]);
+  }
+  for(int j=m;j<B;j++) pack.set(j,h_v[0]);   // valid but unused lanes
+
+  deviceVector<batch_t> partial(sites);
+  batch_t *partial_v = &partial[0];
+  {
+    autoView(right_v,right,AcceleratorRead);
+    accelerator_for(ss,sites,1,{
+	auto r = right_v[ss];
+	batch_t acc;
+	for(int j=0;j<B;j++){
+	  if ( j<m ) acc._internal[j] = innerProductD(pack.v(j)[ss],r);
+	  else       zeroit(acc._internal[j]);
+	}
+	partial_v[ss] = acc;
+    });
+  }
+  for(int j=0;j<m;j++) h_v[j].ViewClose();
+
+  auto res = sum(partial_v,sites);   // one reduction for the whole batch
+  for(int j=0;j<m;j++) out[j] = TensorRemove(res._internal[j]);
+}
+
+template<class vobj>
+void rankInnerProductMulti(std::vector<ComplexD> &out,
+			   const std::vector<const Lattice<vobj>*> &left,
+			   const Lattice<vobj> &right)
+{
+  int m = left.size();
+  out.resize(m);
+  for(int j0=0;j0<m;j0+=16){
+    int mm = std::min(16,m-j0);
+    std::vector<const Lattice<vobj>*> sub(left.begin()+j0,left.begin()+j0+mm);
+    if      ( mm<=2 ) rankInnerProductMultiChunk<2> (&out[j0],mm,sub,right);
+    else if ( mm<=4 ) rankInnerProductMultiChunk<4> (&out[j0],mm,sub,right);
+    else if ( mm<=8 ) rankInnerProductMultiChunk<8> (&out[j0],mm,sub,right);
+    else              rankInnerProductMultiChunk<16>(&out[j0],mm,sub,right);
+  }
+}
+
+template<class vobj>
+void innerProductMulti(std::vector<ComplexD> &out,
+		       const std::vector<const Lattice<vobj>*> &left,
+		       const Lattice<vobj> &right)
+{
+  rankInnerProductMulti(out,left,right);
+  if ( out.size() ) right.Grid()->GlobalSumVector(&out[0],(int)out.size());
+}
+
+// z = z + sum_{j<m} b[j] x[j] for one chunk of at most B vectors; if do_norm,
+// also writes per-site |z|^2 into inner_tmp_v.
+template<int B,class vobj>
+void axpyMultiChunk(Lattice<vobj> &z,const ComplexD *b,
+		    const std::vector<const Lattice<vobj>*> &x,int m,
+		    int do_norm,
+		    decltype(innerProduct(vobj(),vobj())) *inner_tmp_v)
+{
+  typedef decltype(z.View(AcceleratorRead)) View;
+  GRID_ASSERT(m>=1 && m<=B);
+  GridBase *grid = z.Grid();
+  const uint64_t nsimd = grid->Nsimd();
+  const uint64_t sites = grid->oSites();
+
+  std::vector<View> h_v; h_v.reserve(m);
+  ViewPack<View,B> pack;
+  for(int j=0;j<m;j++){
+    conformable(*x[j],z);
+    GRID_ASSERT(x[j]!=&z);            // window must not alias the accumulator
+    h_v.push_back(x[j]->View(AcceleratorRead));
+    pack.set(j,h_v[j]);
+    pack.b[j] = b[j];
+  }
+  for(int j=m;j<B;j++){ pack.set(j,h_v[0]); pack.b[j] = ComplexD(0.0); }
+
+  autoView(z_v,z,AcceleratorWrite);
+  accelerator_for(ss,sites,nsimd,{
+      auto acc = coalescedRead(z_v[ss]);
+      for(int j=0;j<B;j++) if ( j<m ) acc = acc + pack.b[j]*coalescedRead(pack.v(j)[ss]);
+      coalescedWrite(z_v[ss],acc);
+      if ( do_norm ) coalescedWrite(inner_tmp_v[ss],innerProduct(acc,acc));
+  });
+  for(int j=0;j<m;j++) h_v[j].ViewClose();
+}
+
+// z = z + sum_j b[j] x[j];  if do_norm, returns global |z|^2 from the same
+// (last) pass.  Chunks of 16 for windows longer than 16.
+template<class vobj>
+RealD axpyMultiNormImpl(Lattice<vobj> &z,const std::vector<ComplexD> &b,
+			const std::vector<const Lattice<vobj>*> &x,int do_norm)
+{
+  typedef decltype(innerProduct(vobj(),vobj())) inner_t;
+  int m = x.size();
+  GRID_ASSERT((int)b.size()>=m);
+  GridBase *grid = z.Grid();
+  const uint64_t sites = grid->oSites();
+
+  deviceVector<inner_t> inner_tmp(do_norm ? sites : 1);
+  inner_t *inner_tmp_v = &inner_tmp[0];
+
+  if ( m==0 ) {
+    if ( do_norm ) return norm2(z);
+    return 0.0;
+  }
+  for(int j0=0;j0<m;j0+=16){
+    int mm   = std::min(16,m-j0);
+    int last = (j0+mm>=m);
+    std::vector<const Lattice<vobj>*> sub(x.begin()+j0,x.begin()+j0+mm);
+    int dn = do_norm && last;
+    if      ( mm<=2 ) axpyMultiChunk<2> (z,&b[j0],sub,mm,dn,inner_tmp_v);
+    else if ( mm<=4 ) axpyMultiChunk<4> (z,&b[j0],sub,mm,dn,inner_tmp_v);
+    else if ( mm<=8 ) axpyMultiChunk<8> (z,&b[j0],sub,mm,dn,inner_tmp_v);
+    else              axpyMultiChunk<16>(z,&b[j0],sub,mm,dn,inner_tmp_v);
+  }
+
+  RealD nrm = 0.0;
+  if ( do_norm ) {
+    nrm = real(TensorRemove(sumD(inner_tmp_v,sites)));
+    grid->GlobalSum(nrm);
+  }
+  return nrm;
+}
+template<class vobj>
+void axpyMulti(Lattice<vobj> &z,const std::vector<ComplexD> &b,const std::vector<const Lattice<vobj>*> &x)
+{
+  axpyMultiNormImpl(z,b,x,0);
+}
+template<class vobj>
+RealD axpyMultiNorm(Lattice<vobj> &z,const std::vector<ComplexD> &b,const std::vector<const Lattice<vobj>*> &x)
+{
+  return axpyMultiNormImpl(z,b,x,1);
+}
+
 NAMESPACE_END(Grid);
 
 

Grid/lattice/Lattice_reduction_gpu.h

@@ -197,55 +197,205 @@ __global__ void reduceKernel(const vobj *lat, sobj *buffer, Iterator n) {
 /////////////////////////////////////////////////////////////////////////////////////////////////////////
 // Possibly promote to double and sum
 /////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#undef GRID_REDUCTION_TIMING
+
 template <class vobj>
-inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osites) 
+inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osites)
 {
   typedef typename vobj::scalar_objectD sobj;
   typedef decltype(lat) Iterator;
-  
+
   Integer nsimd= vobj::Nsimd();
   Integer size = osites*nsimd;
 
   Integer numThreads, numBlocks;
   int ok = getNumBlocksAndThreads(size, sizeof(sobj), numThreads, numBlocks);
-  assert(ok);
+  GRID_ASSERT(ok);
 
   Integer smemSize = numThreads * sizeof(sobj);
-  // Move out of UVM
-  // Turns out I had messed up the synchronise after move to compute stream
-  // as running this on the default stream fools the synchronise
   deviceVector<sobj> buffer(numBlocks);
   sobj *buffer_v = &buffer[0];
   sobj result;
+
+#ifdef GRID_REDUCTION_TIMING
+  RealD t_kernel = -usecond();
+#endif
   reduceKernel<<< numBlocks, numThreads, smemSize, computeStream >>>(lat, buffer_v, size);
   accelerator_barrier();
+#ifdef GRID_REDUCTION_TIMING
+  t_kernel += usecond();
+  RealD t_d2h = -usecond();
+#endif
   acceleratorCopyFromDevice(buffer_v,&result,sizeof(result));
+#ifdef GRID_REDUCTION_TIMING
+  t_d2h += usecond();
+  std::cout << GridLogDebug << "  sumD_gpu_small"
+            << " sizeof(sobj)=" << sizeof(sobj)
+            << " blocks=" << numBlocks << " threads=" << numThreads
+            << " kernel+barrier=" << t_kernel << " us"
+            << " D2H=" << t_d2h << " us" << std::endl;
+#endif
   return result;
 }
 
+// Fused pack+reduce: reads R words of each vobj at word offset 'base',
+// accumulates directly into iVector<iScalar<scalarD>,R> without staging
+// through an intermediate bundle buffer.  One HBM pass instead of three.
+template <int R, class vobj, class sobj, class Iterator>
+__device__ void packReduceBlocks(
+    const iScalar<typename vobj::vector_type> *idat,
+    sobj *g_odata, Iterator osites, int base, int words)
+{
+  constexpr Iterator nsimd = vobj::Nsimd();
+  Iterator blockSize = blockDim.x;
+
+  extern __shared__ __align__(COALESCE_GRANULARITY) unsigned char shmem_pointer[];
+  sobj *sdata = (sobj *)shmem_pointer;
+
+  Iterator tid      = threadIdx.x;
+  Iterator i        = blockIdx.x * (blockSize * 2) + threadIdx.x;
+  Iterator gridSize = blockSize * 2 * gridDim.x;
+  sobj mySum = Zero();
+
+  while (i < osites * nsimd) {
+    Iterator lane = i % nsimd;
+    Iterator ss   = i / nsimd;
+    sobj tmpD; zeroit(tmpD);
+    for (int k = 0; k < R; k++) {
+      auto w = extractLane(lane, idat[ss * words + base + k]);
+      iScalar<typename vobj::scalar_typeD> wd; wd = w;
+      tmpD._internal[k] = wd;
+    }
+    mySum += tmpD;
+
+    if (i + blockSize < osites * nsimd) {
+      lane = (i + blockSize) % nsimd;
+      ss   = (i + blockSize) / nsimd;
+      sobj tmpD2; zeroit(tmpD2);
+      for (int k = 0; k < R; k++) {
+        auto w = extractLane(lane, idat[ss * words + base + k]);
+        iScalar<typename vobj::scalar_typeD> wd; wd = w;
+        tmpD2._internal[k] = wd;
+      }
+      mySum += tmpD2;
+    }
+    i += gridSize;
+  }
+
+  reduceBlock(sdata, mySum, tid);
+  if (tid == 0) g_odata[blockIdx.x] = sdata[0];
+}
+
+template <int R, class vobj, class sobj, class Iterator>
+__global__ void packReduceKernel(
+    const iScalar<typename vobj::vector_type> *idat,
+    sobj *buffer, Iterator osites, int base, int words)
+{
+  Iterator blockSize = blockDim.x;
+
+  packReduceBlocks<R, vobj, sobj>(idat, buffer, osites, base, words);
+
+  if (gridDim.x > 1) {
+    const Iterator tid = threadIdx.x;
+    __shared__ bool amLast;
+    extern __shared__ __align__(COALESCE_GRANULARITY) unsigned char shmem_pointer[];
+    sobj *smem = (sobj *)shmem_pointer;
+
+    acceleratorFence();
+
+    if (tid == 0) {
+      unsigned int ticket = atomicInc(&retirementCount, gridDim.x);
+      amLast = (ticket == gridDim.x - 1);
+    }
+    acceleratorSynchroniseAll();
+
+    if (amLast) {
+      Iterator i = tid;
+      sobj mySum = Zero();
+      while (i < (Iterator)gridDim.x) {
+        mySum += buffer[i];
+        i += blockSize;
+      }
+      reduceBlock(smem, mySum, tid);
+      if (tid == 0) {
+        buffer[0] = smem[0];
+        retirementCount = 0;
+      }
+    }
+  }
+}
+
+template<int R, class vobj>
+inline void sumD_gpu_reduce_words(const vobj *lat, Integer osites,
+                                   typename vobj::scalar_typeD *ret_p, int base)
+{
+  typedef typename vobj::vector_type  vector;
+  typedef typename vobj::scalar_typeD scalarD;
+  using BundleScalarD = iVector<iScalar<scalarD>, R>;
+
+  constexpr int Nsimd = vobj::Nsimd();
+  const int words = sizeof(vobj) / sizeof(vector);
+  const iScalar<vector> *idat = (const iScalar<vector> *)lat;
+
+  Integer size = (Integer)osites * Nsimd;
+  Integer numThreads, numBlocks;
+  int ok = getNumBlocksAndThreads(size, sizeof(BundleScalarD), numThreads, numBlocks);
+  GRID_ASSERT(ok);
+
+  Integer smemSize = numThreads * sizeof(BundleScalarD);
+  deviceVector<BundleScalarD> buffer(numBlocks);
+  BundleScalarD *buffer_v = &buffer[0];
+  BundleScalarD result;
+
+#ifdef GRID_REDUCTION_TIMING
+  RealD t_kernel = -usecond();
+#endif
+  packReduceKernel<R, vobj, BundleScalarD, Integer>
+    <<<numBlocks, numThreads, smemSize, computeStream>>>
+    (idat, buffer_v, osites, base, words);
+  accelerator_barrier();
+#ifdef GRID_REDUCTION_TIMING
+  t_kernel += usecond();
+  RealD t_d2h = -usecond();
+#endif
+  acceleratorCopyFromDevice(buffer_v, &result, sizeof(result));
+#ifdef GRID_REDUCTION_TIMING
+  t_d2h += usecond();
+  std::cout << GridLogDebug << " sumD_gpu_reduce_words R=" << R
+            << " base=" << base
+            << " kernel=" << t_kernel << " D2H=" << t_d2h << " us" << std::endl;
+#endif
+
+  for (int k = 0; k < R; k++)
+    ret_p[base + k] = TensorRemove(result._internal[k]);
+}
+
 template <class vobj>
 inline typename vobj::scalar_objectD sumD_gpu_large(const vobj *lat, Integer osites)
 {
-  typedef typename vobj::vector_type  vector;
-  typedef typename vobj::scalar_typeD scalarD;
-  typedef typename vobj::scalar_objectD sobj;
-  sobj ret;
+  typedef typename vobj::vector_type    vector;
+  typedef typename vobj::scalar_typeD   scalarD;
+  typedef typename vobj::scalar_objectD sobjD;
+
+  const int words = sizeof(vobj) / sizeof(vector);
+  sobjD ret; zeroit(ret);
   scalarD *ret_p = (scalarD *)&ret;
-  
-  const int words = sizeof(vobj)/sizeof(vector);
 
-  deviceVector<vector> buffer(osites);
-  vector *dat = (vector *)lat;
-  vector *buf = &buffer[0];
-  iScalar<vector> *tbuf =(iScalar<vector> *)  &buffer[0];
-  for(int w=0;w<words;w++) {
+#ifdef GRID_REDUCTION_TIMING
+  RealD t_large = -usecond();
+#endif
+  int w = 0;
+  while (w + 12 <= words) { sumD_gpu_reduce_words<12>(lat, osites, ret_p, w); w += 12; }
+  while (w +  4 <= words) { sumD_gpu_reduce_words< 4>(lat, osites, ret_p, w); w +=  4; }
+  while (w      <  words) { sumD_gpu_reduce_words< 1>(lat, osites, ret_p, w); w +=  1; }
+#ifdef GRID_REDUCTION_TIMING
+  t_large += usecond();
+  std::cout << GridLogDebug << "sumD_gpu_large"
+            << " sizeof(sobjD)=" << sizeof(sobjD)
+            << " words=" << words << " total=" << t_large << " us" << std::endl;
+#endif
 
-    accelerator_for(ss,osites,1,{
-	buf[ss] = dat[ss*words+w];
-      });
-      
-    ret_p[w] = sumD_gpu_small(tbuf,osites);
-  }
   return ret;
 }
 
@@ -288,5 +438,11 @@ inline typename vobj::scalar_object sum_gpu_large(const vobj *lat, Integer osite
   result = sumD_gpu_large(lat,osites);
   return result;
 }
+template<class Word> Word checksum_gpu(Word *vec,uint64_t L)
+{
+  Word w;
+  bzero(&w,sizeof(w));
+  return w;
+}
 
 NAMESPACE_END(Grid);