Merge branch 'develop' of https://github.com/paboyle/Grid into bugfix/dminus

.gitignore

@@ -92,7 +92,6 @@ build*/*
 #####################
 *.xcodeproj/*
 build.sh
-.vscode
 
 # Eigen source #
 ################
@@ -107,10 +106,6 @@ lib/fftw/*
 m4/lt*
 m4/libtool.m4
 
-# github pages #
-################
-gh-pages/
-
 # Buck files #
 ##############
 .buck*
@@ -121,5 +116,4 @@ make-bin-BUCK.sh
 # generated sources #
 #####################
 lib/qcd/spin/gamma-gen/*.h
-lib/qcd/spin/gamma-gen/*.cc
-
+lib/qcd/spin/gamma-gen/*.cc

.travis.yml

@@ -7,11 +7,9 @@ cache:
 matrix:
   include:
     - os:        osx
-      osx_image: xcode8.3
+      osx_image: xcode7.2
       compiler: clang
     - compiler: gcc
-      dist: trusty
-      sudo: required
       addons:
         apt:
           sources:
@@ -26,8 +24,6 @@ matrix:
             - binutils-dev
       env: VERSION=-4.9
     - compiler: gcc
-      dist: trusty
-      sudo: required
       addons:
         apt:
           sources:
@@ -42,7 +38,6 @@ matrix:
             - binutils-dev
       env: VERSION=-5
     - compiler: clang
-      dist: trusty
       addons:
         apt:
           sources:
@@ -57,7 +52,6 @@ matrix:
             - binutils-dev
       env: CLANG_LINK=http://llvm.org/releases/3.8.0/clang+llvm-3.8.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz
     - compiler: clang
-      dist: trusty
       addons:
         apt:
           sources:
@@ -79,15 +73,13 @@ before_install:
     - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then export LD_LIBRARY_PATH="${GRIDDIR}/clang/lib:${LD_LIBRARY_PATH}"; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew update; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install libmpc; fi
+    - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install openmpi; fi
+    - if [[ "$TRAVIS_OS_NAME" == "osx" ]] && [[ "$CC" == "gcc" ]]; then brew install gcc5; fi
     
 install:
     - export CC=$CC$VERSION
     - export CXX=$CXX$VERSION
     - echo $PATH
-    - which autoconf
-    - autoconf  --version
-    - which automake
-    - automake  --version
     - which $CC
     - $CC  --version
     - which $CXX
@@ -100,15 +92,15 @@ script:
     - cd build
     - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=none
     - make -j4 
-    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
+    - ./benchmarks/Benchmark_dwf --threads 1
     - echo make clean
     - ../configure --enable-precision=double --enable-simd=SSE4 --enable-comms=none
     - make -j4
-    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
-    - make check
+    - ./benchmarks/Benchmark_dwf --threads 1
     - echo make clean
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=mpi-auto ; fi
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then make -j4; fi
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then mpirun.openmpi -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
+    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then export CXXFLAGS='-DMPI_UINT32_T=MPI_UNSIGNED -DMPI_UINT64_T=MPI_UNSIGNED_LONG'; fi
+    - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=mpi-auto
+    - make -j4
+    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then mpirun.openmpi -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
 
 

Makefile.am

@@ -3,15 +3,10 @@ SUBDIRS = lib benchmarks tests extras
 
 include $(top_srcdir)/doxygen.inc
 
-bin_SCRIPTS=grid-config
+tests: all
+	$(MAKE) -C tests tests
 
-
-.PHONY: bench check tests doxygen-run doxygen-doc $(DX_PS_GOAL) $(DX_PDF_GOAL)
-
-tests-local: all
-bench-local: all
-check-local: all
+.PHONY: tests doxygen-run doxygen-doc $(DX_PS_GOAL) $(DX_PDF_GOAL)
 
 AM_CXXFLAGS += -I$(top_builddir)/include
-
 ACLOCAL_AMFLAGS = -I m4

README.md

@@ -22,26 +22,6 @@ Last update Nov 2016.
 
 _Please do not send pull requests to the `master` branch which is reserved for releases._
 
-### Compilers
-
-Intel ICPC v16.0.3 and later
-
-Clang v3.5 and later (need 3.8 and later for OpenMP)
-
-GCC   v4.9.x (recommended)
-
-GCC   v6.3 and later
-
-### Important: 
-
-Some versions of GCC appear to have a bug under high optimisation (-O2, -O3).
-
-The safety of these compiler versions cannot be guaranteed at this time. Follow Issue 100 for details and updates.
-
-GCC   v5.x
-
-GCC   v6.1, v6.2
-
 ### Bug report
 
 _To help us tracking and solving more efficiently issues with Grid, please report problems using the issue system of GitHub rather than sending emails to Grid developers._
@@ -52,7 +32,7 @@ When you file an issue, please go though the following checklist:
 2. Give a description of the target platform (CPU, network, compiler). Please give the full CPU part description, using for example `cat /proc/cpuinfo | grep 'model name' | uniq` (Linux) or `sysctl machdep.cpu.brand_string` (macOS) and the full output the `--version` option of your compiler.
 3. Give the exact `configure` command used.
 4. Attach `config.log`.
-5. Attach `grid.config.summary`.
+5. Attach `config.summary`.
 6. Attach the output of `make V=1`.
 7. Describe the issue and any previous attempt to solve it. If relevant, show how to reproduce the issue using a minimal working example.
 
@@ -115,10 +95,10 @@ install Grid. Other options are detailed in the next section, you can also use `
 `CXX`, `CXXFLAGS`, `LDFLAGS`, ... environment variables can be modified to
 customise the build.
 
-Finally, you can build, check, and install Grid:
+Finally, you can build and install Grid:
 
 ``` bash
-make; make check; make install
+make; make install
 ```
 
 To minimise the build time, only the tests at the root of the `tests` directory are built by default. If you want to build tests in the sub-directory `<subdir>` you can execute:
@@ -141,7 +121,7 @@ If you want to build all the tests at once just use `make tests`.
 - `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
 - `--enable-precision={single|double}`: set the default precision (default: `double`).
 - `--enable-precision=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
-- `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
+- `--enable-rng={ranlux48|mt19937}`: choose the RNG (default: `ranlux48 `).
 - `--disable-timers`: disable system dependent high-resolution timers.
 - `--enable-chroma`: enable Chroma regression tests.
 - `--enable-doxygen-doc`: enable the Doxygen documentation generation (build with `make doxygen-doc`)
@@ -179,6 +159,7 @@ Alternatively, some CPU codenames can be directly used:
 
 | `<code>`    | Description                            |
 | ----------- | -------------------------------------- |
+| `KNC`       | [Intel Xeon Phi codename Knights Corner](http://ark.intel.com/products/codename/57721/Knights-Corner) |
 | `KNL`       | [Intel Xeon Phi codename Knights Landing](http://ark.intel.com/products/codename/48999/Knights-Landing) |
 | `BGQ`       | Blue Gene/Q                            |
 

TODO

@@ -1,33 +1,6 @@
 TODO:
 ---------------
 
-Large item work list:
-1)- MultiRHS with spread out extra dim -- Go through filesystem with SciDAC I/O
-
-2)- Christoph's local basis expansion Lanczos
-3)- BG/Q port and check
-4)- Precision conversion and sort out localConvert      <-- partial
-  - Consistent linear solver flop count/rate -- PARTIAL, time but no flop/s yet
-5)- Physical propagator interface
-6)- Conserved currents
-7)- Multigrid Wilson and DWF, compare to other Multigrid implementations
-8)- HDCR resume
-
-Recent DONE 
--- Lanczos Remove DenseVector, DenseMatrix; Use Eigen instead. <-- DONE
--- GaugeFix into central location                      <-- DONE
--- Scidac and Ildg metadata handling                   <-- DONE
--- Binary I/O MPI2 IO                                  <-- DONE
--- Binary I/O speed up & x-strips                      <-- DONE
--- Cut down the exterior overhead                      <-- DONE
--- Interior legs from SHM comms                        <-- DONE
--- Half-precision comms                                <-- DONE
--- Merge high precision reduction into develop         <-- DONE
--- BlockCG, BCGrQ                                      <-- DONE
--- multiRHS DWF; benchmark on Cori/BNL for comms elimination <-- DONE
-   -- slice* linalg routines for multiRHS, BlockCG    
-
------
 * Forces; the UdSdU  term in gauge force term is half of what I think it should
   be. This is a consequence of taking ONLY the first term in:
 
@@ -48,8 +21,16 @@ Recent DONE
   This means we must double the force in the Test_xxx_force routines, and is the origin of the factor of two.
   This 2x is applied by hand in the fermion routines and in the Test_rect_force routine.
 
+
+Policies:
+
+* Link smearing/boundary conds; Policy class based implementation ; framework more in place
+
 * Support different boundary conditions (finite temp, chem. potential ... )
 
+* Support different fermion representations? 
+  - contained entirely within the integrator presently
+
 - Sign of force term.
 
 - Reversibility test.
@@ -60,6 +41,11 @@ Recent DONE
 
 - Audit oIndex usage for cb behaviour
 
+- Rectangle gauge actions.
+  Iwasaki,
+  Symanzik,
+  ... etc...
+
 - Prepare multigrid for HMC. - Alternate setup schemes.
 
 - Support for ILDG --- ugly, not done
@@ -69,11 +55,9 @@ Recent DONE
 - FFTnD ?
 
 - Gparity; hand opt use template specialisation elegance to enable the optimised paths ?
-
 - Gparity force term; Gparity (R)HMC.
-
+- Random number state save restore
 - Mobius implementation clean up to rmove #if 0 stale code sequences
-
 - CG -- profile carefully, kernel fusion, whole CG performance measurements.
 
 ================================================================
@@ -106,7 +90,6 @@ Insert/Extract
 Not sure of status of this -- reverify. Things are working nicely now though.
 
 * Make the Tensor types and Complex etc... play more nicely.
-
   - TensorRemove is a hack, come up with a long term rationalised approach to Complex vs. Scalar<Scalar<Scalar<Complex > > >
     QDP forces use of "toDouble" to get back to non tensor scalar. This role is presently taken TensorRemove, but I
     want to introduce a syntax that does not require this.
@@ -129,8 +112,6 @@ Not sure of status of this -- reverify. Things are working nicely now though.
 RECENT
 ---------------
 
-  - Support different fermion representations? -- DONE
-  - contained entirely within the integrator presently
   - Clean up HMC                                                             -- DONE
   - LorentzScalar<GaugeField> gets Gauge link type (cleaner).                -- DONE
   - Simplified the integrators a bit.                                        -- DONE
@@ -142,26 +123,6 @@ RECENT
   - Parallel io improvements                                  -- DONE
   - Plaquette and link trace checks into nersc reader from the Grid_nersc_io.cc test. -- DONE
 
-
-DONE:
-- MultiArray -- MultiRHS done
-- ConjugateGradientMultiShift -- DONE
-- MCR                         -- DONE
-- Remez -- Mike or Boost?     -- DONE
-- Proto (ET)                  -- DONE
-- uBlas                       -- DONE ; Eigen
-- Potentially Useful Boost libraries -- DONE ; Eigen
-- Aligned allocator; memory pool -- DONE
-- Multiprecision              -- DONE
-- Serialization               -- DONE
-- Regex -- Not needed
-- Tokenize -- Why?
-
-- Random number state save restore -- DONE
-- Rectangle gauge actions. -- DONE
-  Iwasaki,
-  Symanzik,
-  ... etc...
 Done: Cayley, Partial , ContFrac force terms.
 
 DONE
@@ -246,7 +207,6 @@ Done
 FUNCTIONALITY: it pleases me to keep track of things I have done (keeps me arguably sane)
 ======================================================================================================
 
-* Link smearing/boundary conds; Policy class based implementation ; framework more in place -- DONE
 * Command line args for geometry, simd, etc. layout. Is it necessary to have -- DONE
   user pass these? Is this a QCD specific?
 

VERSION

@@ -1,5 +1,6 @@
-Version : 0.7.0
+Version : 0.6.0
 
-- Clang 3.5 and above, ICPC v16 and above, GCC 6.3 and above recommended
-- MPI and MPI3 comms optimisations for KNL and OPA finished
-- Half precision comms
+- AVX512, AVX2, AVX, SSE good
+- Clang 3.5 and above, ICPC v16 and above, GCC 4.9 and above
+- MPI and MPI3
+- HiRep, Smearing, Generic gauge group

benchmarks/Benchmark_comms.cc

@@ -31,32 +31,6 @@ using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
-struct time_statistics{
-  double mean;
-  double err;
-  double min;
-  double max;
-
-  void statistics(std::vector<double> v){
-      double sum = std::accumulate(v.begin(), v.end(), 0.0);
-      mean = sum / v.size();
-
-      std::vector<double> diff(v.size());
-      std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
-      double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
-      err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
-
-      auto result = std::minmax_element(v.begin(), v.end());
-      min = *result.first;
-      max = *result.second;
-}
-};
-
-void header(){
-  std::cout <<GridLogMessage << " L  "<<"\t"<<" Ls  "<<"\t"
-            <<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
-};
-
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
@@ -66,19 +40,15 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  int Nloop=100;
+  int Nloop=10;
   int nmu=0;
-  int maxlat=24;
   for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
 
-  std::cout << GridLogMessage << "Number of iterations to average: "<< Nloop << std::endl;
-  std::vector<double> t_time(Nloop);
-  time_statistics timestat;
-
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking concurrent halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  header();
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  int maxlat=24;
   for(int lat=4;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=32;Ls*=2){
 
@@ -88,9 +58,6 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      RealD Nrank = Grid._Nprocessors;
-      RealD Nnode = Grid.NodeCount();
-      RealD ppn = Nrank/Nnode;
 
       std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
       std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
@@ -98,8 +65,8 @@ int main (int argc, char ** argv)
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      for(int i=0;i<Nloop;i++){
       double start=usecond();
+      for(int i=0;i<Nloop;i++){
 
 	std::vector<CartesianCommunicator::CommsRequest_t> requests;
 
@@ -135,24 +102,18 @@ int main (int argc, char ** argv)
 	}
 	Grid.SendToRecvFromComplete(requests);
 	Grid.Barrier();
-	double stop=usecond();
-	t_time[i] = stop-start; // microseconds
+
       }
+      double stop=usecond();
 
-      timestat.statistics(t_time);
-
-      double dbytes    = bytes*ppn;
-      double xbytes    = dbytes*2.0*ncomm;
+      double dbytes    = bytes;
+      double xbytes    = Nloop*dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
-               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
-               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
-               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
-               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
-               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+      double time = stop-start; // microseconds
 
+      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }    
 
@@ -160,7 +121,8 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking sequential halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  header();
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+
 
   for(int lat=4;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=32;Ls*=2){
@@ -168,9 +130,6 @@ int main (int argc, char ** argv)
       std::vector<int> latt_size  ({lat,lat,lat,lat});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      RealD Nrank = Grid._Nprocessors;
-      RealD Nnode = Grid.NodeCount();
-      RealD ppn = Nrank/Nnode;
 
       std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
       std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
@@ -179,8 +138,8 @@ int main (int argc, char ** argv)
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      for(int i=0;i<Nloop;i++){
       double start=usecond();
+      for(int i=0;i<Nloop;i++){
     
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
@@ -219,34 +178,27 @@ int main (int argc, char ** argv)
 	  }
 	}
 	Grid.Barrier();
-	double stop=usecond();
-	t_time[i] = stop-start; // microseconds
-
       }
 
-      timestat.statistics(t_time);
+      double stop=usecond();
       
-      double dbytes    = bytes*ppn;
-      double xbytes    = dbytes*2.0*ncomm;
+      double dbytes    = bytes;
+      double xbytes    = Nloop*dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-    std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
-               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
-               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
-               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
-               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
-               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+      double time = stop-start;
 
-      
+      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }  
 
 
+  Nloop=10;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking concurrent STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  header();
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
 
   for(int lat=4;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=32;Ls*=2){
@@ -257,9 +209,6 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      RealD Nrank = Grid._Nprocessors;
-      RealD Nnode = Grid.NodeCount();
-      RealD ppn = Nrank/Nnode;
 
       std::vector<HalfSpinColourVectorD *> xbuf(8);
       std::vector<HalfSpinColourVectorD *> rbuf(8);
@@ -267,115 +216,16 @@ int main (int argc, char ** argv)
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
       }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      double dbytes;
+      double start=usecond();
       for(int i=0;i<Nloop;i++){
-	double start=usecond();
-
-	dbytes=0;
-	ncomm=0;
 
 	std::vector<CartesianCommunicator::CommsRequest_t> requests;
 
-	for(int mu=0;mu<4;mu++){
-	
-
-	  if (mpi_layout[mu]>1 ) {
-	  
-	    ncomm++;
-	    int comm_proc=1;
-	    int xmit_to_rank;
-	    int recv_from_rank;
-	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    dbytes+=
-	      Grid.StencilSendToRecvFromBegin(requests,
-					      (void *)&xbuf[mu][0],
-					      xmit_to_rank,
-					      (void *)&rbuf[mu][0],
-					      recv_from_rank,
-					      bytes);
-	
-	    comm_proc = mpi_layout[mu]-1;
-	  
-	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    dbytes+=
-	      Grid.StencilSendToRecvFromBegin(requests,
-					      (void *)&xbuf[mu+4][0],
-					      xmit_to_rank,
-					      (void *)&rbuf[mu+4][0],
-					      recv_from_rank,
-					      bytes);
-	  
-	  }
-	}
-	Grid.StencilSendToRecvFromComplete(requests);
-	Grid.Barrier();
-	double stop=usecond();
-	t_time[i] = stop-start; // microseconds
-	
-      }
-
-      timestat.statistics(t_time);
-
-      dbytes=dbytes*ppn;
-      double xbytes    = dbytes*0.5;
-      double rbytes    = dbytes*0.5;
-      double bidibytes = dbytes;
-
-      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
-               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
-               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
-               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
-               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
-               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
-
-
-    }
-  }    
-
-
-  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "= Benchmarking sequential STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
-  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  header();
-
-  for(int lat=4;lat<=maxlat;lat+=4){
-    for(int Ls=8;Ls<=32;Ls*=2){
-
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-      				    lat*mpi_layout[1],
-      				    lat*mpi_layout[2],
-      				    lat*mpi_layout[3]});
-
-      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      RealD Nrank = Grid._Nprocessors;
-      RealD Nnode = Grid.NodeCount();
-      RealD ppn = Nrank/Nnode;
-
-      std::vector<HalfSpinColourVectorD *> xbuf(8);
-      std::vector<HalfSpinColourVectorD *> rbuf(8);
-      Grid.ShmBufferFreeAll();
-      for(int d=0;d<8;d++){
-	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
-      }
-
-      int ncomm;
-      int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
-      double dbytes;
-      for(int i=0;i<Nloop;i++){
-	double start=usecond();
-
-	std::vector<CartesianCommunicator::CommsRequest_t> requests;
-	dbytes=0;
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
 	
@@ -387,52 +237,123 @@ int main (int argc, char ** argv)
 	    int recv_from_rank;
 	    
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    dbytes+=
-	      Grid.StencilSendToRecvFromBegin(requests,
-					      (void *)&xbuf[mu][0],
-					      xmit_to_rank,
-					      (void *)&rbuf[mu][0],
-					      recv_from_rank,
-					      bytes);
+	    Grid.StencilSendToRecvFromBegin(requests,
+					    (void *)&xbuf[mu][0],
+					    xmit_to_rank,
+					    (void *)&rbuf[mu][0],
+					    recv_from_rank,
+					    bytes);
+	
+	    comm_proc = mpi_layout[mu]-1;
+	  
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    Grid.StencilSendToRecvFromBegin(requests,
+					    (void *)&xbuf[mu+4][0],
+					    xmit_to_rank,
+					    (void *)&rbuf[mu+4][0],
+					    recv_from_rank,
+					    bytes);
+	  
+	  }
+	}
+	Grid.StencilSendToRecvFromComplete(requests);
+	Grid.Barrier();
+
+      }
+      double stop=usecond();
+
+      double dbytes    = bytes;
+      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      double rbytes    = xbytes;
+      double bidibytes = xbytes+rbytes;
+
+      double time = stop-start; // microseconds
+
+      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
+    }
+  }    
+
+
+
+  Nloop=100;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking sequential STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=32;Ls*=2){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],
+      				    lat*mpi_layout[1],
+      				    lat*mpi_layout[2],
+      				    lat*mpi_layout[3]});
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+
+      std::vector<HalfSpinColourVectorD *> xbuf(8);
+      std::vector<HalfSpinColourVectorD *> rbuf(8);
+      Grid.ShmBufferFreeAll();
+      for(int d=0;d<8;d++){
+	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+      }
+
+      int ncomm;
+      int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+
+      double start=usecond();
+      for(int i=0;i<Nloop;i++){
+
+	std::vector<CartesianCommunicator::CommsRequest_t> requests;
+
+	ncomm=0;
+	for(int mu=0;mu<4;mu++){
+	
+	  if (mpi_layout[mu]>1 ) {
+	  
+	    ncomm++;
+	    int comm_proc=1;
+	    int xmit_to_rank;
+	    int recv_from_rank;
+	    
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    Grid.StencilSendToRecvFromBegin(requests,
+					    (void *)&xbuf[mu][0],
+					    xmit_to_rank,
+					    (void *)&rbuf[mu][0],
+					    recv_from_rank,
+					    bytes);
 	    Grid.StencilSendToRecvFromComplete(requests);
 	    requests.resize(0);
 
 	    comm_proc = mpi_layout[mu]-1;
 	  
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    dbytes+=
-	      Grid.StencilSendToRecvFromBegin(requests,
-					      (void *)&xbuf[mu+4][0],
-					      xmit_to_rank,
-					      (void *)&rbuf[mu+4][0],
-					      recv_from_rank,
-					      bytes);
+	    Grid.StencilSendToRecvFromBegin(requests,
+					    (void *)&xbuf[mu+4][0],
+					    xmit_to_rank,
+					    (void *)&rbuf[mu+4][0],
+					    recv_from_rank,
+					    bytes);
 	    Grid.StencilSendToRecvFromComplete(requests);
 	    requests.resize(0);
 	  
 	  }
 	}
 	Grid.Barrier();
-	double stop=usecond();
-	t_time[i] = stop-start; // microseconds
-	
+
       }
+      double stop=usecond();
 
-      timestat.statistics(t_time);
+      double dbytes    = bytes;
+      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      double rbytes    = xbytes;
+      double bidibytes = xbytes+rbytes;
 
-      dbytes=dbytes*ppn;
-      double xbytes    = dbytes*0.5;
-      double rbytes    = dbytes*0.5;
-      double bidibytes = dbytes;
+      double time = stop-start; // microseconds
 
-
-      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
-               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
-               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
-               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
-               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
-               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
- 
+      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }    
 

benchmarks/Benchmark_dwf.cc

@@ -1,22 +1,28 @@
- /*************************************************************************************
+    /*************************************************************************************
+
     Grid physics library, www.github.com/paboyle/Grid 
+
     Source file: ./benchmarks/Benchmark_dwf.cc
+
     Copyright (C) 2015
 
-    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-    Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
 
     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 */
@@ -145,7 +151,9 @@ int main (int argc, char ** argv)
   RealD M5  =1.8;
 
   RealD NP = UGrid->_Nprocessors;
-  RealD NN = UGrid->NodeCount();
+
+  std::cout << GridLogMessage << "Creating action operator " << std::endl;
+  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
@@ -155,22 +163,16 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
   if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
-  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
   int ncall =1000;
   if (1) {
     FGrid->Barrier();
     Dw.ZeroCounters();
     Dw.Dhop(src,result,0);
-    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
@@ -188,7 +190,6 @@ int main (int argc, char ** argv)
     //    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
     err = ref-result; 
     std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
 
@@ -205,34 +206,6 @@ int main (int argc, char ** argv)
     Dw.Report();
   }
 
-  DomainWallFermionRL DwH(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  if (1) {
-    FGrid->Barrier();
-    DwH.ZeroCounters();
-    DwH.Dhop(src,result,0);
-    double t0=usecond();
-    for(int i=0;i<ncall;i++){
-      __SSC_START;
-      DwH.Dhop(src,result,0);
-      __SSC_STOP;
-    }
-    double t1=usecond();
-    FGrid->Barrier();
-    
-    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
-
-    std::cout<<GridLogMessage << "Called half prec comms Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
-    err = ref-result; 
-    std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-
-    assert (norm2(err)< 1.0e-3 );
-    DwH.Report();
-  }
-
   if (1)
   {
 
@@ -241,10 +214,6 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
     if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
     if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
     if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
     if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
     if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
@@ -276,7 +245,6 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "Called Dw s_inner "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
     //    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
     sDw.Report();
     RealD sum=0;
@@ -309,10 +277,6 @@ int main (int argc, char ** argv)
       std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
       if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
       if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
       if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) 
 	std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
       if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) 
@@ -352,7 +316,6 @@ int main (int argc, char ** argv)
 
       std::cout<<GridLogMessage << "sDeo mflop/s =   "<< flops/(t1-t0)<<std::endl;
       std::cout<<GridLogMessage << "sDeo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
-      std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
       sDw.Report();
 
       sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
@@ -431,15 +394,14 @@ int main (int argc, char ** argv)
 
 
   // S-direction is INNERMOST and takes no part in the parity.
+  static int Opt;  // these are a temporary hack
+  static int Comms;  // these are a temporary hack
+
   std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::DhopEO                "<<std::endl;
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
   if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
@@ -460,7 +422,6 @@ int main (int argc, char ** argv)
 
     std::cout<<GridLogMessage << "Deo mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "Deo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
-    std::cout<<GridLogMessage << "Deo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
     Dw.Report();
   }
   Dw.DhopEO(src_o,r_e,DaggerNo);
@@ -492,4 +453,3 @@ int main (int argc, char ** argv)
 
   Grid_finalize();
 }
-

benchmarks/Benchmark_memory_bandwidth.cc

@@ -55,8 +55,8 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
-  uint64_t lmax=64;
-#define NLOOP (100*lmax*lmax*lmax*lmax/vol)
+  uint64_t lmax=44;
+#define NLOOP (1*lmax*lmax*lmax*lmax/vol)
   for(int lat=4;lat<=lmax;lat+=4){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});

benchmarks/Benchmark_su3.cc

@@ -35,9 +35,8 @@ using namespace Grid::QCD;
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
-#define LMAX (64)
 
-  int Nloop=20;
+  int Nloop=1000;
 
   std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
   std::vector<int> mpi_layout  = GridDefaultMpi();
@@ -51,7 +50,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=LMAX;lat+=2){
+  for(int lat=2;lat<=32;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
@@ -83,7 +82,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=LMAX;lat+=2){
+  for(int lat=2;lat<=32;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
@@ -114,7 +113,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=LMAX;lat+=2){
+  for(int lat=2;lat<=32;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
@@ -145,7 +144,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=LMAX;lat+=2){
+  for(int lat=2;lat<=32;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];

benchmarks/Makefile.am

@@ -1,7 +1,11 @@
 include Make.inc
 
-bench-local: all
-	./Benchmark_su3
-	./Benchmark_memory_bandwidth
-	./Benchmark_wilson
-	./Benchmark_dwf --dslash-unroll
+simple: simple_su3_test.o simple_su3_expr.o simple_simd_test.o
+
+EXTRA_LIBRARIES = libsimple_su3_test.a libsimple_su3_expr.a libsimple_simd_test.a
+
+libsimple_su3_test_a_SOURCES = simple_su3_test.cc
+
+libsimple_su3_expr_a_SOURCES = simple_su3_expr.cc
+
+libsimple_simd_test_a_SOURCES = simple_simd_test.cc

bootstrap.sh

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.3.3.tar.bz2'
+EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.2.9.tar.bz2'
 
 echo "-- deploying Eigen source..."
 wget ${EIGEN_URL} --no-check-certificate

configure.ac

@@ -1,19 +1,16 @@
 AC_PREREQ([2.63])
-AC_INIT([Grid], [0.7.0], [https://github.com/paboyle/Grid], [Grid])
+AC_INIT([Grid], [0.6.0], [https://github.com/paboyle/Grid], [Grid])
 AC_CANONICAL_BUILD
 AC_CANONICAL_HOST
 AC_CANONICAL_TARGET
-AM_INIT_AUTOMAKE([subdir-objects 1.13])
-AM_EXTRA_RECURSIVE_TARGETS([tests bench])
+AM_INIT_AUTOMAKE(subdir-objects)
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_SRCDIR([lib/Grid.h])
 AC_CONFIG_HEADERS([lib/Config.h],[sed -i 's|PACKAGE_|GRID_|' lib/Config.h])
 m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
-################ Get git info
-#AC_REVISION([m4_esyscmd_s([./scripts/configure.commit])])
-
 ############### Checks for programs
+CXXFLAGS="-O3 $CXXFLAGS"
 AC_PROG_CXX
 AC_PROG_RANLIB
 
@@ -27,15 +24,12 @@ AX_GXX_VERSION
 AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
       [version of g++ that will compile the code])
 
-CXXFLAGS="-g $CXXFLAGS"
-
-
 ############### Checks for typedefs, structures, and compiler characteristics
 AC_TYPE_SIZE_T
 AC_TYPE_UINT32_T
 AC_TYPE_UINT64_T
 
-############### OpenMP
+############### OpenMP 
 AC_OPENMP
 ac_openmp=no
 if test "${OPENMP_CXXFLAGS}X" != "X"; then
@@ -66,23 +60,16 @@ AC_ARG_WITH([mpfr],
     [AM_CXXFLAGS="-I$with_mpfr/include $AM_CXXFLAGS"]
     [AM_LDFLAGS="-L$with_mpfr/lib $AM_LDFLAGS"])
 
-############### FFTW3
-AC_ARG_WITH([fftw],
+############### FFTW3 
+AC_ARG_WITH([fftw],    
             [AS_HELP_STRING([--with-fftw=prefix],
             [try this for a non-standard install prefix of the FFTW3 library])],
             [AM_CXXFLAGS="-I$with_fftw/include $AM_CXXFLAGS"]
             [AM_LDFLAGS="-L$with_fftw/lib $AM_LDFLAGS"])
 
-############### LIME
-AC_ARG_WITH([lime],
-            [AS_HELP_STRING([--with-lime=prefix],
-            [try this for a non-standard install prefix of the LIME library])],
-            [AM_CXXFLAGS="-I$with_lime/include $AM_CXXFLAGS"]
-            [AM_LDFLAGS="-L$with_lime/lib $AM_LDFLAGS"])
-
-############### lapack
+############### lapack 
 AC_ARG_ENABLE([lapack],
-    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])],
+    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])], 
     [ac_LAPACK=${enable_lapack}], [ac_LAPACK=no])
 
 case ${ac_LAPACK} in
@@ -96,18 +83,6 @@ case ${ac_LAPACK} in
         AC_DEFINE([USE_LAPACK],[1],[use LAPACK]);;
 esac
 
-############### FP16 conversions
-AC_ARG_ENABLE([sfw-fp16],
-    [AC_HELP_STRING([--enable-sfw-fp16=yes|no], [enable software fp16 comms])],
-    [ac_SFW_FP16=${enable_sfw_fp16}], [ac_SFW_FP16=yes])
-case ${ac_SFW_FP16} in
-    yes)
-      AC_DEFINE([SFW_FP16],[1],[software conversion to fp16]);;
-    no);;
-    *)
-      AC_MSG_ERROR(["SFW FP16 option not supported ${ac_SFW_FP16}"]);;
-esac
-
 ############### MKL
 AC_ARG_ENABLE([mkl],
     [AC_HELP_STRING([--enable-mkl=yes|no|prefix], [enable Intel MKL for LAPACK & FFTW])],
@@ -133,7 +108,7 @@ AC_ARG_WITH([hdf5],
 
 ############### first-touch
 AC_ARG_ENABLE([numa],
-    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])],
+    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])], 
     [ac_NUMA=${enable_NUMA}],[ac_NUMA=no])
 
 case ${ac_NUMA} in
@@ -159,8 +134,8 @@ if test "${ac_MKL}x" != "nox"; then
 fi
 
 AC_SEARCH_LIBS([__gmpf_init], [gmp],
-               [AC_SEARCH_LIBS([mpfr_init], [mpfr],
-                               [AC_DEFINE([HAVE_LIBMPFR], [1],
+               [AC_SEARCH_LIBS([mpfr_init], [mpfr], 
+                               [AC_DEFINE([HAVE_LIBMPFR], [1], 
                                           [Define to 1 if you have the `MPFR' library])]
                                [have_mpfr=true], [AC_MSG_ERROR([MPFR library not found])])]
                [AC_DEFINE([HAVE_LIBGMP], [1], [Define to 1 if you have the `GMP' library])]
@@ -169,7 +144,7 @@ AC_SEARCH_LIBS([__gmpf_init], [gmp],
 if test "${ac_LAPACK}x" != "nox"; then
     AC_SEARCH_LIBS([LAPACKE_sbdsdc], [lapack], [],
                    [AC_MSG_ERROR("LAPACK enabled but library not found")])
-fi
+fi   
 
 AC_SEARCH_LIBS([fftw_execute], [fftw3],
                [AC_SEARCH_LIBS([fftwf_execute], [fftw3f], [],
@@ -177,18 +152,6 @@ AC_SEARCH_LIBS([fftw_execute], [fftw3],
                [AC_DEFINE([HAVE_FFTW], [1], [Define to 1 if you have the `FFTW' library])]
                [have_fftw=true])
 
-AC_SEARCH_LIBS([limeCreateReader], [lime],
-               [AC_DEFINE([HAVE_LIME], [1], [Define to 1 if you have the `LIME' library])]
-               [have_lime=true],
-	       [AC_MSG_WARN(C-LIME library was not found in your system.
-In order to use ILGG file format please install or provide the correct path to your installation
-Info at: http://usqcd.jlab.org/usqcd-docs/c-lime/)])
-
-AC_SEARCH_LIBS([crc32], [z],
-               [AC_DEFINE([HAVE_ZLIB], [1], [Define to 1 if you have the `LIBZ' library])]
-               [have_zlib=true],
-	       [AC_MSG_ERROR(zlib library was not found in your system.)])
-
 AC_SEARCH_LIBS([H5Fopen], [hdf5_cpp],
                [AC_DEFINE([HAVE_HDF5], [1], [Define to 1 if you have the `HDF5' library])]
                [have_hdf5=true]
@@ -213,26 +176,19 @@ case ${ax_cv_cxx_compiler_vendor} in
     case ${ac_SIMD} in
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
-	case ${ac_SFW_FP16} in
-	  yes)
-	  SIMD_FLAGS='-msse4.2';;
-	  no)
-	  SIMD_FLAGS='-msse4.2 -mf16c';;
-	  *)
-          AC_MSG_ERROR(["SFW_FP16 must be either yes or no value ${ac_SFW_FP16} "]);;
-	esac;;
+        SIMD_FLAGS='-msse4.2';;
       AVX)
         AC_DEFINE([AVX1],[1],[AVX intrinsics])
-        SIMD_FLAGS='-mavx -mf16c';;
+        SIMD_FLAGS='-mavx';;
       AVXFMA4)
         AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -mfma4 -mf16c';;
+        SIMD_FLAGS='-mavx -mfma4';;
       AVXFMA)
         AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA3])
-        SIMD_FLAGS='-mavx -mfma -mf16c';;
+        SIMD_FLAGS='-mavx -mfma';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
-        SIMD_FLAGS='-mavx2 -mfma -mf16c';;
+        SIMD_FLAGS='-mavx2 -mfma';;
       AVX512)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
         SIMD_FLAGS='-mavx512f -mavx512pf -mavx512er -mavx512cd';;
@@ -341,7 +297,7 @@ case ${ac_COMMS} in
         comms_type='shmem'
      ;;
      *)
-        AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]);
+        AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]); 
      ;;
 esac
 case ${ac_COMMS} in
@@ -378,7 +334,7 @@ case ${ac_RNG} in
       AC_DEFINE([RNG_SITMO],[1],[RNG_SITMO] )
      ;;
      *)
-      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]);
+      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]); 
      ;;
 esac
 
@@ -395,7 +351,7 @@ case ${ac_TIMERS} in
       AC_DEFINE([TIMERS_OFF],[1],[TIMERS_OFF] )
      ;;
      *)
-      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]);
+      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]); 
      ;;
 esac
 
@@ -407,7 +363,7 @@ case ${ac_CHROMA} in
      yes|no)
      ;;
      *)
-       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]);
+       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]); 
      ;;
 esac
 
@@ -428,65 +384,12 @@ DX_INIT_DOXYGEN([$PACKAGE_NAME], [doxygen.cfg])
 
 ############### Ouput
 cwd=`pwd -P`; cd ${srcdir}; abs_srcdir=`pwd -P`; cd ${cwd}
-GRID_CXXFLAGS="$AM_CXXFLAGS $CXXFLAGS"
-GRID_LDFLAGS="$AM_LDFLAGS $LDFLAGS"
-GRID_LIBS=$LIBS
-GRID_SHORT_SHA=`git rev-parse --short HEAD`
-GRID_SHA=`git rev-parse HEAD`
-GRID_BRANCH=`git rev-parse --abbrev-ref HEAD`
 AM_CXXFLAGS="-I${abs_srcdir}/include $AM_CXXFLAGS"
 AM_CFLAGS="-I${abs_srcdir}/include $AM_CFLAGS"
 AM_LDFLAGS="-L${cwd}/lib $AM_LDFLAGS"
 AC_SUBST([AM_CFLAGS])
 AC_SUBST([AM_CXXFLAGS])
 AC_SUBST([AM_LDFLAGS])
-AC_SUBST([GRID_CXXFLAGS])
-AC_SUBST([GRID_LDFLAGS])
-AC_SUBST([GRID_LIBS])
-AC_SUBST([GRID_SHA])
-AC_SUBST([GRID_BRANCH])
-
-git_commit=`cd $srcdir && ./scripts/configure.commit`
-
-echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-Summary of configuration for $PACKAGE v$VERSION
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
------ GIT VERSION -------------------------------------
-$git_commit
------ PLATFORM ----------------------------------------
-architecture (build)        : $build_cpu
-os (build)                  : $build_os
-architecture (target)       : $target_cpu
-os (target)                 : $target_os
-compiler vendor             : ${ax_cv_cxx_compiler_vendor}
-compiler version            : ${ax_cv_gxx_version}
------ BUILD OPTIONS -----------------------------------
-SIMD                        : ${ac_SIMD}${SIMD_GEN_WIDTH_MSG}
-Threading                   : ${ac_openmp}
-Communications type         : ${comms_type}
-Default precision           : ${ac_PRECISION}
-Software FP16 conversion    : ${ac_SFW_FP16}
-RNG choice                  : ${ac_RNG}
-GMP                         : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
-LAPACK                      : ${ac_LAPACK}
-FFTW                        : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
-LIME (ILDG support)         : `if test "x$have_lime" = xtrue; then echo yes; else echo no; fi`
-HDF5                        : `if test "x$have_hdf5" = xtrue; then echo yes; else echo no; fi`
-build DOXYGEN documentation : `if test "$DX_FLAG_doc" = '1'; then echo yes; else echo no; fi`
------ BUILD FLAGS -------------------------------------
-CXXFLAGS:
-`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-LDFLAGS:
-`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-LIBS:
-`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
--------------------------------------------------------" > grid.configure.summary
-
-GRID_SUMMARY="`cat grid.configure.summary`"
-AM_SUBST_NOTMAKE([GRID_SUMMARY])
-AC_SUBST([GRID_SUMMARY])
-
-AC_CONFIG_FILES([grid-config], [chmod +x grid-config])
 AC_CONFIG_FILES(Makefile)
 AC_CONFIG_FILES(lib/Makefile)
 AC_CONFIG_FILES(tests/Makefile)
@@ -497,7 +400,6 @@ AC_CONFIG_FILES(tests/forces/Makefile)
 AC_CONFIG_FILES(tests/hadrons/Makefile)
 AC_CONFIG_FILES(tests/hmc/Makefile)
 AC_CONFIG_FILES(tests/solver/Makefile)
-AC_CONFIG_FILES(tests/smearing/Makefile)
 AC_CONFIG_FILES(tests/qdpxx/Makefile)
 AC_CONFIG_FILES(tests/testu01/Makefile)
 AC_CONFIG_FILES(benchmarks/Makefile)
@@ -505,7 +407,36 @@ AC_CONFIG_FILES(extras/Makefile)
 AC_CONFIG_FILES(extras/Hadrons/Makefile)
 AC_OUTPUT
 
-echo ""
-cat grid.configure.summary
-echo ""
+echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Summary of configuration for $PACKAGE v$VERSION
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
+----- PLATFORM ----------------------------------------
+architecture (build)        : $build_cpu
+os (build)                  : $build_os
+architecture (target)       : $target_cpu
+os (target)                 : $target_os
+compiler vendor             : ${ax_cv_cxx_compiler_vendor}
+compiler version            : ${ax_cv_gxx_version}
+----- BUILD OPTIONS -----------------------------------
+SIMD                        : ${ac_SIMD}${SIMD_GEN_WIDTH_MSG}
+Threading                   : ${ac_openmp} 
+Communications type         : ${comms_type}
+Default precision           : ${ac_PRECISION}
+RNG choice                  : ${ac_RNG} 
+GMP                         : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
+LAPACK                      : ${ac_LAPACK}
+FFTW                        : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
+HDF5                        : `if test "x$have_hdf5" = xtrue; then echo yes; else echo no; fi`
+build DOXYGEN documentation : `if test "$DX_FLAG_doc" = '1'; then echo yes; else echo no; fi`
+----- BUILD FLAGS -------------------------------------
+CXXFLAGS:
+`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LDFLAGS:
+`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LIBS:
+`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+-------------------------------------------------------" > config.summary
+echo ""
+cat config.summary
+echo ""

extras/Hadrons/Application.cc

@@ -162,8 +162,7 @@ void Application::saveParameterFile(const std::string parameterFileName)
 sizeString((size)*locVol_) << " (" << sizeString(size)  << "/site)"
 
 #define DEFINE_MEMPEAK \
-GeneticScheduler<unsigned int>::ObjFunc memPeak = \
-[this](const std::vector<unsigned int> &program)\
+auto memPeak = [this](const std::vector<unsigned int> &program)\
 {\
     unsigned int memPeak;\
     bool         msg;\

extras/Hadrons/Global.hpp

@@ -145,15 +145,6 @@ std::string typeName(void)
     return typeName(typeIdPt<T>());
 }
 
-// default writers/readers
-#ifdef HAVE_HDF5
-typedef Hdf5Reader CorrReader;
-typedef Hdf5Writer CorrWriter;
-#else
-typedef XmlReader CorrReader;
-typedef XmlWriter CorrWriter;
-#endif
-
 END_HADRONS_NAMESPACE
 
 #endif // Hadrons_Global_hpp_

extras/Hadrons/Modules.hpp

@@ -29,20 +29,12 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Grid/Hadrons/Modules/MAction/DWF.hpp>
 #include <Grid/Hadrons/Modules/MAction/Wilson.hpp>
 #include <Grid/Hadrons/Modules/MContraction/Baryon.hpp>
-#include <Grid/Hadrons/Modules/MContraction/DiscLoop.hpp>
-#include <Grid/Hadrons/Modules/MContraction/Gamma3pt.hpp>
 #include <Grid/Hadrons/Modules/MContraction/Meson.hpp>
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
-#include <Grid/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
 #include <Grid/Hadrons/Modules/MGauge/Load.hpp>
 #include <Grid/Hadrons/Modules/MGauge/Random.hpp>
 #include <Grid/Hadrons/Modules/MGauge/Unit.hpp>
-#include <Grid/Hadrons/Modules/MLoop/NoiseLoop.hpp>
 #include <Grid/Hadrons/Modules/MSolver/RBPrecCG.hpp>
 #include <Grid/Hadrons/Modules/MSource/Point.hpp>
 #include <Grid/Hadrons/Modules/MSource/SeqGamma.hpp>
-#include <Grid/Hadrons/Modules/MSource/Wall.hpp>
 #include <Grid/Hadrons/Modules/MSource/Z2.hpp>
 #include <Grid/Hadrons/Modules/Quark.hpp>

extras/Hadrons/Modules/MAction/DWF.hpp

@@ -48,8 +48,7 @@ public:
                                     std::string, gauge,
                                     unsigned int, Ls,
                                     double      , mass,
-                                    double      , M5,
-                                    std::string , boundary);
+                                    double      , M5);
 };
 
 template <typename FImpl>
@@ -117,19 +116,14 @@ void TDWF<FImpl>::execute(void)
                  << par().mass << ", M5= " << par().M5 << " and Ls= "
                  << par().Ls << " using gauge field '" << par().gauge << "'"
                  << std::endl;
-    LOG(Message) << "Fermion boundary conditions: " << par().boundary 
-                 << std::endl;
     env().createGrid(par().Ls);
     auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
     auto &g4     = *env().getGrid();
     auto &grb4   = *env().getRbGrid();
     auto &g5     = *env().getGrid(par().Ls);
     auto &grb5   = *env().getRbGrid(par().Ls);
-    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
-    typename DomainWallFermion<FImpl>::ImplParams implParams(boundary);
     FMat *fMatPt = new DomainWallFermion<FImpl>(U, g5, grb5, g4, grb4,
-                                                par().mass, par().M5,
-                                                implParams);
+                                                par().mass, par().M5);
     env().setObject(getName(), fMatPt);
 }
 

extras/Hadrons/Modules/MAction/Wilson.hpp

@@ -46,8 +46,7 @@ class WilsonPar: Serializable
 public:
     GRID_SERIALIZABLE_CLASS_MEMBERS(WilsonPar,
                                     std::string, gauge,
-                                    double     , mass,
-                                    std::string, boundary);
+                                    double     , mass);
 };
 
 template <typename FImpl>
@@ -113,15 +112,10 @@ void TWilson<FImpl>::execute()
 {
     LOG(Message) << "Setting up TWilson fermion matrix with m= " << par().mass
                  << " using gauge field '" << par().gauge << "'" << std::endl;
-    LOG(Message) << "Fermion boundary conditions: " << par().boundary 
-                 << std::endl;
     auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
     auto &grid   = *env().getGrid();
     auto &gridRb = *env().getRbGrid();
-    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
-    typename WilsonFermion<FImpl>::ImplParams implParams(boundary);
-    FMat *fMatPt = new WilsonFermion<FImpl>(U, grid, gridRb, par().mass,
-                                            implParams);
+    FMat *fMatPt = new WilsonFermion<FImpl>(U, grid, gridRb, par().mass);
     env().setObject(getName(), fMatPt);
 }
 

extras/Hadrons/Modules/MContraction/Baryon.hpp

@@ -112,7 +112,7 @@ void TBaryon<FImpl1, FImpl2, FImpl3>::execute(void)
                  << " quarks '" << par().q1 << "', '" << par().q2 << "', and '"
                  << par().q3 << "'" << std::endl;
     
-    CorrWriter             writer(par().output);
+    XmlWriter             writer(par().output);
     PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
     PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
     PropagatorField3      &q3 = *env().template getObject<PropagatorField3>(par().q2);

extras/Hadrons/Modules/MContraction/DiscLoop.hpp

@@ -1,144 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/DiscLoop.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_DiscLoop_hpp_
-#define Hadrons_DiscLoop_hpp_
-
-#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Module.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/******************************************************************************
- *                                DiscLoop                                    *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-class DiscLoopPar: Serializable
-{
-public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(DiscLoopPar,
-                                    std::string,    q_loop,
-                                    Gamma::Algebra, gamma,
-                                    std::string,    output);
-};
-
-template <typename FImpl>
-class TDiscLoop: public Module<DiscLoopPar>
-{
-    TYPE_ALIASES(FImpl,);
-    class Result: Serializable
-    {
-    public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
-                                        Gamma::Algebra, gamma,
-                                        std::vector<Complex>, corr);
-    };
-public:
-    // constructor
-    TDiscLoop(const std::string name);
-    // destructor
-    virtual ~TDiscLoop(void) = default;
-    // dependency relation
-    virtual std::vector<std::string> getInput(void);
-    virtual std::vector<std::string> getOutput(void);
-    // setup
-    virtual void setup(void);
-    // execution
-    virtual void execute(void);
-};
-
-MODULE_REGISTER_NS(DiscLoop, TDiscLoop<FIMPL>, MContraction);
-
-/******************************************************************************
- *                       TDiscLoop implementation                             *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-template <typename FImpl>
-TDiscLoop<FImpl>::TDiscLoop(const std::string name)
-: Module<DiscLoopPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-template <typename FImpl>
-std::vector<std::string> TDiscLoop<FImpl>::getInput(void)
-{
-    std::vector<std::string> in = {par().q_loop};
-    
-    return in;
-}
-
-template <typename FImpl>
-std::vector<std::string> TDiscLoop<FImpl>::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TDiscLoop<FImpl>::setup(void)
-{
-    
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TDiscLoop<FImpl>::execute(void)
-{
-    LOG(Message) << "Computing disconnected loop contraction '" << getName() 
-                 << "' using '" << par().q_loop << "' with " << par().gamma 
-                 << " insertion." << std::endl;
-
-    CorrWriter            writer(par().output);
-    PropagatorField       &q_loop = *env().template getObject<PropagatorField>(par().q_loop);
-    LatticeComplex        c(env().getGrid());
-    Gamma                 gamma(par().gamma);
-    std::vector<TComplex> buf;
-    Result                result;
-
-    c = trace(gamma*q_loop);
-    sliceSum(c, buf, Tp);
-
-    result.gamma = par().gamma;
-    result.corr.resize(buf.size());
-    for (unsigned int t = 0; t < buf.size(); ++t)
-    {
-        result.corr[t] = TensorRemove(buf[t]);
-    }
-
-    write(writer, "disc", result);
-}
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_DiscLoop_hpp_

extras/Hadrons/Modules/MContraction/Gamma3pt.hpp

@@ -1,170 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/Gamma3pt.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_Gamma3pt_hpp_
-#define Hadrons_Gamma3pt_hpp_
-
-#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Module.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/*
- * 3pt contraction with gamma matrix insertion.
- *
- * Schematic:
- *
- *             q2           q3
- *        /----<------*------<----¬
- *       /          gamma          \
- *      /                           \
- *   i *                            * f
- *      \                          /
- *       \                        /
- *        \----------->----------/
- *                   q1
- *
- *      trace(g5*q1*adj(q2)*g5*gamma*q3)
- */
-
-/******************************************************************************
- *                               Gamma3pt                                     *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-class Gamma3ptPar: Serializable
-{
-public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(Gamma3ptPar,
-                                    std::string,    q1,
-                                    std::string,    q2,
-                                    std::string,    q3,
-                                    Gamma::Algebra, gamma,
-                                    std::string,    output);
-};
-
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-class TGamma3pt: public Module<Gamma3ptPar>
-{
-    TYPE_ALIASES(FImpl1, 1);
-    TYPE_ALIASES(FImpl2, 2);
-    TYPE_ALIASES(FImpl3, 3);
-    class Result: Serializable
-    {
-    public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
-                                        Gamma::Algebra, gamma,
-                                        std::vector<Complex>, corr);
-    };
-public:
-    // constructor
-    TGamma3pt(const std::string name);
-    // destructor
-    virtual ~TGamma3pt(void) = default;
-    // dependency relation
-    virtual std::vector<std::string> getInput(void);
-    virtual std::vector<std::string> getOutput(void);
-    // setup
-    virtual void setup(void);
-    // execution
-    virtual void execute(void);
-};
-
-MODULE_REGISTER_NS(Gamma3pt, ARG(TGamma3pt<FIMPL, FIMPL, FIMPL>), MContraction);
-
-/******************************************************************************
- *                       TGamma3pt implementation                             *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-TGamma3pt<FImpl1, FImpl2, FImpl3>::TGamma3pt(const std::string name)
-: Module<Gamma3ptPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getInput(void)
-{
-    std::vector<std::string> in = {par().q1, par().q2, par().q3};
-    
-    return in;
-}
-
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-void TGamma3pt<FImpl1, FImpl2, FImpl3>::setup(void)
-{
-    
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-template <typename FImpl1, typename FImpl2, typename FImpl3>
-void TGamma3pt<FImpl1, FImpl2, FImpl3>::execute(void)
-{
-    LOG(Message) << "Computing 3pt contractions '" << getName() << "' using"
-                 << " quarks '" << par().q1 << "', '" << par().q2 << "' and '"
-                 << par().q3 << "', with " << par().gamma << " insertion." 
-                 << std::endl;
-
-    CorrWriter            writer(par().output);
-    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
-    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
-    PropagatorField3      &q3 = *env().template getObject<PropagatorField3>(par().q3);
-    LatticeComplex        c(env().getGrid());
-    Gamma                 g5(Gamma::Algebra::Gamma5);
-    Gamma                 gamma(par().gamma);
-    std::vector<TComplex> buf;
-    Result                result;
-
-    c = trace(g5*q1*adj(q2)*(g5*gamma)*q3);
-    sliceSum(c, buf, Tp);
-
-    result.gamma = par().gamma;
-    result.corr.resize(buf.size());
-    for (unsigned int t = 0; t < buf.size(); ++t)
-    {
-        result.corr[t] = TensorRemove(buf[t]);
-    }
-
-    write(writer, "gamma3pt", result);
-}
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_Gamma3pt_hpp_

extras/Hadrons/Modules/MContraction/Meson.hpp

@@ -6,10 +6,8 @@ Source file: extras/Hadrons/Modules/MContraction/Meson.hpp
 
 Copyright (C) 2015
 Copyright (C) 2016
-Copyright (C) 2017
 
 Author: Antonin Portelli <antonin.portelli@me.com>
-        Andrew Lawson    <andrew.lawson1991@gmail.com>
 
 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
@@ -38,39 +36,20 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 BEGIN_HADRONS_NAMESPACE
 
-/*
- 
- Meson contractions
- -----------------------------
- 
- * options:
- - q1: input propagator 1 (string)
- - q2: input propagator 2 (string)
- - gammas: gamma products to insert at sink & source, pairs of gamma matrices 
-           (space-separated strings) in angled brackets (i.e. <g_sink g_src>),
-           in a sequence (e.g. "<Gamma5 Gamma5><Gamma5 GammaT>").
-
-           Special values: "all" - perform all possible contractions.
- - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0."),
-        given as multiples of (2*pi) / L.
-*/
-
 /******************************************************************************
  *                                TMeson                                       *
  ******************************************************************************/
 BEGIN_MODULE_NAMESPACE(MContraction)
 
-typedef std::pair<Gamma::Algebra, Gamma::Algebra> GammaPair;
-
 class MesonPar: Serializable
 {
 public:
     GRID_SERIALIZABLE_CLASS_MEMBERS(MesonPar,
-                                    std::string, q1,
-                                    std::string, q2,
-                                    std::string, gammas,
-                                    std::string, mom,
-                                    std::string, output);
+                                    std::string,    q1,
+                                    std::string,    q2,
+                                    std::string,    output,
+                                    Gamma::Algebra, gammaSource,
+                                    Gamma::Algebra, gammaSink);
 };
 
 template <typename FImpl1, typename FImpl2>
@@ -82,10 +61,7 @@ public:
     class Result: Serializable
     {
     public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
-                                        Gamma::Algebra, gamma_snk,
-                                        Gamma::Algebra, gamma_src,
-                                        std::vector<Complex>, corr);
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result, std::vector<Complex>, corr);
     };
 public:
     // constructor
@@ -95,7 +71,6 @@ public:
     // dependencies/products
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
-    virtual void parseGammaString(std::vector<GammaPair> &gammaList);
     // execution
     virtual void execute(void);
 };
@@ -128,31 +103,6 @@ std::vector<std::string> TMeson<FImpl1, FImpl2>::getOutput(void)
     return output;
 }
 
-template <typename FImpl1, typename FImpl2>
-void TMeson<FImpl1, FImpl2>::parseGammaString(std::vector<GammaPair> &gammaList)
-{
-    gammaList.clear();
-    // Determine gamma matrices to insert at source/sink.
-    if (par().gammas.compare("all") == 0)
-    {
-        // Do all contractions.
-        for (unsigned int i = 1; i < Gamma::nGamma; i += 2)
-        {
-            for (unsigned int j = 1; j < Gamma::nGamma; j += 2)
-            {
-                gammaList.push_back(std::make_pair((Gamma::Algebra)i, 
-                                                   (Gamma::Algebra)j));
-            }
-        }
-    }
-    else
-    {
-        // Parse individual contractions from input string.
-        gammaList = strToVec<GammaPair>(par().gammas);
-    }
-}
-
-
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl1, typename FImpl2>
 void TMeson<FImpl1, FImpl2>::execute(void)
@@ -161,44 +111,21 @@ void TMeson<FImpl1, FImpl2>::execute(void)
                  << " quarks '" << par().q1 << "' and '" << par().q2 << "'"
                  << std::endl;
     
-    CorrWriter              writer(par().output);
-    PropagatorField1       &q1 = *env().template getObject<PropagatorField1>(par().q1);
-    PropagatorField2       &q2 = *env().template getObject<PropagatorField2>(par().q2);
-    LatticeComplex         c(env().getGrid());
-    Gamma                  g5(Gamma::Algebra::Gamma5);
-    std::vector<GammaPair> gammaList;
-    std::vector<TComplex>  buf;
-    std::vector<Result>    result;
-    std::vector<Real>      p;
-
-    p  = strToVec<Real>(par().mom);
-    LatticeComplex         ph(env().getGrid()), coor(env().getGrid());
-    Complex                i(0.0,1.0);
-    ph = zero;
-    for(unsigned int mu = 0; mu < env().getNd(); mu++)
-    {
-        LatticeCoordinate(coor, mu);
-        ph = ph + p[mu]*coor*((1./(env().getGrid()->_fdimensions[mu])));
-    }
-    ph = exp((Real)(2*M_PI)*i*ph);
+    XmlWriter             writer(par().output);
+    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
+    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
+    LatticeComplex        c(env().getGrid());
+    Gamma                 gSrc(par().gammaSource), gSnk(par().gammaSink);
+    Gamma                 g5(Gamma::Algebra::Gamma5);
+    std::vector<TComplex> buf;
+    Result                result;
     
-    parseGammaString(gammaList);
-
-    result.resize(gammaList.size());
-    for (unsigned int i = 0; i < result.size(); ++i)
+    c = trace(gSnk*q1*adj(gSrc)*g5*adj(q2)*g5);
+    sliceSum(c, buf, Tp);
+    result.corr.resize(buf.size());
+    for (unsigned int t = 0; t < buf.size(); ++t)
     {
-        Gamma gSnk(gammaList[i].first);
-        Gamma gSrc(gammaList[i].second);
-        c = trace((g5*gSnk)*q1*(adj(gSrc)*g5)*adj(q2))*ph;
-        sliceSum(c, buf, Tp);
-
-        result[i].gamma_snk = gammaList[i].first;
-        result[i].gamma_src = gammaList[i].second;
-        result[i].corr.resize(buf.size());
-        for (unsigned int t = 0; t < buf.size(); ++t)
-        {
-            result[i].corr[t] = TensorRemove(buf[t]);
-        }
+        result.corr[t] = TensorRemove(buf[t]);
     }
     write(writer, "meson", result);
 }

extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp

@@ -1,114 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_WeakHamiltonian_hpp_
-#define Hadrons_WeakHamiltonian_hpp_
-
-#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Module.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/******************************************************************************
- *                         WeakHamiltonian                                    *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-/*******************************************************************************
- * Utilities for contractions involving the Weak Hamiltonian.
- ******************************************************************************/
-//// Sum and store correlator.
-#define MAKE_DIAG(exp, buf, res, n)\
-sliceSum(exp, buf, Tp);\
-res.name = (n);\
-res.corr.resize(buf.size());\
-for (unsigned int t = 0; t < buf.size(); ++t)\
-{\
-    res.corr[t] = TensorRemove(buf[t]);\
-}
-
-//// Contraction of mu index: use 'mu' variable in exp.
-#define SUM_MU(buf,exp)\
-buf = zero;\
-for (unsigned int mu = 0; mu < ndim; ++mu)\
-{\
-    buf += exp;\
-}
-
-enum 
-{
-  i_V = 0,
-  i_A = 1,
-  n_i = 2
-};
-
-class WeakHamiltonianPar: Serializable
-{
-public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(WeakHamiltonianPar,
-                                    std::string, q1,
-                                    std::string, q2,
-                                    std::string, q3,
-                                    std::string, q4,
-                                    std::string, output);
-};
-
-#define MAKE_WEAK_MODULE(modname)\
-class T##modname: public Module<WeakHamiltonianPar>\
-{\
-public:\
-    TYPE_ALIASES(FIMPL,)\
-    class Result: Serializable\
-    {\
-    public:\
-        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,\
-                                        std::string, name,\
-                                        std::vector<Complex>, corr);\
-    };\
-public:\
-    /* constructor */ \
-    T##modname(const std::string name);\
-    /* destructor */ \
-    virtual ~T##modname(void) = default;\
-    /* dependency relation */ \
-    virtual std::vector<std::string> getInput(void);\
-    virtual std::vector<std::string> getOutput(void);\
-    /* setup */ \
-    virtual void setup(void);\
-    /* execution */ \
-    virtual void execute(void);\
-    std::vector<std::string> VA_label = {"V", "A"};\
-};\
-MODULE_REGISTER_NS(modname, T##modname, MContraction);
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_WeakHamiltonian_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc

@@ -1,137 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
-
-using namespace Grid;
-using namespace Hadrons;
-using namespace MContraction;
-
-/*
- * Weak Hamiltonian current-current contractions, Eye-type.
- * 
- * These contractions are generated by the Q1 and Q2 operators in the physical
- * basis (see e.g. Fig 3 of arXiv:1507.03094).
- * 
- * Schematics:        q4                 |                  
- *                  /-<-¬                |                             
- *                 /     \               |             q2           q3
- *                 \     /               |        /----<------*------<----¬                        
- *            q2    \   /    q3          |       /          /-*-¬          \
- *       /-----<-----* *-----<----¬      |      /          /     \          \
- *    i *            H_W           * f   |   i *           \     /  q4      * f
- *       \                        /      |      \           \->-/          /   
- *        \                      /       |       \                        /       
- *         \---------->---------/        |        \----------->----------/        
- *                   q1                  |                   q1                  
- *                                       |
- *                Saucer (S)             |                  Eye (E)
- * 
- * S: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1]*q4*gL[mu][p_2])
- * E: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1])*trace(q4*gL[mu][p_2])
- */
-
-/******************************************************************************
- *                  TWeakHamiltonianEye implementation                        *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-TWeakHamiltonianEye::TWeakHamiltonianEye(const std::string name)
-: Module<WeakHamiltonianPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-std::vector<std::string> TWeakHamiltonianEye::getInput(void)
-{
-    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
-    
-    return in;
-}
-
-std::vector<std::string> TWeakHamiltonianEye::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-void TWeakHamiltonianEye::setup(void)
-{
-
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-void TWeakHamiltonianEye::execute(void)
-{
-    LOG(Message) << "Computing Weak Hamiltonian (Eye type) contractions '" 
-                 << getName() << "' using quarks '" << par().q1 << "', '" 
-                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
-                 << "'." << std::endl;
-
-    CorrWriter             writer(par().output);
-    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
-    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
-    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
-    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
-    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
-    LatticeComplex        expbuf(env().getGrid());
-    std::vector<TComplex> corrbuf;
-    std::vector<Result>   result(n_eye_diag);
-    unsigned int ndim   = env().getNd();
-
-    PropagatorField              tmp1(env().getGrid());
-    LatticeComplex               tmp2(env().getGrid());
-    std::vector<PropagatorField> S_body(ndim, tmp1);
-    std::vector<PropagatorField> S_loop(ndim, tmp1);
-    std::vector<LatticeComplex>  E_body(ndim, tmp2);
-    std::vector<LatticeComplex>  E_loop(ndim, tmp2);
-
-    // Setup for S-type contractions.
-    for (int mu = 0; mu < ndim; ++mu)
-    {
-        S_body[mu] = MAKE_SE_BODY(q1, q2, q3, GammaL(Gamma::gmu[mu]));
-        S_loop[mu] = MAKE_SE_LOOP(q4, GammaL(Gamma::gmu[mu]));
-    }
-
-    // Perform S-type contractions.    
-    SUM_MU(expbuf, trace(S_body[mu]*S_loop[mu]))
-    MAKE_DIAG(expbuf, corrbuf, result[S_diag], "HW_S")
-
-    // Recycle sub-expressions for E-type contractions.
-    for (unsigned int mu = 0; mu < ndim; ++mu)
-    {
-        E_body[mu] = trace(S_body[mu]);
-        E_loop[mu] = trace(S_loop[mu]);
-    }
-
-    // Perform E-type contractions.
-    SUM_MU(expbuf, E_body[mu]*E_loop[mu])
-    MAKE_DIAG(expbuf, corrbuf, result[E_diag], "HW_E")
-
-    write(writer, "HW_Eye", result);
-}

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp

@@ -1,58 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_WeakHamiltonianEye_hpp_
-#define Hadrons_WeakHamiltonianEye_hpp_
-
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/******************************************************************************
- *                         WeakHamiltonianEye                                 *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-enum
-{
-    S_diag = 0,
-    E_diag = 1,
-    n_eye_diag = 2
-};
-
-// Saucer and Eye subdiagram contractions.
-#define MAKE_SE_BODY(Q_1, Q_2, Q_3, gamma) (Q_3*g5*Q_1*adj(Q_2)*g5*gamma)
-#define MAKE_SE_LOOP(Q_loop, gamma) (Q_loop*gamma)
-
-MAKE_WEAK_MODULE(WeakHamiltonianEye)
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_WeakHamiltonianEye_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc

@@ -1,139 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
-
-using namespace Grid;
-using namespace Hadrons;
-using namespace MContraction;
-
-/*
- * Weak Hamiltonian current-current contractions, Non-Eye-type.
- * 
- * These contractions are generated by the Q1 and Q2 operators in the physical
- * basis (see e.g. Fig 3 of arXiv:1507.03094).
- * 
- * Schematic:     
- *            q2             q3          |           q2              q3
- *          /--<--¬       /--<--¬        |        /--<--¬         /--<--¬       
- *         /       \     /       \       |       /       \       /       \      
- *        /         \   /         \      |      /         \     /         \     
- *       /           \ /           \     |     /           \   /           \    
- *    i *             * H_W         *  f |  i *             * * H_W         * f 
- *      \             *             |    |     \           /   \           /
- *       \           / \           /     |      \         /     \         /    
- *        \         /   \         /      |       \       /       \       /  
- *         \       /     \       /       |        \-->--/         \-->--/      
- *          \-->--/       \-->--/        |          q1               q4 
- *            q1             q4          |
- *                Connected (C)          |                 Wing (W)
- *
- * C: trace(q1*adj(q2)*g5*gL[mu]*q3*adj(q4)*g5*gL[mu])
- * W: trace(q1*adj(q2)*g5*gL[mu])*trace(q3*adj(q4)*g5*gL[mu])
- * 
- */
-
-/******************************************************************************
- *                  TWeakHamiltonianNonEye implementation                     *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-TWeakHamiltonianNonEye::TWeakHamiltonianNonEye(const std::string name)
-: Module<WeakHamiltonianPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-std::vector<std::string> TWeakHamiltonianNonEye::getInput(void)
-{
-    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
-    
-    return in;
-}
-
-std::vector<std::string> TWeakHamiltonianNonEye::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-void TWeakHamiltonianNonEye::setup(void)
-{
-
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-void TWeakHamiltonianNonEye::execute(void)
-{
-    LOG(Message) << "Computing Weak Hamiltonian (Non-Eye type) contractions '" 
-                 << getName() << "' using quarks '" << par().q1 << "', '" 
-                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
-                 << "'." << std::endl;
-    
-    CorrWriter             writer(par().output);
-    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
-    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
-    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
-    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
-    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
-    LatticeComplex        expbuf(env().getGrid());
-    std::vector<TComplex> corrbuf;
-    std::vector<Result>   result(n_noneye_diag); 
-    unsigned int ndim   = env().getNd();
-
-    PropagatorField              tmp1(env().getGrid());
-    LatticeComplex               tmp2(env().getGrid());
-    std::vector<PropagatorField> C_i_side_loop(ndim, tmp1);
-    std::vector<PropagatorField> C_f_side_loop(ndim, tmp1);
-    std::vector<LatticeComplex>  W_i_side_loop(ndim, tmp2);
-    std::vector<LatticeComplex>  W_f_side_loop(ndim, tmp2);
-
-    // Setup for C-type contractions.
-    for (int mu = 0; mu < ndim; ++mu)
-    {
-        C_i_side_loop[mu] = MAKE_CW_SUBDIAG(q1, q2, GammaL(Gamma::gmu[mu]));
-        C_f_side_loop[mu] = MAKE_CW_SUBDIAG(q3, q4, GammaL(Gamma::gmu[mu]));
-    }
-
-    // Perform C-type contractions.    
-    SUM_MU(expbuf, trace(C_i_side_loop[mu]*C_f_side_loop[mu]))
-    MAKE_DIAG(expbuf, corrbuf, result[C_diag], "HW_C")
-
-    // Recycle sub-expressions for W-type contractions.
-    for (unsigned int mu = 0; mu < ndim; ++mu)
-    {
-        W_i_side_loop[mu] = trace(C_i_side_loop[mu]);
-        W_f_side_loop[mu] = trace(C_f_side_loop[mu]);
-    }
-
-    // Perform W-type contractions.
-    SUM_MU(expbuf, W_i_side_loop[mu]*W_f_side_loop[mu])
-    MAKE_DIAG(expbuf, corrbuf, result[W_diag], "HW_W")
-
-    write(writer, "HW_NonEye", result);
-}

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp

@@ -1,57 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_WeakHamiltonianNonEye_hpp_
-#define Hadrons_WeakHamiltonianNonEye_hpp_
-
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/******************************************************************************
- *                         WeakHamiltonianNonEye                              *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-enum
-{
-    W_diag = 0,
-    C_diag = 1,
-    n_noneye_diag = 2
-};
-
-// Wing and Connected subdiagram contractions
-#define MAKE_CW_SUBDIAG(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
-
-MAKE_WEAK_MODULE(WeakHamiltonianNonEye)
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_WeakHamiltonianNonEye_hpp_

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc

@@ -1,135 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#include <Grid/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
-
-using namespace Grid;
-using namespace Hadrons;
-using namespace MContraction;
-
-/*
- * Weak Hamiltonian + current contractions, disconnected topology for neutral 
- * mesons.
- * 
- * These contractions are generated by operators Q_1,...,10 of the dS=1 Weak
- * Hamiltonian in the physical basis and an additional current J (see e.g. 
- * Fig 11 of arXiv:1507.03094).
- * 
- * Schematic:
- *                        
- *           q2          q4             q3
- *       /--<--¬     /---<--¬       /---<--¬
- *     /         \ /         \     /        \
- *  i *           * H_W      |  J *          * f
- *     \         / \         /     \        /
- *      \--->---/   \-------/       \------/
- *          q1 
- * 
- * options
- * - q1: input propagator 1 (string)
- * - q2: input propagator 2 (string)
- * - q3: input propagator 3 (string), assumed to be sequential propagator 
- * - q4: input propagator 4 (string), assumed to be a loop
- * 
- * type 1: trace(q1*adj(q2)*g5*gL[mu])*trace(loop*gL[mu])*trace(q3*g5)
- * type 2: trace(q1*adj(q2)*g5*gL[mu]*loop*gL[mu])*trace(q3*g5)
- */
-
-/*******************************************************************************
- *                  TWeakNeutral4ptDisc implementation                         *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-TWeakNeutral4ptDisc::TWeakNeutral4ptDisc(const std::string name)
-: Module<WeakHamiltonianPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-std::vector<std::string> TWeakNeutral4ptDisc::getInput(void)
-{
-    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
-    
-    return in;
-}
-
-std::vector<std::string> TWeakNeutral4ptDisc::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-void TWeakNeutral4ptDisc::setup(void)
-{
-
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-void TWeakNeutral4ptDisc::execute(void)
-{
-    LOG(Message) << "Computing Weak Hamiltonian neutral disconnected contractions '" 
-                 << getName() << "' using quarks '" << par().q1 << "', '" 
-                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
-                 << "'." << std::endl;
-
-    CorrWriter             writer(par().output);
-    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
-    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
-    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
-    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
-    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
-    LatticeComplex        expbuf(env().getGrid());
-    std::vector<TComplex> corrbuf;
-    std::vector<Result>   result(n_neut_disc_diag);
-    unsigned int ndim   = env().getNd();
-
-    PropagatorField              tmp(env().getGrid());
-    std::vector<PropagatorField> meson(ndim, tmp);
-    std::vector<PropagatorField> loop(ndim, tmp);
-    LatticeComplex               curr(env().getGrid());
-
-    // Setup for type 1 contractions.
-    for (int mu = 0; mu < ndim; ++mu)
-    {
-        meson[mu] = MAKE_DISC_MESON(q1, q2, GammaL(Gamma::gmu[mu]));
-        loop[mu] = MAKE_DISC_LOOP(q4, GammaL(Gamma::gmu[mu]));
-    }
-    curr = MAKE_DISC_CURR(q3, GammaL(Gamma::Algebra::Gamma5));
-
-    // Perform type 1 contractions.    
-    SUM_MU(expbuf, trace(meson[mu]*loop[mu]))
-    expbuf *= curr;
-    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_1_diag], "HW_disc0_1")
-
-    // Perform type 2 contractions.
-    SUM_MU(expbuf, trace(meson[mu])*trace(loop[mu]))
-    expbuf *= curr;
-    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_2_diag], "HW_disc0_2")
-
-    write(writer, "HW_disc0", result);
-}

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp

@@ -1,59 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_WeakNeutral4ptDisc_hpp_
-#define Hadrons_WeakNeutral4ptDisc_hpp_
-
-#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/******************************************************************************
- *                         WeakNeutral4ptDisc                                 *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MContraction)
-
-enum
-{
-    neut_disc_1_diag = 0,
-    neut_disc_2_diag = 1,
-    n_neut_disc_diag = 2
-};
-
-// Neutral 4pt disconnected subdiagram contractions.
-#define MAKE_DISC_MESON(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
-#define MAKE_DISC_LOOP(Q_LOOP, gamma) (Q_LOOP*gamma)
-#define MAKE_DISC_CURR(Q_c, gamma) (trace(Q_c*gamma))
-
-MAKE_WEAK_MODULE(WeakNeutral4ptDisc)
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_WeakNeutral4ptDisc_hpp_

extras/Hadrons/Modules/MGauge/Load.cc

@@ -65,7 +65,7 @@ void TLoad::setup(void)
 // execution ///////////////////////////////////////////////////////////////////
 void TLoad::execute(void)
 {
-    FieldMetaData  header;
+    NerscField  header;
     std::string fileName = par().file + "."
                            + std::to_string(env().getTrajectory());
     
@@ -74,5 +74,5 @@ void TLoad::execute(void)
     LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
     NerscIO::readConfiguration(U, header, fileName);
     LOG(Message) << "NERSC header:" << std::endl;
-    dump_meta_data(header, LOG(Message));
+    dump_nersc_header(header, LOG(Message));
 }

extras/Hadrons/Modules/MLoop/NoiseLoop.hpp

@@ -1,132 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MLoop/NoiseLoop.hpp
-
-Copyright (C) 2016
-
-Author: Andrew Lawson <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_NoiseLoop_hpp_
-#define Hadrons_NoiseLoop_hpp_
-
-#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Module.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/*
- 
- Noise loop propagator
- -----------------------------
- * loop_x = q_x * adj(eta_x)
- 
- * options:
- - q = Result of inversion on noise source.
- - eta = noise source.
-
- */
-
-
-/******************************************************************************
- *                         NoiseLoop                                          *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MLoop)
-
-class NoiseLoopPar: Serializable
-{
-public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(NoiseLoopPar,
-                                    std::string, q,
-                                    std::string, eta);
-};
-
-template <typename FImpl>
-class TNoiseLoop: public Module<NoiseLoopPar>
-{
-public:
-    TYPE_ALIASES(FImpl,);
-public:
-    // constructor
-    TNoiseLoop(const std::string name);
-    // destructor
-    virtual ~TNoiseLoop(void) = default;
-    // dependency relation
-    virtual std::vector<std::string> getInput(void);
-    virtual std::vector<std::string> getOutput(void);
-    // setup
-    virtual void setup(void);
-    // execution
-    virtual void execute(void);
-};
-
-MODULE_REGISTER_NS(NoiseLoop, TNoiseLoop<FIMPL>, MLoop);
-
-/******************************************************************************
- *                 TNoiseLoop implementation                                  *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-template <typename FImpl>
-TNoiseLoop<FImpl>::TNoiseLoop(const std::string name)
-: Module<NoiseLoopPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-template <typename FImpl>
-std::vector<std::string> TNoiseLoop<FImpl>::getInput(void)
-{
-    std::vector<std::string> in = {par().q, par().eta};
-    
-    return in;
-}
-
-template <typename FImpl>
-std::vector<std::string> TNoiseLoop<FImpl>::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TNoiseLoop<FImpl>::setup(void)
-{
-    env().template registerLattice<PropagatorField>(getName());
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TNoiseLoop<FImpl>::execute(void)
-{
-    PropagatorField &loop = *env().template createLattice<PropagatorField>(getName());
-    PropagatorField &q    = *env().template getObject<PropagatorField>(par().q);
-    PropagatorField &eta  = *env().template getObject<PropagatorField>(par().eta);
-    loop = q*adj(eta);
-}
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_NoiseLoop_hpp_

extras/Hadrons/Modules/MSource/SeqGamma.hpp

@@ -6,7 +6,6 @@ Source file: extras/Hadrons/Modules/MSource/SeqGamma.hpp
 
 Copyright (C) 2015
 Copyright (C) 2016
-Copyright (C) 2017
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -150,9 +149,9 @@ void TSeqGamma<FImpl>::execute(void)
     for(unsigned int mu = 0; mu < env().getNd(); mu++)
     {
         LatticeCoordinate(coor, mu);
-        ph = ph + p[mu]*coor*((1./(env().getGrid()->_fdimensions[mu])));
+        ph = ph + p[mu]*coor;
     }
-    ph = exp((Real)(2*M_PI)*i*ph);
+    ph = exp(i*ph);
     LatticeCoordinate(t, Tp);
     src = where((t >= par().tA) and (t <= par().tB), ph*(g*q), 0.*q);
 }

extras/Hadrons/Modules/MSource/Wall.hpp

@@ -1,147 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid 
-
-Source file: extras/Hadrons/Modules/MSource/Wall.hpp
-
-Copyright (C) 2017
-
-Author: Andrew Lawson <andrew.lawson1991@gmail.com>
-
-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 */
-
-#ifndef Hadrons_WallSource_hpp_
-#define Hadrons_WallSource_hpp_
-
-#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Module.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
-
-BEGIN_HADRONS_NAMESPACE
-
-/*
- 
- Wall source
- -----------------------------
- * src_x = delta(x_3 - tW) * exp(i x.mom)
- 
- * options:
- - tW: source timeslice (integer)
- - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0.")
- 
- */
-
-/******************************************************************************
- *                         Wall                                               *
- ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MSource)
-
-class WallPar: Serializable
-{
-public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(WallPar,
-                                    unsigned int, tW,
-                                    std::string, mom);
-};
-
-template <typename FImpl>
-class TWall: public Module<WallPar>
-{
-public:
-    TYPE_ALIASES(FImpl,);
-public:
-    // constructor
-    TWall(const std::string name);
-    // destructor
-    virtual ~TWall(void) = default;
-    // dependency relation
-    virtual std::vector<std::string> getInput(void);
-    virtual std::vector<std::string> getOutput(void);
-    // setup
-    virtual void setup(void);
-    // execution
-    virtual void execute(void);
-};
-
-MODULE_REGISTER_NS(Wall, TWall<FIMPL>, MSource);
-
-/******************************************************************************
- *                 TWall implementation                                       *
- ******************************************************************************/
-// constructor /////////////////////////////////////////////////////////////////
-template <typename FImpl>
-TWall<FImpl>::TWall(const std::string name)
-: Module<WallPar>(name)
-{}
-
-// dependencies/products ///////////////////////////////////////////////////////
-template <typename FImpl>
-std::vector<std::string> TWall<FImpl>::getInput(void)
-{
-    std::vector<std::string> in;
-    
-    return in;
-}
-
-template <typename FImpl>
-std::vector<std::string> TWall<FImpl>::getOutput(void)
-{
-    std::vector<std::string> out = {getName()};
-    
-    return out;
-}
-
-// setup ///////////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TWall<FImpl>::setup(void)
-{
-    env().template registerLattice<PropagatorField>(getName());
-}
-
-// execution ///////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TWall<FImpl>::execute(void)
-{    
-    LOG(Message) << "Generating wall source at t = " << par().tW 
-                 << " with momentum " << par().mom << std::endl;
-    
-    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
-    Lattice<iScalar<vInteger>> t(env().getGrid());
-    LatticeComplex             ph(env().getGrid()), coor(env().getGrid());
-    std::vector<Real>          p;
-    Complex                    i(0.0,1.0);
-    
-    p  = strToVec<Real>(par().mom);
-    ph = zero;
-    for(unsigned int mu = 0; mu < Nd; mu++)
-    {
-        LatticeCoordinate(coor, mu);
-        ph = ph + p[mu]*coor*((1./(env().getGrid()->_fdimensions[mu])));
-    }
-    ph = exp((Real)(2*M_PI)*i*ph);
-    LatticeCoordinate(t, Tp);
-    src = 1.;
-    src = where((t == par().tW), src*ph, 0.*src);
-}
-
-END_MODULE_NAMESPACE
-
-END_HADRONS_NAMESPACE
-
-#endif // Hadrons_WallSource_hpp_

extras/Hadrons/Modules/Quark.hpp

@@ -173,7 +173,7 @@ void TQuark<FImpl>::execute(void)
                 *env().template getObject<PropagatorField>(getName());
             
             axpby_ssp_pminus(sol, 0., sol, 1., sol, 0, 0);
-            axpby_ssp_pplus(sol, 1., sol, 1., sol, 0, Ls_-1);
+            axpby_ssp_pplus(sol, 0., sol, 1., sol, 0, Ls_-1);
             ExtractSlice(tmp, sol, 0, 0);
             FermToProp(p4d, tmp, s, c);
         }

extras/Hadrons/modules.inc

@@ -1,7 +1,4 @@
 modules_cc =\
-  Modules/MContraction/WeakHamiltonianEye.cc \
-  Modules/MContraction/WeakHamiltonianNonEye.cc \
-  Modules/MContraction/WeakNeutral4ptDisc.cc \
   Modules/MGauge/Load.cc \
   Modules/MGauge/Random.cc \
   Modules/MGauge/Unit.cc
@@ -10,21 +7,13 @@ modules_hpp =\
   Modules/MAction/DWF.hpp \
   Modules/MAction/Wilson.hpp \
   Modules/MContraction/Baryon.hpp \
-  Modules/MContraction/DiscLoop.hpp \
-  Modules/MContraction/Gamma3pt.hpp \
   Modules/MContraction/Meson.hpp \
-  Modules/MContraction/WeakHamiltonian.hpp \
-  Modules/MContraction/WeakHamiltonianEye.hpp \
-  Modules/MContraction/WeakHamiltonianNonEye.hpp \
-  Modules/MContraction/WeakNeutral4ptDisc.hpp \
   Modules/MGauge/Load.hpp \
   Modules/MGauge/Random.hpp \
   Modules/MGauge/Unit.hpp \
-  Modules/MLoop/NoiseLoop.hpp \
   Modules/MSolver/RBPrecCG.hpp \
   Modules/MSource/Point.hpp \
   Modules/MSource/SeqGamma.hpp \
-  Modules/MSource/Wall.hpp \
   Modules/MSource/Z2.hpp \
   Modules/Quark.hpp
 

gcc-bug-report/README

@@ -20,17 +20,4 @@ The simple testcase in this directory is the submitted bug report that encapsula
 problem. The test case works with icpc and with clang++, but fails consistently on g++
 current variants.
 
-Peter
-
-
-************
-
-Second GCC bug reported, see Issue 100.
-
-https://wandbox.org/permlink/tzssJza6R9XnqANw
-https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80652
-
-Getting Travis fails under gcc-5 for Test_simd, now that I added more comprehensive testing to the
-CI test suite. The limitations of Travis runtime limits & weak cores are being shown.
-
-Travis uses 5.4.1 for g++-5.
+Peter

grid-config.in

@@ -1,86 +0,0 @@
-#! /bin/sh
-
-prefix=@prefix@
-exec_prefix=@exec_prefix@
-includedir=@includedir@
-
-usage()
-{
-  cat <<EOF
-Usage: grid-config [OPTION]
-
-Known values for OPTION are:
-
-  --prefix     show Grid installation prefix
-  --cxxflags   print pre-processor and compiler flags
-  --ldflags    print library linking flags
-  --libs       print library linking information
-  --summary    print full build summary
-  --help       display this help and exit
-  --version    output version information
-  --git        print git revision
-
-EOF
-  
-  exit $1
-}
-
-if test $# -eq 0; then
-  usage 1
-fi
-
-cflags=false
-libs=false
-
-while test $# -gt 0; do
-  case "$1" in
-    -*=*) optarg=`echo "$1" | sed 's/[-_a-zA-Z0-9]*=//'` ;;
-    *) optarg= ;;
-  esac
-  
-  case "$1" in
-    --prefix)
-      echo $prefix
-    ;;
-    
-    --version)
-      echo @VERSION@
-      exit 0
-    ;;
-    
-    --git)
-      echo "@GRID_BRANCH@ @GRID_SHA@"
-      exit 0
-    ;;
-    
-    --help)
-      usage 0
-    ;;
-    
-    --cxxflags)
-      echo @GRID_CXXFLAGS@
-    ;;
-    
-    --ldflags)
-      echo @GRID_LDFLAGS@
-    ;;
-    
-    --libs)
-      echo @GRID_LIBS@
-    ;;
-    
-    --summary)
-      echo ""
-      echo "@GRID_SUMMARY@"
-      echo ""
-    ;;
-    
-    *)
-      usage
-      exit 1
-    ;;
-  esac
-  shift
-done
-
-exit 0

lib/DisableWarnings.h

@@ -1,37 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/DisableWarnings.h
-
-Copyright (C) 2016
-
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
-
-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 */
-
-#ifndef DISABLE_WARNINGS_H
-#define DISABLE_WARNINGS_H
-
- //disables and intel compiler specific warning (in json.hpp)
-#pragma warning disable 488  
-
-
-#endif

lib/Grid.h

@@ -41,9 +41,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/GridCore.h>
 #include <Grid/GridQCDcore.h>
 #include <Grid/qcd/action/Action.h>
-#include <Grid/qcd/utils/GaugeFix.h>
 #include <Grid/qcd/smearing/Smearing.h>
-#include <Grid/parallelIO/MetaData.h>
 #include <Grid/qcd/hmc/HMC_aggregate.h>
 
 #endif

lib/GridCore.h

@@ -38,7 +38,28 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef GRID_BASE_H
 #define GRID_BASE_H
 
-#include <Grid/GridStd.h>
+///////////////////
+// Std C++ dependencies
+///////////////////
+#include <cassert>
+#include <complex>
+#include <vector>
+#include <iostream>
+#include <iomanip>
+#include <random>
+#include <functional>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <signal.h>
+#include <ctime>
+#include <sys/time.h>
+#include <chrono>
+
+///////////////////
+// Grid headers
+///////////////////
+#include "Config.h"
 
 #include <Grid/perfmon/Timer.h>
 #include <Grid/perfmon/PerfCount.h>

lib/GridStd.h

@@ -1,29 +0,0 @@
-#ifndef GRID_STD_H
-#define GRID_STD_H
-
-///////////////////
-// Std C++ dependencies
-///////////////////
-#include <cassert>
-#include <complex>
-#include <vector>
-#include <string>
-#include <iostream>
-#include <iomanip>
-#include <random>
-#include <functional>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <signal.h>
-#include <ctime>
-#include <sys/time.h>
-#include <chrono>
-#include <zlib.h>
-
-///////////////////
-// Grid config
-///////////////////
-#include "Config.h"
-
-#endif /* GRID_STD_H */

lib/Grid_Eigen_Dense.h

@@ -1,9 +0,0 @@
-#pragma once
-#if defined __GNUC__
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
-#endif
-#include <Grid/Eigen/Dense>
-#if defined __GNUC__
-#pragma GCC diagnostic pop
-#endif

lib/algorithms/Algorithms.h

@@ -46,7 +46,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/algorithms/iterative/ConjugateGradientMixedPrec.h>
 
 // Lanczos support
-//#include <Grid/algorithms/iterative/MatrixUtils.h>
+#include <Grid/algorithms/iterative/MatrixUtils.h>
 #include <Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h>
 #include <Grid/algorithms/CoarsenedMatrix.h>
 #include <Grid/algorithms/FFT.h>

lib/algorithms/LinearOperator.h

@@ -235,7 +235,7 @@ namespace Grid {
 	Field tmp(in._grid);
 
 	_Mat.MeooeDag(in,tmp);
-        _Mat.MooeeInvDag(tmp,out);
+	_Mat.MooeeInvDag(tmp,out);
 	_Mat.MeooeDag(out,tmp);
 
 	_Mat.MooeeDag(in,out);

lib/algorithms/approx/Chebyshev.h

@@ -197,9 +197,8 @@ namespace Grid {
     void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
 
       GridBase *grid=in._grid;
-
-      // std::cout << "Chevyshef(): in._grid="<<in._grid<<std::endl;
-      //std::cout <<" Linop.Grid()="<<Linop.Grid()<<"Linop.RedBlackGrid()="<<Linop.RedBlackGrid()<<std::endl;
+//std::cout << "Chevyshef(): in._grid="<<in._grid<<std::endl;
+//<<" Linop.Grid()="<<Linop.Grid()<<"Linop.RedBlackGrid()="<<Linop.RedBlackGrid()<<std::endl;
 
       int vol=grid->gSites();
 

lib/algorithms/approx/Remez.h

@@ -16,7 +16,7 @@
 #define INCLUDED_ALG_REMEZ_H
 
 #include <stddef.h>
-#include <Grid/GridStd.h>
+#include <Config.h>
 
 #ifdef HAVE_LIBGMP
 #include "bigfloat.h"

lib/algorithms/iterative/BlockConjugateGradient.h

@@ -1,593 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/algorithms/iterative/BlockConjugateGradient.h
-
-Copyright (C) 2017
-
-Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-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 */
-#ifndef GRID_BLOCK_CONJUGATE_GRADIENT_H
-#define GRID_BLOCK_CONJUGATE_GRADIENT_H
-
-
-namespace Grid {
-
-enum BlockCGtype { BlockCG, BlockCGrQ, CGmultiRHS };
-
-//////////////////////////////////////////////////////////////////////////
-// Block conjugate gradient. Dimension zero should be the block direction
-//////////////////////////////////////////////////////////////////////////
-template <class Field>
-class BlockConjugateGradient : public OperatorFunction<Field> {
- public:
-
-
-  typedef typename Field::scalar_type scomplex;
-
-  int blockDim ;
-  int Nblock;
-
-  BlockCGtype CGtype;
-  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
-                           // Defaults true.
-  RealD Tolerance;
-  Integer MaxIterations;
-  Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
-  
-  BlockConjugateGradient(BlockCGtype cgtype,int _Orthog,RealD tol, Integer maxit, bool err_on_no_conv = true)
-    : Tolerance(tol), CGtype(cgtype),   blockDim(_Orthog),  MaxIterations(maxit), ErrorOnNoConverge(err_on_no_conv)
-  {};
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Thin QR factorisation (google it)
-////////////////////////////////////////////////////////////////////////////////////////////////////
-void ThinQRfact (Eigen::MatrixXcd &m_rr,
-		 Eigen::MatrixXcd &C,
-		 Eigen::MatrixXcd &Cinv,
-		 Field & Q,
-		 const Field & R)
-{
-  int Orthog = blockDim; // First dimension is block dim; this is an assumption
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  //Dimensions
-  // R_{ferm x Nblock} =  Q_{ferm x Nblock} x  C_{Nblock x Nblock} -> ferm x Nblock
-  //
-  // Rdag R = m_rr = Herm = L L^dag        <-- Cholesky decomposition (LLT routine in Eigen)
-  //
-  //   Q  C = R => Q = R C^{-1}
-  //
-  // Want  Ident = Q^dag Q = C^{-dag} R^dag R C^{-1} = C^{-dag} L L^dag C^{-1} = 1_{Nblock x Nblock} 
-  //
-  // Set C = L^{dag}, and then Q^dag Q = ident 
-  //
-  // Checks:
-  // Cdag C = Rdag R ; passes.
-  // QdagQ  = 1      ; passes
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  sliceInnerProductMatrix(m_rr,R,R,Orthog);
-
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Cholesky from Eigen
-  // There exists a ldlt that is documented as more stable
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  Eigen::MatrixXcd L    = m_rr.llt().matrixL(); 
-
-  C    = L.adjoint();
-  Cinv = C.inverse();
-
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Q = R C^{-1}
-  //
-  // Q_j  = R_i Cinv(i,j) 
-  //
-  // NB maddMatrix conventions are Right multiplication X[j] a[j,i] already
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // FIXME:: make a sliceMulMatrix to avoid zero vector
-  sliceMulMatrix(Q,Cinv,R,Orthog);
-}
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Call one of several implementations
-////////////////////////////////////////////////////////////////////////////////////////////////////
-void operator()(LinearOperatorBase<Field> &Linop, const Field &Src, Field &Psi) 
-{
-  if ( CGtype == BlockCGrQ ) {
-    BlockCGrQsolve(Linop,Src,Psi);
-  } else if (CGtype == BlockCG ) {
-    BlockCGsolve(Linop,Src,Psi);
-  } else if (CGtype == CGmultiRHS ) {
-    CGmultiRHSsolve(Linop,Src,Psi);
-  } else {
-    assert(0);
-  }
-}
-
-////////////////////////////////////////////////////////////////////////////
-// BlockCGrQ implementation:
-//--------------------------
-// X is guess/Solution
-// B is RHS
-// Solve A X_i = B_i    ;        i refers to Nblock index
-////////////////////////////////////////////////////////////////////////////
-void BlockCGrQsolve(LinearOperatorBase<Field> &Linop, const Field &B, Field &X) 
-{
-  int Orthog = blockDim; // First dimension is block dim; this is an assumption
-  Nblock = B._grid->_fdimensions[Orthog];
-
-  std::cout<<GridLogMessage<<" Block Conjugate Gradient : Orthog "<<Orthog<<" Nblock "<<Nblock<<std::endl;
-
-  X.checkerboard = B.checkerboard;
-  conformable(X, B);
-
-  Field tmp(B);
-  Field Q(B);
-  Field D(B);
-  Field Z(B);
-  Field AD(B);
-
-  Eigen::MatrixXcd m_DZ     = Eigen::MatrixXcd::Identity(Nblock,Nblock);
-  Eigen::MatrixXcd m_M      = Eigen::MatrixXcd::Identity(Nblock,Nblock);
-  Eigen::MatrixXcd m_rr     = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-  Eigen::MatrixXcd m_C      = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-  Eigen::MatrixXcd m_Cinv   = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-  Eigen::MatrixXcd m_S      = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-  Eigen::MatrixXcd m_Sinv   = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-  Eigen::MatrixXcd m_tmp    = Eigen::MatrixXcd::Identity(Nblock,Nblock);
-  Eigen::MatrixXcd m_tmp1   = Eigen::MatrixXcd::Identity(Nblock,Nblock);
-
-  // Initial residual computation & set up
-  std::vector<RealD> residuals(Nblock);
-  std::vector<RealD> ssq(Nblock);
-
-  sliceNorm(ssq,B,Orthog);
-  RealD sssum=0;
-  for(int b=0;b<Nblock;b++) sssum+=ssq[b];
-
-  sliceNorm(residuals,B,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
-
-  sliceNorm(residuals,X,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
-
-  /************************************************************************
-   * Block conjugate gradient rQ (Sebastien Birk Thesis, after Dubrulle 2001)
-   ************************************************************************
-   * Dimensions:
-   *
-   *   X,B==(Nferm x Nblock)
-   *   A==(Nferm x Nferm)
-   *  
-   * Nferm = Nspin x Ncolour x Ncomplex x Nlattice_site
-   * 
-   * QC = R = B-AX, D = Q     ; QC => Thin QR factorisation (google it)
-   * for k: 
-   *   Z  = AD
-   *   M  = [D^dag Z]^{-1}
-   *   X  = X + D MC
-   *   QS = Q - ZM
-   *   D  = Q + D S^dag
-   *   C  = S C
-   */
-  ///////////////////////////////////////
-  // Initial block: initial search dir is guess
-  ///////////////////////////////////////
-  std::cout << GridLogMessage<<"BlockCGrQ algorithm initialisation " <<std::endl;
-
-  //1.  QC = R = B-AX, D = Q     ; QC => Thin QR factorisation (google it)
-
-  Linop.HermOp(X, AD);
-  tmp = B - AD;  
-  ThinQRfact (m_rr, m_C, m_Cinv, Q, tmp);
-  D=Q;
-
-  std::cout << GridLogMessage<<"BlockCGrQ computed initial residual and QR fact " <<std::endl;
-
-  ///////////////////////////////////////
-  // Timers
-  ///////////////////////////////////////
-  GridStopWatch sliceInnerTimer;
-  GridStopWatch sliceMaddTimer;
-  GridStopWatch QRTimer;
-  GridStopWatch MatrixTimer;
-  GridStopWatch SolverTimer;
-  SolverTimer.Start();
-
-  int k;
-  for (k = 1; k <= MaxIterations; k++) {
-
-    //3. Z  = AD
-    MatrixTimer.Start();
-    Linop.HermOp(D, Z);      
-    MatrixTimer.Stop();
-
-    //4. M  = [D^dag Z]^{-1}
-    sliceInnerTimer.Start();
-    sliceInnerProductMatrix(m_DZ,D,Z,Orthog);
-    sliceInnerTimer.Stop();
-    m_M       = m_DZ.inverse();
-
-    //5. X  = X + D MC
-    m_tmp     = m_M * m_C;
-    sliceMaddTimer.Start();
-    sliceMaddMatrix(X,m_tmp, D,X,Orthog);     
-    sliceMaddTimer.Stop();
-
-    //6. QS = Q - ZM
-    sliceMaddTimer.Start();
-    sliceMaddMatrix(tmp,m_M,Z,Q,Orthog,-1.0);
-    sliceMaddTimer.Stop();
-    QRTimer.Start();
-    ThinQRfact (m_rr, m_S, m_Sinv, Q, tmp);
-    QRTimer.Stop();
-    
-    //7. D  = Q + D S^dag
-    m_tmp = m_S.adjoint();
-    sliceMaddTimer.Start();
-    sliceMaddMatrix(D,m_tmp,D,Q,Orthog);
-    sliceMaddTimer.Stop();
-
-    //8. C  = S C
-    m_C = m_S*m_C;
-    
-    /*********************
-     * convergence monitor
-     *********************
-     */
-    m_rr = m_C.adjoint() * m_C;
-
-    RealD max_resid=0;
-    RealD rrsum=0;
-    RealD rr;
-
-    for(int b=0;b<Nblock;b++) {
-      rrsum+=real(m_rr(b,b));
-      rr = real(m_rr(b,b))/ssq[b];
-      if ( rr > max_resid ) max_resid = rr;
-    }
-
-    std::cout << GridLogIterative << "\titeration "<<k<<" rr_sum "<<rrsum<<" ssq_sum "<< sssum
-	      <<" ave "<<std::sqrt(rrsum/sssum) << " max "<< max_resid <<std::endl;
-
-    if ( max_resid < Tolerance*Tolerance ) { 
-
-      SolverTimer.Stop();
-
-      std::cout << GridLogMessage<<"BlockCGrQ converged in "<<k<<" iterations"<<std::endl;
-
-      for(int b=0;b<Nblock;b++){
-	std::cout << GridLogMessage<< "\t\tblock "<<b<<" computed resid "
-		  << std::sqrt(real(m_rr(b,b))/ssq[b])<<std::endl;
-      }
-      std::cout << GridLogMessage<<"\tMax residual is "<<std::sqrt(max_resid)<<std::endl;
-
-      Linop.HermOp(X, AD);
-      AD = AD-B;
-      std::cout << GridLogMessage <<"\t True residual is " << std::sqrt(norm2(AD)/norm2(B)) <<std::endl;
-
-      std::cout << GridLogMessage << "Time Breakdown "<<std::endl;
-      std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tInnerProd  " << sliceInnerTimer.Elapsed() <<std::endl;
-      std::cout << GridLogMessage << "\tMaddMatrix " << sliceMaddTimer.Elapsed()  <<std::endl;
-      std::cout << GridLogMessage << "\tThinQRfact " << QRTimer.Elapsed()  <<std::endl;
-	    
-      IterationsToComplete = k;
-      return;
-    }
-
-  }
-  std::cout << GridLogMessage << "BlockConjugateGradient(rQ) did NOT converge" << std::endl;
-
-  if (ErrorOnNoConverge) assert(0);
-  IterationsToComplete = k;
-}
-//////////////////////////////////////////////////////////////////////////
-// Block conjugate gradient; Original O'Leary Dimension zero should be the block direction
-//////////////////////////////////////////////////////////////////////////
-void BlockCGsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &Psi) 
-{
-  int Orthog = blockDim; // First dimension is block dim; this is an assumption
-  Nblock = Src._grid->_fdimensions[Orthog];
-
-  std::cout<<GridLogMessage<<" Block Conjugate Gradient : Orthog "<<Orthog<<" Nblock "<<Nblock<<std::endl;
-
-  Psi.checkerboard = Src.checkerboard;
-  conformable(Psi, Src);
-
-  Field P(Src);
-  Field AP(Src);
-  Field R(Src);
-  
-  Eigen::MatrixXcd m_pAp    = Eigen::MatrixXcd::Identity(Nblock,Nblock);
-  Eigen::MatrixXcd m_pAp_inv= Eigen::MatrixXcd::Identity(Nblock,Nblock);
-  Eigen::MatrixXcd m_rr     = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-  Eigen::MatrixXcd m_rr_inv = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-  Eigen::MatrixXcd m_alpha      = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-  Eigen::MatrixXcd m_beta   = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-  // Initial residual computation & set up
-  std::vector<RealD> residuals(Nblock);
-  std::vector<RealD> ssq(Nblock);
-
-  sliceNorm(ssq,Src,Orthog);
-  RealD sssum=0;
-  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); }
-
-  sliceNorm(residuals,Psi,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
-
-  // Initial search dir is guess
-  Linop.HermOp(Psi, AP);
-  
-
-  /************************************************************************
-   * Block conjugate gradient (Stephen Pickles, thesis 1995, pp 71, O Leary 1980)
-   ************************************************************************
-   * O'Leary : R = B - A X
-   * O'Leary : P = M R ; preconditioner M = 1
-   * O'Leary : alpha = PAP^{-1} RMR
-   * O'Leary : beta  = RMR^{-1}_old RMR_new
-   * O'Leary : X=X+Palpha
-   * O'Leary : R_new=R_old-AP alpha
-   * O'Leary : P=MR_new+P beta
-   */
-
-  R = Src - AP;  
-  P = R;
-  sliceInnerProductMatrix(m_rr,R,R,Orthog);
-
-  GridStopWatch sliceInnerTimer;
-  GridStopWatch sliceMaddTimer;
-  GridStopWatch MatrixTimer;
-  GridStopWatch SolverTimer;
-  SolverTimer.Start();
-
-  int k;
-  for (k = 1; k <= MaxIterations; k++) {
-
-    RealD rrsum=0;
-    for(int b=0;b<Nblock;b++) rrsum+=real(m_rr(b,b));
-
-    std::cout << GridLogIterative << "\titeration "<<k<<" rr_sum "<<rrsum<<" ssq_sum "<< sssum
-	      <<" / "<<std::sqrt(rrsum/sssum) <<std::endl;
-
-    MatrixTimer.Start();
-    Linop.HermOp(P, AP);
-    MatrixTimer.Stop();
-
-    // Alpha
-    sliceInnerTimer.Start();
-    sliceInnerProductMatrix(m_pAp,P,AP,Orthog);
-    sliceInnerTimer.Stop();
-    m_pAp_inv = m_pAp.inverse();
-    m_alpha   = m_pAp_inv * m_rr ;
-
-    // Psi, R update
-    sliceMaddTimer.Start();
-    sliceMaddMatrix(Psi,m_alpha, P,Psi,Orthog);     // add alpha *  P to psi
-    sliceMaddMatrix(R  ,m_alpha,AP,  R,Orthog,-1.0);// sub alpha * AP to resid
-    sliceMaddTimer.Stop();
-
-    // Beta
-    m_rr_inv = m_rr.inverse();
-    sliceInnerTimer.Start();
-    sliceInnerProductMatrix(m_rr,R,R,Orthog);
-    sliceInnerTimer.Stop();
-    m_beta = m_rr_inv *m_rr;
-
-    // Search update
-    sliceMaddTimer.Start();
-    sliceMaddMatrix(AP,m_beta,P,R,Orthog);
-    sliceMaddTimer.Stop();
-    P= AP;
-
-    /*********************
-     * convergence monitor
-     *********************
-     */
-    RealD max_resid=0;
-    RealD rr;
-    for(int b=0;b<Nblock;b++){
-      rr = real(m_rr(b,b))/ssq[b];
-      if ( rr > max_resid ) max_resid = rr;
-    }
-    
-    if ( max_resid < Tolerance*Tolerance ) { 
-
-      SolverTimer.Stop();
-
-      std::cout << GridLogMessage<<"BlockCG converged in "<<k<<" iterations"<<std::endl;
-      for(int b=0;b<Nblock;b++){
-	std::cout << GridLogMessage<< "\t\tblock "<<b<<" computed resid "
-		  << std::sqrt(real(m_rr(b,b))/ssq[b])<<std::endl;
-      }
-      std::cout << GridLogMessage<<"\tMax residual is "<<std::sqrt(max_resid)<<std::endl;
-
-      Linop.HermOp(Psi, AP);
-      AP = AP-Src;
-      std::cout << GridLogMessage <<"\t True residual is " << std::sqrt(norm2(AP)/norm2(Src)) <<std::endl;
-
-      std::cout << GridLogMessage << "Time Breakdown "<<std::endl;
-      std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tInnerProd  " << sliceInnerTimer.Elapsed() <<std::endl;
-      std::cout << GridLogMessage << "\tMaddMatrix " << sliceMaddTimer.Elapsed()  <<std::endl;
-	    
-      IterationsToComplete = k;
-      return;
-    }
-
-  }
-  std::cout << GridLogMessage << "BlockConjugateGradient did NOT converge" << std::endl;
-
-  if (ErrorOnNoConverge) assert(0);
-  IterationsToComplete = k;
-}
-//////////////////////////////////////////////////////////////////////////
-// multiRHS conjugate gradient. Dimension zero should be the block direction
-// Use this for spread out across nodes
-//////////////////////////////////////////////////////////////////////////
-void CGmultiRHSsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &Psi) 
-{
-  int Orthog = blockDim; // First dimension is block dim
-  Nblock = Src._grid->_fdimensions[Orthog];
-
-  std::cout<<GridLogMessage<<"MultiRHS Conjugate Gradient : Orthog "<<Orthog<<" Nblock "<<Nblock<<std::endl;
-
-  Psi.checkerboard = Src.checkerboard;
-  conformable(Psi, Src);
-
-  Field P(Src);
-  Field AP(Src);
-  Field R(Src);
-  
-  std::vector<ComplexD> v_pAp(Nblock);
-  std::vector<RealD> v_rr (Nblock);
-  std::vector<RealD> v_rr_inv(Nblock);
-  std::vector<RealD> v_alpha(Nblock);
-  std::vector<RealD> v_beta(Nblock);
-
-  // Initial residual computation & set up
-  std::vector<RealD> residuals(Nblock);
-  std::vector<RealD> ssq(Nblock);
-
-  sliceNorm(ssq,Src,Orthog);
-  RealD sssum=0;
-  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); }
-
-  sliceNorm(residuals,Psi,Orthog);
-  for(int b=0;b<Nblock;b++){ assert(std::isnan(residuals[b])==0); }
-
-  // Initial search dir is guess
-  Linop.HermOp(Psi, AP);
-
-  R = Src - AP;  
-  P = R;
-  sliceNorm(v_rr,R,Orthog);
-
-  GridStopWatch sliceInnerTimer;
-  GridStopWatch sliceMaddTimer;
-  GridStopWatch sliceNormTimer;
-  GridStopWatch MatrixTimer;
-  GridStopWatch SolverTimer;
-
-  SolverTimer.Start();
-  int k;
-  for (k = 1; k <= MaxIterations; k++) {
-
-    RealD rrsum=0;
-    for(int b=0;b<Nblock;b++) rrsum+=real(v_rr[b]);
-
-    std::cout << GridLogIterative << "\titeration "<<k<<" rr_sum "<<rrsum<<" ssq_sum "<< sssum
-	      <<" / "<<std::sqrt(rrsum/sssum) <<std::endl;
-
-    MatrixTimer.Start();
-    Linop.HermOp(P, AP);
-    MatrixTimer.Stop();
-
-    // Alpha
-    sliceInnerTimer.Start();
-    sliceInnerProductVector(v_pAp,P,AP,Orthog);
-    sliceInnerTimer.Stop();
-    for(int b=0;b<Nblock;b++){
-      v_alpha[b] = v_rr[b]/real(v_pAp[b]);
-    }
-
-    // Psi, R update
-    sliceMaddTimer.Start();
-    sliceMaddVector(Psi,v_alpha, P,Psi,Orthog);     // add alpha *  P to psi
-    sliceMaddVector(R  ,v_alpha,AP,  R,Orthog,-1.0);// sub alpha * AP to resid
-    sliceMaddTimer.Stop();
-
-    // Beta
-    for(int b=0;b<Nblock;b++){
-      v_rr_inv[b] = 1.0/v_rr[b];
-    }
-    sliceNormTimer.Start();
-    sliceNorm(v_rr,R,Orthog);
-    sliceNormTimer.Stop();
-    for(int b=0;b<Nblock;b++){
-      v_beta[b] = v_rr_inv[b] *v_rr[b];
-    }
-
-    // Search update
-    sliceMaddTimer.Start();
-    sliceMaddVector(P,v_beta,P,R,Orthog);
-    sliceMaddTimer.Stop();
-
-    /*********************
-     * convergence monitor
-     *********************
-     */
-    RealD max_resid=0;
-    for(int b=0;b<Nblock;b++){
-      RealD rr = v_rr[b]/ssq[b];
-      if ( rr > max_resid ) max_resid = rr;
-    }
-    
-    if ( max_resid < Tolerance*Tolerance ) { 
-
-      SolverTimer.Stop();
-
-      std::cout << GridLogMessage<<"MultiRHS solver converged in " <<k<<" iterations"<<std::endl;
-      for(int b=0;b<Nblock;b++){
-	std::cout << GridLogMessage<< "\t\tBlock "<<b<<" computed resid "<< std::sqrt(v_rr[b]/ssq[b])<<std::endl;
-      }
-      std::cout << GridLogMessage<<"\tMax residual is "<<std::sqrt(max_resid)<<std::endl;
-
-      Linop.HermOp(Psi, AP);
-      AP = AP-Src;
-      std::cout <<GridLogMessage << "\tTrue residual is " << std::sqrt(norm2(AP)/norm2(Src)) <<std::endl;
-
-      std::cout << GridLogMessage << "Time Breakdown "<<std::endl;
-      std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed()     <<std::endl;
-      std::cout << GridLogMessage << "\tInnerProd  " << sliceInnerTimer.Elapsed() <<std::endl;
-      std::cout << GridLogMessage << "\tNorm       " << sliceNormTimer.Elapsed() <<std::endl;
-      std::cout << GridLogMessage << "\tMaddMatrix " << sliceMaddTimer.Elapsed()  <<std::endl;
-
-
-      IterationsToComplete = k;
-      return;
-    }
-
-  }
-  std::cout << GridLogMessage << "MultiRHSConjugateGradient did NOT converge" << std::endl;
-
-  if (ErrorOnNoConverge) assert(0);
-  IterationsToComplete = k;
-}
-
-};
-
-}
-#endif

lib/algorithms/iterative/ConjugateGradient.h

@@ -78,12 +78,18 @@ class ConjugateGradient : public OperatorFunction<Field> {
     cp = a;
     ssq = norm2(src);
 
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient: guess " << guess << std::endl;
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient:   src " << ssq << std::endl;
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient:    mp " << d << std::endl;
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient:   mmp " << b << std::endl;
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient:  cp,r " << cp << std::endl;
-    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradient:     p " << a << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient: guess " << guess << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:   src " << ssq << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:    mp " << d << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:   mmp " << b << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:  cp,r " << cp << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:     p " << a << std::endl;
 
     RealD rsq = Tolerance * Tolerance * ssq;
 
@@ -93,7 +99,8 @@ class ConjugateGradient : public OperatorFunction<Field> {
     }
 
     std::cout << GridLogIterative << std::setprecision(4)
-              << "ConjugateGradient: k=0 residual " << cp << " target " << rsq << std::endl;
+              << "ConjugateGradient: k=0 residual " << cp << " target " << rsq
+              << std::endl;
 
     GridStopWatch LinalgTimer;
     GridStopWatch MatrixTimer;
@@ -123,11 +130,8 @@ class ConjugateGradient : public OperatorFunction<Field> {
       p = p * b + r;
 
       LinalgTimer.Stop();
-
       std::cout << GridLogIterative << "ConjugateGradient: Iteration " << k
                 << " residual " << cp << " target " << rsq << std::endl;
-      std::cout << GridLogDebug << "a = "<< a << " b_pred = "<< b_pred << "  b = "<< b << std::endl;
-      std::cout << GridLogDebug << "qq = "<< qq << " d = "<< d << "  c = "<< c << std::endl;
 
       // Stopping condition
       if (cp <= rsq) {
@@ -135,33 +139,32 @@ class ConjugateGradient : public OperatorFunction<Field> {
         Linop.HermOpAndNorm(psi, mmp, d, qq);
         p = mmp - src;
 
+        RealD mmpnorm = sqrt(norm2(mmp));
+        RealD psinorm = sqrt(norm2(psi));
         RealD srcnorm = sqrt(norm2(src));
         RealD resnorm = sqrt(norm2(p));
         RealD true_residual = resnorm / srcnorm;
 
-        std::cout << GridLogMessage << "ConjugateGradient Converged on iteration " << k << std::endl;
-        std::cout << GridLogMessage << "\tComputed residual " << sqrt(cp / ssq)<<std::endl;
-	std::cout << GridLogMessage << "\tTrue residual " << true_residual<<std::endl;
-	std::cout << GridLogMessage << "\tTarget " << Tolerance << std::endl;
-
-        std::cout << GridLogMessage << "Time breakdown "<<std::endl;
-	std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed() <<std::endl;
-	std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
-	std::cout << GridLogMessage << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
+        std::cout << GridLogMessage
+                  << "ConjugateGradient: Converged on iteration " << k << std::endl;
+        std::cout << GridLogMessage << "Computed residual " << sqrt(cp / ssq)
+                  << " true residual " << true_residual << " target "
+                  << Tolerance << std::endl;
+        std::cout << GridLogMessage << "Time elapsed: Iterations "
+                  << SolverTimer.Elapsed() << " Matrix  "
+                  << MatrixTimer.Elapsed() << " Linalg "
+                  << LinalgTimer.Elapsed();
+        std::cout << std::endl;
 
         if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
-
 	IterationsToComplete = k;	
-
         return;
       }
     }
     std::cout << GridLogMessage << "ConjugateGradient did NOT converge"
               << std::endl;
-
     if (ErrorOnNoConverge) assert(0);
     IterationsToComplete = k;
-
   }
 };
 }

lib/algorithms/iterative/DenseMatrix.h

@@ -0,0 +1,137 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/DenseMatrix.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef GRID_DENSE_MATRIX_H
+#define GRID_DENSE_MATRIX_H
+
+namespace Grid {
+    /////////////////////////////////////////////////////////////
+    // Matrix untils
+    /////////////////////////////////////////////////////////////
+
+template<class T> using DenseVector = std::vector<T>;
+template<class T> using DenseMatrix = DenseVector<DenseVector<T> >;
+
+template<class T> void Size(DenseVector<T> & vec, int &N) 
+{ 
+  N= vec.size();
+}
+template<class T> void Size(DenseMatrix<T> & mat, int &N,int &M) 
+{ 
+  N= mat.size();
+  M= mat[0].size();
+}
+
+template<class T> void SizeSquare(DenseMatrix<T> & mat, int &N) 
+{ 
+  int M; Size(mat,N,M);
+  assert(N==M);
+}
+
+template<class T> void Resize(DenseVector<T > & mat, int N) { 
+  mat.resize(N);
+}
+template<class T> void Resize(DenseMatrix<T > & mat, int N, int M) { 
+  mat.resize(N);
+  for(int i=0;i<N;i++){
+    mat[i].resize(M);
+  }
+}
+template<class T> void Fill(DenseMatrix<T> & mat, T&val) { 
+  int N,M;
+  Size(mat,N,M);
+  for(int i=0;i<N;i++){
+  for(int j=0;j<M;j++){
+    mat[i][j] = val;
+  }}
+}
+
+/** Transpose of a matrix **/
+template<class T> DenseMatrix<T> Transpose(DenseMatrix<T> & mat){
+  int N,M;
+  Size(mat,N,M);
+  DenseMatrix<T> C; Resize(C,M,N);
+  for(int i=0;i<M;i++){
+  for(int j=0;j<N;j++){
+    C[i][j] = mat[j][i];
+  }} 
+  return C;
+}
+/** Set DenseMatrix to unit matrix **/
+template<class T> void Unity(DenseMatrix<T> &A){
+  int N;  SizeSquare(A,N);
+  for(int i=0;i<N;i++){
+    for(int j=0;j<N;j++){
+      if ( i==j ) A[i][j] = 1;
+      else        A[i][j] = 0;
+    } 
+  } 
+}
+
+/** Add C * I to matrix **/
+template<class T>
+void PlusUnit(DenseMatrix<T> & A,T c){
+  int dim;  SizeSquare(A,dim);
+  for(int i=0;i<dim;i++){A[i][i] = A[i][i] + c;} 
+}
+
+/** return the Hermitian conjugate of matrix **/
+template<class T>
+DenseMatrix<T> HermitianConj(DenseMatrix<T> &mat){
+
+  int dim; SizeSquare(mat,dim);
+
+  DenseMatrix<T> C; Resize(C,dim,dim);
+
+  for(int i=0;i<dim;i++){
+    for(int j=0;j<dim;j++){
+      C[i][j] = conj(mat[j][i]);
+    } 
+  } 
+  return C;
+}
+/**Get a square submatrix**/
+template <class T>
+DenseMatrix<T> GetSubMtx(DenseMatrix<T> &A,int row_st, int row_end, int col_st, int col_end)
+{
+  DenseMatrix<T> H; Resize(H,row_end - row_st,col_end-col_st);
+
+  for(int i = row_st; i<row_end; i++){
+  for(int j = col_st; j<col_end; j++){
+    H[i-row_st][j-col_st]=A[i][j];
+  }}
+  return H;
+}
+
+}
+
+#include "Householder.h"
+#include "Francis.h"
+
+#endif
+

lib/algorithms/iterative/EigenSort.h

@@ -0,0 +1,81 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/EigenSort.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef GRID_EIGENSORT_H
+#define GRID_EIGENSORT_H
+
+
+namespace Grid {
+    /////////////////////////////////////////////////////////////
+    // Eigen sorter to begin with
+    /////////////////////////////////////////////////////////////
+
+template<class Field>
+class SortEigen {
+ private:
+  
+//hacking for testing for now
+ private:
+  static bool less_lmd(RealD left,RealD right){
+    return left > right;
+  }  
+  static bool less_pair(std::pair<RealD,Field const*>& left,
+                        std::pair<RealD,Field const*>& right){
+    return left.first > (right.first);
+  }  
+  
+  
+ public:
+
+  void push(DenseVector<RealD>& lmd,
+            DenseVector<Field>& evec,int N) {
+    DenseVector<Field> cpy(lmd.size(),evec[0]._grid);
+    for(int i=0;i<lmd.size();i++) cpy[i] = evec[i];
+    
+    DenseVector<std::pair<RealD, Field const*> > emod(lmd.size());    
+    for(int i=0;i<lmd.size();++i)
+      emod[i] = std::pair<RealD,Field const*>(lmd[i],&cpy[i]);
+
+    partial_sort(emod.begin(),emod.begin()+N,emod.end(),less_pair);
+
+    typename DenseVector<std::pair<RealD, Field const*> >::iterator it = emod.begin();
+    for(int i=0;i<N;++i){
+      lmd[i]=it->first;
+      evec[i]=*(it->second);
+      ++it;
+    }
+  }
+  void push(DenseVector<RealD>& lmd,int N) {
+    std::partial_sort(lmd.begin(),lmd.begin()+N,lmd.end(),less_lmd);
+  }
+  bool saturated(RealD lmd, RealD thrs) {
+    return fabs(lmd) > fabs(thrs);
+  }
+};
+
+}
+#endif

lib/algorithms/iterative/Francis.h

@@ -0,0 +1,525 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/Francis.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef FRANCIS_H
+#define FRANCIS_H
+
+#include <cstdlib>
+#include <string>
+#include <cmath>
+#include <iostream>
+#include <sstream>
+#include <stdexcept>
+#include <fstream>
+#include <complex>
+#include <algorithm>
+
+//#include <timer.h>
+//#include <lapacke.h>
+//#include <Eigen/Dense>
+
+namespace Grid {
+
+template <class T> int SymmEigensystem(DenseMatrix<T > &Ain, DenseVector<T> &evals, DenseMatrix<T> &evecs, RealD small);
+template <class T> int     Eigensystem(DenseMatrix<T > &Ain, DenseVector<T> &evals, DenseMatrix<T> &evecs, RealD small);
+
+/**
+  Find the eigenvalues of an upper hessenberg matrix using the Francis QR algorithm.
+H =
+      x  x  x  x  x  x  x  x  x
+      x  x  x  x  x  x  x  x  x
+      0  x  x  x  x  x  x  x  x
+      0  0  x  x  x  x  x  x  x
+      0  0  0  x  x  x  x  x  x
+      0  0  0  0  x  x  x  x  x
+      0  0  0  0  0  x  x  x  x
+      0  0  0  0  0  0  x  x  x
+      0  0  0  0  0  0  0  x  x
+Factorization is P T P^H where T is upper triangular (mod cc blocks) and P is orthagonal/unitary.
+**/
+template <class T>
+int QReigensystem(DenseMatrix<T> &Hin, DenseVector<T> &evals, DenseMatrix<T> &evecs, RealD small)
+{
+  DenseMatrix<T> H = Hin; 
+
+  int N ; SizeSquare(H,N);
+  int M = N;
+
+  Fill(evals,0);
+  Fill(evecs,0);
+
+  T s,t,x=0,y=0,z=0;
+  T u,d;
+  T apd,amd,bc;
+  DenseVector<T> p(N,0);
+  T nrm = Norm(H);    ///DenseMatrix Norm
+  int n, m;
+  int e = 0;
+  int it = 0;
+  int tot_it = 0;
+  int l = 0;
+  int r = 0;
+  DenseMatrix<T> P; Resize(P,N,N); Unity(P);
+  DenseVector<int> trows(N,0);
+
+  /// Check if the matrix is really hessenberg, if not abort
+  RealD sth = 0;
+  for(int j=0;j<N;j++){
+    for(int i=j+2;i<N;i++){
+      sth = abs(H[i][j]);
+      if(sth > small){
+	std::cout << "Non hessenberg H = " << sth << " > " << small << std::endl;
+	exit(1);
+      }
+    }
+  }
+
+  do{
+    std::cout << "Francis QR Step N = " << N << std::endl;
+    /** Check for convergence
+      x  x  x  x  x
+      0  x  x  x  x
+      0  0  x  x  x
+      0  0  x  x  x
+      0  0  0  0  x
+      for this matrix l = 4
+     **/
+    do{
+      l = Chop_subdiag(H,nrm,e,small);
+      r = 0;    ///May have converged on more than one eval
+      ///Single eval
+      if(l == N-1){
+        evals[e] = H[l][l];
+        N--; e++; r++; it = 0;
+      }
+      ///RealD eval
+      if(l == N-2){
+        trows[l+1] = 1;    ///Needed for UTSolve
+        apd = H[l][l] + H[l+1][l+1];
+        amd = H[l][l] - H[l+1][l+1];
+        bc =  (T)4.0*H[l+1][l]*H[l][l+1];
+        evals[e]   = (T)0.5*( apd + sqrt(amd*amd + bc) );
+        evals[e+1] = (T)0.5*( apd - sqrt(amd*amd + bc) );
+        N-=2; e+=2; r++; it = 0;
+      }
+    } while(r>0);
+
+    if(N ==0) break;
+
+    DenseVector<T > ck; Resize(ck,3);
+    DenseVector<T> v;   Resize(v,3);
+
+    for(int m = N-3; m >= l; m--){
+      ///Starting vector essentially random shift.
+      if(it%10 == 0 && N >= 3 && it > 0){
+        s = (T)1.618033989*( abs( H[N-1][N-2] ) + abs( H[N-2][N-3] ) );
+        t = (T)0.618033989*( abs( H[N-1][N-2] ) + abs( H[N-2][N-3] ) );
+        x = H[m][m]*H[m][m] + H[m][m+1]*H[m+1][m] - s*H[m][m] + t;
+        y = H[m+1][m]*(H[m][m] + H[m+1][m+1] - s);
+        z = H[m+1][m]*H[m+2][m+1];
+      }
+      ///Starting vector implicit Q theorem
+      else{
+        s = (H[N-2][N-2] + H[N-1][N-1]);
+        t = (H[N-2][N-2]*H[N-1][N-1] - H[N-2][N-1]*H[N-1][N-2]);
+        x = H[m][m]*H[m][m] + H[m][m+1]*H[m+1][m] - s*H[m][m] + t;
+        y = H[m+1][m]*(H[m][m] + H[m+1][m+1] - s);
+        z = H[m+1][m]*H[m+2][m+1];
+      }
+      ck[0] = x; ck[1] = y; ck[2] = z;
+
+      if(m == l) break;
+
+      /** Some stupid thing from numerical recipies, seems to work**/
+      // PAB.. for heaven's sake quote page, purpose, evidence it works.
+      //       what sort of comment is that!?!?!?
+      u=abs(H[m][m-1])*(abs(y)+abs(z));
+      d=abs(x)*(abs(H[m-1][m-1])+abs(H[m][m])+abs(H[m+1][m+1]));
+      if ((T)abs(u+d) == (T)abs(d) ){
+	l = m; break;
+      }
+
+      //if (u < small){l = m; break;}
+    }
+    if(it > 100000){
+     std::cout << "QReigensystem: bugger it got stuck after 100000 iterations" << std::endl;
+     std::cout << "got " << e << " evals " << l << " " << N << std::endl;
+      exit(1);
+    }
+    normalize(ck);    ///Normalization cancels in PHP anyway
+    T beta;
+    Householder_vector<T >(ck, 0, 2, v, beta);
+    Householder_mult<T >(H,v,beta,0,l,l+2,0);
+    Householder_mult<T >(H,v,beta,0,l,l+2,1);
+    ///Accumulate eigenvector
+    Householder_mult<T >(P,v,beta,0,l,l+2,1);
+    int sw = 0;      ///Are we on the last row?
+    for(int k=l;k<N-2;k++){
+      x = H[k+1][k];
+      y = H[k+2][k];
+      z = (T)0.0;
+      if(k+3 <= N-1){
+	z = H[k+3][k];
+      } else{
+	sw = 1; 
+	v[2] = (T)0.0;
+      }
+      ck[0] = x; ck[1] = y; ck[2] = z;
+      normalize(ck);
+      Householder_vector<T >(ck, 0, 2-sw, v, beta);
+      Householder_mult<T >(H,v, beta,0,k+1,k+3-sw,0);
+      Householder_mult<T >(H,v, beta,0,k+1,k+3-sw,1);
+      ///Accumulate eigenvector
+      Householder_mult<T >(P,v, beta,0,k+1,k+3-sw,1);
+    }
+    it++;
+    tot_it++;
+  }while(N > 1);
+  N = evals.size();
+  ///Annoying - UT solves in reverse order;
+  DenseVector<T> tmp; Resize(tmp,N);
+  for(int i=0;i<N;i++){
+    tmp[i] = evals[N-i-1];
+  } 
+  evals = tmp;
+  UTeigenvectors(H, trows, evals, evecs);
+  for(int i=0;i<evals.size();i++){evecs[i] = P*evecs[i]; normalize(evecs[i]);}
+  return tot_it;
+}
+
+template <class T>
+int my_Wilkinson(DenseMatrix<T> &Hin, DenseVector<T> &evals, DenseMatrix<T> &evecs, RealD small)
+{
+  /**
+  Find the eigenvalues of an upper Hessenberg matrix using the Wilkinson QR algorithm.
+  H =
+  x  x  0  0  0  0
+  x  x  x  0  0  0
+  0  x  x  x  0  0
+  0  0  x  x  x  0
+  0  0  0  x  x  x
+  0  0  0  0  x  x
+  Factorization is P T P^H where T is upper triangular (mod cc blocks) and P is orthagonal/unitary.  **/
+  return my_Wilkinson(Hin, evals, evecs, small, small);
+}
+
+template <class T>
+int my_Wilkinson(DenseMatrix<T> &Hin, DenseVector<T> &evals, DenseMatrix<T> &evecs, RealD small, RealD tol)
+{
+  int N; SizeSquare(Hin,N);
+  int M = N;
+
+  ///I don't want to modify the input but matricies must be passed by reference
+  //Scale a matrix by its "norm"
+  //RealD Hnorm = abs( Hin.LargestDiag() ); H =  H*(1.0/Hnorm);
+  DenseMatrix<T> H;  H = Hin;
+  
+  RealD Hnorm = abs(Norm(Hin));
+  H = H * (1.0 / Hnorm);
+
+  // TODO use openmp and memset
+  Fill(evals,0);
+  Fill(evecs,0);
+
+  T s, t, x = 0, y = 0, z = 0;
+  T u, d;
+  T apd, amd, bc;
+  DenseVector<T> p; Resize(p,N); Fill(p,0);
+
+  T nrm = Norm(H);    ///DenseMatrix Norm
+  int n, m;
+  int e = 0;
+  int it = 0;
+  int tot_it = 0;
+  int l = 0;
+  int r = 0;
+  DenseMatrix<T> P; Resize(P,N,N);
+  Unity(P);
+  DenseVector<int> trows(N, 0);
+  /// Check if the matrix is really symm tridiag
+  RealD sth = 0;
+  for(int j = 0; j < N; ++j)
+  {
+    for(int i = j + 2; i < N; ++i)
+    {
+      if(abs(H[i][j]) > tol || abs(H[j][i]) > tol)
+      {
+	std::cout << "Non Tridiagonal H(" << i << ","<< j << ") = |" << Real( real( H[j][i] ) ) << "| > " << tol << std::endl;
+	std::cout << "Warning tridiagonalize and call again" << std::endl;
+        // exit(1); // see what is going on
+        //return;
+      }
+    }
+  }
+
+  do{
+    do{
+      //Jasper
+      //Check if the subdiagonal term is small enough (<small)
+      //if true then it is converged.
+      //check start from H.dim - e - 1
+      //How to deal with more than 2 are converged?
+      //What if Chop_symm_subdiag return something int the middle?
+      //--------------
+      l = Chop_symm_subdiag(H,nrm, e, small);
+      r = 0;    ///May have converged on more than one eval
+      //Jasper
+      //In this case
+      // x  x  0  0  0  0
+      // x  x  x  0  0  0
+      // 0  x  x  x  0  0
+      // 0  0  x  x  x  0
+      // 0  0  0  x  x  0
+      // 0  0  0  0  0  x  <- l
+      //--------------
+      ///Single eval
+      if(l == N - 1)
+      {
+        evals[e] = H[l][l];
+        N--;
+        e++;
+        r++;
+        it = 0;
+      }
+      //Jasper
+      // x  x  0  0  0  0
+      // x  x  x  0  0  0
+      // 0  x  x  x  0  0
+      // 0  0  x  x  0  0
+      // 0  0  0  0  x  x  <- l
+      // 0  0  0  0  x  x
+      //--------------
+      ///RealD eval
+      if(l == N - 2)
+      {
+        trows[l + 1] = 1;    ///Needed for UTSolve
+        apd = H[l][l] + H[l + 1][ l + 1];
+        amd = H[l][l] - H[l + 1][l + 1];
+        bc =  (T) 4.0 * H[l + 1][l] * H[l][l + 1];
+        evals[e] = (T) 0.5 * (apd + sqrt(amd * amd + bc));
+        evals[e + 1] = (T) 0.5 * (apd - sqrt(amd * amd + bc));
+        N -= 2;
+        e += 2;
+        r++;
+        it = 0;
+      }
+    }while(r > 0);
+    //Jasper
+    //Already converged
+    //--------------
+    if(N == 0) break;
+
+    DenseVector<T> ck,v; Resize(ck,2); Resize(v,2);
+
+    for(int m = N - 3; m >= l; m--)
+    {
+      ///Starting vector essentially random shift.
+      if(it%10 == 0 && N >= 3 && it > 0)
+      {
+        t = abs(H[N - 1][N - 2]) + abs(H[N - 2][N - 3]);
+        x = H[m][m] - t;
+        z = H[m + 1][m];
+      } else {
+      ///Starting vector implicit Q theorem
+        d = (H[N - 2][N - 2] - H[N - 1][N - 1]) * (T) 0.5;
+        t =  H[N - 1][N - 1] - H[N - 1][N - 2] * H[N - 1][N - 2] 
+	  / (d + sign(d) * sqrt(d * d + H[N - 1][N - 2] * H[N - 1][N - 2]));
+        x = H[m][m] - t;
+        z = H[m + 1][m];
+      }
+      //Jasper
+      //why it is here????
+      //-----------------------
+      if(m == l)
+        break;
+
+      u = abs(H[m][m - 1]) * (abs(y) + abs(z));
+      d = abs(x) * (abs(H[m - 1][m - 1]) + abs(H[m][m]) + abs(H[m + 1][m + 1]));
+      if ((T)abs(u + d) == (T)abs(d))
+      {
+        l = m;
+        break;
+      }
+    }
+    //Jasper
+    if(it > 1000000)
+    {
+      std::cout << "Wilkinson: bugger it got stuck after 100000 iterations" << std::endl;
+      std::cout << "got " << e << " evals " << l << " " << N << std::endl;
+      exit(1);
+    }
+    //
+    T s, c;
+    Givens_calc<T>(x, z, c, s);
+    Givens_mult<T>(H, l, l + 1, c, -s, 0);
+    Givens_mult<T>(H, l, l + 1, c,  s, 1);
+    Givens_mult<T>(P, l, l + 1, c,  s, 1);
+    //
+    for(int k = l; k < N - 2; ++k)
+    {
+      x = H.A[k + 1][k];
+      z = H.A[k + 2][k];
+      Givens_calc<T>(x, z, c, s);
+      Givens_mult<T>(H, k + 1, k + 2, c, -s, 0);
+      Givens_mult<T>(H, k + 1, k + 2, c,  s, 1);
+      Givens_mult<T>(P, k + 1, k + 2, c,  s, 1);
+    }
+    it++;
+    tot_it++;
+  }while(N > 1);
+
+  N = evals.size();
+  ///Annoying - UT solves in reverse order;
+  DenseVector<T> tmp(N);
+  for(int i = 0; i < N; ++i)
+    tmp[i] = evals[N-i-1];
+  evals = tmp;
+  //
+  UTeigenvectors(H, trows, evals, evecs);
+  //UTSymmEigenvectors(H, trows, evals, evecs);
+  for(int i = 0; i < evals.size(); ++i)
+  {
+    evecs[i] = P * evecs[i];
+    normalize(evecs[i]);
+    evals[i] = evals[i] * Hnorm;
+  }
+  // // FIXME this is to test
+  // Hin.write("evecs3", evecs);
+  // Hin.write("evals3", evals);
+  // // check rsd
+  // for(int i = 0; i < M; i++) {
+  //   vector<T> Aevec = Hin * evecs[i];
+  //   RealD norm2(0.);
+  //   for(int j = 0; j < M; j++) {
+  //     norm2 += (Aevec[j] - evals[i] * evecs[i][j]) * (Aevec[j] - evals[i] * evecs[i][j]);
+  //   }
+  // }
+  return tot_it;
+}
+
+template <class T>
+void Hess(DenseMatrix<T > &A, DenseMatrix<T> &Q, int start){
+
+  /**
+  turn a matrix A =
+  x  x  x  x  x
+  x  x  x  x  x
+  x  x  x  x  x
+  x  x  x  x  x
+  x  x  x  x  x
+  into
+  x  x  x  x  x
+  x  x  x  x  x
+  0  x  x  x  x
+  0  0  x  x  x
+  0  0  0  x  x
+  with householder rotations
+  Slow.
+  */
+  int N ; SizeSquare(A,N);
+  DenseVector<T > p; Resize(p,N); Fill(p,0);
+
+  for(int k=start;k<N-2;k++){
+    //cerr << "hess" << k << std::endl;
+    DenseVector<T > ck,v; Resize(ck,N-k-1); Resize(v,N-k-1);
+    for(int i=k+1;i<N;i++){ck[i-k-1] = A(i,k);}  ///kth column
+    normalize(ck);    ///Normalization cancels in PHP anyway
+    T beta;
+    Householder_vector<T >(ck, 0, ck.size()-1, v, beta);  ///Householder vector
+    Householder_mult<T>(A,v,beta,start,k+1,N-1,0);  ///A -> PA
+    Householder_mult<T >(A,v,beta,start,k+1,N-1,1);  ///PA -> PAP^H
+    ///Accumulate eigenvector
+    Householder_mult<T >(Q,v,beta,start,k+1,N-1,1);  ///Q -> QP^H
+  }
+  /*for(int l=0;l<N-2;l++){
+    for(int k=l+2;k<N;k++){
+    A(0,k,l);
+    }
+    }*/
+}
+
+template <class T>
+void Tri(DenseMatrix<T > &A, DenseMatrix<T> &Q, int start){
+///Tridiagonalize a matrix
+  int N; SizeSquare(A,N);
+  Hess(A,Q,start);
+  /*for(int l=0;l<N-2;l++){
+    for(int k=l+2;k<N;k++){
+    A(0,l,k);
+    }
+    }*/
+}
+
+template <class T>
+void ForceTridiagonal(DenseMatrix<T> &A){
+///Tridiagonalize a matrix
+  int N ; SizeSquare(A,N);
+  for(int l=0;l<N-2;l++){
+    for(int k=l+2;k<N;k++){
+      A[l][k]=0;
+      A[k][l]=0;
+    }
+  }
+}
+
+template <class T>
+int my_SymmEigensystem(DenseMatrix<T > &Ain, DenseVector<T> &evals, DenseVector<DenseVector<T> > &evecs, RealD small){
+  ///Solve a symmetric eigensystem, not necessarily in tridiagonal form
+  int N; SizeSquare(Ain,N);
+  DenseMatrix<T > A; A = Ain;
+  DenseMatrix<T > Q; Resize(Q,N,N); Unity(Q);
+  Tri(A,Q,0);
+  int it = my_Wilkinson<T>(A, evals, evecs, small);
+  for(int k=0;k<N;k++){evecs[k] = Q*evecs[k];}
+  return it;
+}
+
+
+template <class T>
+int Wilkinson(DenseMatrix<T> &Ain, DenseVector<T> &evals, DenseVector<DenseVector<T> > &evecs, RealD small){
+  return my_Wilkinson(Ain, evals, evecs, small);
+}
+
+template <class T>
+int SymmEigensystem(DenseMatrix<T> &Ain, DenseVector<T> &evals, DenseVector<DenseVector<T> > &evecs, RealD small){
+  return my_SymmEigensystem(Ain, evals, evecs, small);
+}
+
+template <class T>
+int Eigensystem(DenseMatrix<T > &Ain, DenseVector<T> &evals, DenseVector<DenseVector<T> > &evecs, RealD small){
+///Solve a general eigensystem, not necessarily in tridiagonal form
+  int N = Ain.dim;
+  DenseMatrix<T > A(N); A = Ain;
+  DenseMatrix<T > Q(N);Q.Unity();
+  Hess(A,Q,0);
+  int it = QReigensystem<T>(A, evals, evecs, small);
+  for(int k=0;k<N;k++){evecs[k] = Q*evecs[k];}
+  return it;
+}
+
+}
+#endif

lib/algorithms/iterative/Householder.h

@@ -0,0 +1,242 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/Householder.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef HOUSEHOLDER_H
+#define HOUSEHOLDER_H
+
+#define TIMER(A) std::cout << GridLogMessage << __FUNC__ << " file "<< __FILE__ <<" line " << __LINE__ << std::endl;
+#define ENTER()  std::cout << GridLogMessage << "ENTRY "<<__FUNC__ << " file "<< __FILE__ <<" line " << __LINE__ << std::endl;
+#define LEAVE()  std::cout << GridLogMessage << "EXIT  "<<__FUNC__ << " file "<< __FILE__ <<" line " << __LINE__ << std::endl;
+
+#include <cstdlib>
+#include <string>
+#include <cmath>
+#include <iostream>
+#include <sstream>
+#include <stdexcept>
+#include <fstream>
+#include <complex>
+#include <algorithm>
+
+namespace Grid {
+/** Comparison function for finding the max element in a vector **/
+template <class T> bool cf(T i, T j) { 
+  return abs(i) < abs(j); 
+}
+
+/** 
+	Calculate a real Givens angle 
+ **/
+template <class T> inline void Givens_calc(T y, T z, T &c, T &s){
+
+  RealD mz = (RealD)abs(z);
+  
+  if(mz==0.0){
+    c = 1; s = 0;
+  }
+  if(mz >= (RealD)abs(y)){
+    T t = -y/z;
+    s = (T)1.0 / sqrt ((T)1.0 + t * t);
+    c = s * t;
+  } else {
+    T t = -z/y;
+    c = (T)1.0 / sqrt ((T)1.0 + t * t);
+    s = c * t;
+  }
+}
+
+template <class T> inline void Givens_mult(DenseMatrix<T> &A,  int i, int k, T c, T s, int dir)
+{
+  int q ; SizeSquare(A,q);
+
+  if(dir == 0){
+    for(int j=0;j<q;j++){
+      T nu = A[i][j];
+      T w  = A[k][j];
+      A[i][j] = (c*nu + s*w);
+      A[k][j] = (-s*nu + c*w);
+    }
+  }
+
+  if(dir == 1){
+    for(int j=0;j<q;j++){
+      T nu = A[j][i];
+      T w  = A[j][k];
+      A[j][i] = (c*nu - s*w);
+      A[j][k] = (s*nu + c*w);
+    }
+  }
+}
+
+/**
+	from input = x;
+	Compute the complex Householder vector, v, such that
+	P = (I - b v transpose(v) )
+	b = 2/v.v
+
+	P | x |    | x | k = 0
+	| x |    | 0 | 
+	| x | =  | 0 |
+	| x |    | 0 | j = 3
+	| x |	   | x |
+
+	These are the "Unreduced" Householder vectors.
+
+ **/
+template <class T> inline void Householder_vector(DenseVector<T> input, int k, int j, DenseVector<T> &v, T &beta)
+{
+  int N ; Size(input,N);
+  T m = *max_element(input.begin() + k, input.begin() + j + 1, cf<T> );
+
+  if(abs(m) > 0.0){
+    T alpha = 0;
+
+    for(int i=k; i<j+1; i++){
+      v[i] = input[i]/m;
+      alpha = alpha + v[i]*conj(v[i]);
+    }
+    alpha = sqrt(alpha);
+    beta = (T)1.0/(alpha*(alpha + abs(v[k]) ));
+
+    if(abs(v[k]) > 0.0)  v[k] = v[k] + (v[k]/abs(v[k]))*alpha;
+    else                 v[k] = -alpha;
+  } else{
+    for(int i=k; i<j+1; i++){
+      v[i] = 0.0;
+    } 
+  }
+}
+
+/**
+	from input = x;
+	Compute the complex Householder vector, v, such that
+	P = (I - b v transpose(v) )
+	b = 2/v.v
+
+	Px = alpha*e_dir
+
+	These are the "Unreduced" Householder vectors.
+
+ **/
+
+template <class T> inline void Householder_vector(DenseVector<T> input, int k, int j, int dir, DenseVector<T> &v, T &beta)
+{
+  int N = input.size();
+  T m = *max_element(input.begin() + k, input.begin() + j + 1, cf);
+  
+  if(abs(m) > 0.0){
+    T alpha = 0;
+
+    for(int i=k; i<j+1; i++){
+      v[i] = input[i]/m;
+      alpha = alpha + v[i]*conj(v[i]);
+    }
+    
+    alpha = sqrt(alpha);
+    beta = 1.0/(alpha*(alpha + abs(v[dir]) ));
+	
+    if(abs(v[dir]) > 0.0) v[dir] = v[dir] + (v[dir]/abs(v[dir]))*alpha;
+    else                  v[dir] = -alpha;
+  }else{
+    for(int i=k; i<j+1; i++){
+      v[i] = 0.0;
+    } 
+  }
+}
+
+/**
+	Compute the product PA if trans = 0
+	AP if trans = 1
+	P = (I - b v transpose(v) )
+	b = 2/v.v
+	start at element l of matrix A
+	v is of length j - k + 1 of v are nonzero
+ **/
+
+template <class T> inline void Householder_mult(DenseMatrix<T> &A , DenseVector<T> v, T beta, int l, int k, int j, int trans)
+{
+  int N ; SizeSquare(A,N);
+
+  if(abs(beta) > 0.0){
+    for(int p=l; p<N; p++){
+      T s = 0;
+      if(trans==0){
+	for(int i=k;i<j+1;i++) s += conj(v[i-k])*A[i][p];
+	s *= beta;
+	for(int i=k;i<j+1;i++){ A[i][p] = A[i][p]-s*conj(v[i-k]);}
+      } else {
+	for(int i=k;i<j+1;i++){ s += conj(v[i-k])*A[p][i];}
+	s *= beta;
+	for(int i=k;i<j+1;i++){ A[p][i]=A[p][i]-s*conj(v[i-k]);}
+      }
+    }
+  }
+}
+
+/**
+	Compute the product PA if trans = 0
+	AP if trans = 1
+	P = (I - b v transpose(v) )
+	b = 2/v.v
+	start at element l of matrix A
+	v is of length j - k + 1 of v are nonzero
+	A is tridiagonal
+ **/
+template <class T> inline void Householder_mult_tri(DenseMatrix<T> &A , DenseVector<T> v, T beta, int l, int M, int k, int j, int trans)
+{
+  if(abs(beta) > 0.0){
+
+    int N ; SizeSquare(A,N);
+
+    DenseMatrix<T> tmp; Resize(tmp,N,N); Fill(tmp,0); 
+
+    T s;
+    for(int p=l; p<M; p++){
+      s = 0;
+      if(trans==0){
+	for(int i=k;i<j+1;i++) s = s + conj(v[i-k])*A[i][p];
+      }else{
+	for(int i=k;i<j+1;i++) s = s + v[i-k]*A[p][i];
+      }
+      s = beta*s;
+      if(trans==0){
+	for(int i=k;i<j+1;i++) tmp[i][p] = tmp(i,p) - s*v[i-k];
+      }else{
+	for(int i=k;i<j+1;i++) tmp[p][i] = tmp[p][i] - s*conj(v[i-k]);
+      }
+    }
+    for(int p=l; p<M; p++){
+      if(trans==0){
+	for(int i=k;i<j+1;i++) A[i][p] = A[i][p] + tmp[i][p];
+      }else{
+	for(int i=k;i<j+1;i++) A[p][i] = A[p][i] + tmp[p][i];
+      }
+    }
+  }
+}
+}
+#endif

lib/algorithms/iterative/Matrix.h

@@ -0,0 +1,453 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/Matrix.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef MATRIX_H
+#define MATRIX_H
+
+#include <cstdlib>
+#include <string>
+#include <cmath>
+#include <vector>
+#include <iostream>
+#include <iomanip>
+#include <complex>
+#include <typeinfo>
+#include <Grid/Grid.h>
+
+
+/** Sign function **/
+template <class T> T sign(T p){return ( p/abs(p) );}
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+///////////////////// Hijack STL containers for our wicked means /////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class T> using Vector = Vector<T>;
+template<class T> using Matrix = Vector<Vector<T> >;
+
+template<class T> void Resize(Vector<T > & vec, int N) { vec.resize(N); }
+
+template<class T> void Resize(Matrix<T > & mat, int N, int M) { 
+  mat.resize(N);
+  for(int i=0;i<N;i++){
+    mat[i].resize(M);
+  }
+}
+template<class T> void Size(Vector<T> & vec, int &N) 
+{ 
+  N= vec.size();
+}
+template<class T> void Size(Matrix<T> & mat, int &N,int &M) 
+{ 
+  N= mat.size();
+  M= mat[0].size();
+}
+template<class T> void SizeSquare(Matrix<T> & mat, int &N) 
+{ 
+  int M; Size(mat,N,M);
+  assert(N==M);
+}
+template<class T> void SizeSame(Matrix<T> & mat1,Matrix<T> &mat2, int &N1,int &M1) 
+{ 
+  int N2,M2;
+  Size(mat1,N1,M1);
+  Size(mat2,N2,M2);
+  assert(N1==N2);
+  assert(M1==M2);
+}
+
+//*****************************************
+//*	(Complex) Vector operations	*
+//*****************************************
+
+/**Conj of a Vector **/
+template <class T> Vector<T> conj(Vector<T> p){
+	Vector<T> q(p.size());
+	for(int i=0;i<p.size();i++){q[i] = conj(p[i]);}
+	return q;
+}
+
+/** Norm of a Vector**/
+template <class T> T norm(Vector<T> p){
+	T sum = 0;
+	for(int i=0;i<p.size();i++){sum = sum + p[i]*conj(p[i]);}
+	return abs(sqrt(sum));
+}
+
+/** Norm squared of a Vector **/
+template <class T> T norm2(Vector<T> p){
+	T sum = 0;
+	for(int i=0;i<p.size();i++){sum = sum + p[i]*conj(p[i]);}
+	return abs((sum));
+}
+
+/** Sum elements of a Vector **/
+template <class T> T trace(Vector<T> p){
+	T sum = 0;
+	for(int i=0;i<p.size();i++){sum = sum + p[i];}
+	return sum;
+}
+
+/** Fill a Vector with constant c **/
+template <class T> void Fill(Vector<T> &p, T c){
+	for(int i=0;i<p.size();i++){p[i] = c;}
+}
+/** Normalize a Vector **/
+template <class T> void normalize(Vector<T> &p){
+	T m = norm(p);
+	if( abs(m) > 0.0) for(int i=0;i<p.size();i++){p[i] /= m;}
+}
+/** Vector by scalar **/
+template <class T, class U> Vector<T> times(Vector<T> p, U s){
+	for(int i=0;i<p.size();i++){p[i] *= s;}
+	return p;
+}
+template <class T, class U> Vector<T> times(U s, Vector<T> p){
+	for(int i=0;i<p.size();i++){p[i] *= s;}
+	return p;
+}
+/** inner product of a and b = conj(a) . b **/
+template <class T> T inner(Vector<T> a, Vector<T> b){
+	T m = 0.;
+	for(int i=0;i<a.size();i++){m = m + conj(a[i])*b[i];}
+	return m;
+}
+/** sum of a and b = a + b **/
+template <class T> Vector<T> add(Vector<T> a, Vector<T> b){
+	Vector<T> m(a.size());
+	for(int i=0;i<a.size();i++){m[i] = a[i] + b[i];}
+	return m;
+}
+/** sum of a and b = a - b **/
+template <class T> Vector<T> sub(Vector<T> a, Vector<T> b){
+	Vector<T> m(a.size());
+	for(int i=0;i<a.size();i++){m[i] = a[i] - b[i];}
+	return m;
+}
+
+/** 
+ *********************************
+ *	Matrices	         *
+ *********************************
+ **/
+
+template<class T> void Fill(Matrix<T> & mat, T&val) { 
+  int N,M;
+  Size(mat,N,M);
+  for(int i=0;i<N;i++){
+  for(int j=0;j<M;j++){
+    mat[i][j] = val;
+  }}
+}
+
+/** Transpose of a matrix **/
+Matrix<T> Transpose(Matrix<T> & mat){
+  int N,M;
+  Size(mat,N,M);
+  Matrix C; Resize(C,M,N);
+  for(int i=0;i<M;i++){
+  for(int j=0;j<N;j++){
+    C[i][j] = mat[j][i];
+  }} 
+  return C;
+}
+/** Set Matrix to unit matrix **/
+template<class T> void Unity(Matrix<T> &mat){
+  int N;  SizeSquare(mat,N);
+  for(int i=0;i<N;i++){
+    for(int j=0;j<N;j++){
+      if ( i==j ) A[i][j] = 1;
+      else        A[i][j] = 0;
+    } 
+  } 
+}
+/** Add C * I to matrix **/
+template<class T>
+void PlusUnit(Matrix<T> & A,T c){
+  int dim;  SizeSquare(A,dim);
+  for(int i=0;i<dim;i++){A[i][i] = A[i][i] + c;} 
+}
+
+/** return the Hermitian conjugate of matrix **/
+Matrix<T> HermitianConj(Matrix<T> &mat){
+
+  int dim; SizeSquare(mat,dim);
+
+  Matrix<T> C; Resize(C,dim,dim);
+
+  for(int i=0;i<dim;i++){
+    for(int j=0;j<dim;j++){
+      C[i][j] = conj(mat[j][i]);
+    } 
+  } 
+  return C;
+}
+
+/** return diagonal entries as a Vector **/
+Vector<T> diag(Matrix<T> &A)
+{
+  int dim; SizeSquare(A,dim);
+  Vector<T> d; Resize(d,dim);
+
+  for(int i=0;i<dim;i++){
+    d[i] = A[i][i];
+  }
+  return d;
+}
+
+/** Left multiply by a Vector **/
+Vector<T> operator *(Vector<T> &B,Matrix<T> &A)
+{
+  int K,M,N; 
+  Size(B,K);
+  Size(A,M,N);
+  assert(K==M);
+  
+  Vector<T> C; Resize(C,N);
+
+  for(int j=0;j<N;j++){
+    T sum = 0.0;
+    for(int i=0;i<M;i++){
+      sum += B[i] * A[i][j];
+    }
+    C[j] =  sum;
+  }
+  return C; 
+}
+
+/** return 1/diagonal entries as a Vector **/
+Vector<T> inv_diag(Matrix<T> & A){
+  int dim; SizeSquare(A,dim);
+  Vector<T> d; Resize(d,dim);
+  for(int i=0;i<dim;i++){
+    d[i] = 1.0/A[i][i];
+  }
+  return d;
+}
+/** Matrix Addition **/
+inline Matrix<T> operator + (Matrix<T> &A,Matrix<T> &B)
+{
+  int N,M  ; SizeSame(A,B,N,M);
+  Matrix C; Resize(C,N,M);
+  for(int i=0;i<N;i++){
+    for(int j=0;j<M;j++){
+      C[i][j] = A[i][j] +  B[i][j];
+    } 
+  } 
+  return C;
+} 
+/** Matrix Subtraction **/
+inline Matrix<T> operator- (Matrix<T> & A,Matrix<T> &B){
+  int N,M  ; SizeSame(A,B,N,M);
+  Matrix C; Resize(C,N,M);
+  for(int i=0;i<N;i++){
+  for(int j=0;j<M;j++){
+    C[i][j] = A[i][j] -  B[i][j];
+  }}
+  return C;
+} 
+
+/** Matrix scalar multiplication **/
+inline Matrix<T> operator* (Matrix<T> & A,T c){
+  int N,M; Size(A,N,M);
+  Matrix C; Resize(C,N,M);
+  for(int i=0;i<N;i++){
+  for(int j=0;j<M;j++){
+    C[i][j] = A[i][j]*c;
+  }} 
+  return C;
+} 
+/** Matrix Matrix multiplication **/
+inline Matrix<T> operator* (Matrix<T> &A,Matrix<T> &B){
+  int K,L,N,M;
+  Size(A,K,L);
+  Size(B,N,M); assert(L==N);
+  Matrix C; Resize(C,K,M);
+
+  for(int i=0;i<K;i++){
+    for(int j=0;j<M;j++){
+      T sum = 0.0;
+      for(int k=0;k<N;k++) sum += A[i][k]*B[k][j];
+      C[i][j] =sum;
+    }
+  }
+  return C; 
+} 
+/** Matrix Vector multiplication **/
+inline Vector<T> operator* (Matrix<T> &A,Vector<T> &B){
+  int M,N,K;
+  Size(A,N,M);
+  Size(B,K); assert(K==M);
+  Vector<T> C; Resize(C,N);
+  for(int i=0;i<N;i++){
+    T sum = 0.0;
+    for(int j=0;j<M;j++) sum += A[i][j]*B[j];
+    C[i] =  sum;
+  }
+  return C; 
+} 
+
+/** Some version of Matrix norm **/
+/*
+inline T Norm(){ // this is not a usual L2 norm
+    T norm = 0;
+    for(int i=0;i<dim;i++){
+      for(int j=0;j<dim;j++){
+	norm += abs(A[i][j]);
+    }}
+    return norm;
+  }
+*/
+
+/** Some version of Matrix norm **/
+template<class T> T LargestDiag(Matrix<T> &A)
+{
+  int dim ; SizeSquare(A,dim); 
+
+  T ld = abs(A[0][0]);
+  for(int i=1;i<dim;i++){
+    T cf = abs(A[i][i]);
+    if(abs(cf) > abs(ld) ){ld = cf;}
+  }
+  return ld;
+}
+
+/** Look for entries on the leading subdiagonal that are smaller than 'small' **/
+template <class T,class U> int Chop_subdiag(Matrix<T> &A,T norm, int offset, U small)
+{
+  int dim; SizeSquare(A,dim);
+  for(int l = dim - 1 - offset; l >= 1; l--) {             		
+    if((U)abs(A[l][l - 1]) < (U)small) {
+      A[l][l-1]=(U)0.0;
+      return l;
+    }
+  }
+  return 0;
+}
+
+/** Look for entries on the leading subdiagonal that are smaller than 'small' **/
+template <class T,class U> int Chop_symm_subdiag(Matrix<T> & A,T norm, int offset, U small) 
+{
+  int dim; SizeSquare(A,dim);
+  for(int l = dim - 1 - offset; l >= 1; l--) {
+    if((U)abs(A[l][l - 1]) < (U)small) {
+      A[l][l - 1] = (U)0.0;
+      A[l - 1][l] = (U)0.0;
+      return l;
+    }
+  }
+  return 0;
+}
+/**Assign a submatrix to a larger one**/
+template<class T>
+void AssignSubMtx(Matrix<T> & A,int row_st, int row_end, int col_st, int col_end, Matrix<T> &S)
+{
+  for(int i = row_st; i<row_end; i++){
+    for(int j = col_st; j<col_end; j++){
+      A[i][j] = S[i - row_st][j - col_st];
+    }
+  }
+}
+
+/**Get a square submatrix**/
+template <class T>
+Matrix<T> GetSubMtx(Matrix<T> &A,int row_st, int row_end, int col_st, int col_end)
+{
+  Matrix<T> H; Resize(row_end - row_st,col_end-col_st);
+
+  for(int i = row_st; i<row_end; i++){
+  for(int j = col_st; j<col_end; j++){
+    H[i-row_st][j-col_st]=A[i][j];
+  }}
+  return H;
+}
+  
+ /**Assign a submatrix to a larger one NB remember Vector Vectors are transposes of the matricies they represent**/
+template<class T>
+void AssignSubMtx(Matrix<T> & A,int row_st, int row_end, int col_st, int col_end, Matrix<T> &S)
+{
+  for(int i = row_st; i<row_end; i++){
+  for(int j = col_st; j<col_end; j++){
+    A[i][j] = S[i - row_st][j - col_st];
+  }}
+}
+  
+/** compute b_i A_ij b_j **/ // surprised no Conj
+template<class T> T proj(Matrix<T> A, Vector<T> B){
+  int dim; SizeSquare(A,dim);
+  int dimB; Size(B,dimB);
+  assert(dimB==dim);
+  T C = 0;
+  for(int i=0;i<dim;i++){
+    T sum = 0.0;
+    for(int j=0;j<dim;j++){
+      sum += A[i][j]*B[j];
+    }
+    C +=  B[i]*sum; // No conj?
+  }
+  return C; 
+}
+
+
+/*
+ *************************************************************
+ *
+ * Matrix Vector products
+ *
+ *************************************************************
+ */
+// Instead make a linop and call my CG;
+
+/// q -> q Q
+template <class T,class Fermion> void times(Vector<Fermion> &q, Matrix<T> &Q)
+{
+  int M; SizeSquare(Q,M);
+  int N; Size(q,N); 
+  assert(M==N);
+
+  times(q,Q,N);
+}
+
+/// q -> q Q
+template <class T> void times(multi1d<LatticeFermion> &q, Matrix<T> &Q, int N)
+{
+  GridBase *grid = q[0]._grid;
+  int M; SizeSquare(Q,M);
+  int K; Size(q,K); 
+  assert(N<M);
+  assert(N<K);
+  Vector<Fermion> S(N,grid );
+  for(int j=0;j<N;j++){
+    S[j] = zero;
+    for(int k=0;k<N;k++){
+      S[j] = S[j] +  q[k]* Q[k][j]; 
+    }
+  }
+  for(int j=0;j<q.size();j++){
+    q[j] = S[j];
+  }
+}
+#endif

lib/algorithms/iterative/MatrixUtils.h

@@ -0,0 +1,75 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/MatrixUtils.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    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 */
+#ifndef GRID_MATRIX_UTILS_H
+#define GRID_MATRIX_UTILS_H
+
+namespace Grid {
+
+  namespace MatrixUtils { 
+
+    template<class T> inline void Size(Matrix<T>& A,int &N,int &M){
+      N=A.size(); assert(N>0);
+      M=A[0].size();
+      for(int i=0;i<N;i++){
+	assert(A[i].size()==M);
+      }
+    }
+
+    template<class T> inline void SizeSquare(Matrix<T>& A,int &N)
+    {
+      int M;
+      Size(A,N,M);
+      assert(N==M);
+    }
+
+    template<class T> inline void Fill(Matrix<T>& A,T & val)
+    { 
+      int N,M;
+      Size(A,N,M);
+      for(int i=0;i<N;i++){
+      for(int j=0;j<M;j++){
+	A[i][j]=val;
+      }}
+    }
+    template<class T> inline void Diagonal(Matrix<T>& A,T & val)
+    { 
+      int N;
+      SizeSquare(A,N);
+      for(int i=0;i<N;i++){
+	A[i][i]=val;
+      }
+    }
+    template<class T> inline void Identity(Matrix<T>& A)
+    {
+      Fill(A,0.0);
+      Diagonal(A,1.0);
+    }
+
+  };
+}
+#endif

lib/algorithms/iterative/SchurRedBlack.h

@@ -141,5 +141,85 @@ namespace Grid {
     }     
   };
 
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////
+  // Take a matrix and form a Red Black solver calling a Herm solver
+  // Use of RB info prevents making SchurRedBlackSolve conform to standard interface
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////
+  template<class Field> class SchurRedBlackDiagTwoSolve {
+  private:
+    OperatorFunction<Field> & _HermitianRBSolver;
+    int CBfactorise;
+  public:
+
+    /////////////////////////////////////////////////////
+    // Wrap the usual normal equations Schur trick
+    /////////////////////////////////////////////////////
+  SchurRedBlackDiagTwoSolve(OperatorFunction<Field> &HermitianRBSolver)  :
+     _HermitianRBSolver(HermitianRBSolver) 
+    { 
+      CBfactorise=0;
+    };
+
+    template<class Matrix>
+      void operator() (Matrix & _Matrix,const Field &in, Field &out){
+
+      // FIXME CGdiagonalMee not implemented virtual function
+      // FIXME use CBfactorise to control schur decomp
+      GridBase *grid = _Matrix.RedBlackGrid();
+      GridBase *fgrid= _Matrix.Grid();
+
+      SchurDiagTwoOperator<Matrix,Field> _HermOpEO(_Matrix);
+ 
+      Field src_e(grid);
+      Field src_o(grid);
+      Field sol_e(grid);
+      Field sol_o(grid);
+      Field   tmp(grid);
+      Field  Mtmp(grid);
+      Field resid(fgrid);
+
+      pickCheckerboard(Even,src_e,in);
+      pickCheckerboard(Odd ,src_o,in);
+      pickCheckerboard(Even,sol_e,out);
+      pickCheckerboard(Odd ,sol_o,out);
+    
+      /////////////////////////////////////////////////////
+      // 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);     
+
+      // get the right MpcDag
+      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.checkerboard ==Odd);       
+
+      //////////////////////////////////////////////////////////////
+      // Call the red-black solver
+      //////////////////////////////////////////////////////////////
+      std::cout<<GridLogMessage << "SchurRedBlack solver calling the MpcDagMp solver" <<std::endl;
+//      _HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.checkerboard==Odd);
+      _HermitianRBSolver(_HermOpEO,src_o,tmp);  assert(tmp.checkerboard==Odd);
+      _Matrix.MooeeInv(tmp,sol_o);        assert(  sol_o.checkerboard   ==Odd);
+
+      ///////////////////////////////////////////////////
+      // 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);
+     
+      setCheckerboard(out,sol_e); assert(  sol_e.checkerboard ==Even);
+      setCheckerboard(out,sol_o); assert(  sol_o.checkerboard ==Odd );
+
+      // Verify the unprec residual
+      _Matrix.M(out,resid); 
+      resid = resid-in;
+      RealD ns = norm2(in);
+      RealD nr = norm2(resid);
+
+      std::cout<<GridLogMessage << "SchurRedBlackDiagTwo solver true unprec resid "<< std::sqrt(nr/ns) <<" nr "<< nr <<" ns "<<ns << std::endl;
+    }     
+  };
+
 }
 #endif

lib/algorithms/iterative/TODO

@@ -0,0 +1,15 @@
+- ConjugateGradientMultiShift
+- MCR
+
+- Potentially Useful Boost libraries
+
+- MultiArray
+- Aligned allocator; memory pool
+- Remez -- Mike or Boost?
+- Multiprecision
+- quaternians
+- Tokenize
+- Serialization
+- Regex
+- Proto (ET)
+- uBlas

lib/algorithms/iterative/bisec.c

@@ -0,0 +1,122 @@
+#include <math.h>
+#include <stdlib.h>
+#include <vector>
+
+struct Bisection {
+
+static void get_eig2(int row_num,std::vector<RealD> &ALPHA,std::vector<RealD> &BETA, std::vector<RealD> & eig)
+{
+  int i,j;
+  std::vector<RealD> evec1(row_num+3);
+  std::vector<RealD> evec2(row_num+3);
+  RealD eps2;
+  ALPHA[1]=0.;
+  BETHA[1]=0.;
+  for(i=0;i<row_num-1;i++) {
+    ALPHA[i+1] = A[i*(row_num+1)].real();
+    BETHA[i+2] = A[i*(row_num+1)+1].real();
+  }
+  ALPHA[row_num] = A[(row_num-1)*(row_num+1)].real();
+  bisec(ALPHA,BETHA,row_num,1,row_num,1e-10,1e-10,evec1,eps2);
+  bisec(ALPHA,BETHA,row_num,1,row_num,1e-16,1e-16,evec2,eps2);
+
+  // Do we really need to sort here?
+  int begin=1;
+  int end = row_num;
+  int swapped=1;
+  while(swapped) {
+    swapped=0;
+    for(i=begin;i<end;i++){
+      if(mag(evec2[i])>mag(evec2[i+1]))	{
+	swap(evec2+i,evec2+i+1);
+	swapped=1;
+      }
+    }
+    end--;
+    for(i=end-1;i>=begin;i--){
+      if(mag(evec2[i])>mag(evec2[i+1]))	{
+	swap(evec2+i,evec2+i+1);
+	swapped=1;
+      }
+    }
+    begin++;
+  }
+
+  for(i=0;i<row_num;i++){
+    for(j=0;j<row_num;j++) {
+      if(i==j) H[i*row_num+j]=evec2[i+1];
+      else H[i*row_num+j]=0.;
+    }
+  }
+}
+
+static void bisec(std::vector<RealD> &c,   
+		  std::vector<RealD> &b,
+		  int n,
+		  int m1,
+		  int m2,
+		  RealD eps1,
+		  RealD relfeh,
+		  std::vector<RealD> &x,
+		  RealD &eps2)
+{
+  std::vector<RealD> wu(n+2);
+
+  RealD h,q,x1,xu,x0,xmin,xmax; 
+  int i,a,k;
+
+  b[1]=0.0;
+  xmin=c[n]-fabs(b[n]);
+  xmax=c[n]+fabs(b[n]);
+  for(i=1;i<n;i++){
+    h=fabs(b[i])+fabs(b[i+1]);
+    if(c[i]+h>xmax) xmax= c[i]+h;
+    if(c[i]-h<xmin) xmin= c[i]-h;
+  }
+  xmax *=2.;
+
+  eps2=relfeh*((xmin+xmax)>0.0 ? xmax : -xmin);
+  if(eps1<=0.0) eps1=eps2;
+  eps2=0.5*eps1+7.0*(eps2);
+  x0=xmax;
+  for(i=m1;i<=m2;i++){
+    x[i]=xmax;
+    wu[i]=xmin;
+  }
+
+  for(k=m2;k>=m1;k--){
+    xu=xmin;
+    i=k;
+    do{
+      if(xu<wu[i]){
+	xu=wu[i];
+	i=m1-1;
+      }
+      i--;
+    }while(i>=m1);
+    if(x0>x[k]) x0=x[k];
+    while((x0-xu)>2*relfeh*(fabs(xu)+fabs(x0))+eps1){
+      x1=(xu+x0)/2;
+
+      a=0;
+      q=1.0;
+      for(i=1;i<=n;i++){
+	q=c[i]-x1-((q!=0.0)? b[i]*b[i]/q:fabs(b[i])/relfeh);
+	if(q<0) a++;
+      }
+      //			printf("x1=%e a=%d\n",x1,a);
+      if(a<k){
+	if(a<m1){
+	  xu=x1;
+	  wu[m1]=x1;
+	}else {
+	  xu=x1;
+	  wu[a+1]=x1;
+	  if(x[a]>x1) x[a]=x1;
+	}
+      }else x0=x1;
+    }
+    x[k]=(x0+xu)/2;
+  }
+}
+}

lib/algorithms/iterative/get_eig.c

@@ -0,0 +1 @@
+

lib/cartesian/Cartesian_base.h

@@ -6,9 +6,8 @@
 
     Copyright (C) 2015
 
-    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-    Author: paboyle <paboyle@ph.ed.ac.uk>
-    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
 
     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
@@ -50,6 +49,7 @@ public:
 
     GridBase(const std::vector<int> & processor_grid) : CartesianCommunicator(processor_grid) {};
 
+
     // Physics Grid information.
     std::vector<int> _simd_layout;// Which dimensions get relayed out over simd lanes.
     std::vector<int> _fdimensions;// (full) Global dimensions of array prior to cb removal
@@ -62,12 +62,13 @@ public:
     int _isites;
     int _fsites;                  // _isites*_osites = product(dimensions).
     int _gsites;
-    std::vector<int> _slice_block;// subslice information
+    std::vector<int> _slice_block;   // subslice information
     std::vector<int> _slice_stride;
     std::vector<int> _slice_nblock;
 
-    std::vector<int> _lstart;     // local start of array in gcoors _processor_coor[d]*_ldimensions[d]
-    std::vector<int> _lend  ;     // local end of array in gcoors   _processor_coor[d]*_ldimensions[d]+_ldimensions_[d]-1
+    // Might need these at some point
+    //    std::vector<int> _lstart;     // local start of array in gcoors. _processor_coor[d]*_ldimensions[d]
+    //    std::vector<int> _lend;       // local end of array in gcoors    _processor_coor[d]*_ldimensions[d]+_ldimensions_[d]-1
 
 public:
 
@@ -98,7 +99,7 @@ public:
     virtual int oIndex(std::vector<int> &coor)
     {
         int idx=0;
-        // Works with either global or local coordinates
+	// Works with either global or local coordinates
         for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
         return idx;
     }
@@ -120,12 +121,6 @@ public:
       Lexicographic::CoorFromIndex(coor,Oindex,_rdimensions);
     }
 
-    inline void InOutCoorToLocalCoor (std::vector<int> &ocoor, std::vector<int> &icoor, std::vector<int> &lcoor) {
-      lcoor.resize(_ndimension);
-      for (int d = 0; d < _ndimension; d++)
-        lcoor[d] = ocoor[d] + _rdimensions[d] * icoor[d];
-    }
-
     //////////////////////////////////////////////////////////
     // SIMD lane addressing
     //////////////////////////////////////////////////////////
@@ -133,7 +128,6 @@ public:
     {
       Lexicographic::CoorFromIndex(coor,lane,_simd_layout);
     }
-
     inline int PermuteDim(int dimension){
       return _simd_layout[dimension]>1;
     }
@@ -151,15 +145,15 @@ public:
       // Distance should be either 0,1,2..
       //
       if ( _simd_layout[dimension] > 2 ) { 
-        for(int d=0;d<_ndimension;d++){
-          if ( d != dimension ) assert ( (_simd_layout[d]==1)  );
-        }
-        permute_type = RotateBit; // How to specify distance; this is not just direction.
-        return permute_type;
+	for(int d=0;d<_ndimension;d++){
+	  if ( d != dimension ) assert ( (_simd_layout[d]==1)  );
+	}
+	permute_type = RotateBit; // How to specify distance; this is not just direction.
+	return permute_type;
       }
 
       for(int d=_ndimension-1;d>dimension;d--){
-        if (_simd_layout[d]>1 ) permute_type++;
+	if (_simd_layout[d]>1 ) permute_type++;
       }
       return permute_type;
     }
@@ -174,30 +168,11 @@ public:
     inline int gSites(void) const { return _isites*_osites*_Nprocessors; }; 
     inline int Nd    (void) const { return _ndimension;};
 
-    inline const std::vector<int> LocalStarts(void)             { return _lstart;    };
     inline const std::vector<int> &FullDimensions(void)         { return _fdimensions;};
     inline const std::vector<int> &GlobalDimensions(void)       { return _gdimensions;};
     inline const std::vector<int> &LocalDimensions(void)        { return _ldimensions;};
     inline const std::vector<int> &VirtualLocalDimensions(void) { return _ldimensions;};
 
-    ////////////////////////////////////////////////////////////////
-    // Utility to print the full decomposition details 
-    ////////////////////////////////////////////////////////////////
-
-    void show_decomposition(){
-      std::cout << GridLogMessage << "Full Dimensions    : " << _fdimensions << std::endl;
-      std::cout << GridLogMessage << "Global Dimensions  : " << _gdimensions << std::endl;
-      std::cout << GridLogMessage << "Local Dimensions   : " << _ldimensions << std::endl;
-      std::cout << GridLogMessage << "Reduced Dimensions : " << _rdimensions << std::endl;
-      std::cout << GridLogMessage << "Outer strides      : " << _ostride << std::endl;
-      std::cout << GridLogMessage << "Inner strides      : " << _istride << std::endl;
-      std::cout << GridLogMessage << "iSites             : " << _isites << std::endl;
-      std::cout << GridLogMessage << "oSites             : " << _osites << std::endl;
-      std::cout << GridLogMessage << "lSites             : " << lSites() << std::endl;        
-      std::cout << GridLogMessage << "gSites             : " << gSites() << std::endl;
-      std::cout << GridLogMessage << "Nd                 : " << _ndimension << std::endl;             
-    } 
-
     ////////////////////////////////////////////////////////////////
     // Global addressing
     ////////////////////////////////////////////////////////////////
@@ -209,15 +184,12 @@ public:
       assert(lidx<lSites());
       Lexicographic::CoorFromIndex(lcoor,lidx,_ldimensions);
     }
-
-
-
     void GlobalCoorToGlobalIndex(const std::vector<int> & gcoor,int & gidx){
       gidx=0;
       int mult=1;
       for(int mu=0;mu<_ndimension;mu++) {
-        gidx+=mult*gcoor[mu];
-        mult*=_gdimensions[mu];
+	gidx+=mult*gcoor[mu];
+	mult*=_gdimensions[mu];
       }
     }
     void GlobalCoorToProcessorCoorLocalCoor(std::vector<int> &pcoor,std::vector<int> &lcoor,const std::vector<int> &gcoor)
@@ -225,9 +197,9 @@ public:
       pcoor.resize(_ndimension);
       lcoor.resize(_ndimension);
       for(int mu=0;mu<_ndimension;mu++){
-        int _fld  = _fdimensions[mu]/_processors[mu];
-        pcoor[mu] = gcoor[mu]/_fld;
-        lcoor[mu] = gcoor[mu]%_fld;
+	int _fld  = _fdimensions[mu]/_processors[mu];
+	pcoor[mu] = gcoor[mu]/_fld;
+	lcoor[mu] = gcoor[mu]%_fld;
       }
     }
     void GlobalCoorToRankIndex(int &rank, int &o_idx, int &i_idx ,const std::vector<int> &gcoor)
@@ -239,9 +211,9 @@ public:
       /*
       std::vector<int> cblcoor(lcoor);
       for(int d=0;d<cblcoor.size();d++){
-        if( this->CheckerBoarded(d) ) {
-          cblcoor[d] = lcoor[d]/2;
-        }
+	if( this->CheckerBoarded(d) ) {
+	  cblcoor[d] = lcoor[d]/2;
+	}
       }
       */
       i_idx= iIndex(lcoor);
@@ -267,7 +239,7 @@ public:
     {
       RankIndexToGlobalCoor(rank,o_idx,i_idx ,fcoor);
       if(CheckerBoarded(0)){
-        fcoor[0] = fcoor[0]*2+cb;
+	fcoor[0] = fcoor[0]*2+cb;
       }
     }
     void ProcessorCoorLocalCoorToGlobalCoor(std::vector<int> &Pcoor,std::vector<int> &Lcoor,std::vector<int> &gcoor)

lib/cartesian/Cartesian_full.h

@@ -76,8 +76,6 @@ public:
         _ldimensions.resize(_ndimension);
         _rdimensions.resize(_ndimension);
         _simd_layout.resize(_ndimension);
-	_lstart.resize(_ndimension);
-	_lend.resize(_ndimension);
             
         _ostride.resize(_ndimension);
         _istride.resize(_ndimension);
@@ -96,10 +94,8 @@ public:
 	  // Use a reduced simd grid
 	  _ldimensions[d]= _gdimensions[d]/_processors[d];  //local dimensions
 	  _rdimensions[d]= _ldimensions[d]/_simd_layout[d]; //overdecomposition
-	  _lstart[d]     = _processor_coor[d]*_ldimensions[d];
-	  _lend[d]       = _processor_coor[d]*_ldimensions[d]+_ldimensions[d]-1;
-	  _osites  *= _rdimensions[d];
-	  _isites  *= _simd_layout[d];
+	  _osites *= _rdimensions[d];
+	  _isites *= _simd_layout[d];
                 
 	  // Addressing support
 	  if ( d==0 ) {

lib/cartesian/Cartesian_red_black.h

@@ -151,8 +151,6 @@ public:
       _ldimensions.resize(_ndimension);
       _rdimensions.resize(_ndimension);
       _simd_layout.resize(_ndimension);
-      _lstart.resize(_ndimension);
-      _lend.resize(_ndimension);
       
       _ostride.resize(_ndimension);
       _istride.resize(_ndimension);
@@ -171,8 +169,6 @@ public:
 	  _gdimensions[d] = _gdimensions[d]/2; // Remove a checkerboard
 	}
 	_ldimensions[d] = _gdimensions[d]/_processors[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];

lib/communicator/Communicator_base.cc

@@ -60,7 +60,6 @@ void CartesianCommunicator::ShmBufferFreeAll(void) {
 /////////////////////////////////
 // Grid information queries
 /////////////////////////////////
-int                      CartesianCommunicator::Dimensions(void)         { return _ndimension; };
 int                      CartesianCommunicator::IsBoss(void)            { return _processor==0; };
 int                      CartesianCommunicator::BossRank(void)          { return 0; };
 int                      CartesianCommunicator::ThisRank(void)          { return _processor; };
@@ -92,7 +91,6 @@ void CartesianCommunicator::GlobalSumVector(ComplexD *c,int N)
 #if !defined( GRID_COMMS_MPI3) && !defined (GRID_COMMS_MPI3L)
 
 int                      CartesianCommunicator::NodeCount(void)    { return ProcessorCount();};
-int                      CartesianCommunicator::RankCount(void)    { return ProcessorCount();};
 
 double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsRequest_t> &list,
 						       void *xmit,

lib/communicator/Communicator_base.h

@@ -148,7 +148,6 @@ class CartesianCommunicator {
   int  RankFromProcessorCoor(std::vector<int> &coor);
   void ProcessorCoorFromRank(int rank,std::vector<int> &coor);
   
-  int                      Dimensions(void)        ;
   int                      IsBoss(void)            ;
   int                      BossRank(void)          ;
   int                      ThisRank(void)          ;
@@ -156,7 +155,6 @@ class CartesianCommunicator {
   const std::vector<int> & ProcessorGrid(void)     ;
   int                      ProcessorCount(void)    ;
   int                      NodeCount(void)    ;
-  int                      RankCount(void)    ;
 
   ////////////////////////////////////////////////////////////////////////////////
   // very VERY rarely (Log, serial RNG) we need world without a grid
@@ -177,8 +175,6 @@ class CartesianCommunicator {
   void GlobalSumVector(ComplexF *c,int N);
   void GlobalSum(ComplexD &c);
   void GlobalSumVector(ComplexD *c,int N);
-  void GlobalXOR(uint32_t &);
-  void GlobalXOR(uint64_t &);
   
   template<class obj> void GlobalSum(obj &o){
     typedef typename obj::scalar_type scalar_type;

lib/communicator/Communicator_mpi.cc

@@ -83,14 +83,6 @@ void CartesianCommunicator::GlobalSum(uint64_t &u){
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_SUM,communicator);
   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);
-}
-void CartesianCommunicator::GlobalXOR(uint64_t &u){
-  int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_BXOR,communicator);
-  assert(ierr==0);
-}
 void CartesianCommunicator::GlobalSum(float &f){
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&f,1,MPI_FLOAT,MPI_SUM,communicator);
   assert(ierr==0);

lib/communicator/Communicator_mpi3.cc

@@ -65,7 +65,6 @@ std::vector<int> CartesianCommunicator::MyGroup;
 std::vector<void *> CartesianCommunicator::ShmCommBufs;
 
 int CartesianCommunicator::NodeCount(void)    { return GroupSize;};
-int CartesianCommunicator::RankCount(void)    { return WorldSize;};
 
 
 #undef FORCE_COMMS
@@ -510,14 +509,6 @@ void CartesianCommunicator::GlobalSum(uint64_t &u){
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_SUM,communicator);
   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);
-}
-void CartesianCommunicator::GlobalXOR(uint64_t &u){
-  int ierr=MPI_Allreduce(MPI_IN_PLACE,&u,1,MPI_UINT64_T,MPI_BXOR,communicator);
-  assert(ierr==0);
-}
 void CartesianCommunicator::GlobalSum(float &f){
   int ierr=MPI_Allreduce(MPI_IN_PLACE,&f,1,MPI_FLOAT,MPI_SUM,communicator);
   assert(ierr==0);

lib/communicator/Communicator_none.cc

@@ -59,8 +59,6 @@ void CartesianCommunicator::GlobalSum(double &){}
 void CartesianCommunicator::GlobalSum(uint32_t &){}
 void CartesianCommunicator::GlobalSum(uint64_t &){}
 void CartesianCommunicator::GlobalSumVector(double *,int N){}
-void CartesianCommunicator::GlobalXOR(uint32_t &){}
-void CartesianCommunicator::GlobalXOR(uint64_t &){}
 
 void CartesianCommunicator::SendRecvPacket(void *xmit,
 					   void *recv,

lib/cshift/Cshift_common.h

@@ -30,11 +30,21 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 namespace Grid {
 
+template<class vobj>
+class SimpleCompressor {
+public:
+  void Point(int) {};
+
+  vobj operator() (const vobj &arg) {
+    return arg;
+  }
+};
+
 ///////////////////////////////////////////////////////////////////
-// Gather for when there is no need to SIMD split 
+// Gather for when there is no need to SIMD split with compression
 ///////////////////////////////////////////////////////////////////
-template<class vobj> void 
-Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer,int dimension,int plane,int cbmask, int off=0)
+template<class vobj,class cobj,class compressor> void 
+Gather_plane_simple (const Lattice<vobj> &rhs,commVector<cobj> &buffer,int dimension,int plane,int cbmask,compressor &compress, int off=0)
 {
   int rd = rhs._grid->_rdimensions[dimension];
 
@@ -52,7 +62,7 @@ Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer,int dimen
       for(int b=0;b<e2;b++){
 	int o  = n*stride;
 	int bo = n*e2;
-	buffer[off+bo+b]=rhs._odata[so+o+b];
+	buffer[off+bo+b]=compress(rhs._odata[so+o+b]);
       }
     }
   } else { 
@@ -68,16 +78,17 @@ Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer,int dimen
        }
      }
      parallel_for(int i=0;i<table.size();i++){
-       buffer[off+table[i].first]=rhs._odata[so+table[i].second];
+       buffer[off+table[i].first]=compress(rhs._odata[so+table[i].second]);
      }
   }
 }
 
+
 ///////////////////////////////////////////////////////////////////
-// Gather for when there *is* need to SIMD split 
+// Gather for when there *is* need to SIMD split with compression
 ///////////////////////////////////////////////////////////////////
-template<class vobj> void 
-Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_object *> pointers,int dimension,int plane,int cbmask)
+template<class cobj,class vobj,class compressor> void 
+Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename cobj::scalar_object *> pointers,int dimension,int plane,int cbmask,compressor &compress)
 {
   int rd = rhs._grid->_rdimensions[dimension];
 
@@ -98,8 +109,8 @@ Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_
 	int o      =   n*n1;
 	int offset = b+n*e2;
 	
-	vobj temp =rhs._odata[so+o+b];
-	extract<vobj>(temp,pointers,offset);
+	cobj temp =compress(rhs._odata[so+o+b]);
+	extract<cobj>(temp,pointers,offset);
 
       }
     }
@@ -116,14 +127,32 @@ Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_
 	int offset = b+n*e2;
 
 	if ( ocb & cbmask ) {
-	  vobj temp =rhs._odata[so+o+b];
-	  extract<vobj>(temp,pointers,offset);
+	  cobj temp =compress(rhs._odata[so+o+b]);
+	  extract<cobj>(temp,pointers,offset);
 	}
       }
     }
   }
 }
 
+//////////////////////////////////////////////////////
+// Gather for when there is no need to SIMD split
+//////////////////////////////////////////////////////
+template<class vobj> void Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer, int dimension,int plane,int cbmask)
+{
+  SimpleCompressor<vobj> dontcompress;
+  Gather_plane_simple (rhs,buffer,dimension,plane,cbmask,dontcompress);
+}
+
+//////////////////////////////////////////////////////
+// Gather for when there *is* need to SIMD split
+//////////////////////////////////////////////////////
+template<class vobj> void Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_object *> pointers,int dimension,int plane,int cbmask)
+{
+  SimpleCompressor<vobj> dontcompress;
+  Gather_plane_extract<vobj,vobj,decltype(dontcompress)>(rhs,pointers,dimension,plane,cbmask,dontcompress);
+}
+
 //////////////////////////////////////////////////////
 // Scatter for when there is no need to SIMD split
 //////////////////////////////////////////////////////
@@ -171,7 +200,7 @@ template<class vobj> void Scatter_plane_simple (Lattice<vobj> &rhs,commVector<vo
 //////////////////////////////////////////////////////
 // Scatter for when there *is* need to SIMD split
 //////////////////////////////////////////////////////
-template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,std::vector<typename vobj::scalar_object *> pointers,int dimension,int plane,int cbmask)
+ template<class vobj,class cobj> void Scatter_plane_merge(Lattice<vobj> &rhs,std::vector<cobj *> pointers,int dimension,int plane,int cbmask)
 {
   int rd = rhs._grid->_rdimensions[dimension];
 

lib/cshift/Cshift_mpi.h

@@ -154,7 +154,13 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
 			   recv_from_rank,
 			   bytes);
       grid->Barrier();
-
+      /*
+      for(int i=0;i<send_buf.size();i++){
+	assert(recv_buf.size()==buffer_size);
+	assert(send_buf.size()==buffer_size);
+	std::cout << "SendRecv_Cshift_comms ["<<i<<" "<< dimension<<"] snd "<<send_buf[i]<<" rcv " << recv_buf[i] << "  0x" << cbmask<<std::endl;
+      }
+      */
       Scatter_plane_simple (ret,recv_buf,dimension,x,cbmask);
     }
   }
@@ -240,6 +246,13 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
 			     (void *)&recv_buf_extract[i][0],
 			     recv_from_rank,
 			     bytes);
+	/*
+	for(int w=0;w<recv_buf_extract[i].size();w++){
+	  assert(recv_buf_extract[i].size()==buffer_size);
+	  assert(send_buf_extract[i].size()==buffer_size);
+	  std::cout << "SendRecv_Cshift_comms ["<<w<<" "<< dimension<<"] recv "<<recv_buf_extract[i][w]<<" send " << send_buf_extract[nbr_lane][w]  << cbmask<<std::endl;
+	}
+	*/	
 	grid->Barrier();
 	rpointers[i] = &recv_buf_extract[i][0];
       } else { 

lib/lattice/Lattice_base.h

@@ -235,74 +235,64 @@ public:
     }
   };
 
-  //////////////////////////////////////////////////////////////////
-  // Constructor requires "grid" passed.
-  // what about a default grid?
-  //////////////////////////////////////////////////////////////////
-  Lattice(GridBase *grid) : _odata(grid->oSites()) {
-    _grid = grid;
+    //////////////////////////////////////////////////////////////////
+    // Constructor requires "grid" passed.
+    // what about a default grid?
+    //////////////////////////////////////////////////////////////////
+    Lattice(GridBase *grid) : _odata(grid->oSites()) {
+        _grid = grid;
     //        _odata.reserve(_grid->oSites());
     //        _odata.resize(_grid->oSites());
     //      std::cout << "Constructing lattice object with Grid pointer "<<_grid<<std::endl;
-    assert((((uint64_t)&_odata[0])&0xF) ==0);
-    checkerboard=0;
-  }
-  
-  Lattice(const Lattice& r){ // copy constructor
-    _grid = r._grid;
-    checkerboard = r.checkerboard;
-    _odata.resize(_grid->oSites());// essential
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      _odata[ss]=r._odata[ss];
-    }  	
-  }
-  
-  
-  
-  virtual ~Lattice(void) = default;
-    
-  void reset(GridBase* grid) {
-    if (_grid != grid) {
-      _grid = grid;
-      _odata.resize(grid->oSites());
-      checkerboard = 0;
+        assert((((uint64_t)&_odata[0])&0xF) ==0);
+        checkerboard=0;
     }
-  }
-  
 
-  template<class sobj> strong_inline Lattice<vobj> & operator = (const sobj & r){
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      this->_odata[ss]=r;
+    Lattice(const Lattice& r){ // copy constructor
+    	_grid = r._grid;
+    	checkerboard = r.checkerboard;
+    	_odata.resize(_grid->oSites());// essential
+	parallel_for(int ss=0;ss<_grid->oSites();ss++){
+            _odata[ss]=r._odata[ss];
+        }  	
     }
-    return *this;
-  }
-  
-  template<class robj> strong_inline Lattice<vobj> & operator = (const Lattice<robj> & r){
-    this->checkerboard = r.checkerboard;
-    conformable(*this,r);
+
+
+
+    virtual ~Lattice(void) = default;
     
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      this->_odata[ss]=r._odata[ss];
+    template<class sobj> strong_inline Lattice<vobj> & operator = (const sobj & r){
+      parallel_for(int ss=0;ss<_grid->oSites();ss++){
+            this->_odata[ss]=r;
+        }
+        return *this;
     }
-    return *this;
-  }
-  
-  // *=,+=,-= operators inherit behvour from correspond */+/- operation
-  template<class T> strong_inline Lattice<vobj> &operator *=(const T &r) {
-    *this = (*this)*r;
-    return *this;
-  }
-  
-  template<class T> strong_inline Lattice<vobj> &operator -=(const T &r) {
-    *this = (*this)-r;
-    return *this;
-  }
-  template<class T> strong_inline Lattice<vobj> &operator +=(const T &r) {
-    *this = (*this)+r;
-    return *this;
-  }
-}; // class Lattice
-  
+    template<class robj> strong_inline Lattice<vobj> & operator = (const Lattice<robj> & r){
+      this->checkerboard = r.checkerboard;
+      conformable(*this,r);
+      
+      parallel_for(int ss=0;ss<_grid->oSites();ss++){
+            this->_odata[ss]=r._odata[ss];
+        }
+        return *this;
+    }
+
+    // *=,+=,-= operators inherit behvour from correspond */+/- operation
+    template<class T> strong_inline Lattice<vobj> &operator *=(const T &r) {
+        *this = (*this)*r;
+        return *this;
+    }
+
+    template<class T> strong_inline Lattice<vobj> &operator -=(const T &r) {
+        *this = (*this)-r;
+        return *this;
+    }
+    template<class T> strong_inline Lattice<vobj> &operator +=(const T &r) {
+        *this = (*this)+r;
+        return *this;
+    }
+ }; // class Lattice
+
   template<class vobj> std::ostream& operator<< (std::ostream& stream, const Lattice<vobj> &o){
     std::vector<int> gcoor;
     typedef typename vobj::scalar_object sobj;
@@ -320,7 +310,7 @@ public:
     }
     return stream;
   }
-  
+
 }
 
 

lib/lattice/Lattice_reduction.h

@@ -1,154 +1,157 @@
-/*************************************************************************************
+    /*************************************************************************************
+
     Grid physics library, www.github.com/paboyle/Grid 
+
     Source file: ./lib/lattice/Lattice_reduction.h
+
     Copyright (C) 2015
+
 Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
+
     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 */
 #ifndef GRID_LATTICE_REDUCTION_H
 #define GRID_LATTICE_REDUCTION_H
 
-#include <Grid/Grid_Eigen_Dense.h>
-
 namespace Grid {
 #ifdef GRID_WARN_SUBOPTIMAL
 #warning "Optimisation alert all these reduction loops are NOT threaded "
 #endif     
 
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Deterministic Reduction operations
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-template<class vobj> inline RealD norm2(const Lattice<vobj> &arg){
-  ComplexD nrm = innerProduct(arg,arg);
-  return std::real(nrm); 
-}
-
-// Double inner product
-template<class vobj>
-inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &right) 
-{
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_typeD vector_type;
-  scalar_type  nrm;
-  
-  GridBase *grid = left._grid;
-  
-  std::vector<vector_type,alignedAllocator<vector_type> > sumarray(grid->SumArraySize());
-  
-  parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
-    int nwork, mywork, myoff;
-    GridThread::GetWork(left._grid->oSites(),thr,mywork,myoff);
-    
-    decltype(innerProductD(left._odata[0],right._odata[0])) vnrm=zero; // private to thread; sub summation
-    for(int ss=myoff;ss<mywork+myoff; ss++){
-      vnrm = vnrm + innerProductD(left._odata[ss],right._odata[ss]);
-    }
-    sumarray[thr]=TensorRemove(vnrm) ;
+    ////////////////////////////////////////////////////////////////////////////////////////////////////
+    // Deterministic Reduction operations
+    ////////////////////////////////////////////////////////////////////////////////////////////////////
+  template<class vobj> inline RealD norm2(const Lattice<vobj> &arg){
+    ComplexD nrm = innerProduct(arg,arg);
+    return std::real(nrm); 
   }
-  
-  vector_type vvnrm; vvnrm=zero;  // sum across threads
-  for(int i=0;i<grid->SumArraySize();i++){
-    vvnrm = vvnrm+sumarray[i];
-  } 
-  nrm = Reduce(vvnrm);// sum across simd
-  right._grid->GlobalSum(nrm);
-  return nrm;
-}
- 
-template<class Op,class T1>
-inline auto sum(const LatticeUnaryExpression<Op,T1> & expr)
-  ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second))))::scalar_object
-{
-  return sum(closure(expr));
-}
 
-template<class Op,class T1,class T2>
-inline auto sum(const LatticeBinaryExpression<Op,T1,T2> & expr)
+    template<class vobj>
+    inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &right) 
+    {
+      typedef typename vobj::scalar_type scalar_type;
+      typedef typename vobj::vector_type vector_type;
+      scalar_type  nrm;
+
+      GridBase *grid = left._grid;
+
+      std::vector<vector_type,alignedAllocator<vector_type> > sumarray(grid->SumArraySize());
+      for(int i=0;i<grid->SumArraySize();i++){
+	sumarray[i]=zero;
+      }
+
+      parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
+	int nwork, mywork, myoff;
+	GridThread::GetWork(left._grid->oSites(),thr,mywork,myoff);
+	
+	decltype(innerProduct(left._odata[0],right._odata[0])) vnrm=zero; // private to thread; sub summation
+        for(int ss=myoff;ss<mywork+myoff; ss++){
+	  vnrm = vnrm + innerProduct(left._odata[ss],right._odata[ss]);
+	}
+	sumarray[thr]=TensorRemove(vnrm) ;
+      }
+      
+      vector_type vvnrm; vvnrm=zero;  // sum across threads
+      for(int i=0;i<grid->SumArraySize();i++){
+	vvnrm = vvnrm+sumarray[i];
+      } 
+      nrm = Reduce(vvnrm);// sum across simd
+      right._grid->GlobalSum(nrm);
+      return nrm;
+    }
+
+    template<class Op,class T1>
+      inline auto sum(const LatticeUnaryExpression<Op,T1> & expr)
+      ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second))))::scalar_object
+    {
+      return sum(closure(expr));
+    }
+
+    template<class Op,class T1,class T2>
+      inline auto sum(const LatticeBinaryExpression<Op,T1,T2> & expr)
       ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second)),eval(0,std::get<1>(expr.second))))::scalar_object
-{
-  return sum(closure(expr));
-}
-
-
-template<class Op,class T1,class T2,class T3>
-inline auto sum(const LatticeTrinaryExpression<Op,T1,T2,T3> & expr)
-  ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second)),
-				      eval(0,std::get<1>(expr.second)),
-				      eval(0,std::get<2>(expr.second))
-				      ))::scalar_object
-{
-  return sum(closure(expr));
-}
-
-template<class vobj>
-inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
-{
-  GridBase *grid=arg._grid;
-  int Nsimd = grid->Nsimd();
-  
-  std::vector<vobj,alignedAllocator<vobj> > sumarray(grid->SumArraySize());
-  for(int i=0;i<grid->SumArraySize();i++){
-    sumarray[i]=zero;
-  }
-  
-  parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
-    int nwork, mywork, myoff;
-    GridThread::GetWork(grid->oSites(),thr,mywork,myoff);
-    
-    vobj vvsum=zero;
-    for(int ss=myoff;ss<mywork+myoff; ss++){
-      vvsum = vvsum + arg._odata[ss];
+    {
+      return sum(closure(expr));
+    }
+
+
+    template<class Op,class T1,class T2,class T3>
+      inline auto sum(const LatticeTrinaryExpression<Op,T1,T2,T3> & expr)
+      ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second)),
+				 eval(0,std::get<1>(expr.second)),
+				 eval(0,std::get<2>(expr.second))
+				 ))::scalar_object
+    {
+      return sum(closure(expr));
+    }
+
+    template<class vobj>
+    inline typename vobj::scalar_object sum(const Lattice<vobj> &arg){
+
+      GridBase *grid=arg._grid;
+      int Nsimd = grid->Nsimd();
+
+      std::vector<vobj,alignedAllocator<vobj> > sumarray(grid->SumArraySize());
+      for(int i=0;i<grid->SumArraySize();i++){
+	sumarray[i]=zero;
+      }
+
+      parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
+	int nwork, mywork, myoff;
+	GridThread::GetWork(grid->oSites(),thr,mywork,myoff);
+	
+	vobj vvsum=zero;
+        for(int ss=myoff;ss<mywork+myoff; ss++){
+	  vvsum = vvsum + arg._odata[ss];
+	}
+	sumarray[thr]=vvsum;
+      }
+      
+      vobj vsum=zero;  // sum across threads
+      for(int i=0;i<grid->SumArraySize();i++){
+	vsum = vsum+sumarray[i];
+      } 
+
+      typedef typename vobj::scalar_object sobj;
+      sobj ssum=zero;
+
+      std::vector<sobj>               buf(Nsimd);
+      extract(vsum,buf);
+
+      for(int i=0;i<Nsimd;i++) ssum = ssum + buf[i];
+      arg._grid->GlobalSum(ssum);
+
+      return ssum;
     }
-    sumarray[thr]=vvsum;
-  }
-  
-  vobj vsum=zero;  // sum across threads
-  for(int i=0;i<grid->SumArraySize();i++){
-    vsum = vsum+sumarray[i];
-  } 
-  
-  typedef typename vobj::scalar_object sobj;
-  sobj ssum=zero;
-  
-  std::vector<sobj>               buf(Nsimd);
-  extract(vsum,buf);
-  
-  for(int i=0;i<Nsimd;i++) ssum = ssum + buf[i];
-  arg._grid->GlobalSum(ssum);
-  
-  return ssum;
-}
 
 
-//////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// sliceSum, sliceInnerProduct, sliceAxpy, sliceNorm etc...
-//////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
 template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<typename vobj::scalar_object> &result,int orthogdim)
 {
-  ///////////////////////////////////////////////////////
-  // FIXME precision promoted summation
-  // may be important for correlation functions
-  // But easily avoided by using double precision fields
-  ///////////////////////////////////////////////////////
   typedef typename vobj::scalar_object sobj;
   GridBase  *grid = Data._grid;
   assert(grid!=NULL);
 
+  // FIXME
+  // std::cout<<GridLogMessage<<"WARNING ! SliceSum is unthreaded "<<grid->SumArraySize()<<" threads "<<std::endl;
+
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
 
@@ -160,31 +163,23 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
   int rd=grid->_rdimensions[orthogdim];
 
   std::vector<vobj,alignedAllocator<vobj> > lvSum(rd); // will locally sum vectors first
-  std::vector<sobj> lsSum(ld,zero);                    // sum across these down to scalars
-  std::vector<sobj> extracted(Nsimd);                  // splitting the SIMD
+  std::vector<sobj> lsSum(ld,zero); // sum across these down to scalars
+  std::vector<sobj> extracted(Nsimd);     // splitting the SIMD
 
-  result.resize(fd); // And then global sum to return the same vector to every node 
+  result.resize(fd); // And then global sum to return the same vector to every node for IO to file
   for(int r=0;r<rd;r++){
     lvSum[r]=zero;
   }
 
-  int e1=    grid->_slice_nblock[orthogdim];
-  int e2=    grid->_slice_block [orthogdim];
-  int stride=grid->_slice_stride[orthogdim];
+  std::vector<int>  coor(Nd);  
 
   // sum over reduced dimension planes, breaking out orthog dir
-  // Parallel over orthog direction
-  parallel_for(int r=0;r<rd;r++){
 
-    int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
-
-    for(int n=0;n<e1;n++){
-      for(int b=0;b<e2;b++){
-	int ss= so+n*stride+b;
-	lvSum[r]=lvSum[r]+Data._odata[ss];
-      }
-    }
-  }
+  for(int ss=0;ss<grid->oSites();ss++){
+    Lexicographic::CoorFromIndex(coor,ss,grid->_rdimensions);
+    int r = coor[orthogdim];
+    lvSum[r]=lvSum[r]+Data._odata[ss];
+  }  
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
   std::vector<int> icoor(Nd);
@@ -219,341 +214,10 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
 
     result[t]=gsum;
   }
+
 }
 
-template<class vobj>
-static void sliceInnerProductVector( std::vector<ComplexD> & result, const Lattice<vobj> &lhs,const Lattice<vobj> &rhs,int orthogdim) 
-{
-  typedef typename vobj::vector_type   vector_type;
-  typedef typename vobj::scalar_type   scalar_type;
-  GridBase  *grid = lhs._grid;
-  assert(grid!=NULL);
-  conformable(grid,rhs._grid);
 
-  const int    Nd = grid->_ndimension;
-  const int Nsimd = grid->Nsimd();
-
-  assert(orthogdim >= 0);
-  assert(orthogdim < Nd);
-
-  int fd=grid->_fdimensions[orthogdim];
-  int ld=grid->_ldimensions[orthogdim];
-  int rd=grid->_rdimensions[orthogdim];
-
-  std::vector<vector_type,alignedAllocator<vector_type> > lvSum(rd); // will locally sum vectors first
-  std::vector<scalar_type > lsSum(ld,scalar_type(0.0));                    // sum across these down to scalars
-  std::vector<iScalar<scalar_type> > extracted(Nsimd);                  // splitting the SIMD
-
-  result.resize(fd); // And then global sum to return the same vector to every node for IO to file
-  for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
-  }
-
-  int e1=    grid->_slice_nblock[orthogdim];
-  int e2=    grid->_slice_block [orthogdim];
-  int stride=grid->_slice_stride[orthogdim];
-
-  parallel_for(int r=0;r<rd;r++){
-
-    int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
-
-    for(int n=0;n<e1;n++){
-      for(int b=0;b<e2;b++){
-	int ss= so+n*stride+b;
-	vector_type vv = TensorRemove(innerProduct(lhs._odata[ss],rhs._odata[ss]));
-	lvSum[r]=lvSum[r]+vv;
-      }
-    }
-  }
-
-  // Sum across simd lanes in the plane, breaking out orthog dir.
-  std::vector<int> icoor(Nd);
-  for(int rt=0;rt<rd;rt++){
-
-    iScalar<vector_type> temp; 
-    temp._internal = lvSum[rt];
-    extract(temp,extracted);
-
-    for(int idx=0;idx<Nsimd;idx++){
-
-      grid->iCoorFromIindex(icoor,idx);
-
-      int ldx =rt+icoor[orthogdim]*rd;
-
-      lsSum[ldx]=lsSum[ldx]+extracted[idx]._internal;
-
-    }
-  }
-  
-  // sum over nodes.
-  scalar_type gsum;
-  for(int t=0;t<fd;t++){
-    int pt = t/ld; // processor plane
-    int lt = t%ld;
-    if ( pt == grid->_processor_coor[orthogdim] ) {
-      gsum=lsSum[lt];
-    } else {
-      gsum=scalar_type(0.0);
-    }
-
-    grid->GlobalSum(gsum);
-
-    result[t]=gsum;
-  }
 }
-template<class vobj>
-static void sliceNorm (std::vector<RealD> &sn,const Lattice<vobj> &rhs,int Orthog) 
-{
-  typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-  
-  int Nblock = rhs._grid->GlobalDimensions()[Orthog];
-  std::vector<ComplexD> ip(Nblock);
-  sn.resize(Nblock);
-  
-  sliceInnerProductVector(ip,rhs,rhs,Orthog);
-  for(int ss=0;ss<Nblock;ss++){
-    sn[ss] = real(ip[ss]);
-  }
-};
-
-
-template<class vobj>
-static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice<vobj> &X,const Lattice<vobj> &Y,
-			    int orthogdim,RealD scale=1.0) 
-{    
-  typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-  typedef typename vobj::tensor_reduced tensor_reduced;
-  
-  scalar_type zscale(scale);
-
-  GridBase *grid  = X._grid;
-
-  int Nsimd  =grid->Nsimd();
-  int Nblock =grid->GlobalDimensions()[orthogdim];
-
-  int fd     =grid->_fdimensions[orthogdim];
-  int ld     =grid->_ldimensions[orthogdim];
-  int rd     =grid->_rdimensions[orthogdim];
-
-  int e1     =grid->_slice_nblock[orthogdim];
-  int e2     =grid->_slice_block [orthogdim];
-  int stride =grid->_slice_stride[orthogdim];
-
-  std::vector<int> icoor;
-
-  for(int r=0;r<rd;r++){
-
-    int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
-
-    vector_type    av;
-
-    for(int l=0;l<Nsimd;l++){
-      grid->iCoorFromIindex(icoor,l);
-      int ldx =r+icoor[orthogdim]*rd;
-      scalar_type *as =(scalar_type *)&av;
-      as[l] = scalar_type(a[ldx])*zscale;
-    }
-
-    tensor_reduced at; at=av;
-
-    parallel_for_nest2(int n=0;n<e1;n++){
-      for(int b=0;b<e2;b++){
-	int ss= so+n*stride+b;
-	R._odata[ss] = at*X._odata[ss]+Y._odata[ss];
-      }
-    }
-  }
-};
-
-inline GridBase         *makeSubSliceGrid(const GridBase *BlockSolverGrid,int Orthog)
-{
-  int NN    = BlockSolverGrid->_ndimension;
-  int nsimd = BlockSolverGrid->Nsimd();
-  
-  std::vector<int> latt_phys(0);
-  std::vector<int> simd_phys(0);
-  std::vector<int>  mpi_phys(0);
-  
-  for(int d=0;d<NN;d++){
-    if( d!=Orthog ) { 
-      latt_phys.push_back(BlockSolverGrid->_fdimensions[d]);
-      simd_phys.push_back(BlockSolverGrid->_simd_layout[d]);
-      mpi_phys.push_back(BlockSolverGrid->_processors[d]);
-    }
-  }
-  return (GridBase *)new GridCartesian(latt_phys,simd_phys,mpi_phys); 
-}
-
-template<class vobj>
-static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,const Lattice<vobj> &Y,int Orthog,RealD scale=1.0) 
-{    
-  typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-
-  int Nblock = X._grid->GlobalDimensions()[Orthog];
-
-  GridBase *FullGrid  = X._grid;
-  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
-
-  Lattice<vobj> Xslice(SliceGrid);
-  Lattice<vobj> Rslice(SliceGrid);
-
-  assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
-  int nl = SliceGrid->_ndimension;
-
-  //FIXME package in a convenient iterator
-  //Should loop over a plane orthogonal to direction "Orthog"
-  int stride=FullGrid->_slice_stride[Orthog];
-  int block =FullGrid->_slice_block [Orthog];
-  int nblock=FullGrid->_slice_nblock[Orthog];
-  int ostride=FullGrid->_ostride[Orthog];
-#pragma omp parallel 
-  {
-    std::vector<vobj> s_x(Nblock);
-
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
-    for(int b=0;b<block;b++){
-      int o  = n*stride + b;
-
-      for(int i=0;i<Nblock;i++){
-	s_x[i] = X[o+i*ostride];
-      }
-
-      vobj dot;
-      for(int i=0;i<Nblock;i++){
-	dot = Y[o+i*ostride];
-	for(int j=0;j<Nblock;j++){
-	  dot = dot + s_x[j]*(scale*aa(j,i));
-	}
-	R[o+i*ostride]=dot;
-      }
-    }}
-  }
-};
-
-template<class vobj>
-static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,int Orthog,RealD scale=1.0) 
-{    
-  typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-
-  int Nblock = X._grid->GlobalDimensions()[Orthog];
-
-  GridBase *FullGrid  = X._grid;
-  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
-
-  Lattice<vobj> Xslice(SliceGrid);
-  Lattice<vobj> Rslice(SliceGrid);
-
-  assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
-  int nl = SliceGrid->_ndimension;
-
-  //FIXME package in a convenient iterator
-  //Should loop over a plane orthogonal to direction "Orthog"
-  int stride=FullGrid->_slice_stride[Orthog];
-  int block =FullGrid->_slice_block [Orthog];
-  int nblock=FullGrid->_slice_nblock[Orthog];
-  int ostride=FullGrid->_ostride[Orthog];
-#pragma omp parallel 
-  {
-    std::vector<vobj> s_x(Nblock);
-
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
-    for(int b=0;b<block;b++){
-      int o  = n*stride + b;
-
-      for(int i=0;i<Nblock;i++){
-	s_x[i] = X[o+i*ostride];
-      }
-
-      vobj dot;
-      for(int i=0;i<Nblock;i++){
-	dot = s_x[0]*(scale*aa(0,i));
-	for(int j=1;j<Nblock;j++){
-	  dot = dot + s_x[j]*(scale*aa(j,i));
-	}
-	R[o+i*ostride]=dot;
-      }
-    }}
-  }
-
-};
-
-
-template<class vobj>
-static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj> &lhs,const Lattice<vobj> &rhs,int Orthog) 
-{
-  typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-  
-  GridBase *FullGrid  = lhs._grid;
-  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
-  
-  int Nblock = FullGrid->GlobalDimensions()[Orthog];
-  
-  Lattice<vobj> Lslice(SliceGrid);
-  Lattice<vobj> Rslice(SliceGrid);
-  
-  mat = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-  assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
-  int nl = SliceGrid->_ndimension;
-
-  //FIXME package in a convenient iterator
-  //Should loop over a plane orthogonal to direction "Orthog"
-  int stride=FullGrid->_slice_stride[Orthog];
-  int block =FullGrid->_slice_block [Orthog];
-  int nblock=FullGrid->_slice_nblock[Orthog];
-  int ostride=FullGrid->_ostride[Orthog];
-
-  typedef typename vobj::vector_typeD vector_typeD;
-
-#pragma omp parallel 
-  {
-    std::vector<vobj> Left(Nblock);
-    std::vector<vobj> Right(Nblock);
-    Eigen::MatrixXcd  mat_thread = Eigen::MatrixXcd::Zero(Nblock,Nblock);
-
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
-    for(int b=0;b<block;b++){
-
-      int o  = n*stride + b;
-
-      for(int i=0;i<Nblock;i++){
-	Left [i] = lhs[o+i*ostride];
-	Right[i] = rhs[o+i*ostride];
-      }
-
-      for(int i=0;i<Nblock;i++){
-      for(int j=0;j<Nblock;j++){
-	auto tmp = innerProduct(Left[i],Right[j]);
-	vector_typeD rtmp = TensorRemove(tmp);
-	mat_thread(i,j) += Reduce(rtmp);
-      }}
-    }}
-#pragma omp critical
-    {
-      mat += mat_thread;
-    }  
-  }
-  return;
-}
-
-} /*END NAMESPACE GRID*/
 #endif
 
-
-

lib/lattice/Lattice_rng.h

@@ -6,8 +6,8 @@
 
     Copyright (C) 2015
 
-    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
 
     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
@@ -75,55 +75,6 @@ namespace Grid {
     return multiplicity;
   }
 
-  
-// merge of April 11 2017
-//<<<<<<< HEAD
-
-
-  // this function is necessary for the LS vectorised field
-  inline int RNGfillable_general(GridBase *coarse,GridBase *fine)
-  {
-    int rngdims = coarse->_ndimension;
-    
-    // trivially extended in higher dims, with locality guaranteeing RNG state is local to node
-    int lowerdims   = fine->_ndimension - coarse->_ndimension;  assert(lowerdims >= 0);
-    // assumes that the higher dimensions are not using more processors
-    // all further divisions are local
-    for(int d=0;d<lowerdims;d++) assert(fine->_processors[d]==1);
-    for(int d=0;d<rngdims;d++) assert(coarse->_processors[d] == fine->_processors[d+lowerdims]);
-    
-
-    // then divide the number of local sites
-    // check that the total number of sims agree, meanse the iSites are the same
-    assert(fine->Nsimd() == coarse->Nsimd());
-
-    // check that the two grids divide cleanly
-    assert( (fine->lSites() / coarse->lSites() ) * coarse->lSites() == fine->lSites() );
-
-    return fine->lSites() / coarse->lSites();
-  }
-
-  /*
-  // Wrap seed_seq to give common interface with random_device
-  class fixedSeed {
-  public:
-    typedef std::seed_seq::result_type result_type;
-    std::seed_seq src;
-    
-    fixedSeed(const std::vector<int> &seeds) : src(seeds.begin(),seeds.end()) {};
-
-    result_type operator () (void){
-      std::vector<result_type> list(1);
-      src.generate(list.begin(),list.end());
-      return list[0];
-    }
-
-  };
-
-=======
->>>>>>> develop
-  */
-  
   // real scalars are one component
   template<class scalar,class distribution,class generator> 
   void fillScalar(scalar &s,distribution &dist,generator & gen)
@@ -158,7 +109,7 @@ namespace Grid {
 #ifdef RNG_SITMO
     typedef sitmo::prng_engine 	RngEngine;
     typedef uint64_t    	RngStateType;
-    static const int    	RngStateCount = 13;
+    static const int    	RngStateCount = 4;
 #endif
 
     std::vector<RngEngine>                             _generators;
@@ -213,7 +164,7 @@ namespace Grid {
       ss<<eng;
       ss.seekg(0,ss.beg);
       for(int i=0;i<RngStateCount;i++){
-        ss>>saved[i];
+	ss>>saved[i];
       }
     }
     void GetState(std::vector<RngStateType> & saved,int gen) {
@@ -223,7 +174,7 @@ namespace Grid {
       assert(saved.size()==RngStateCount);
       std::stringstream ss;
       for(int i=0;i<RngStateCount;i++){
-        ss<< saved[i]<<" ";
+	ss<< saved[i]<<" ";
       }
       ss.seekg(0,ss.beg);
       ss>>eng;
@@ -264,7 +215,7 @@ namespace Grid {
 
       dist[0].reset();
       for(int idx=0;idx<words;idx++){
-  fillScalar(buf[idx],dist[0],_generators[0]);
+	fillScalar(buf[idx],dist[0],_generators[0]);
       }
 
       CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
@@ -296,7 +247,7 @@ namespace Grid {
       RealF *pointer=(RealF *)&l;
       dist[0].reset();
       for(int i=0;i<2*vComplexF::Nsimd();i++){
-  fillScalar(pointer[i],dist[0],_generators[0]);
+	fillScalar(pointer[i],dist[0],_generators[0]);
       }
       CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
     }
@@ -304,7 +255,7 @@ namespace Grid {
       RealD *pointer=(RealD *)&l;
       dist[0].reset();
       for(int i=0;i<2*vComplexD::Nsimd();i++){
-  fillScalar(pointer[i],dist[0],_generators[0]);
+	fillScalar(pointer[i],dist[0],_generators[0]);
       }
       CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
     }
@@ -312,7 +263,7 @@ namespace Grid {
       RealF *pointer=(RealF *)&l;
       dist[0].reset();
       for(int i=0;i<vRealF::Nsimd();i++){
-  fillScalar(pointer[i],dist[0],_generators[0]);
+	fillScalar(pointer[i],dist[0],_generators[0]);
       }
       CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
     }
@@ -324,7 +275,7 @@ namespace Grid {
       }
       CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
     }
-    
+
     void SeedFixedIntegers(const std::vector<int> &seeds){
       CartesianCommunicator::BroadcastWorld(0,(void *)&seeds[0],sizeof(int)*seeds.size());
       std::seed_seq src(seeds.begin(),seeds.end());
@@ -333,20 +284,18 @@ namespace Grid {
   };
 
   class GridParallelRNG : public GridRNGbase {
-
-    double _time_counter;
-
   public:
     GridBase *_grid;
-    unsigned int _vol;
+    int _vol;
+  public:
 
-    int generator_idx(int os,int is) {
+    int generator_idx(int os,int is){
       return is*_grid->oSites()+os;
     }
 
     GridParallelRNG(GridBase *grid) : GridRNGbase() {
-      _grid = grid;
-      _vol  =_grid->iSites()*_grid->oSites();
+      _grid=grid;
+      _vol =_grid->iSites()*_grid->oSites();
 
       _generators.resize(_vol);
       _uniform.resize(_vol,std::uniform_real_distribution<RealD>{0,1});
@@ -360,34 +309,33 @@ namespace Grid {
       typedef typename vobj::scalar_object scalar_object;
       typedef typename vobj::scalar_type scalar_type;
       typedef typename vobj::vector_type vector_type;
+      
+      int multiplicity = RNGfillable(_grid,l._grid);
 
-      double inner_time_counter = usecond();
-
-      int multiplicity = RNGfillable_general(_grid, l._grid); // l has finer or same grid
-      int Nsimd  = _grid->Nsimd();  // guaranteed to be the same for l._grid too
-      int osites = _grid->oSites();  // guaranteed to be <= l._grid->oSites() by a factor multiplicity
-      int words  = sizeof(scalar_object) / sizeof(scalar_type);
+      int     Nsimd =_grid->Nsimd();
+      int     osites=_grid->oSites();
+      int words=sizeof(scalar_object)/sizeof(scalar_type);
 
       parallel_for(int ss=0;ss<osites;ss++){
-        std::vector<scalar_object> buf(Nsimd);
-        for (int m = 0; m < multiplicity; m++) {  // Draw from same generator multiplicity times
 
-          int sm = multiplicity * ss + m;  // Maps the generator site to the fine site
+	std::vector<scalar_object> buf(Nsimd);
+	for(int m=0;m<multiplicity;m++) {// Draw from same generator multiplicity times
 
-          for (int si = 0; si < Nsimd; si++) {
-            
-            int gdx = generator_idx(ss, si);  // index of generator state
-            scalar_type *pointer = (scalar_type *)&buf[si];
-            dist[gdx].reset();
-            for (int idx = 0; idx < words; idx++) 
-              fillScalar(pointer[idx], dist[gdx], _generators[gdx]);
-          }
-          // merge into SIMD lanes, FIXME suboptimal implementation
-          merge(l._odata[sm], buf);
-        }
+	  int sm=multiplicity*ss+m;      // Maps the generator site to the fine site
+
+	  for(int si=0;si<Nsimd;si++){
+	    int gdx = generator_idx(ss,si); // index of generator state
+	    scalar_type *pointer = (scalar_type *)&buf[si];
+	    dist[gdx].reset();
+	    for(int idx=0;idx<words;idx++){
+	      fillScalar(pointer[idx],dist[gdx],_generators[gdx]);
+	    }
+	  }
+
+	  // merge into SIMD lanes
+	  merge(l._odata[sm],buf);
+	}
       }
-
-      _time_counter += usecond()- inner_time_counter;
     };
 
     void SeedFixedIntegers(const std::vector<int> &seeds){
@@ -464,12 +412,6 @@ namespace Grid {
       }
 #endif
     }
-
-    void Report(){
-      std::cout << GridLogMessage << "Time spent in the fill() routine by GridParallelRNG: "<< _time_counter/1e3 << " ms" << std::endl;
-    }
-
-
     ////////////////////////////////////////////////////////////////////////
     // Support for rigorous test of RNG's
     // Return uniform random uint32_t from requested site generator
@@ -477,6 +419,7 @@ namespace Grid {
     uint32_t GlobalU01(int gsite){
 
       uint32_t the_number;
+
       // who
       std::vector<int> gcoor;
       int rank,o_idx,i_idx;

lib/lattice/Lattice_transfer.h

@@ -1,4 +1,4 @@
-/*************************************************************************************
+    /*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -359,7 +359,7 @@ void localConvert(const Lattice<vobj> &in,Lattice<vvobj> &out)
 
 
 template<class vobj>
-void InsertSlice(const Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice, int orthog)
+void InsertSlice(Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice, int orthog)
 {
   typedef typename vobj::scalar_object sobj;
 
@@ -401,7 +401,7 @@ void InsertSlice(const Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice
 }
 
 template<class vobj>
-void ExtractSlice(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int slice, int orthog)
+void ExtractSlice(Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int slice, int orthog)
 {
   typedef typename vobj::scalar_object sobj;
 
@@ -444,7 +444,7 @@ void ExtractSlice(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int slic
 
 
 template<class vobj>
-void InsertSliceLocal(const Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int slice_lo,int slice_hi, int orthog)
+void InsertSliceLocal(Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int slice_lo,int slice_hi, int orthog)
 {
   typedef typename vobj::scalar_object sobj;
 
@@ -551,10 +551,7 @@ void Replicate(Lattice<vobj> &coarse,Lattice<vobj> & fine)
 
 //Copy SIMD-vectorized lattice to array of scalar objects in lexicographic order
 template<typename vobj, typename sobj>
-typename std::enable_if<isSIMDvectorized<vobj>::value && !isSIMDvectorized<sobj>::value, void>::type 
-unvectorizeToLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in)
-{
-
+typename std::enable_if<isSIMDvectorized<vobj>::value && !isSIMDvectorized<sobj>::value, void>::type unvectorizeToLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in){
   typedef typename vobj::vector_type vtype;
   
   GridBase* in_grid = in._grid;
@@ -593,54 +590,6 @@ unvectorizeToLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in)
     extract1(in_vobj, out_ptrs, 0);
   }
 }
-//Copy SIMD-vectorized lattice to array of scalar objects in lexicographic order
-template<typename vobj, typename sobj>
-typename std::enable_if<isSIMDvectorized<vobj>::value 
-                    && !isSIMDvectorized<sobj>::value, void>::type 
-vectorizeFromLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
-{
-
-  typedef typename vobj::vector_type vtype;
-  
-  GridBase* grid = out._grid;
-  assert(in.size()==grid->lSites());
-  
-  int ndim     = grid->Nd();
-  int nsimd    = vtype::Nsimd();
-
-  std::vector<std::vector<int> > icoor(nsimd);
-      
-  for(int lane=0; lane < nsimd; lane++){
-    icoor[lane].resize(ndim);
-    grid->iCoorFromIindex(icoor[lane],lane);
-  }
-  
-  parallel_for(uint64_t oidx = 0; oidx < grid->oSites(); oidx++){ //loop over outer index
-    //Assemble vector of pointers to output elements
-    std::vector<sobj*> ptrs(nsimd);
-
-    std::vector<int> ocoor(ndim);
-    grid->oCoorFromOindex(ocoor, oidx);
-
-    std::vector<int> lcoor(grid->Nd());
-      
-    for(int lane=0; lane < nsimd; lane++){
-
-      for(int mu=0;mu<ndim;mu++){
-	lcoor[mu] = ocoor[mu] + grid->_rdimensions[mu]*icoor[lane][mu];
-      }
-
-      int lex;
-      Lexicographic::IndexFromCoor(lcoor, lex, grid->_ldimensions);
-      ptrs[lane] = &in[lex];
-    }
-    
-    //pack from those ptrs
-    vobj vecobj;
-    merge1(vecobj, ptrs, 0);
-    out._odata[oidx] = vecobj; 
-  }
-}
 
 //Convert a Lattice from one precision to another
 template<class VobjOut, class VobjIn>
@@ -666,7 +615,7 @@ void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in){
   std::vector<SobjOut> in_slex_conv(in_grid->lSites());
   unvectorizeToLexOrdArray(in_slex_conv, in);
     
-  parallel_for(uint64_t out_oidx=0;out_oidx<out_grid->oSites();out_oidx++){
+  parallel_for(int out_oidx=0;out_oidx<out_grid->oSites();out_oidx++){
     std::vector<int> out_ocoor(ndim);
     out_grid->oCoorFromOindex(out_ocoor, out_oidx);
 

lib/lattice/Lattice_unary.h

@@ -62,20 +62,14 @@ namespace Grid {
     return ret;
   }
 
-  template<class obj> Lattice<obj> expMat(const Lattice<obj> &rhs, RealD alpha, Integer Nexp = DEFAULT_MAT_EXP){
+  template<class obj> Lattice<obj> expMat(const Lattice<obj> &rhs, ComplexD alpha, Integer Nexp = DEFAULT_MAT_EXP){
     Lattice<obj> ret(rhs._grid);
     ret.checkerboard = rhs.checkerboard;
     conformable(ret,rhs);
     parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
       ret._odata[ss]=Exponentiate(rhs._odata[ss],alpha, Nexp);
     }
-
     return ret;
-
-    
-    
-
-    
   }
 
 

lib/log/Log.cc

@@ -30,7 +30,6 @@ directory
 *************************************************************************************/
 /*  END LEGAL */
 #include <Grid/GridCore.h>
-#include <Grid/util/CompilerCompatible.h>
 
 #include <cxxabi.h>
 #include <memory>

lib/log/Log.h

@@ -110,8 +110,8 @@ public:
   friend std::ostream& operator<< (std::ostream& stream, Logger& log){
 
     if ( log.active ) {
-      stream << log.background()<< std::setw(8) << std::left << log.topName << log.background()<< " : ";
-      stream << log.colour() << std::setw(10) << std::left << log.name << log.background() << " : ";
+      stream << log.background()<< std::setw(10) << std::left << log.topName << log.background()<< " : ";
+      stream << log.colour() << std::setw(14) << std::left << log.name << log.background() << " : ";
       if ( log.timestamp ) {
 	StopWatch.Stop();
 	GridTime now = StopWatch.Elapsed();

lib/parallelIO/IldgIO.h

@@ -1,716 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/parallelIO/IldgIO.h
-
-Copyright (C) 2015
-
-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 */
-#ifndef GRID_ILDG_IO_H
-#define GRID_ILDG_IO_H
-
-#ifdef HAVE_LIME
-#include <algorithm>
-#include <fstream>
-#include <iomanip>
-#include <iostream>
-#include <map>
-
-#include <pwd.h>
-#include <sys/utsname.h>
-#include <unistd.h>
-
-//C-Lime is a must have for this functionality
-extern "C" {  
-#include "lime.h"
-}
-
-namespace Grid {
-namespace QCD {
-
-  /////////////////////////////////
-  // Encode word types as strings
-  /////////////////////////////////
- template<class word> inline std::string ScidacWordMnemonic(void){ return std::string("unknown"); }
- template<> inline std::string ScidacWordMnemonic<double>  (void){ return std::string("D"); }
- template<> inline std::string ScidacWordMnemonic<float>   (void){ return std::string("F"); }
- template<> inline std::string ScidacWordMnemonic< int32_t>(void){ return std::string("I32_t"); }
- template<> inline std::string ScidacWordMnemonic<uint32_t>(void){ return std::string("U32_t"); }
- template<> inline std::string ScidacWordMnemonic< int64_t>(void){ return std::string("I64_t"); }
- template<> inline std::string ScidacWordMnemonic<uint64_t>(void){ return std::string("U64_t"); }
-
-  /////////////////////////////////////////
-  // Encode a generic tensor as a string
-  /////////////////////////////////////////
- template<class vobj> std::string ScidacRecordTypeString(int &colors, int &spins, int & typesize,int &datacount) { 
-
-   typedef typename getPrecision<vobj>::real_scalar_type stype;
-
-   int _ColourN       = indexRank<ColourIndex,vobj>();
-   int _ColourScalar  =  isScalar<ColourIndex,vobj>();
-   int _ColourVector  =  isVector<ColourIndex,vobj>();
-   int _ColourMatrix  =  isMatrix<ColourIndex,vobj>();
-
-   int _SpinN       = indexRank<SpinIndex,vobj>();
-   int _SpinScalar  =  isScalar<SpinIndex,vobj>();
-   int _SpinVector  =  isVector<SpinIndex,vobj>();
-   int _SpinMatrix  =  isMatrix<SpinIndex,vobj>();
-
-   int _LorentzN       = indexRank<LorentzIndex,vobj>();
-   int _LorentzScalar  =  isScalar<LorentzIndex,vobj>();
-   int _LorentzVector  =  isVector<LorentzIndex,vobj>();
-   int _LorentzMatrix  =  isMatrix<LorentzIndex,vobj>();
-
-   std::stringstream stream;
-
-   stream << "GRID_";
-   stream << ScidacWordMnemonic<stype>();
-
-   //   std::cout << " Lorentz N/S/V/M : " << _LorentzN<<" "<<_LorentzScalar<<"/"<<_LorentzVector<<"/"<<_LorentzMatrix<<std::endl;
-   //   std::cout << " Spin    N/S/V/M : " << _SpinN   <<" "<<_SpinScalar   <<"/"<<_SpinVector   <<"/"<<_SpinMatrix<<std::endl;
-   //   std::cout << " Colour  N/S/V/M : " << _ColourN <<" "<<_ColourScalar <<"/"<<_ColourVector <<"/"<<_ColourMatrix<<std::endl;
-
-   if ( _LorentzVector )   stream << "_LorentzVector"<<_LorentzN;
-   if ( _LorentzMatrix )   stream << "_LorentzMatrix"<<_LorentzN;
-
-   if ( _SpinVector )   stream << "_SpinVector"<<_SpinN;
-   if ( _SpinMatrix )   stream << "_SpinMatrix"<<_SpinN;
-
-   if ( _ColourVector )   stream << "_ColourVector"<<_ColourN;
-   if ( _ColourMatrix )   stream << "_ColourMatrix"<<_ColourN;
-
-   if ( _ColourScalar && _LorentzScalar && _SpinScalar )   stream << "_Complex";
-
-
-   typesize = sizeof(typename vobj::scalar_type);
-
-   if ( _ColourMatrix ) typesize*= _ColourN*_ColourN;
-   else                 typesize*= _ColourN;
-
-   if ( _SpinMatrix )   typesize*= _SpinN*_SpinN;
-   else                 typesize*= _SpinN;
-
-   colors    = _ColourN;
-   spins     = _SpinN;
-   datacount = _LorentzN;
-
-   return stream.str();
- }
- 
- template<class vobj> std::string ScidacRecordTypeString(Lattice<vobj> & lat,int &colors, int &spins, int & typesize,int &datacount) { 
-   return ScidacRecordTypeString<vobj>(colors,spins,typesize,datacount);
- };
-
-
- ////////////////////////////////////////////////////////////
- // Helper to fill out metadata
- ////////////////////////////////////////////////////////////
- template<class vobj> void ScidacMetaData(Lattice<vobj> & field,
-					  FieldMetaData &header,
-					  scidacRecord & _scidacRecord,
-					  scidacFile   & _scidacFile) 
- {
-   typedef typename getPrecision<vobj>::real_scalar_type stype;
-
-   /////////////////////////////////////
-   // Pull Grid's metadata
-   /////////////////////////////////////
-   PrepareMetaData(field,header);
-
-   /////////////////////////////////////
-   // Scidac Private File structure
-   /////////////////////////////////////
-   _scidacFile              = scidacFile(field._grid);
-
-   /////////////////////////////////////
-   // Scidac Private Record structure
-   /////////////////////////////////////
-   scidacRecord sr;
-   sr.datatype   = ScidacRecordTypeString(field,sr.colors,sr.spins,sr.typesize,sr.datacount);
-   sr.date       = header.creation_date;
-   sr.precision  = ScidacWordMnemonic<stype>();
-   sr.recordtype = GRID_IO_FIELD;
-
-   _scidacRecord = sr;
-
-   std::cout << GridLogMessage << "Build SciDAC datatype " <<sr.datatype<<std::endl;
- }
- 
- ///////////////////////////////////////////////////////
- // Scidac checksum
- ///////////////////////////////////////////////////////
- static int scidacChecksumVerify(scidacChecksum &scidacChecksum_,uint32_t scidac_csuma,uint32_t scidac_csumb)
- {
-   uint32_t scidac_checksuma = stoull(scidacChecksum_.suma,0,16);
-   uint32_t scidac_checksumb = stoull(scidacChecksum_.sumb,0,16);
-   if ( scidac_csuma !=scidac_checksuma) return 0;
-   if ( scidac_csumb !=scidac_checksumb) return 0;
-    return 1;
- }
-
-////////////////////////////////////////////////////////////////////////////////////
-// Lime, ILDG and Scidac I/O classes
-////////////////////////////////////////////////////////////////////////////////////
-class GridLimeReader : public BinaryIO {
- public:
-   ///////////////////////////////////////////////////
-   // FIXME: format for RNG? Now just binary out instead
-   ///////////////////////////////////////////////////
-
-   FILE       *File;
-   LimeReader *LimeR;
-   std::string filename;
-
-   /////////////////////////////////////////////
-   // Open the file
-   /////////////////////////////////////////////
-   void open(std::string &_filename) 
-   {
-     filename= _filename;
-     File = fopen(filename.c_str(), "r");
-     LimeR = limeCreateReader(File);
-   }
-   /////////////////////////////////////////////
-   // Close the file
-   /////////////////////////////////////////////
-   void close(void){
-     fclose(File);
-     //     limeDestroyReader(LimeR);
-   }
-
-  ////////////////////////////////////////////
-  // Read a generic lattice field and verify checksum
-  ////////////////////////////////////////////
-  template<class vobj>
-  void readLimeLatticeBinaryObject(Lattice<vobj> &field,std::string record_name)
-  {
-    typedef typename vobj::scalar_object sobj;
-    scidacChecksum scidacChecksum_;
-    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-
-    std::string format = getFormatString<vobj>();
-
-    while ( limeReaderNextRecord(LimeR) == LIME_SUCCESS ) { 
-
-      std::cout << GridLogMessage << limeReaderType(LimeR) <<std::endl;
-	
-      if ( strncmp(limeReaderType(LimeR), record_name.c_str(),strlen(record_name.c_str()) )  ) {
-
-
-	off_t offset= ftell(File);
-	BinarySimpleMunger<sobj,sobj> munge;
-	BinaryIO::readLatticeObject< sobj, sobj >(field, filename, munge, offset, format,nersc_csum,scidac_csuma,scidac_csumb);
-
-	/////////////////////////////////////////////
-	// Insist checksum is next record
-	/////////////////////////////////////////////
-	readLimeObject(scidacChecksum_,std::string("scidacChecksum"),record_name);
-
-	/////////////////////////////////////////////
-	// Verify checksums
-	/////////////////////////////////////////////
-	scidacChecksumVerify(scidacChecksum_,scidac_csuma,scidac_csumb);
-	return;
-      }
-    }
-  }
-  ////////////////////////////////////////////
-  // Read a generic serialisable object
-  ////////////////////////////////////////////
-  template<class serialisable_object>
-  void readLimeObject(serialisable_object &object,std::string object_name,std::string record_name)
-  {
-    std::string xmlstring;
-    // should this be a do while; can we miss a first record??
-    while ( limeReaderNextRecord(LimeR) == LIME_SUCCESS ) { 
-
-      uint64_t nbytes = limeReaderBytes(LimeR);//size of this record (configuration)
-
-      if ( strncmp(limeReaderType(LimeR), record_name.c_str(),strlen(record_name.c_str()) )  ) {
-	std::vector<char> xmlc(nbytes+1,'\0');
-	limeReaderReadData((void *)&xmlc[0], &nbytes, LimeR);    
-	XmlReader RD(&xmlc[0],"");
-	read(RD,object_name,object);
-	return;
-      }
-
-    }  
-    assert(0);
-  }
-};
-
-class GridLimeWriter : public BinaryIO {
- public:
-   ///////////////////////////////////////////////////
-   // FIXME: format for RNG? Now just binary out instead
-   ///////////////////////////////////////////////////
-
-   FILE       *File;
-   LimeWriter *LimeW;
-   std::string filename;
-
-   void open(std::string &_filename) { 
-     filename= _filename;
-     File = fopen(filename.c_str(), "w");
-     LimeW = limeCreateWriter(File); assert(LimeW != NULL );
-   }
-   /////////////////////////////////////////////
-   // Close the file
-   /////////////////////////////////////////////
-   void close(void) {
-     fclose(File);
-     //  limeDestroyWriter(LimeW);
-   }
-  ///////////////////////////////////////////////////////
-  // Lime utility functions
-  ///////////////////////////////////////////////////////
-  int createLimeRecordHeader(std::string message, int MB, int ME, size_t PayloadSize)
-  {
-    LimeRecordHeader *h;
-    h = limeCreateHeader(MB, ME, const_cast<char *>(message.c_str()), PayloadSize);
-    assert(limeWriteRecordHeader(h, LimeW) >= 0);
-    limeDestroyHeader(h);
-    return LIME_SUCCESS;
-  }
-  ////////////////////////////////////////////
-  // Write a generic serialisable object
-  ////////////////////////////////////////////
-  template<class serialisable_object>
-  void writeLimeObject(int MB,int ME,serialisable_object &object,std::string object_name,std::string record_name)
-  {
-    std::string xmlstring;
-    {
-      XmlWriter WR("","");
-      write(WR,object_name,object);
-      xmlstring = WR.XmlString();
-    }
-    uint64_t nbytes = xmlstring.size();
-    int err;
-    LimeRecordHeader *h = limeCreateHeader(MB, ME,(char *)record_name.c_str(), nbytes); assert(h!= NULL);
-
-    err=limeWriteRecordHeader(h, LimeW);                    assert(err>=0);
-    err=limeWriteRecordData(&xmlstring[0], &nbytes, LimeW); assert(err>=0);
-    err=limeWriterCloseRecord(LimeW);                       assert(err>=0);
-    limeDestroyHeader(h);
-  }
-  ////////////////////////////////////////////
-  // Write a generic lattice field and csum
-  ////////////////////////////////////////////
-  template<class vobj>
-  void writeLimeLatticeBinaryObject(Lattice<vobj> &field,std::string record_name)
-  {
-    ////////////////////////////////////////////
-    // Create record header
-    ////////////////////////////////////////////
-    typedef typename vobj::scalar_object sobj;
-    int err;
-    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-    uint64_t PayloadSize = sizeof(sobj) * field._grid->_gsites;
-    createLimeRecordHeader(record_name, 0, 0, PayloadSize);
-
-    ////////////////////////////////////////////////////////////////////
-    // NB: FILE and iostream are jointly writing disjoint sequences in the
-    // the same file through different file handles (integer units).
-    // 
-    // These are both buffered, so why I think this code is right is as follows.
-    //
-    // i)  write record header to FILE *File, telegraphing the size. 
-    // ii) ftell reads the offset from FILE *File .
-    // iii) iostream / MPI Open independently seek this offset. Write sequence direct to disk.
-    //      Closes iostream and flushes.
-    // iv) fseek on FILE * to end of this disjoint section.
-    //  v) Continue writing scidac record.
-    ////////////////////////////////////////////////////////////////////
-    off_t offset = ftell(File);
-    std::string format = getFormatString<vobj>();
-    BinarySimpleMunger<sobj,sobj> munge;
-    BinaryIO::writeLatticeObject<vobj,sobj>(field, filename, munge, offset, format,nersc_csum,scidac_csuma,scidac_csumb);
-    err=limeWriterCloseRecord(LimeW);  assert(err>=0);
-    ////////////////////////////////////////
-    // Write checksum element, propagaing forward from the BinaryIO
-    // Always pair a checksum with a binary object, and close message
-    ////////////////////////////////////////
-    scidacChecksum checksum;
-    std::stringstream streama; streama << std::hex << scidac_csuma;
-    std::stringstream streamb; streamb << std::hex << scidac_csumb;
-    checksum.suma= streama.str();
-    checksum.sumb= streamb.str();
-    std::cout << GridLogMessage<<" writing scidac checksums "<<std::hex<<scidac_csuma<<"/"<<scidac_csumb<<std::dec<<std::endl;
-    writeLimeObject(0,1,checksum,std::string("scidacChecksum"    ),std::string(SCIDAC_CHECKSUM));
-  }
-};
-
-class ScidacWriter : public GridLimeWriter {
- public:
-
-   template<class SerialisableUserFile>
-   void writeScidacFileRecord(GridBase *grid,SerialisableUserFile &_userFile)
-   {
-     scidacFile    _scidacFile(grid);
-     writeLimeObject(1,0,_scidacFile,_scidacFile.SerialisableClassName(),std::string(SCIDAC_PRIVATE_FILE_XML));
-     writeLimeObject(0,1,_userFile,_userFile.SerialisableClassName(),std::string(SCIDAC_FILE_XML));
-   }
-  ////////////////////////////////////////////////
-  // Write generic lattice field in scidac format
-  ////////////////////////////////////////////////
-   template <class vobj, class userRecord>
-  void writeScidacFieldRecord(Lattice<vobj> &field,userRecord _userRecord) 
-  {
-    typedef typename vobj::scalar_object sobj;
-    uint64_t nbytes;
-    GridBase * grid = field._grid;
-
-    ////////////////////////////////////////
-    // fill the Grid header
-    ////////////////////////////////////////
-    FieldMetaData header;
-    scidacRecord  _scidacRecord;
-    scidacFile    _scidacFile;
-
-    ScidacMetaData(field,header,_scidacRecord,_scidacFile);
-
-    //////////////////////////////////////////////
-    // Fill the Lime file record by record
-    //////////////////////////////////////////////
-    writeLimeObject(1,0,header ,std::string("FieldMetaData"),std::string(GRID_FORMAT)); // Open message 
-    writeLimeObject(0,0,_userRecord,_userRecord.SerialisableClassName(),std::string(SCIDAC_RECORD_XML));
-    writeLimeObject(0,0,_scidacRecord,_scidacRecord.SerialisableClassName(),std::string(SCIDAC_PRIVATE_RECORD_XML));
-    writeLimeLatticeBinaryObject(field,std::string(ILDG_BINARY_DATA));      // Closes message with checksum
-  }
-};
-
-class IldgWriter : public ScidacWriter {
- public:
-
-  ///////////////////////////////////
-  // A little helper
-  ///////////////////////////////////
-  void writeLimeIldgLFN(std::string &LFN)
-  {
-    uint64_t PayloadSize = LFN.size();
-    int err;
-    createLimeRecordHeader(ILDG_DATA_LFN, 0 , 0, PayloadSize);
-    err=limeWriteRecordData(const_cast<char*>(LFN.c_str()), &PayloadSize,LimeW); assert(err>=0);
-    err=limeWriterCloseRecord(LimeW); assert(err>=0);
-  }
-
-  ////////////////////////////////////////////////////////////////
-  // Special ILDG operations ; gauge configs only.
-  // Don't require scidac records EXCEPT checksum
-  // Use Grid MetaData object if present.
-  ////////////////////////////////////////////////////////////////
-  template <class vsimd>
-  void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,int sequence,std::string LFN,std::string description) 
-  {
-    GridBase * grid = Umu._grid;
-    typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
-    typedef iLorentzColourMatrix<vsimd> vobj;
-    typedef typename vobj::scalar_object sobj;
-
-    uint64_t nbytes;
-
-    ////////////////////////////////////////
-    // fill the Grid header
-    ////////////////////////////////////////
-    FieldMetaData header;
-    scidacRecord  _scidacRecord;
-    scidacFile    _scidacFile;
-
-    ScidacMetaData(Umu,header,_scidacRecord,_scidacFile);
-
-    std::string format = header.floating_point;
-    header.ensemble_id    = description;
-    header.ensemble_label = description;
-    header.sequence_number = sequence;
-    header.ildg_lfn = LFN;
-
-    assert ( (format == std::string("IEEE32BIG"))  
-           ||(format == std::string("IEEE64BIG")) );
-
-    //////////////////////////////////////////////////////
-    // Fill ILDG header data struct
-    //////////////////////////////////////////////////////
-    ildgFormat ildgfmt ;
-    ildgfmt.field     = std::string("su3gauge");
-
-    if ( format == std::string("IEEE32BIG") ) { 
-      ildgfmt.precision = 32;
-    } else { 
-      ildgfmt.precision = 64;
-    }
-    ildgfmt.version = 1.0;
-    ildgfmt.lx = header.dimension[0];
-    ildgfmt.ly = header.dimension[1];
-    ildgfmt.lz = header.dimension[2];
-    ildgfmt.lt = header.dimension[3];
-    assert(header.nd==4);
-    assert(header.nd==header.dimension.size());
-
-    //////////////////////////////////////////////////////////////////////////////
-    // Fill the USQCD info field
-    //////////////////////////////////////////////////////////////////////////////
-    usqcdInfo info;
-    info.version=1.0;
-    info.plaq   = header.plaquette;
-    info.linktr = header.link_trace;
-
-    std::cout << GridLogMessage << " Writing config; IldgIO "<<std::endl;
-    //////////////////////////////////////////////
-    // Fill the Lime file record by record
-    //////////////////////////////////////////////
-    writeLimeObject(1,0,header ,std::string("FieldMetaData"),std::string(GRID_FORMAT)); // Open message 
-    writeLimeObject(0,0,_scidacFile,_scidacFile.SerialisableClassName(),std::string(SCIDAC_PRIVATE_FILE_XML));
-    writeLimeObject(0,1,info,info.SerialisableClassName(),std::string(SCIDAC_FILE_XML));
-    writeLimeObject(1,0,_scidacRecord,_scidacRecord.SerialisableClassName(),std::string(SCIDAC_PRIVATE_RECORD_XML));
-    writeLimeObject(0,0,info,info.SerialisableClassName(),std::string(SCIDAC_RECORD_XML));
-    writeLimeObject(0,0,ildgfmt,std::string("ildgFormat")   ,std::string(ILDG_FORMAT)); // rec
-    writeLimeIldgLFN(header.ildg_lfn);                                                 // rec
-    writeLimeLatticeBinaryObject(Umu,std::string(ILDG_BINARY_DATA));      // Closes message with checksum
-    //    limeDestroyWriter(LimeW);
-    fclose(File);
-  }
-};
-
-class IldgReader : public GridLimeReader {
- public:
-
-  ////////////////////////////////////////////////////////////////
-  // Read either Grid/SciDAC/ILDG configuration
-  // Don't require scidac records EXCEPT checksum
-  // Use Grid MetaData object if present.
-  // Else use ILDG MetaData object if present.
-  // Else use SciDAC MetaData object if present.
-  ////////////////////////////////////////////////////////////////
-  template <class vsimd>
-  void readConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu, FieldMetaData &FieldMetaData_) {
-
-    typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
-    typedef typename GaugeField::vector_object  vobj;
-    typedef typename vobj::scalar_object sobj;
-
-    typedef LorentzColourMatrixF fobj;
-    typedef LorentzColourMatrixD dobj;
-
-    GridBase *grid = Umu._grid;
-
-    std::vector<int> dims = Umu._grid->FullDimensions();
-
-    assert(dims.size()==4);
-
-    // Metadata holders
-    ildgFormat     ildgFormat_    ;
-    std::string    ildgLFN_       ;
-    scidacChecksum scidacChecksum_; 
-    usqcdInfo      usqcdInfo_     ;
-
-    // track what we read from file
-    int found_ildgFormat    =0;
-    int found_ildgLFN       =0;
-    int found_scidacChecksum=0;
-    int found_usqcdInfo     =0;
-    int found_ildgBinary =0;
-    int found_FieldMetaData =0;
-
-    uint32_t nersc_csum;
-    uint32_t scidac_csuma;
-    uint32_t scidac_csumb;
-
-    // Binary format
-    std::string format;
-
-    //////////////////////////////////////////////////////////////////////////
-    // Loop over all records
-    // -- Order is poorly guaranteed except ILDG header preceeds binary section.
-    // -- Run like an event loop.
-    // -- Impose trust hierarchy. Grid takes precedence & look for ILDG, and failing
-    //    that Scidac. 
-    // -- Insist on Scidac checksum record.
-    //////////////////////////////////////////////////////////////////////////
-
-    while ( limeReaderNextRecord(LimeR) == LIME_SUCCESS ) { 
-
-      uint64_t nbytes = limeReaderBytes(LimeR);//size of this record (configuration)
-      
-      //////////////////////////////////////////////////////////////////
-      // If not BINARY_DATA read a string and parse
-      //////////////////////////////////////////////////////////////////
-      if ( strncmp(limeReaderType(LimeR), ILDG_BINARY_DATA,strlen(ILDG_BINARY_DATA) )  ) {
-	
-	// Copy out the string
-	std::vector<char> xmlc(nbytes+1,'\0');
-	limeReaderReadData((void *)&xmlc[0], &nbytes, LimeR);    
-	std::cout << GridLogMessage<< "Non binary record :" <<limeReaderType(LimeR) <<std::endl; //<<"\n"<<(&xmlc[0])<<std::endl;
-
-	//////////////////////////////////
-	// ILDG format record
-	if ( !strncmp(limeReaderType(LimeR), ILDG_FORMAT,strlen(ILDG_FORMAT)) ) { 
-
-	  XmlReader RD(&xmlc[0],"");
-	  read(RD,"ildgFormat",ildgFormat_);
-
-	  if ( ildgFormat_.precision == 64 ) format = std::string("IEEE64BIG");
-	  if ( ildgFormat_.precision == 32 ) format = std::string("IEEE32BIG");
-
-	  assert( ildgFormat_.lx == dims[0]);
-	  assert( ildgFormat_.ly == dims[1]);
-	  assert( ildgFormat_.lz == dims[2]);
-	  assert( ildgFormat_.lt == dims[3]);
-
-	  found_ildgFormat = 1;
-	}
-
-	if ( !strncmp(limeReaderType(LimeR), ILDG_DATA_LFN,strlen(ILDG_DATA_LFN)) ) {
-	  FieldMetaData_.ildg_lfn = std::string(&xmlc[0]);
-	  found_ildgLFN = 1;
-	}
-
-	if ( !strncmp(limeReaderType(LimeR), GRID_FORMAT,strlen(ILDG_FORMAT)) ) { 
-
-	  XmlReader RD(&xmlc[0],"");
-	  read(RD,"FieldMetaData",FieldMetaData_);
-
-	  format = FieldMetaData_.floating_point;
-
-	  assert(FieldMetaData_.dimension[0] == dims[0]);
-	  assert(FieldMetaData_.dimension[1] == dims[1]);
-	  assert(FieldMetaData_.dimension[2] == dims[2]);
-	  assert(FieldMetaData_.dimension[3] == dims[3]);
-
-	  found_FieldMetaData = 1;
-	}
-
-	if ( !strncmp(limeReaderType(LimeR), SCIDAC_RECORD_XML,strlen(SCIDAC_RECORD_XML)) ) { 
-	  std::string xmls(&xmlc[0]);
-	  // is it a USQCD info field
-	  if ( xmls.find(std::string("usqcdInfo")) != std::string::npos ) { 
-	    std::cout << GridLogMessage<<"...found a usqcdInfo field"<<std::endl;
-	    XmlReader RD(&xmlc[0],"");
-	    read(RD,"usqcdInfo",usqcdInfo_);
-	    found_usqcdInfo = 1;
-	  }
-	}
-
-	if ( !strncmp(limeReaderType(LimeR), SCIDAC_CHECKSUM,strlen(SCIDAC_CHECKSUM)) ) { 
-	  XmlReader RD(&xmlc[0],"");
-	  read(RD,"scidacChecksum",scidacChecksum_);
-	  found_scidacChecksum = 1;
-	}
-
-      } else {  
-	/////////////////////////////////
-	// Binary data
-	/////////////////////////////////
-	std::cout << GridLogMessage << "ILDG Binary record found : "  ILDG_BINARY_DATA << std::endl;
-	off_t offset= ftell(File);
-
-	if ( format == std::string("IEEE64BIG") ) {
-	  GaugeSimpleMunger<dobj, sobj> munge;
-	  BinaryIO::readLatticeObject< vobj, dobj >(Umu, filename, munge, offset, format,nersc_csum,scidac_csuma,scidac_csumb);
-	} else { 
-	  GaugeSimpleMunger<fobj, sobj> munge;
-	  BinaryIO::readLatticeObject< vobj, fobj >(Umu, filename, munge, offset, format,nersc_csum,scidac_csuma,scidac_csumb);
-	}
-
-	found_ildgBinary = 1;
-      }
-
-    }
-
-    //////////////////////////////////////////////////////
-    // Minimally must find binary segment and checksum
-    // Since this is an ILDG reader require ILDG format
-    //////////////////////////////////////////////////////
-    assert(found_ildgBinary);
-    assert(found_ildgFormat);
-    assert(found_scidacChecksum);
-
-    // Must find something with the lattice dimensions
-    assert(found_FieldMetaData||found_ildgFormat);
-
-    if ( found_FieldMetaData ) {
-
-      std::cout << GridLogMessage<<"Grid MetaData was record found: configuration was probably written by Grid ! Yay ! "<<std::endl;
-
-    } else { 
-
-      assert(found_ildgFormat);
-      assert ( ildgFormat_.field == std::string("su3gauge") );
-
-      ///////////////////////////////////////////////////////////////////////////////////////
-      // Populate our Grid metadata as best we can
-      ///////////////////////////////////////////////////////////////////////////////////////
-
-      std::ostringstream vers; vers << ildgFormat_.version;
-      FieldMetaData_.hdr_version = vers.str();
-      FieldMetaData_.data_type = std::string("4D_SU3_GAUGE_3X3");
-
-      FieldMetaData_.nd=4;
-      FieldMetaData_.dimension.resize(4);
-
-      FieldMetaData_.dimension[0] = ildgFormat_.lx ;
-      FieldMetaData_.dimension[1] = ildgFormat_.ly ;
-      FieldMetaData_.dimension[2] = ildgFormat_.lz ;
-      FieldMetaData_.dimension[3] = ildgFormat_.lt ;
-
-      if ( found_usqcdInfo ) { 
-	FieldMetaData_.plaquette = usqcdInfo_.plaq;
-	FieldMetaData_.link_trace= usqcdInfo_.linktr;
-	std::cout << GridLogMessage <<"This configuration was probably written by USQCD "<<std::endl;
-	std::cout << GridLogMessage <<"USQCD xml record Plaquette : "<<FieldMetaData_.plaquette<<std::endl;
-	std::cout << GridLogMessage <<"USQCD xml record LinkTrace : "<<FieldMetaData_.link_trace<<std::endl;
-      } else { 
-	FieldMetaData_.plaquette = 0.0;
-	FieldMetaData_.link_trace= 0.0;
-	std::cout << GridLogWarning << "This configuration is unsafe with no plaquette records that can verify it !!! "<<std::endl;
-      }
-    }
-
-    ////////////////////////////////////////////////////////////
-    // Really really want to mandate a scidac checksum
-    ////////////////////////////////////////////////////////////
-    if ( found_scidacChecksum ) {
-      FieldMetaData_.scidac_checksuma = stoull(scidacChecksum_.suma,0,16);
-      FieldMetaData_.scidac_checksumb = stoull(scidacChecksum_.sumb,0,16);
-      scidacChecksumVerify(scidacChecksum_,scidac_csuma,scidac_csumb);
-      assert( scidac_csuma ==FieldMetaData_.scidac_checksuma);
-      assert( scidac_csumb ==FieldMetaData_.scidac_checksumb);
-      std::cout << GridLogMessage<<"SciDAC checksums match " << std::endl;
-    } else { 
-      std::cout << GridLogWarning<<"SciDAC checksums not found. This is unsafe. " << std::endl;
-      assert(0); // Can I insist always checksum ?
-    }
-
-    if ( found_FieldMetaData || found_usqcdInfo ) {
-      FieldMetaData checker;
-      GaugeStatistics(Umu,checker);
-      assert(fabs(checker.plaquette  - FieldMetaData_.plaquette )<1.0e-5);
-      assert(fabs(checker.link_trace - FieldMetaData_.link_trace)<1.0e-5);
-      std::cout << GridLogMessage<<"Plaquette and link trace match " << std::endl;
-    }
-  }
- };
-
-}}
-
-//HAVE_LIME
-#endif
-
-#endif

lib/parallelIO/IldgIOtypes.h

@@ -1,231 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/parallelIO/IldgIO.h
-
-Copyright (C) 2015
-
-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 */
-#ifndef GRID_ILDGTYPES_IO_H
-#define GRID_ILDGTYPES_IO_H
-
-#ifdef HAVE_LIME
-extern "C" { // for linkage
-#include "lime.h"
-}
-
-namespace Grid {
-
-/////////////////////////////////////////////////////////////////////////////////
-// Data representation of records that enter ILDG and SciDac formats
-/////////////////////////////////////////////////////////////////////////////////
-
-#define GRID_FORMAT      "grid-format"
-#define ILDG_FORMAT      "ildg-format"
-#define ILDG_BINARY_DATA "ildg-binary-data"
-#define ILDG_DATA_LFN    "ildg-data-lfn"
-#define SCIDAC_CHECKSUM           "scidac-checksum"
-#define SCIDAC_PRIVATE_FILE_XML   "scidac-private-file-xml"
-#define SCIDAC_FILE_XML           "scidac-file-xml"
-#define SCIDAC_PRIVATE_RECORD_XML "scidac-private-record-xml"
-#define SCIDAC_RECORD_XML         "scidac-record-xml"
-#define SCIDAC_BINARY_DATA        "scidac-binary-data"
-// Unused SCIDAC records names; could move to support this functionality
-#define SCIDAC_SITELIST           "scidac-sitelist"
-
-  ////////////////////////////////////////////////////////////
-  const int GRID_IO_SINGLEFILE = 0; // hardcode lift from QIO compat
-  const int GRID_IO_MULTIFILE  = 1; // hardcode lift from QIO compat
-  const int GRID_IO_FIELD      = 0; // hardcode lift from QIO compat
-  const int GRID_IO_GLOBAL     = 1; // hardcode lift from QIO compat
-  ////////////////////////////////////////////////////////////
-
-/////////////////////////////////////////////////////////////////////////////////
-// QIO uses mandatory "private" records fixed format
-// Private is in principle "opaque" however it can't be changed now because that would break existing 
-// file compatability, so should be correct to assume the undocumented but defacto file structure.
-/////////////////////////////////////////////////////////////////////////////////
-
-////////////////////////
-// Scidac private file xml
-// <?xml version="1.0" encoding="UTF-8"?><scidacFile><version>1.1</version><spacetime>4</spacetime><dims>16 16 16 32 </dims><volfmt>0</volfmt></scidacFile>
-////////////////////////
-struct scidacFile : Serializable {
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(scidacFile,
-                                  double, version,
-                                  int, spacetime,
-				  std::string, dims, // must convert to int
-                                  int, volfmt);
-
-  std::vector<int> getDimensions(void) { 
-    std::stringstream stream(dims);
-    std::vector<int> dimensions;
-    int n;
-    while(stream >> n){
-      dimensions.push_back(n);
-    }
-    return dimensions;
-  }
-
-  void setDimensions(std::vector<int> dimensions) { 
-    char delimiter = ' ';
-    std::stringstream stream;
-    for(int i=0;i<dimensions.size();i++){ 
-      stream << dimensions[i];
-      if ( i != dimensions.size()-1) { 
-	stream << delimiter <<std::endl;
-      }
-    }
-    dims = stream.str();
-  }
-
-  // Constructor provides Grid
-  scidacFile() =default; // default constructor
-  scidacFile(GridBase * grid){
-    version      = 1.0;
-    spacetime    = grid->_ndimension;
-    setDimensions(grid->FullDimensions()); 
-    volfmt       = GRID_IO_SINGLEFILE;
-  }
-
-};
-
-///////////////////////////////////////////////////////////////////////
-// scidac-private-record-xml : example
-// <scidacRecord>
-// <version>1.1</version><date>Tue Jul 26 21:14:44 2011 UTC</date><recordtype>0</recordtype>
-// <datatype>QDP_D3_ColorMatrix</datatype><precision>D</precision><colors>3</colors><spins>4</spins>
-// <typesize>144</typesize><datacount>4</datacount>
-// </scidacRecord>
-///////////////////////////////////////////////////////////////////////
-
-struct scidacRecord : Serializable {
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(scidacRecord,
-                                  double, version,
-                                  std::string, date,
-				  int, recordtype,
-				  std::string, datatype,
-				  std::string, precision,
-				  int, colors,
-				  int, spins,
-				  int, typesize,
-				  int, datacount);
-
-  scidacRecord() { version =1.0; }
-
-};
-
-////////////////////////
-// ILDG format
-////////////////////////
-struct ildgFormat : Serializable {
-public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(ildgFormat,
-				  double, version,
-				  std::string, field,
-				  int, precision,
-				  int, lx,
-				  int, ly,
-				  int, lz,
-				  int, lt);
-  ildgFormat() { version=1.0; };
-};
-////////////////////////
-// USQCD info
-////////////////////////
-struct usqcdInfo : Serializable { 
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdInfo,
-				  double, version,
-				  double, plaq,
-				  double, linktr,
-				  std::string, info);
-  usqcdInfo() { 
-    version=1.0; 
-  };
-};
-////////////////////////
-// Scidac Checksum
-////////////////////////
-struct scidacChecksum : Serializable { 
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(scidacChecksum,
-				  double, version,
-				  std::string, suma,
-				  std::string, sumb);
-  scidacChecksum() { 
-    version=1.0; 
-  };
-};
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// Type:           scidac-file-xml         <title>MILC ILDG archival gauge configuration</title>
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// Type:           
-////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-////////////////////////
-// Scidac private file xml 
-// <?xml version="1.0" encoding="UTF-8"?><scidacFile><version>1.1</version><spacetime>4</spacetime><dims>16 16 16 32 </dims><volfmt>0</volfmt></scidacFile> 
-////////////////////////                                                                                                                                                                              
-
-#if 0
-////////////////////////////////////////////////////////////////////////////////////////
-// From http://www.physics.utah.edu/~detar/scidac/qio_2p3.pdf
-////////////////////////////////////////////////////////////////////////////////////////
-struct usqcdPropFile : Serializable { 
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdPropFile,
-				  double, version,
-				  std::string, type,
-				  std::string, info);
-  usqcdPropFile() { 
-    version=1.0; 
-  };
-};
-struct usqcdSourceInfo : Serializable { 
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdSourceInfo,
-				  double, version,
-				  std::string, info);
-  usqcdSourceInfo() { 
-    version=1.0; 
-  };
-};
-struct usqcdPropInfo : Serializable { 
- public:
-  GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdPropInfo,
-				  double, version,
-				  int, spin,
-				  int, color,
-				  std::string, info);
-  usqcdPropInfo() { 
-    version=1.0; 
-  };
-};
-#endif
-
-}
-#endif
-#endif

lib/parallelIO/MetaData.h

@@ -1,325 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/parallelIO/NerscIO.h
-
-    Copyright (C) 2015
-
-
-    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-    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 */
-
-#include <algorithm>
-#include <iostream>
-#include <iomanip>
-#include <fstream>
-#include <map>
-#include <unistd.h>
-#include <sys/utsname.h>
-#include <pwd.h>
-
-namespace Grid {
-
-  ///////////////////////////////////////////////////////
-  // Precision mapping
-  ///////////////////////////////////////////////////////
-  template<class vobj> static std::string getFormatString (void)
-  {
-    std::string format;
-    typedef typename getPrecision<vobj>::real_scalar_type stype;
-    if ( sizeof(stype) == sizeof(float) ) {
-      format = std::string("IEEE32BIG");
-    }
-    if ( sizeof(stype) == sizeof(double) ) {
-      format = std::string("IEEE64BIG");
-    }
-    return format;
-  }
-  ////////////////////////////////////////////////////////////////////////////////
-  // header specification/interpretation
-  ////////////////////////////////////////////////////////////////////////////////
-    class FieldMetaData : Serializable {
-    public:
-
-      GRID_SERIALIZABLE_CLASS_MEMBERS(FieldMetaData,
-				      int, nd,
-				      std::vector<int>, dimension,
-				      std::vector<std::string>, boundary,
-				      int, data_start,
-				      std::string, hdr_version,
-				      std::string, storage_format,
-				      double, link_trace,
-				      double, plaquette,
-				      uint32_t, checksum,
-				      uint32_t, scidac_checksuma,
-				      uint32_t, scidac_checksumb,
-				      unsigned int, sequence_number,
-				      std::string, data_type,
-				      std::string, ensemble_id,
-				      std::string, ensemble_label,
-				      std::string, ildg_lfn,
-				      std::string, creator,
-				      std::string, creator_hardware,
-				      std::string, creation_date,
-				      std::string, archive_date,
-				      std::string, floating_point);
-      FieldMetaData(void) { 
-	nd=4;
-	dimension.resize(4);
-	boundary.resize(4);
-      }
-    };
-
-
-
-  namespace QCD {
-
-    using namespace Grid;
-
-
-    //////////////////////////////////////////////////////////////////////
-    // Bit and Physical Checksumming and QA of data
-    //////////////////////////////////////////////////////////////////////
-    inline void GridMetaData(GridBase *grid,FieldMetaData &header)
-    {
-      int nd = grid->_ndimension;
-      header.nd = nd;
-      header.dimension.resize(nd);
-      header.boundary.resize(nd);
-      for(int d=0;d<nd;d++) {
-	header.dimension[d] = grid->_fdimensions[d];
-      }
-      for(int d=0;d<nd;d++) {
-	header.boundary[d] = std::string("PERIODIC");
-      }
-    }
-
-    inline void MachineCharacteristics(FieldMetaData &header)
-    {
-      // Who
-      struct passwd *pw = getpwuid (getuid());
-      if (pw) header.creator = std::string(pw->pw_name); 
-
-      // When
-      std::time_t t = std::time(nullptr);
-      std::tm tm_ = *std::localtime(&t);
-      std::ostringstream oss; 
-      //      oss << std::put_time(&tm_, "%c %Z");
-      header.creation_date = oss.str();
-      header.archive_date  = header.creation_date;
-
-      // What
-      struct utsname name;  uname(&name);
-      header.creator_hardware = std::string(name.nodename)+"-";
-      header.creator_hardware+= std::string(name.machine)+"-";
-      header.creator_hardware+= std::string(name.sysname)+"-";
-      header.creator_hardware+= std::string(name.release);
-    }
-
-#define dump_meta_data(field, s)					\
-      s << "BEGIN_HEADER"      << std::endl;				\
-      s << "HDR_VERSION = "    << field.hdr_version    << std::endl;	\
-      s << "DATATYPE = "       << field.data_type      << std::endl;	\
-      s << "STORAGE_FORMAT = " << field.storage_format << std::endl;	\
-      for(int i=0;i<4;i++){						\
-	s << "DIMENSION_" << i+1 << " = " << field.dimension[i] << std::endl ; \
-      }									\
-      s << "LINK_TRACE = " << std::setprecision(10) << field.link_trace << std::endl; \
-      s << "PLAQUETTE  = " << std::setprecision(10) << field.plaquette  << std::endl; \
-      for(int i=0;i<4;i++){						\
-	s << "BOUNDARY_"<<i+1<<" = " << field.boundary[i] << std::endl;	\
-      }									\
-									\
-      s << "CHECKSUM = "<< std::hex << std::setw(10) << field.checksum << std::dec<<std::endl; \
-      s << "SCIDAC_CHECKSUMA = "<< std::hex << std::setw(10) << field.scidac_checksuma << std::dec<<std::endl; \
-      s << "SCIDAC_CHECKSUMB = "<< std::hex << std::setw(10) << field.scidac_checksumb << std::dec<<std::endl; \
-      s << "ENSEMBLE_ID = "     << field.ensemble_id      << std::endl;	\
-      s << "ENSEMBLE_LABEL = "  << field.ensemble_label   << std::endl;	\
-      s << "SEQUENCE_NUMBER = " << field.sequence_number  << std::endl;	\
-      s << "CREATOR = "         << field.creator          << std::endl;	\
-      s << "CREATOR_HARDWARE = "<< field.creator_hardware << std::endl;	\
-      s << "CREATION_DATE = "   << field.creation_date    << std::endl;	\
-      s << "ARCHIVE_DATE = "    << field.archive_date     << std::endl;	\
-      s << "FLOATING_POINT = "  << field.floating_point   << std::endl;	\
-      s << "END_HEADER"         << std::endl;
-
-template<class vobj> inline void PrepareMetaData(Lattice<vobj> & field, FieldMetaData &header)
-{
-  GridBase *grid = field._grid;
-  std::string format = getFormatString<vobj>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   MachineCharacteristics(header);
- }
- inline void GaugeStatistics(Lattice<vLorentzColourMatrixF> & data,FieldMetaData &header)
- {
-   // How to convert data precision etc...
-   header.link_trace=Grid::QCD::WilsonLoops<PeriodicGimplF>::linkTrace(data);
-   header.plaquette =Grid::QCD::WilsonLoops<PeriodicGimplF>::avgPlaquette(data);
- }
- inline void GaugeStatistics(Lattice<vLorentzColourMatrixD> & data,FieldMetaData &header)
- {
-   // How to convert data precision etc...
-   header.link_trace=Grid::QCD::WilsonLoops<PeriodicGimplD>::linkTrace(data);
-   header.plaquette =Grid::QCD::WilsonLoops<PeriodicGimplD>::avgPlaquette(data);
- }
- template<> inline void PrepareMetaData<vLorentzColourMatrixF>(Lattice<vLorentzColourMatrixF> & field, FieldMetaData &header)
- {
-   
-   GridBase *grid = field._grid;
-   std::string format = getFormatString<vLorentzColourMatrixF>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   GaugeStatistics(field,header);
-   MachineCharacteristics(header);
- }
- template<> inline void PrepareMetaData<vLorentzColourMatrixD>(Lattice<vLorentzColourMatrixD> & field, FieldMetaData &header)
- {
-   GridBase *grid = field._grid;
-   std::string format = getFormatString<vLorentzColourMatrixD>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   GaugeStatistics(field,header);
-   MachineCharacteristics(header);
- }
-
-    //////////////////////////////////////////////////////////////////////
-    // Utilities ; these are QCD aware
-    //////////////////////////////////////////////////////////////////////
-    inline void reconstruct3(LorentzColourMatrix & cm)
-    {
-      const int x=0;
-      const int y=1;
-      const int z=2;
-      for(int mu=0;mu<Nd;mu++){
-	cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
-	cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
-	cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
-      }
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Some data types for intermediate storage
-    ////////////////////////////////////////////////////////////////////////////////
-    template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, 2>, Nd >;
-
-    typedef iLorentzColour2x3<Complex>  LorentzColour2x3;
-    typedef iLorentzColour2x3<ComplexF> LorentzColour2x3F;
-    typedef iLorentzColour2x3<ComplexD> LorentzColour2x3D;
-
-/////////////////////////////////////////////////////////////////////////////////
-// Simple classes for precision conversion
-/////////////////////////////////////////////////////////////////////////////////
-template <class fobj, class sobj>
-struct BinarySimpleUnmunger {
-  typedef typename getPrecision<fobj>::real_scalar_type fobj_stype;
-  typedef typename getPrecision<sobj>::real_scalar_type sobj_stype;
-  
-  void operator()(sobj &in, fobj &out) {
-    // take word by word and transform accoding to the status
-    fobj_stype *out_buffer = (fobj_stype *)&out;
-    sobj_stype *in_buffer = (sobj_stype *)&in;
-    size_t fobj_words = sizeof(out) / sizeof(fobj_stype);
-    size_t sobj_words = sizeof(in) / sizeof(sobj_stype);
-    assert(fobj_words == sobj_words);
-    
-    for (unsigned int word = 0; word < sobj_words; word++)
-      out_buffer[word] = in_buffer[word];  // type conversion on the fly
-    
-  }
-};
-
-template <class fobj, class sobj>
-struct BinarySimpleMunger {
-  typedef typename getPrecision<fobj>::real_scalar_type fobj_stype;
-  typedef typename getPrecision<sobj>::real_scalar_type sobj_stype;
-
-  void operator()(fobj &in, sobj &out) {
-    // take word by word and transform accoding to the status
-    fobj_stype *in_buffer = (fobj_stype *)&in;
-    sobj_stype *out_buffer = (sobj_stype *)&out;
-    size_t fobj_words = sizeof(in) / sizeof(fobj_stype);
-    size_t sobj_words = sizeof(out) / sizeof(sobj_stype);
-    assert(fobj_words == sobj_words);
-    
-    for (unsigned int word = 0; word < sobj_words; word++)
-      out_buffer[word] = in_buffer[word];  // type conversion on the fly
-    
-  }
-};
-
-
-    template<class fobj,class sobj>
-    struct GaugeSimpleMunger{
-      void operator()(fobj &in, sobj &out) {
-        for (int mu = 0; mu < Nd; mu++) {
-          for (int i = 0; i < Nc; i++) {
-          for (int j = 0; j < Nc; j++) {
-	    out(mu)()(i, j) = in(mu)()(i, j);
-	  }}
-        }
-      };
-    };
-
-    template <class fobj, class sobj>
-    struct GaugeSimpleUnmunger {
-
-      void operator()(sobj &in, fobj &out) {
-        for (int mu = 0; mu < Nd; mu++) {
-          for (int i = 0; i < Nc; i++) {
-          for (int j = 0; j < Nc; j++) {
-	    out(mu)()(i, j) = in(mu)()(i, j);
-	  }}
-        }
-      };
-    };
-
-    template<class fobj,class sobj>
-    struct Gauge3x2munger{
-      void operator() (fobj &in,sobj &out){
-	for(int mu=0;mu<Nd;mu++){
-	  for(int i=0;i<2;i++){
-	  for(int j=0;j<3;j++){
-	    out(mu)()(i,j) = in(mu)(i)(j);
-	  }}
-	}
-	reconstruct3(out);
-      }
-    };
-
-    template<class fobj,class sobj>
-    struct Gauge3x2unmunger{
-      void operator() (sobj &in,fobj &out){
-	for(int mu=0;mu<Nd;mu++){
-	  for(int i=0;i<2;i++){
-	  for(int j=0;j<3;j++){
-	    out(mu)(i)(j) = in(mu)()(i,j);
-	  }}
-	}
-      }
-    };
-  }
-
-
-}

lib/parallelIO/NerscIO.h

@@ -1,4 +1,4 @@
-/*************************************************************************************
+    /*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -6,9 +6,9 @@
 
     Copyright (C) 2015
 
-    Author: Matt Spraggs <matthew.spraggs@gmail.com>
-    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-    Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: Matt Spraggs <matthew.spraggs@gmail.com>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
 
     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
@@ -25,59 +25,253 @@
     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 */
+    *************************************************************************************/
+    /*  END LEGAL */
 #ifndef GRID_NERSC_IO_H
 #define GRID_NERSC_IO_H
 
+#include <algorithm>
+#include <iostream>
+#include <iomanip>
+#include <fstream>
+#include <map>
+
+#include <unistd.h>
+#include <sys/utsname.h>
+#include <pwd.h>
+
 namespace Grid {
-  namespace QCD {
+namespace QCD {
 
-    using namespace Grid;
+using namespace Grid;
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Write and read from fstream; comput header offset for payload
-    ////////////////////////////////////////////////////////////////////////////////
-    class NerscIO : public BinaryIO { 
-    public:
+////////////////////////////////////////////////////////////////////////////////
+// Some data types for intermediate storage
+////////////////////////////////////////////////////////////////////////////////
+  template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, 2>, 4 >;
 
-      static inline void truncate(std::string file){
-	std::ofstream fout(file,std::ios::out);
+  typedef iLorentzColour2x3<Complex>  LorentzColour2x3;
+  typedef iLorentzColour2x3<ComplexF> LorentzColour2x3F;
+  typedef iLorentzColour2x3<ComplexD> LorentzColour2x3D;
+
+////////////////////////////////////////////////////////////////////////////////
+// header specification/interpretation
+////////////////////////////////////////////////////////////////////////////////
+class NerscField {
+ public:
+    // header strings (not in order)
+    int dimension[4];
+    std::string boundary[4]; 
+    int data_start;
+    std::string hdr_version;
+    std::string storage_format;
+    // Checks on data
+    double link_trace;
+    double plaquette;
+    uint32_t checksum;
+    unsigned int sequence_number;
+    std::string data_type;
+    std::string ensemble_id ;
+    std::string ensemble_label ;
+    std::string creator ;
+    std::string creator_hardware ;
+    std::string creation_date ;
+    std::string archive_date ;
+    std::string floating_point;
+};
+
+//////////////////////////////////////////////////////////////////////
+// Bit and Physical Checksumming and QA of data
+//////////////////////////////////////////////////////////////////////
+
+inline void NerscGrid(GridBase *grid,NerscField &header)
+{
+  assert(grid->_ndimension==4);
+  for(int d=0;d<4;d++) {
+    header.dimension[d] = grid->_fdimensions[d];
+  }
+  for(int d=0;d<4;d++) {
+    header.boundary[d] = std::string("PERIODIC");
+  }
+}
+template<class GaugeField>
+inline void NerscStatistics(GaugeField & data,NerscField &header)
+{
+  // How to convert data precision etc...
+  header.link_trace=Grid::QCD::WilsonLoops<PeriodicGimplR>::linkTrace(data);
+  header.plaquette =Grid::QCD::WilsonLoops<PeriodicGimplR>::avgPlaquette(data);
+}
+
+inline void NerscMachineCharacteristics(NerscField &header)
+{
+  // Who
+  struct passwd *pw = getpwuid (getuid());
+  if (pw) header.creator = std::string(pw->pw_name); 
+
+  // When
+  std::time_t t = std::time(nullptr);
+  std::tm tm = *std::localtime(&t);
+  std::ostringstream oss; 
+  //  oss << std::put_time(&tm, "%c %Z");
+  header.creation_date = oss.str();
+  header.archive_date  = header.creation_date;
+
+  // What
+  struct utsname name;  uname(&name);
+  header.creator_hardware = std::string(name.nodename)+"-";
+  header.creator_hardware+= std::string(name.machine)+"-";
+  header.creator_hardware+= std::string(name.sysname)+"-";
+  header.creator_hardware+= std::string(name.release);
+
+}
+//////////////////////////////////////////////////////////////////////
+// Utilities ; these are QCD aware
+//////////////////////////////////////////////////////////////////////
+    inline void NerscChecksum(uint32_t *buf,uint32_t buf_size_bytes,uint32_t &csum)
+    {
+      BinaryIO::Uint32Checksum(buf,buf_size_bytes,csum);
+    }
+    inline void reconstruct3(LorentzColourMatrix & cm)
+    {
+      const int x=0;
+      const int y=1;
+      const int z=2;
+      for(int mu=0;mu<4;mu++){
+	cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
+	cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
+	cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
       }
-  
-      static inline unsigned int writeHeader(FieldMetaData &field,std::string file)
-      {
-      std::ofstream fout(file,std::ios::out|std::ios::in);
-      fout.seekp(0,std::ios::beg);
-      dump_meta_data(field, fout);
-      field.data_start = fout.tellp();
-      return field.data_start;
     }
 
-      // for the header-reader
-      static inline int readHeader(std::string file,GridBase *grid,  FieldMetaData &field)
-      {
-      int offset=0;
-      std::map<std::string,std::string> header;
-      std::string line;
+    template<class fobj,class sobj>
+    struct NerscSimpleMunger{
 
-      //////////////////////////////////////////////////
-      // read the header
-      //////////////////////////////////////////////////
-      std::ifstream fin(file);
+      void operator() (fobj &in,sobj &out,uint32_t &csum){
 
-      getline(fin,line); // read one line and insist is 
+      for(int mu=0;mu<4;mu++){
+      for(int i=0;i<3;i++){
+      for(int j=0;j<3;j++){
+	out(mu)()(i,j) = in(mu)()(i,j);
+      }}}
+      NerscChecksum((uint32_t *)&in,sizeof(in),csum); 
+      };
+    };
 
-      removeWhitespace(line);
-      std::cout << GridLogMessage << "* " << line << std::endl;
+    template<class fobj,class sobj>
+    struct NerscSimpleUnmunger{
+      void operator() (sobj &in,fobj &out,uint32_t &csum){
+	for(int mu=0;mu<Nd;mu++){
+	for(int i=0;i<Nc;i++){
+	for(int j=0;j<Nc;j++){
+	  out(mu)()(i,j) = in(mu)()(i,j);
+	}}}
+	NerscChecksum((uint32_t *)&out,sizeof(out),csum); 
+      };
+    };
+ 
+    template<class fobj,class sobj>
+    struct Nersc3x2munger{
+      void operator() (fobj &in,sobj &out,uint32_t &csum){
+     
+	NerscChecksum((uint32_t *)&in,sizeof(in),csum); 
 
-      assert(line==std::string("BEGIN_HEADER"));
+	for(int mu=0;mu<4;mu++){
+	  for(int i=0;i<2;i++){
+	    for(int j=0;j<3;j++){
+	      out(mu)()(i,j) = in(mu)(i)(j);
+	    }}
+	}
+	reconstruct3(out);
+      }
+    };
 
-      do {
-      getline(fin,line); // read one line
-      std::cout << GridLogMessage << "* "<<line<< std::endl;
-      int eq = line.find("=");
-      if(eq >0) {
+    template<class fobj,class sobj>
+    struct Nersc3x2unmunger{
+
+      void operator() (sobj &in,fobj &out,uint32_t &csum){
+
+
+	for(int mu=0;mu<4;mu++){
+	  for(int i=0;i<2;i++){
+	    for(int j=0;j<3;j++){
+	      out(mu)(i)(j) = in(mu)()(i,j);
+	    }}
+	}
+
+	NerscChecksum((uint32_t *)&out,sizeof(out),csum); 
+
+      }
+    };
+
+
+////////////////////////////////////////////////////////////////////////////////
+// Write and read from fstream; comput header offset for payload
+////////////////////////////////////////////////////////////////////////////////
+class NerscIO : public BinaryIO { 
+ public:
+
+  static inline void truncate(std::string file){
+    std::ofstream fout(file,std::ios::out);
+  }
+  
+  #define dump_nersc_header(field, s)\
+  s << "BEGIN_HEADER"      << std::endl;\
+  s << "HDR_VERSION = "    << field.hdr_version    << std::endl;\
+  s << "DATATYPE = "       << field.data_type      << std::endl;\
+  s << "STORAGE_FORMAT = " << field.storage_format << std::endl;\
+  for(int i=0;i<4;i++){\
+    s << "DIMENSION_" << i+1 << " = " << field.dimension[i] << std::endl ;\
+  }\
+  s << "LINK_TRACE = " << std::setprecision(10) << field.link_trace << std::endl;\
+  s << "PLAQUETTE  = " << std::setprecision(10) << field.plaquette  << std::endl;\
+  for(int i=0;i<4;i++){\
+    s << "BOUNDARY_"<<i+1<<" = " << field.boundary[i] << std::endl;\
+  }\
+  \
+  s << "CHECKSUM = "<< std::hex << std::setw(10) << field.checksum << std::dec<<std::endl;\
+  s << "ENSEMBLE_ID = "     << field.ensemble_id      << std::endl;\
+  s << "ENSEMBLE_LABEL = "  << field.ensemble_label   << std::endl;\
+  s << "SEQUENCE_NUMBER = " << field.sequence_number  << std::endl;\
+  s << "CREATOR = "         << field.creator          << std::endl;\
+  s << "CREATOR_HARDWARE = "<< field.creator_hardware << std::endl;\
+  s << "CREATION_DATE = "   << field.creation_date    << std::endl;\
+  s << "ARCHIVE_DATE = "    << field.archive_date     << std::endl;\
+  s << "FLOATING_POINT = "  << field.floating_point   << std::endl;\
+  s << "END_HEADER"         << std::endl;
+  
+  static inline unsigned int writeHeader(NerscField &field,std::string file)
+  {
+    std::ofstream fout(file,std::ios::out|std::ios::in);
+    fout.seekp(0,std::ios::beg);
+    dump_nersc_header(field, fout);
+    field.data_start = fout.tellp();
+    return field.data_start;
+}
+
+// for the header-reader
+static inline int readHeader(std::string file,GridBase *grid,  NerscField &field)
+{
+  int offset=0;
+  std::map<std::string,std::string> header;
+  std::string line;
+
+  //////////////////////////////////////////////////
+  // read the header
+  //////////////////////////////////////////////////
+  std::ifstream fin(file);
+
+  getline(fin,line); // read one line and insist is 
+
+  removeWhitespace(line);
+  std::cout << GridLogMessage << "* " << line << std::endl;
+
+  assert(line==std::string("BEGIN_HEADER"));
+
+  do {
+    getline(fin,line); // read one line
+    std::cout << GridLogMessage << "* "<<line<< std::endl;
+    int eq = line.find("=");
+    if(eq >0) {
       std::string key=line.substr(0,eq);
       std::string val=line.substr(eq+1);
       removeWhitespace(key);
@@ -85,269 +279,275 @@ namespace Grid {
       
       header[key] = val;
     }
-    } while( line.find("END_HEADER") == std::string::npos );
+  } while( line.find("END_HEADER") == std::string::npos );
 
-      field.data_start = fin.tellg();
+  field.data_start = fin.tellg();
 
-      //////////////////////////////////////////////////
-      // chomp the values
-      //////////////////////////////////////////////////
-      field.hdr_version    = header["HDR_VERSION"];
-      field.data_type      = header["DATATYPE"];
-      field.storage_format = header["STORAGE_FORMAT"];
+  //////////////////////////////////////////////////
+  // chomp the values
+  //////////////////////////////////////////////////
+  field.hdr_version    = header["HDR_VERSION"];
+  field.data_type      = header["DATATYPE"];
+  field.storage_format = header["STORAGE_FORMAT"];
   
-      field.dimension[0] = std::stol(header["DIMENSION_1"]);
-      field.dimension[1] = std::stol(header["DIMENSION_2"]);
-      field.dimension[2] = std::stol(header["DIMENSION_3"]);
-      field.dimension[3] = std::stol(header["DIMENSION_4"]);
+  field.dimension[0] = std::stol(header["DIMENSION_1"]);
+  field.dimension[1] = std::stol(header["DIMENSION_2"]);
+  field.dimension[2] = std::stol(header["DIMENSION_3"]);
+  field.dimension[3] = std::stol(header["DIMENSION_4"]);
 
-      assert(grid->_ndimension == 4);
-      for(int d=0;d<4;d++){
-      assert(grid->_fdimensions[d]==field.dimension[d]);
+  assert(grid->_ndimension == 4);
+  for(int d=0;d<4;d++){
+    assert(grid->_fdimensions[d]==field.dimension[d]);
+  }
+
+  field.link_trace = std::stod(header["LINK_TRACE"]);
+  field.plaquette  = std::stod(header["PLAQUETTE"]);
+
+  field.boundary[0] = header["BOUNDARY_1"];
+  field.boundary[1] = header["BOUNDARY_2"];
+  field.boundary[2] = header["BOUNDARY_3"];
+  field.boundary[3] = header["BOUNDARY_4"];
+
+  field.checksum = std::stoul(header["CHECKSUM"],0,16);
+  field.ensemble_id      = header["ENSEMBLE_ID"];
+  field.ensemble_label   = header["ENSEMBLE_LABEL"];
+  field.sequence_number  = std::stol(header["SEQUENCE_NUMBER"]);
+  field.creator          = header["CREATOR"];
+  field.creator_hardware = header["CREATOR_HARDWARE"];
+  field.creation_date    = header["CREATION_DATE"];
+  field.archive_date     = header["ARCHIVE_DATE"];
+  field.floating_point   = header["FLOATING_POINT"];
+
+  return field.data_start;
+}
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Now the meat: the object readers
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define PARALLEL_READ
+#define PARALLEL_WRITE
+
+template<class vsimd>
+static inline void readConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,NerscField& header,std::string file)
+{
+  typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+
+  GridBase *grid = Umu._grid;
+  int offset = readHeader(file,Umu._grid,header);
+
+  NerscField clone(header);
+
+  std::string format(header.floating_point);
+
+  int ieee32big = (format == std::string("IEEE32BIG"));
+  int ieee32    = (format == std::string("IEEE32"));
+  int ieee64big = (format == std::string("IEEE64BIG"));
+  int ieee64    = (format == std::string("IEEE64"));
+
+  uint32_t csum;
+  // depending on datatype, set up munger;
+  // munger is a function of <floating point, Real, data_type>
+  if ( header.data_type == std::string("4D_SU3_GAUGE") ) {
+    if ( ieee32 || ieee32big ) {
+#ifdef PARALLEL_READ
+      csum=BinaryIO::readObjectParallel<iLorentzColourMatrix<vsimd>, LorentzColour2x3F> 
+	(Umu,file,Nersc3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format);
+#else
+      csum=BinaryIO::readObjectSerial<iLorentzColourMatrix<vsimd>, LorentzColour2x3F> 
+	(Umu,file,Nersc3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format);
+#endif
     }
-
-      field.link_trace = std::stod(header["LINK_TRACE"]);
-      field.plaquette  = std::stod(header["PLAQUETTE"]);
-
-      field.boundary[0] = header["BOUNDARY_1"];
-      field.boundary[1] = header["BOUNDARY_2"];
-      field.boundary[2] = header["BOUNDARY_3"];
-      field.boundary[3] = header["BOUNDARY_4"];
-
-      field.checksum = std::stoul(header["CHECKSUM"],0,16);
-      field.ensemble_id      = header["ENSEMBLE_ID"];
-      field.ensemble_label   = header["ENSEMBLE_LABEL"];
-      field.sequence_number  = std::stol(header["SEQUENCE_NUMBER"]);
-      field.creator          = header["CREATOR"];
-      field.creator_hardware = header["CREATOR_HARDWARE"];
-      field.creation_date    = header["CREATION_DATE"];
-      field.archive_date     = header["ARCHIVE_DATE"];
-      field.floating_point   = header["FLOATING_POINT"];
-
-      return field.data_start;
+    if ( ieee64 || ieee64big ) {
+#ifdef PARALLEL_READ
+      csum=BinaryIO::readObjectParallel<iLorentzColourMatrix<vsimd>, LorentzColour2x3D> 
+      	(Umu,file,Nersc3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format);
+#else 
+      csum=BinaryIO::readObjectSerial<iLorentzColourMatrix<vsimd>, LorentzColour2x3D> 
+      	(Umu,file,Nersc3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format);
+#endif
     }
-
-    /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Now the meat: the object readers
-    /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-
-    template<class vsimd>
-    static inline void readConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
-					 FieldMetaData& header,
-					 std::string file)
-    {
-      typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
-
-      GridBase *grid = Umu._grid;
-      int offset = readHeader(file,Umu._grid,header);
-
-      FieldMetaData clone(header);
-
-      std::string format(header.floating_point);
-
-      int ieee32big = (format == std::string("IEEE32BIG"));
-      int ieee32    = (format == std::string("IEEE32"));
-      int ieee64big = (format == std::string("IEEE64BIG"));
-      int ieee64    = (format == std::string("IEEE64"));
-
-      uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-      // depending on datatype, set up munger;
-      // munger is a function of <floating point, Real, data_type>
-      if ( header.data_type == std::string("4D_SU3_GAUGE") ) {
-	if ( ieee32 || ieee32big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3F> 
-	    (Umu,file,Gauge3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-	if ( ieee64 || ieee64big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3D> 
-	    (Umu,file,Gauge3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-      } else if ( header.data_type == std::string("4D_SU3_GAUGE_3x3") ) {
-	if ( ieee32 || ieee32big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixF>
-	    (Umu,file,GaugeSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-	if ( ieee64 || ieee64big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixD>
-	    (Umu,file,GaugeSimpleMunger<LorentzColourMatrixD,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-      } else {
-	assert(0);
-      }
-
-      GaugeStatistics(Umu,clone);
-
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" checksum "<<std::hex<<nersc_csum<< std::dec
-	       <<" header   "<<std::hex<<header.checksum<<std::dec <<std::endl;
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" plaquette "<<clone.plaquette
-	       <<" header    "<<header.plaquette<<std::endl;
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" link_trace "<<clone.link_trace
-	       <<" header    "<<header.link_trace<<std::endl;
-
-      if ( fabs(clone.plaquette -header.plaquette ) >=  1.0e-5 ) { 
-	std::cout << " Plaquette mismatch "<<std::endl;
-	std::cout << Umu[0]<<std::endl;
-	std::cout << Umu[1]<<std::endl;
-      }
-      if ( nersc_csum != header.checksum ) { 
-	std::cerr << " checksum mismatch " << std::endl;
-	std::cerr << " plaqs " << clone.plaquette << " " << header.plaquette << std::endl;
-	std::cerr << " trace " << clone.link_trace<< " " << header.link_trace<< std::endl;
-	std::cerr << " nersc_csum  " <<std::hex<< nersc_csum << " " << header.checksum<< std::dec<< std::endl;
-	exit(0);
-      }
-      assert(fabs(clone.plaquette -header.plaquette ) < 1.0e-5 );
-      assert(fabs(clone.link_trace-header.link_trace) < 1.0e-6 );
-      assert(nersc_csum == header.checksum );
-      
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<< " and plaquette, link trace, and checksum agree"<<std::endl;
+  } else if ( header.data_type == std::string("4D_SU3_GAUGE_3x3") ) {
+    if ( ieee32 || ieee32big ) {
+#ifdef PARALLEL_READ
+      csum=BinaryIO::readObjectParallel<iLorentzColourMatrix<vsimd>,LorentzColourMatrixF>
+	(Umu,file,NerscSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format);
+#else
+      csum=BinaryIO::readObjectSerial<iLorentzColourMatrix<vsimd>,LorentzColourMatrixF>
+	(Umu,file,NerscSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format);
+#endif
     }
+    if ( ieee64 || ieee64big ) {
+#ifdef PARALLEL_READ
+      csum=BinaryIO::readObjectParallel<iLorentzColourMatrix<vsimd>,LorentzColourMatrixD>
+	(Umu,file,NerscSimpleMunger<LorentzColourMatrixD,LorentzColourMatrix>(),offset,format);
+#else
+      csum=BinaryIO::readObjectSerial<iLorentzColourMatrix<vsimd>,LorentzColourMatrixD>
+	(Umu,file,NerscSimpleMunger<LorentzColourMatrixD,LorentzColourMatrix>(),offset,format);
+#endif
+    }
+  } else {
+    assert(0);
+  }
 
-      template<class vsimd>
-      static inline void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
-					    std::string file, 
-					    int two_row,
-					    int bits32)
-      {
-	typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+  NerscStatistics<GaugeField>(Umu,clone);
 
-	typedef iLorentzColourMatrix<vsimd> vobj;
-	typedef typename vobj::scalar_object sobj;
+  std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" checksum "<<std::hex<<            csum<< std::dec
+	                                                  <<" header   "<<std::hex<<header.checksum<<std::dec <<std::endl;
+  std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" plaquette "<<clone.plaquette
+	                                                  <<" header    "<<header.plaquette<<std::endl;
+  std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" link_trace "<<clone.link_trace
+	                                                  <<" header    "<<header.link_trace<<std::endl;
+  assert(fabs(clone.plaquette -header.plaquette ) < 1.0e-5 );
+  assert(fabs(clone.link_trace-header.link_trace) < 1.0e-6 );
+  assert(csum == header.checksum );
 
-	FieldMetaData header;
-	///////////////////////////////////////////
-	// Following should become arguments
-	///////////////////////////////////////////
-	header.sequence_number = 1;
-	header.ensemble_id     = "UKQCD";
-	header.ensemble_label  = "DWF";
+  std::cout<<GridLogMessage <<"NERSC Configuration "<<file<< " and plaquette, link trace, and checksum agree"<<std::endl;
+}
 
-	typedef LorentzColourMatrixD fobj3D;
-	typedef LorentzColour2x3D    fobj2D;
+template<class vsimd>
+static inline void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,std::string file, int two_row,int bits32)
+{
+  typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+
+  typedef iLorentzColourMatrix<vsimd> vobj;
+  typedef typename vobj::scalar_object sobj;
+
+  // Following should become arguments
+  NerscField header;
+  header.sequence_number = 1;
+  header.ensemble_id     = "UKQCD";
+  header.ensemble_label  = "DWF";
+
+  typedef LorentzColourMatrixD fobj3D;
+  typedef LorentzColour2x3D    fobj2D;
+  typedef LorentzColourMatrixF fobj3f;
+  typedef LorentzColour2x3F    fobj2f;
+
+  GridBase *grid = Umu._grid;
+
+  NerscGrid(grid,header);
+  NerscStatistics<GaugeField>(Umu,header);
+  NerscMachineCharacteristics(header);
+
+  uint32_t csum;
+  int offset;
   
-	GridBase *grid = Umu._grid;
+  truncate(file);
 
-	GridMetaData(grid,header);
-	assert(header.nd==4);
-	GaugeStatistics(Umu,header);
-	MachineCharacteristics(header);
+  if ( two_row ) { 
 
-	int offset;
-  
-	truncate(file);
+    header.floating_point = std::string("IEEE64BIG");
+    header.data_type      = std::string("4D_SU3_GAUGE");
+    Nersc3x2unmunger<fobj2D,sobj> munge;
+    BinaryIO::Uint32Checksum<vobj,fobj2D>(Umu, munge,header.checksum);
+    offset = writeHeader(header,file);
+#ifdef PARALLEL_WRITE
+    csum=BinaryIO::writeObjectParallel<vobj,fobj2D>(Umu,file,munge,offset,header.floating_point);
+#else
+    csum=BinaryIO::writeObjectSerial<vobj,fobj2D>(Umu,file,munge,offset,header.floating_point);
+#endif
 
-	// Sod it -- always write 3x3 double
-	header.floating_point = std::string("IEEE64BIG");
-	header.data_type      = std::string("4D_SU3_GAUGE_3x3");
-	GaugeSimpleUnmunger<fobj3D,sobj> munge;
-	offset = writeHeader(header,file);
+  } else { 
+    header.floating_point = std::string("IEEE64BIG");
+    header.data_type      = std::string("4D_SU3_GAUGE_3x3");
+    NerscSimpleUnmunger<fobj3D,sobj> munge;
+    BinaryIO::Uint32Checksum<vobj,fobj3D>(Umu, munge,header.checksum);
+    offset = writeHeader(header,file);
+#ifdef PARALLEL_WRITE
+    csum=BinaryIO::writeObjectParallel<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point);
+#else
+    csum=BinaryIO::writeObjectSerial<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point);
+#endif
+  }
 
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::writeLatticeObject<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point,
-								  nersc_csum,scidac_csuma,scidac_csumb);
-	header.checksum = nersc_csum;
-	writeHeader(header,file);
+  std::cout<<GridLogMessage <<"Written NERSC Configuration "<<file<< " checksum "<<std::hex<<csum<< std::dec<<" plaq "<< header.plaquette <<std::endl;
 
-	std::cout<<GridLogMessage <<"Written NERSC Configuration on "<< file << " checksum "
-		 <<std::hex<<header.checksum
-		 <<std::dec<<" plaq "<< header.plaquette <<std::endl;
+ }
 
-      }
-      ///////////////////////////////
-      // RNG state
-      ///////////////////////////////
-      static inline void writeRNGState(GridSerialRNG &serial,GridParallelRNG &parallel,std::string file)
-      {
-	typedef typename GridParallelRNG::RngStateType RngStateType;
 
-	// Following should become arguments
-	FieldMetaData header;
-	header.sequence_number = 1;
-	header.ensemble_id     = "UKQCD";
-	header.ensemble_label  = "DWF";
+    ///////////////////////////////
+    // RNG state
+    ///////////////////////////////
+static inline void writeRNGState(GridSerialRNG &serial,GridParallelRNG &parallel,std::string file)
+{
+  typedef typename GridParallelRNG::RngStateType RngStateType;
 
-	GridBase *grid = parallel._grid;
+  // Following should become arguments
+  NerscField header;
+  header.sequence_number = 1;
+  header.ensemble_id     = "UKQCD";
+  header.ensemble_label  = "DWF";
 
-	GridMetaData(grid,header);
-	assert(header.nd==4);
-	header.link_trace=0.0;
-	header.plaquette=0.0;
-	MachineCharacteristics(header);
+  GridBase *grid = parallel._grid;
 
-	int offset;
+  NerscGrid(grid,header);
+  header.link_trace=0.0;
+  header.plaquette=0.0;
+  NerscMachineCharacteristics(header);
+
+  uint32_t csum;
+  int offset;
   
 #ifdef RNG_RANLUX
-	header.floating_point = std::string("UINT64");
-	header.data_type      = std::string("RANLUX48");
+    header.floating_point = std::string("UINT64");
+    header.data_type      = std::string("RANLUX48");
 #endif
 #ifdef RNG_MT19937
-	header.floating_point = std::string("UINT32");
-	header.data_type      = std::string("MT19937");
+    header.floating_point = std::string("UINT32");
+    header.data_type      = std::string("MT19937");
 #endif
 #ifdef RNG_SITMO
-	header.floating_point = std::string("UINT64");
-	header.data_type      = std::string("SITMO");
+    header.floating_point = std::string("UINT64");
+    header.data_type      = std::string("SITMO");
 #endif
 
-	truncate(file);
-	offset = writeHeader(header,file);
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::writeRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
-	header.checksum = nersc_csum;
-	offset = writeHeader(header,file);
+  truncate(file);
+  offset = writeHeader(header,file);
+  csum=BinaryIO::writeRNGSerial(serial,parallel,file,offset);
+  header.checksum = csum;
+  offset = writeHeader(header,file);
 
-	std::cout<<GridLogMessage 
-		 <<"Written NERSC RNG STATE "<<file<< " checksum "
-		 <<std::hex<<header.checksum
-		 <<std::dec<<std::endl;
+  std::cout<<GridLogMessage <<"Written NERSC RNG STATE "<<file<< " checksum "<<std::hex<<csum<<std::dec<<std::endl;
 
-      }
+ }
     
-      static inline void readRNGState(GridSerialRNG &serial,GridParallelRNG & parallel,FieldMetaData& header,std::string file)
-      {
-	typedef typename GridParallelRNG::RngStateType RngStateType;
+static inline void readRNGState(GridSerialRNG &serial,GridParallelRNG & parallel,NerscField& header,std::string file)
+{
+  typedef typename GridParallelRNG::RngStateType RngStateType;
 
-	GridBase *grid = parallel._grid;
+  GridBase *grid = parallel._grid;
 
-	int offset = readHeader(file,grid,header);
+  int offset = readHeader(file,grid,header);
 
-	FieldMetaData clone(header);
+  NerscField clone(header);
 
-	std::string format(header.floating_point);
-	std::string data_type(header.data_type);
+  std::string format(header.floating_point);
+  std::string data_type(header.data_type);
 
 #ifdef RNG_RANLUX
-	assert(format == std::string("UINT64"));
-	assert(data_type == std::string("RANLUX48"));
+  assert(format == std::string("UINT64"));
+  assert(data_type == std::string("RANLUX48"));
 #endif
 #ifdef RNG_MT19937
-	assert(format == std::string("UINT32"));
-	assert(data_type == std::string("MT19937"));
+  assert(format == std::string("UINT32"));
+  assert(data_type == std::string("MT19937"));
 #endif
 #ifdef RNG_SITMO
-	assert(format == std::string("UINT64"));
-	assert(data_type == std::string("SITMO"));
+  assert(format == std::string("UINT64"));
+  assert(data_type == std::string("SITMO"));
 #endif
 
-	// depending on datatype, set up munger;
-	// munger is a function of <floating point, Real, data_type>
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::readRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
+  // depending on datatype, set up munger;
+  // munger is a function of <floating point, Real, data_type>
+  uint32_t csum=BinaryIO::readRNGSerial(serial,parallel,file,offset);
 
-	if ( nersc_csum != header.checksum ) { 
-	  std::cerr << "checksum mismatch "<<std::hex<< nersc_csum <<" "<<header.checksum<<std::dec<<std::endl;
-	  exit(0);
-	}
-	assert(nersc_csum == header.checksum );
+  assert(csum == header.checksum );
 
-	std::cout<<GridLogMessage <<"Read NERSC RNG file "<<file<< " format "<< data_type <<std::endl;
-      }
+  std::cout<<GridLogMessage <<"Read NERSC RNG file "<<file<< " format "<< data_type <<std::endl;
+}
 
-    };
+};
 
-  }}
+
+}}
 #endif

lib/perfmon/PerfCount.h

@@ -205,14 +205,13 @@ public:
   void Stop(void) {
     count=0;
     cycles=0;
+    size_t ign;
 #ifdef __linux__
-    ssize_t ign;
     if ( fd!= -1) {
       ::ioctl(fd, PERF_EVENT_IOC_DISABLE, 0);
       ::ioctl(cyclefd, PERF_EVENT_IOC_DISABLE, 0);
       ign=::read(fd, &count, sizeof(long long));
-      ign+=::read(cyclefd, &cycles, sizeof(long long));
-      assert(ign=2*sizeof(long long));
+      ign=::read(cyclefd, &cycles, sizeof(long long));
     }
     elapsed = cyclecount() - begin;
 #else

lib/qcd/LatticeTheories.h

@@ -1,124 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/QCD.h
-
-Copyright (C) 2015
-
-Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: neo <cossu@post.kek.jp>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-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 */
-#ifndef GRID_LT_H
-#define GRID_LT_H
-namespace Grid{
-
-// First steps in the complete generalization of the Physics part
-// Design not final
-namespace LatticeTheories {
-
-template <int Dimensions>
-struct LatticeTheory {
-  static const int Nd = Dimensions;
-  static const int Nds = Dimensions * 2;  // double stored field
-  template <typename vtype>
-  using iSinglet = iScalar<iScalar<iScalar<vtype> > >;
-};
-
-template <int Dimensions, int Colours>
-struct LatticeGaugeTheory : public LatticeTheory<Dimensions> {
-  static const int Nds = Dimensions * 2;
-  static const int Nd = Dimensions;
-  static const int Nc = Colours;
-
-  template <typename vtype> 
-  using iColourMatrix = iScalar<iScalar<iMatrix<vtype, Nc> > >;
-  template <typename vtype>
-  using iLorentzColourMatrix = iVector<iScalar<iMatrix<vtype, Nc> >, Nd>;
-  template <typename vtype>
-  using iDoubleStoredColourMatrix = iVector<iScalar<iMatrix<vtype, Nc> >, Nds>;
-  template <typename vtype>
-  using iColourVector = iScalar<iScalar<iVector<vtype, Nc> > >;
-};
-
-template <int Dimensions, int Colours, int Spin>
-struct FermionicLatticeGaugeTheory
-    : public LatticeGaugeTheory<Dimensions, Colours> {
-  static const int Nd = Dimensions;
-  static const int Nds = Dimensions * 2;
-  static const int Nc = Colours;
-  static const int Ns = Spin;
-
-  template <typename vtype>
-  using iSpinMatrix = iScalar<iMatrix<iScalar<vtype>, Ns> >;
-  template <typename vtype>
-  using iSpinColourMatrix = iScalar<iMatrix<iMatrix<vtype, Nc>, Ns> >;
-  template <typename vtype>
-  using iSpinVector = iScalar<iVector<iScalar<vtype>, Ns> >;
-  template <typename vtype>
-  using iSpinColourVector = iScalar<iVector<iVector<vtype, Nc>, Ns> >;
-  // These 2 only if Spin is a multiple of 2
-  static const int Nhs = Spin / 2;
-  template <typename vtype>
-  using iHalfSpinVector = iScalar<iVector<iScalar<vtype>, Nhs> >;
-  template <typename vtype>
-  using iHalfSpinColourVector = iScalar<iVector<iVector<vtype, Nc>, Nhs> >;
-
-  //tests
-  typedef iColourMatrix<Complex> ColourMatrix;
-  typedef iColourMatrix<ComplexF> ColourMatrixF;
-  typedef iColourMatrix<ComplexD> ColourMatrixD;
-
-
-};
-
-// Examples, not complete now.
-struct QCD : public FermionicLatticeGaugeTheory<4, 3, 4> {
-    static const int Xp = 0;
-    static const int Yp = 1;
-    static const int Zp = 2;
-    static const int Tp = 3;
-    static const int Xm = 4;
-    static const int Ym = 5;
-    static const int Zm = 6;
-    static const int Tm = 7;
-
-    typedef FermionicLatticeGaugeTheory FLGT;
-
-    typedef FLGT::iSpinMatrix<Complex  >          SpinMatrix;
-    typedef FLGT::iSpinMatrix<ComplexF >          SpinMatrixF;
-    typedef FLGT::iSpinMatrix<ComplexD >          SpinMatrixD;
-
-};
-struct QED : public FermionicLatticeGaugeTheory<4, 1, 4> {//fill
-};
-
-template <int Dimensions>
-struct Scalar : public LatticeTheory<Dimensions> {};
-
-};  // LatticeTheories
-
-} // Grid
-
-#endif

lib/qcd/QCD.h

@@ -32,12 +32,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef GRID_QCD_BASE_H
 #define GRID_QCD_BASE_H
 namespace Grid{
+
 namespace QCD {
 
-    static const int Xdir = 0;
-    static const int Ydir = 1;
-    static const int Zdir = 2;
-    static const int Tdir = 3;
 
     static const int Xp = 0;
     static const int Yp = 1;
@@ -357,36 +354,36 @@ namespace QCD {
     //////////////////////////////////////////////
     template<class vobj> 
       void pokeColour(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0))> & rhs,
-              int i)
+		      const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0))> & rhs,
+		      int i)
     {
       PokeIndex<ColourIndex>(lhs,rhs,i);
     }
     template<class vobj> 
       void pokeColour(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0,0))> & rhs,
-              int i,int j)
+		      const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0,0))> & rhs,
+		      int i,int j)
     {
       PokeIndex<ColourIndex>(lhs,rhs,i,j);
     }
     template<class vobj> 
       void pokeSpin(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0))> & rhs,
-              int i)
+		      const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0))> & rhs,
+		      int i)
     {
       PokeIndex<SpinIndex>(lhs,rhs,i);
     }
     template<class vobj> 
       void pokeSpin(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0,0))> & rhs,
-              int i,int j)
+		      const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0,0))> & rhs,
+		      int i,int j)
     {
       PokeIndex<SpinIndex>(lhs,rhs,i,j);
     }
     template<class vobj> 
       void pokeLorentz(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<LorentzIndex>(lhs._odata[0],0))> & rhs,
-              int i)
+		      const Lattice<decltype(peekIndex<LorentzIndex>(lhs._odata[0],0))> & rhs,
+		      int i)
     {
       PokeIndex<LorentzIndex>(lhs,rhs,i);
     }
@@ -495,38 +492,6 @@ namespace QCD {
 }   //namespace QCD
 } // Grid
 
-/*
-<<<<<<< HEAD
-#include <Grid/qcd/utils/SpaceTimeGrid.h>
-#include <Grid/qcd/spin/Dirac.h>
-#include <Grid/qcd/spin/TwoSpinor.h>
-#include <Grid/qcd/utils/LinalgUtils.h>
-#include <Grid/qcd/utils/CovariantCshift.h>
-
-// Include representations  
-#include <Grid/qcd/utils/SUn.h>
-#include <Grid/qcd/utils/SUnAdjoint.h>
-#include <Grid/qcd/utils/SUnTwoIndex.h>
-#include <Grid/qcd/representations/hmc_types.h>
-
-// Scalar field
-#include <Grid/qcd/utils/ScalarObjs.h>
-
-#include <Grid/qcd/action/Actions.h>
-
-#include <Grid/qcd/smearing/Smearing.h>
-
-#include <Grid/qcd/hmc/integrators/Integrator.h>
-#include <Grid/qcd/hmc/integrators/Integrator_algorithm.h>
-#include <Grid/qcd/observables/hmc_observable.h>
-#include <Grid/qcd/hmc/HMC.h>
-
-
-//#include <Grid/qcd/modules/mods.h>
-=======
-
->>>>>>> develop
-*/
 
 
 #endif

lib/qcd/action/ActionBase.h

@@ -4,11 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: ./lib/qcd/action/ActionBase.h
 
-Copyright (C) 2015-2016
+Copyright (C) 2015
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: neo <cossu@post.kek.jp>
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
 
 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
@@ -28,29 +27,128 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 /*  END LEGAL */
-
-#ifndef ACTION_BASE_H
-#define ACTION_BASE_H
-
+#ifndef QCD_ACTION_BASE
+#define QCD_ACTION_BASE
 namespace Grid {
 namespace QCD {
 
-template <class GaugeField >
-class Action 
-{
-
+template <class GaugeField>
+class Action {
  public:
   bool is_smeared = false;
-  // Heatbath?
-  virtual void refresh(const GaugeField& U, GridParallelRNG& pRNG) = 0; // refresh pseudofermions
-  virtual RealD S(const GaugeField& U) = 0;                             // evaluate the action
-  virtual void deriv(const GaugeField& U, GaugeField& dSdU) = 0;        // evaluate the action derivative
-  virtual std::string action_name()    = 0;                             // return the action name
-  virtual std::string LogParameters()  = 0;                             // prints action parameters
-  virtual ~Action(){}
+  // Boundary conditions? // Heatbath?
+  virtual void refresh(const GaugeField& U,
+                       GridParallelRNG& pRNG) = 0;  // refresh pseudofermions
+  virtual RealD S(const GaugeField& U) = 0;         // evaluate the action
+  virtual void deriv(const GaugeField& U,
+                     GaugeField& dSdU) = 0;  // evaluate the action derivative
+  virtual ~Action(){};
 };
 
+// Indexing of tuple types
+template <class T, class Tuple>
+struct Index;
+
+template <class T, class... Types>
+struct Index<T, std::tuple<T, Types...>> {
+  static const std::size_t value = 0;
+};
+
+template <class T, class U, class... Types>
+struct Index<T, std::tuple<U, Types...>> {
+  static const std::size_t value = 1 + Index<T, std::tuple<Types...>>::value;
+};
+
+/*
+template <class GaugeField>
+struct ActionLevel {
+ public:
+  typedef Action<GaugeField>*
+      ActPtr;  // now force the same colours as the rest of the code
+
+  //Add supported representations here
+
+
+  unsigned int multiplier;
+
+  std::vector<ActPtr> actions;
+
+  ActionLevel(unsigned int mul = 1) : actions(0), multiplier(mul) {
+    assert(mul >= 1);
+  };
+
+  void push_back(ActPtr ptr) { actions.push_back(ptr); }
+};
+*/
+
+template <class GaugeField, class Repr = NoHirep >
+struct ActionLevel {
+ public:
+  unsigned int multiplier; 
+
+  // Fundamental repr actions separated because of the smearing
+  typedef Action<GaugeField>* ActPtr;
+
+  // construct a tuple of vectors of the actions for the corresponding higher
+  // representation fields
+  typedef typename AccessTypes<Action, Repr>::VectorCollection action_collection;
+  action_collection actions_hirep;
+  typedef typename  AccessTypes<Action, Repr>::FieldTypeCollection action_hirep_types;
+
+  std::vector<ActPtr>& actions;
+
+  // Temporary conversion between ActionLevel and ActionLevelHirep
+  //ActionLevelHirep(ActionLevel<GaugeField>& AL ):actions(AL.actions), multiplier(AL.multiplier){}
+
+  ActionLevel(unsigned int mul = 1) : actions(std::get<0>(actions_hirep)), multiplier(mul) {
+    // initialize the hirep vectors to zero.
+    //apply(this->resize, actions_hirep, 0); //need a working resize
+    assert(mul >= 1);
+  };
+
+  //void push_back(ActPtr ptr) { actions.push_back(ptr); }
+
+
+
+  template < class Field >
+  void push_back(Action<Field>* ptr) {
+    // insert only in the correct vector
+    std::get< Index < Field, action_hirep_types>::value >(actions_hirep).push_back(ptr);
+  };
+
+ 
+
+  template < class ActPtr>
+  static void resize(ActPtr ap, unsigned int n){
+    ap->resize(n);
+
+  }
+
+  //template <std::size_t I>
+  //auto getRepresentation(Repr& R)->decltype(std::get<I>(R).U)  {return std::get<I>(R).U;}
+
+  // Loop on tuple for a callable function
+  template <std::size_t I = 1, typename Callable, typename ...Args>
+  inline typename std::enable_if<I == std::tuple_size<action_collection>::value, void>::type apply(
+      Callable, Repr& R,Args&...) const {}
+
+  template <std::size_t I = 1, typename Callable, typename ...Args>
+  inline typename std::enable_if<I < std::tuple_size<action_collection>::value, void>::type apply(
+      Callable fn, Repr& R, Args&... arguments) const {
+    fn(std::get<I>(actions_hirep), std::get<I>(R.rep), arguments...);
+    apply<I + 1>(fn, R, arguments...);
+  }  
+
+};
+
+
+//template <class GaugeField>
+//using ActionSet = std::vector<ActionLevel<GaugeField> >;
+
+template <class GaugeField, class R>
+using ActionSet = std::vector<ActionLevel<GaugeField, R> >;
+
 }
 }
 
-#endif // ACTION_BASE_H
+#endif

lib/qcd/action/ActionCore.h

@@ -31,31 +31,15 @@ directory
 #define QCD_ACTION_CORE
 
 #include <Grid/qcd/action/ActionBase.h>
-#include <Grid/qcd/action/ActionSet.h>
 #include <Grid/qcd/action/ActionParams.h>
 
 ////////////////////////////////////////////
 // Gauge Actions
 ////////////////////////////////////////////
 #include <Grid/qcd/action/gauge/Gauge.h>
-
 ////////////////////////////////////////////
 // Fermion prereqs
 ////////////////////////////////////////////
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-////////////////////////////////////////////
-// Scalar Actions
-////////////////////////////////////////////
-#include <Grid/qcd/action/scalar/Scalar.h>
-
-////////////////////////////////////////////
-// Utility functions
-////////////////////////////////////////////
-#include <Grid/qcd/utils/Metric.h>
-#include <Grid/qcd/utils/CovariantLaplacian.h>
-
-
-
-
 #endif

lib/qcd/action/ActionParams.h

@@ -1,92 +1,67 @@
-/*************************************************************************************
+    /*************************************************************************************
 
-Grid physics library, www.github.com/paboyle/Grid
+    Grid physics library, www.github.com/paboyle/Grid 
 
-Source file: ./lib/qcd/action/ActionParams.h
+    Source file: ./lib/qcd/action/ActionParams.h
 
-Copyright (C) 2015
+    Copyright (C) 2015
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
 
-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 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.
+    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 */
+    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 */
 #ifndef GRID_QCD_ACTION_PARAMS_H
 #define GRID_QCD_ACTION_PARAMS_H
 
 namespace Grid {
 namespace QCD {
 
-  // These can move into a params header and be given MacroMagic serialisation
-  struct GparityWilsonImplParams {
-    bool overlapCommsCompute;
-    std::vector<int> twists;
-    GparityWilsonImplParams() : twists(Nd, 0), overlapCommsCompute(false){};
-  };
-  
-  struct WilsonImplParams {
-    bool overlapCommsCompute;
-    std::vector<Complex> boundary_phases;
-    WilsonImplParams() : overlapCommsCompute(false) {
-      boundary_phases.resize(Nd, 1.0);
+    // These can move into a params header and be given MacroMagic serialisation
+    struct GparityWilsonImplParams {
+      bool overlapCommsCompute;
+      std::vector<int> twists; 
+      GparityWilsonImplParams () : twists(Nd,0), overlapCommsCompute(false) {};
+
     };
-    WilsonImplParams(const std::vector<Complex> phi)
-      : boundary_phases(phi), overlapCommsCompute(false) {}
-  };
 
-  struct StaggeredImplParams {
-    StaggeredImplParams()  {};
-  };
-  
-  struct OneFlavourRationalParams : Serializable {
-    GRID_SERIALIZABLE_CLASS_MEMBERS(OneFlavourRationalParams, 
-				    RealD, lo, 
-				    RealD, hi, 
-				    int,   MaxIter, 
-				    RealD, tolerance, 
-				    int,   degree, 
-				    int,   precision);
-    
-    // MaxIter and tolerance, vectors??
-    
-    // constructor 
-    OneFlavourRationalParams(	RealD _lo      = 0.0, 
-				RealD _hi      = 1.0, 
-				int _maxit     = 1000,
-				RealD tol      = 1.0e-8, 
-                           	int _degree    = 10,
-				int _precision = 64)
-      : lo(_lo),
-	hi(_hi),
-	MaxIter(_maxit),
-	tolerance(tol),
-	degree(_degree),
-	precision(_precision){};
-  };
-  
-  
-}
-}
+    struct WilsonImplParams {
+      bool overlapCommsCompute;
+      WilsonImplParams() : overlapCommsCompute(false) {};
+    };
 
+    struct StaggeredImplParams {
+      StaggeredImplParams()  {};
+    };
 
+    struct OneFlavourRationalParams { 
+      RealD  lo;
+      RealD  hi;
+      int MaxIter;   // Vector?
+      RealD tolerance; // Vector? 
+      int    degree=10;
+      int precision=64;
 
+      OneFlavourRationalParams (RealD _lo,RealD _hi,int _maxit,RealD tol=1.0e-8,int _degree = 10,int _precision=64) :
+        lo(_lo), hi(_hi), MaxIter(_maxit), tolerance(tol), degree(_degree), precision(_precision)
+      {};
+    };
+
+}}
 
 #endif

lib/qcd/action/ActionSet.h

@@ -1,116 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/ActionSet.h
-
-Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: neo <cossu@post.kek.jp>
-
-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 */
-#ifndef ACTION_SET_H
-#define ACTION_SET_H
-
-namespace Grid {
-
-// Should drop this namespace here
-namespace QCD {
-
-//////////////////////////////////
-// Indexing of tuple types
-//////////////////////////////////
-
-template <class T, class Tuple>
-struct Index;
-
-template <class T, class... Types>
-struct Index<T, std::tuple<T, Types...>> {
-  static const std::size_t value = 0;
-};
-
-template <class T, class U, class... Types>
-struct Index<T, std::tuple<U, Types...>> {
-  static const std::size_t value = 1 + Index<T, std::tuple<Types...>>::value;
-};
-
-
-////////////////////////////////////////////
-// Action Level
-// Action collection 
-// in a integration level
-// (for multilevel integration schemes)
-////////////////////////////////////////////
-
-template <class Field, class Repr = NoHirep >
-struct ActionLevel {
- public:
-  unsigned int multiplier;
-
-  // Fundamental repr actions separated because of the smearing
-  typedef Action<Field>* ActPtr;
-
-  // construct a tuple of vectors of the actions for the corresponding higher
-  // representation fields
-  typedef typename AccessTypes<Action, Repr>::VectorCollection action_collection;
-  typedef typename  AccessTypes<Action, Repr>::FieldTypeCollection action_hirep_types;
-
-  action_collection actions_hirep;
-  std::vector<ActPtr>& actions;
-
-  explicit ActionLevel(unsigned int mul = 1) : 
-  actions(std::get<0>(actions_hirep)), multiplier(mul) {
-    // initialize the hirep vectors to zero.
-    // apply(this->resize, actions_hirep, 0); //need a working resize
-    assert(mul >= 1);
-  }
-
-  template < class GenField >
-  void push_back(Action<GenField>* ptr) {
-    // insert only in the correct vector
-    std::get< Index < GenField, action_hirep_types>::value >(actions_hirep).push_back(ptr);
-  };
-
-  template <class ActPtr>
-  static void resize(ActPtr ap, unsigned int n) {
-    ap->resize(n);
-  }
-
-  // Loop on tuple for a callable function
-  template <std::size_t I = 1, typename Callable, typename ...Args>
-  inline typename std::enable_if<I == std::tuple_size<action_collection>::value, void>::type apply(Callable, Repr& R,Args&...) const {}
-
-  template <std::size_t I = 1, typename Callable, typename ...Args>
-  inline typename std::enable_if<I < std::tuple_size<action_collection>::value, void>::type apply(Callable fn, Repr& R, Args&... arguments) const {
-    fn(std::get<I>(actions_hirep), std::get<I>(R.rep), arguments...);
-    apply<I + 1>(fn, R, arguments...);
-  }  
-
-};
-
-// Define the ActionSet
-template <class GaugeField, class R>
-using ActionSet = std::vector<ActionLevel<GaugeField, R> >;
-
-} // QCD
-} // Grid
-
-#endif  // ACTION_SET_H

lib/qcd/action/fermion/CayleyFermion5D.cc

@@ -29,7 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     *************************************************************************************/
     /*  END LEGAL */
 
-#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/Eigen/Dense>
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
@@ -320,7 +320,7 @@ void CayleyFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const
     this->DhopDeriv(mat,U,Din,dag);
   } else {
     //      U d/du [D_w D5]^dag V = U D5^dag d/du DW^dag Y // implicit adj on U in call
-    Meooe5D(U,Din);
+    MeooeDag5D(U,Din);
     this->DhopDeriv(mat,Din,V,dag);
   }
 };
@@ -335,8 +335,8 @@ void CayleyFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const
     this->DhopDerivOE(mat,U,Din,dag);
   } else {
     //      U d/du [D_w D5]^dag V = U D5^dag d/du DW^dag Y // implicit adj on U in call
-    Meooe5D(U,Din);
-    this->DhopDerivOE(mat,Din,V,dag);
+      MeooeDag5D(U,Din);
+      this->DhopDerivOE(mat,Din,V,dag);
   }
 };
 template<class Impl>
@@ -350,7 +350,7 @@ void CayleyFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const
     this->DhopDerivEO(mat,U,Din,dag);
   } else {
     //      U d/du [D_w D5]^dag V = U D5^dag d/du DW^dag Y // implicit adj on U in call
-    Meooe5D(U,Din);
+    MeooeDag5D(U,Din);
     this->DhopDerivEO(mat,Din,V,dag);
   }
 };
@@ -380,8 +380,6 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Co
   ///////////////////////////////////////////////////////////
   // The Cayley coeffs (unprec)
   ///////////////////////////////////////////////////////////
-  assert(gamma.size()==Ls);
-
   omega.resize(Ls);
   bs.resize(Ls);
   cs.resize(Ls);
@@ -414,11 +412,12 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Co
   for(int i=0; i < Ls; i++){
     as[i] = 1.0;
     omega[i] = gamma[i]*zolo_hi; //NB reciprocal relative to Chroma NEF code
-    //    assert(fabs(omega[i])>0.0);
     bs[i] = 0.5*(bpc/omega[i] + bmc);
     cs[i] = 0.5*(bpc/omega[i] - bmc);
-  }
+    std::cout<<GridLogMessage << "CayleyFermion5D "<<i<<" bs="<<bs[i]<<" cs="<<cs[i]<< std::endl;
 
+  }
+  
   ////////////////////////////////////////////////////////
   // Constants for the preconditioned matrix Cayley form
   ////////////////////////////////////////////////////////
@@ -428,12 +427,12 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Co
   ceo.resize(Ls);
   
   for(int i=0;i<Ls;i++){
-    bee[i]=as[i]*(bs[i]*(4.0-this->M5) +1.0);     
-    //    assert(fabs(bee[i])>0.0);
+    bee[i]=as[i]*(bs[i]*(4.0-this->M5) +1.0);
     cee[i]=as[i]*(1.0-cs[i]*(4.0-this->M5));
     beo[i]=as[i]*bs[i];
     ceo[i]=-as[i]*cs[i];
   }
+  
   aee.resize(Ls);
   aeo.resize(Ls);
   for(int i=0;i<Ls;i++){
@@ -477,16 +476,14 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Co
 	
   { 
     Coeff_t delta_d=mass*cee[Ls-1];
-    for(int j=0;j<Ls-1;j++) {
-      //      assert(fabs(bee[j])>0.0);
-      delta_d *= cee[j]/bee[j];
-    }
+    for(int j=0;j<Ls-1;j++) delta_d *= cee[j]/bee[j];
     dee[Ls-1] += delta_d;
   }  
 
   int inv=1;
   this->MooeeInternalCompute(0,inv,MatpInv,MatmInv);
   this->MooeeInternalCompute(1,inv,MatpInvDag,MatmInvDag);
+
 }
 
 
@@ -500,9 +497,7 @@ void CayleyFermion5D<Impl>::MooeeInternalCompute(int dag, int inv,
   GridBase *grid = this->FermionRedBlackGrid();
   int LLs = grid->_rdimensions[0];
 
-  if ( LLs == Ls ) {
-    return; // Not vectorised in 5th direction
-  }
+  if ( LLs == Ls ) return; // Not vectorised in 5th direction
 
   Eigen::MatrixXcd Pplus  = Eigen::MatrixXcd::Zero(Ls,Ls);
   Eigen::MatrixXcd Pminus = Eigen::MatrixXcd::Zero(Ls,Ls);

lib/qcd/action/fermion/CayleyFermion5Dcache.cc

@@ -237,13 +237,6 @@ void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &
   INSTANTIATE_DPERP(GparityWilsonImplD);
   INSTANTIATE_DPERP(ZWilsonImplF);
   INSTANTIATE_DPERP(ZWilsonImplD);
-
-  INSTANTIATE_DPERP(WilsonImplFH);
-  INSTANTIATE_DPERP(WilsonImplDF);
-  INSTANTIATE_DPERP(GparityWilsonImplFH);
-  INSTANTIATE_DPERP(GparityWilsonImplDF);
-  INSTANTIATE_DPERP(ZWilsonImplFH);
-  INSTANTIATE_DPERP(ZWilsonImplDF);
 #endif
 
 }}

lib/qcd/action/fermion/CayleyFermion5Ddense.cc

@@ -29,7 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     *************************************************************************************/
     /*  END LEGAL */
 
-#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/Eigen/Dense>
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
@@ -137,20 +137,6 @@ template void CayleyFermion5D<WilsonImplF>::MooeeInternal(const FermionField &ps
 template void CayleyFermion5D<WilsonImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 template void CayleyFermion5D<ZWilsonImplF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 template void CayleyFermion5D<ZWilsonImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-
-INSTANTIATE_DPERP(GparityWilsonImplFH);
-INSTANTIATE_DPERP(GparityWilsonImplDF);
-INSTANTIATE_DPERP(WilsonImplFH);
-INSTANTIATE_DPERP(WilsonImplDF);
-INSTANTIATE_DPERP(ZWilsonImplFH);
-INSTANTIATE_DPERP(ZWilsonImplDF);
-
-template void CayleyFermion5D<GparityWilsonImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<GparityWilsonImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<WilsonImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<WilsonImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZWilsonImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZWilsonImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 #endif
 
 }}

lib/qcd/action/fermion/CayleyFermion5Dssp.cc

@@ -37,6 +37,7 @@ namespace Grid {
 namespace QCD {
 
   // FIXME -- make a version of these routines with site loop outermost for cache reuse.
+
   // Pminus fowards
   // Pplus  backwards
 template<class Impl>  
@@ -151,13 +152,6 @@ void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &
   INSTANTIATE_DPERP(GparityWilsonImplD);
   INSTANTIATE_DPERP(ZWilsonImplF);
   INSTANTIATE_DPERP(ZWilsonImplD);
-
-  INSTANTIATE_DPERP(WilsonImplFH);
-  INSTANTIATE_DPERP(WilsonImplDF);
-  INSTANTIATE_DPERP(GparityWilsonImplFH);
-  INSTANTIATE_DPERP(GparityWilsonImplDF);
-  INSTANTIATE_DPERP(ZWilsonImplFH);
-  INSTANTIATE_DPERP(ZWilsonImplDF);
 #endif
 
 }

lib/qcd/action/fermion/CayleyFermion5Dvec.cc

@@ -808,21 +808,10 @@ INSTANTIATE_DPERP(DomainWallVec5dImplF);
 INSTANTIATE_DPERP(ZDomainWallVec5dImplD);
 INSTANTIATE_DPERP(ZDomainWallVec5dImplF);
 
-INSTANTIATE_DPERP(DomainWallVec5dImplDF);
-INSTANTIATE_DPERP(DomainWallVec5dImplFH);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplDF);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplFH);
-
 template void CayleyFermion5D<DomainWallVec5dImplF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 template void CayleyFermion5D<DomainWallVec5dImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 template void CayleyFermion5D<ZDomainWallVec5dImplF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 template void CayleyFermion5D<ZDomainWallVec5dImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 
-template void CayleyFermion5D<DomainWallVec5dImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<DomainWallVec5dImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-
-
 
 }}

lib/qcd/action/fermion/DomainWallFermion.h

@@ -68,7 +68,7 @@ namespace Grid {
 	Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);// eps is ignored for higham
 	assert(zdata->n==this->Ls);
 	
-	std::cout<<GridLogMessage << "DomainWallFermion with Ls="<<this->Ls<<std::endl;
+	//	std::cout<<GridLogMessage << "DomainWallFermion with Ls="<<this->Ls<<std::endl;
 	// Call base setter
 	this->SetCoefficientsTanh(zdata,1.0,0.0);
 

lib/qcd/action/fermion/Fermion.h

@@ -1,4 +1,4 @@
-/*************************************************************************************
+    /*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -58,7 +58,6 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 #include <Grid/qcd/action/fermion/DomainWallFermion.h>
 #include <Grid/qcd/action/fermion/MobiusFermion.h>
 #include <Grid/qcd/action/fermion/ZMobiusFermion.h>
-#include <Grid/qcd/action/fermion/SchurDiagTwoKappa.h>
 #include <Grid/qcd/action/fermion/ScaledShamirFermion.h>
 #include <Grid/qcd/action/fermion/MobiusZolotarevFermion.h>
 #include <Grid/qcd/action/fermion/ShamirZolotarevFermion.h>
@@ -89,10 +88,6 @@ typedef WilsonFermion<WilsonImplR> WilsonFermionR;
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
 typedef WilsonFermion<WilsonImplD> WilsonFermionD;
 
-typedef WilsonFermion<WilsonImplRL> WilsonFermionRL;
-typedef WilsonFermion<WilsonImplFH> WilsonFermionFH;
-typedef WilsonFermion<WilsonImplDF> WilsonFermionDF;
-
 typedef WilsonFermion<WilsonAdjImplR> WilsonAdjFermionR;
 typedef WilsonFermion<WilsonAdjImplF> WilsonAdjFermionF;
 typedef WilsonFermion<WilsonAdjImplD> WilsonAdjFermionD;
@@ -109,50 +104,27 @@ typedef DomainWallFermion<WilsonImplR> DomainWallFermionR;
 typedef DomainWallFermion<WilsonImplF> DomainWallFermionF;
 typedef DomainWallFermion<WilsonImplD> DomainWallFermionD;
 
-typedef DomainWallFermion<WilsonImplRL> DomainWallFermionRL;
-typedef DomainWallFermion<WilsonImplFH> DomainWallFermionFH;
-typedef DomainWallFermion<WilsonImplDF> DomainWallFermionDF;
-
 typedef MobiusFermion<WilsonImplR> MobiusFermionR;
 typedef MobiusFermion<WilsonImplF> MobiusFermionF;
 typedef MobiusFermion<WilsonImplD> MobiusFermionD;
 
-typedef MobiusFermion<WilsonImplRL> MobiusFermionRL;
-typedef MobiusFermion<WilsonImplFH> MobiusFermionFH;
-typedef MobiusFermion<WilsonImplDF> MobiusFermionDF;
-
 typedef ZMobiusFermion<ZWilsonImplR> ZMobiusFermionR;
 typedef ZMobiusFermion<ZWilsonImplF> ZMobiusFermionF;
 typedef ZMobiusFermion<ZWilsonImplD> ZMobiusFermionD;
 
-typedef ZMobiusFermion<ZWilsonImplRL> ZMobiusFermionRL;
-typedef ZMobiusFermion<ZWilsonImplFH> ZMobiusFermionFH;
-typedef ZMobiusFermion<ZWilsonImplDF> ZMobiusFermionDF;
-
 // Ls vectorised 
 typedef DomainWallFermion<DomainWallVec5dImplR> DomainWallFermionVec5dR;
 typedef DomainWallFermion<DomainWallVec5dImplF> DomainWallFermionVec5dF;
 typedef DomainWallFermion<DomainWallVec5dImplD> DomainWallFermionVec5dD;
 
-typedef DomainWallFermion<DomainWallVec5dImplRL> DomainWallFermionVec5dRL;
-typedef DomainWallFermion<DomainWallVec5dImplFH> DomainWallFermionVec5dFH;
-typedef DomainWallFermion<DomainWallVec5dImplDF> DomainWallFermionVec5dDF;
-
 typedef MobiusFermion<DomainWallVec5dImplR> MobiusFermionVec5dR;
 typedef MobiusFermion<DomainWallVec5dImplF> MobiusFermionVec5dF;
 typedef MobiusFermion<DomainWallVec5dImplD> MobiusFermionVec5dD;
 
-typedef MobiusFermion<DomainWallVec5dImplRL> MobiusFermionVec5dRL;
-typedef MobiusFermion<DomainWallVec5dImplFH> MobiusFermionVec5dFH;
-typedef MobiusFermion<DomainWallVec5dImplDF> MobiusFermionVec5dDF;
-
 typedef ZMobiusFermion<ZDomainWallVec5dImplR> ZMobiusFermionVec5dR;
 typedef ZMobiusFermion<ZDomainWallVec5dImplF> ZMobiusFermionVec5dF;
 typedef ZMobiusFermion<ZDomainWallVec5dImplD> ZMobiusFermionVec5dD;
 
-typedef ZMobiusFermion<ZDomainWallVec5dImplRL> ZMobiusFermionVec5dRL;
-typedef ZMobiusFermion<ZDomainWallVec5dImplFH> ZMobiusFermionVec5dFH;
-typedef ZMobiusFermion<ZDomainWallVec5dImplDF> ZMobiusFermionVec5dDF;
 
 typedef ScaledShamirFermion<WilsonImplR> ScaledShamirFermionR;
 typedef ScaledShamirFermion<WilsonImplF> ScaledShamirFermionF;
@@ -193,35 +165,17 @@ typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplD> OverlapWilsonP
 typedef WilsonFermion<GparityWilsonImplR>     GparityWilsonFermionR;
 typedef WilsonFermion<GparityWilsonImplF>     GparityWilsonFermionF;
 typedef WilsonFermion<GparityWilsonImplD>     GparityWilsonFermionD;
-
-typedef WilsonFermion<GparityWilsonImplRL>     GparityWilsonFermionRL;
-typedef WilsonFermion<GparityWilsonImplFH>     GparityWilsonFermionFH;
-typedef WilsonFermion<GparityWilsonImplDF>     GparityWilsonFermionDF;
-
 typedef DomainWallFermion<GparityWilsonImplR> GparityDomainWallFermionR;
 typedef DomainWallFermion<GparityWilsonImplF> GparityDomainWallFermionF;
 typedef DomainWallFermion<GparityWilsonImplD> GparityDomainWallFermionD;
 
-typedef DomainWallFermion<GparityWilsonImplRL> GparityDomainWallFermionRL;
-typedef DomainWallFermion<GparityWilsonImplFH> GparityDomainWallFermionFH;
-typedef DomainWallFermion<GparityWilsonImplDF> GparityDomainWallFermionDF;
-
 typedef WilsonTMFermion<GparityWilsonImplR> GparityWilsonTMFermionR;
 typedef WilsonTMFermion<GparityWilsonImplF> GparityWilsonTMFermionF;
 typedef WilsonTMFermion<GparityWilsonImplD> GparityWilsonTMFermionD;
-
-typedef WilsonTMFermion<GparityWilsonImplRL> GparityWilsonTMFermionRL;
-typedef WilsonTMFermion<GparityWilsonImplFH> GparityWilsonTMFermionFH;
-typedef WilsonTMFermion<GparityWilsonImplDF> GparityWilsonTMFermionDF;
-
 typedef MobiusFermion<GparityWilsonImplR> GparityMobiusFermionR;
 typedef MobiusFermion<GparityWilsonImplF> GparityMobiusFermionF;
 typedef MobiusFermion<GparityWilsonImplD> GparityMobiusFermionD;
 
-typedef MobiusFermion<GparityWilsonImplRL> GparityMobiusFermionRL;
-typedef MobiusFermion<GparityWilsonImplFH> GparityMobiusFermionFH;
-typedef MobiusFermion<GparityWilsonImplDF> GparityMobiusFermionDF;
-
 typedef ImprovedStaggeredFermion<StaggeredImplR> ImprovedStaggeredFermionR;
 typedef ImprovedStaggeredFermion<StaggeredImplF> ImprovedStaggeredFermionF;
 typedef ImprovedStaggeredFermion<StaggeredImplD> ImprovedStaggeredFermionD;

lib/qcd/action/fermion/FermionCore.h

@@ -55,14 +55,7 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
   template class A<ZWilsonImplF>;		\
   template class A<ZWilsonImplD>;		\
   template class A<GparityWilsonImplF>;		\
-  template class A<GparityWilsonImplD>;		\
-  template class A<WilsonImplFH>;		\
-  template class A<WilsonImplDF>;		\
-  template class A<ZWilsonImplFH>;		\
-  template class A<ZWilsonImplDF>;		\
-  template class A<GparityWilsonImplFH>;		\
-  template class A<GparityWilsonImplDF>;		
-
+  template class A<GparityWilsonImplD>;		
 
 #define AdjointFermOpTemplateInstantiate(A) \
   template class A<WilsonAdjImplF>; \
@@ -76,11 +69,7 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
   template class A<DomainWallVec5dImplF>;	\
   template class A<DomainWallVec5dImplD>;	\
   template class A<ZDomainWallVec5dImplF>;	\
-  template class A<ZDomainWallVec5dImplD>;	\
-  template class A<DomainWallVec5dImplFH>;	\
-  template class A<DomainWallVec5dImplDF>;	\
-  template class A<ZDomainWallVec5dImplFH>;	\
-  template class A<ZDomainWallVec5dImplDF>;	
+  template class A<ZDomainWallVec5dImplD>;	
 
 #define FermOpTemplateInstantiate(A) \
  FermOp4dVecTemplateInstantiate(A) \