Modify move assignment operator to be noexcept

Add noexcept specifier to move assignment operator.

Grid/lattice/Lattice_base.h

@@ -289,7 +289,7 @@ public:
   ///////////////////////////////////////////
   // move constructor
   ///////////////////////////////////////////
-  Lattice(Lattice && r){ 
+  Lattice(Lattice && r) noexcept { 
     this->_grid = r.Grid();
     this->_odata      = r._odata;
     this->_odata_size = r._odata_size;
@@ -330,7 +330,7 @@ public:
   ///////////////////////////////////////////
   // Move assignment possible if same type
   ///////////////////////////////////////////
-  inline Lattice<vobj> & operator = (Lattice<vobj> && r){
+  inline Lattice<vobj> & operator = (Lattice<vobj> && r) noexcept {
 
     resize(0); // deletes if appropriate
     this->_grid       = r.Grid();

Grid/lattice/Lattice_slicesum_core.h

@@ -1,7 +1,6 @@
 #pragma once
 
 #if defined(GRID_CUDA)
-
 #include <cub/cub.cuh>
 #define gpucub cub
 #define gpuError_t cudaError_t
@@ -57,8 +56,13 @@ inline void sliceSumReduction_cub_small(const vobj *Data,
   //copy offsets to device
   acceleratorCopyToDeviceAsynch(&offsets[0],d_offsets,sizeof(int)*(rd+1),computeStream);
   
+#if defined(__CUDACC__) && (__CUDACC_VER_MAJOR__ >= 13)
+  #define GRID_CUB_SUM_OP ::cuda::std::plus<>{}
+#else
+  #define GRID_CUB_SUM_OP ::gpucub::Sum()
+#endif
   
-  gpuError_t gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p,d_out, rd, d_offsets, d_offsets+1, ::gpucub::Sum(), zero_init, computeStream);
+  gpuError_t gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p,d_out, rd, d_offsets, d_offsets+1, GRID_CUB_SUM_OP, zero_init, computeStream);
   if (gpuErr!=gpuSuccess) {
     std::cout << GridLogError << "Lattice_slicesum_gpu.h: Encountered error during gpucub::DeviceSegmentedReduce::Reduce (setup)! Error: " << gpuErr <<std::endl;
     exit(EXIT_FAILURE);
@@ -82,11 +86,13 @@ inline void sliceSumReduction_cub_small(const vobj *Data,
   });
   
   //issue segmented reductions in computeStream
-  gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p, d_out, rd, d_offsets, d_offsets+1,::gpucub::Sum(), zero_init, computeStream);
+  gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p, d_out, rd, d_offsets, d_offsets+1, GRID_CUB_SUM_OP, zero_init, computeStream);
   if (gpuErr!=gpuSuccess) {
     std::cout << GridLogError << "Lattice_slicesum_gpu.h: Encountered error during gpucub::DeviceSegmentedReduce::Reduce! Error: " << gpuErr <<std::endl;
     exit(EXIT_FAILURE);
   }
+
+#undef GRID_CUB_SUM_OP
   
   acceleratorCopyFromDeviceAsynch(d_out,&lvSum[0],rd*sizeof(vobj),computeStream);
   

Grid/simd/Simd.h

@@ -113,6 +113,14 @@ accelerator_inline RealD    adj(const RealD  & r){ return r; }
 accelerator_inline ComplexD adj(const ComplexD& r){ return(conjugate(r)); }
 accelerator_inline ComplexF adj(const ComplexF& r ){ return(conjugate(r)); }
 
+#if defined(GRID_CUDA) || defined(GRID_HIP)
+//Provide for convenience
+accelerator_inline std::complex<double> conjugate(const std::complex<double>& r){ return(conj(r)); }
+accelerator_inline std::complex<float>  conjugate(const std::complex<float>& r) { return(conj(r)); }
+accelerator_inline std::complex<double> adj(const std::complex<double>& r)      { return(conj(r)); }
+accelerator_inline std::complex<float>  adj(const std::complex<float>& r)       { return(conj(r)); }
+#endif
+
 accelerator_inline RealF real(const RealF  & r){ return r; }
 accelerator_inline RealD real(const RealD  & r){ return r; }
 accelerator_inline RealF real(const ComplexF  & r){ return r.real(); }

Grid/threads/Accelerator.h

@@ -96,7 +96,9 @@ void     acceleratorInit(void);
 
 #ifdef GRID_CUDA
 
+NAMESPACE_END(Grid);
 #include <cuda.h>
+NAMESPACE_BEGIN(Grid);
 
 #ifdef __CUDA_ARCH__
 #define GRID_SIMT

systems/Perlmutter/config-command

@@ -1,12 +1,12 @@
 DIR=`pwd`
-PREFIX=$HOME/DDHMC/Grid/systems/Prerequisites/install/
+
 ../../configure \
     --enable-comms=mpi \
     --enable-simd=GPU \
     --enable-shm=nvlink \
     --enable-gen-simd-width=64 \
-    --with-gmp=$PREFIX \
+    --with-gmp=$GMP \
-    --with-mpfr=$PREFIX \
+    --with-mpfr=$MPFR \
     --enable-accelerator=cuda \
     --disable-fermion-reps \
     --disable-unified \

systems/Perlmutter/sourceme.sh

@@ -1,4 +1,6 @@
-
 export CRAY_ACCEL_TARGET=nvidia80
+source /global/homes/p/pboyle/spack/share/spack/setup-env.sh
+export MPFR=`spack find --paths mpfr | grep mpfr | cut -c 13-`
+export GMP=`spack find --paths gmp   | grep gmp  | cut -c 12-`
 
-module load PrgEnv-gnu cpe-cuda cudatoolkit/11.4
+module load PrgEnv-gnu cpe-cuda cudatoolkit/12.0

systems/mac-arm/config-command-mpi

@@ -3,7 +3,10 @@
 CXX=mpicxx ../../configure \
 	   --enable-simd=GEN \
 	   --enable-comms=mpi-auto \
-	   --enable-Sp=yes \
+	   --enable-Sp=no \
+	   --disable-fermion-reps \
+	   --disable-gparity \
+	   --with-fftw=$FFTW \
 	   --enable-unified=yes \
 	   --prefix /Users/peterboyle/QCD/vtk/Grid/install \
 	   --with-lime=$CLIME \

systems/mac-arm/sourceme.sh

@@ -0,0 +1,11 @@
+source /Users/peterboyle/QCD//Spack/spack//share/spack/setup-env.sh
+
+export FFTW=`spack find --paths fftw    | grep ^fftw   | awk '{print $2}' `
+#export HDF5=`spack find --paths hdf5+cxx   | grep ^hdf5   | awk '{print $2}' `
+export CLIME=`spack find --paths c-lime | grep ^c-lime | awk '{print $2}' `
+export MPFR=`spack find --paths mpfr    | grep ^mpfr  | awk '{print $2}' `
+export OPENSSL=`spack find --paths openssl | grep openssl | awk  '{print $2}' `
+export GMP=`spack find --paths gmp      | grep ^gmp | awk '{print $2}' `
+
+export LD_LIBRARY_PATH=$MPFR/lib:$LD_LIBRARY_PATH
+export LD_LIBRARY_PATH=$GMP/lib:$LD_LIBRARY_PATH