portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-19 16:55:37 +01:00
Code Releases Activity

Booster update

Browse Source
This commit is contained in:
Peter Boyle 2024-03-06 19:03:45 +01:00
parent 228bbb9d81
commit 7e5bd46dd3
4 changed files with 13 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
Grid/algorithms/blas/BatchedBlas.h
View File

@ -31,7 +31,7 @@ Author: Peter Boyle <pboyle@bnl.gov>
#include <hipblas/hipblas.h>
#endif
#ifdef GRID_CUDA
#include <hipblas/hipblas.h>
#include <cublas_v2.h>
#endif
#ifdef GRID_SYCL
#include <oneapi/mkl.hpp>
@ -51,7 +51,7 @@ NAMESPACE_BEGIN(Grid);
typedef hipblasHandle_t gridblasHandle_t;
#endif
#ifdef GRID_CUDA
typedef cudablasHandle_t gridblasHandle_t;
typedef cublasHandle_t gridblasHandle_t;
#endif
#ifdef GRID_SYCL
typedef cl::sycl::queue *gridblasHandle_t;
@ -78,6 +78,7 @@ public:
#ifdef GRID_CUDA
std::cout << "cublasCreate"<<std::endl;
cublasCreate(&gridblasHandle);
cublasSetPointerMode(gridblasHandle, CUBLAS_POINTER_MODE_DEVICE);
#endif
#ifdef GRID_HIP
std::cout << "hipblasCreate"<<std::endl;

12
Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h
View File

@ -280,20 +280,16 @@ void StaggeredKernels<Impl>::DhopImproved(StencilImpl &st, LebesgueOrder &lo,
if( interior && exterior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGeneric,1); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHand,1); return;}
#ifndef GRID_CUDA
if (Opt == OptInlineAsm ) { ASM_CALL(DhopSiteAsm); return;}
#endif
} else if( interior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGenericInt,1); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHandInt,1); return;}
#endif
} else if( exterior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGenericExt,1); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHandExt,1); return;}
#endif
}
assert(0 && " Kernel optimisation case not covered ");
}
@ -322,19 +318,13 @@ void StaggeredKernels<Impl>::DhopNaive(StencilImpl &st, LebesgueOrder &lo,
if( interior && exterior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGeneric,0); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHand,0); return;}
#endif
} else if( interior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGenericInt,0); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHandInt,0); return;}
#endif
} else if( exterior ) {
if (Opt == OptGeneric ) { KERNEL_CALL(DhopSiteGenericExt,0); return;}
#ifndef GRID_CUDA
if (Opt == OptHandUnroll ) { KERNEL_CALL(DhopSiteHandExt,0); return;}
#endif
}
}

6
systems/Booster/config-command
View File

@ -5,10 +5,12 @@ LIME=/p/home/jusers/boyle2/juwels/gm2dwf/boyle/
--enable-gen-simd-width=64 \
--enable-shm=nvlink \
--enable-accelerator=cuda \
--disable-gparity \
--disable-fermion-reps \
--with-lime=$LIME \
--disable-accelerator-cshift \
--enable-accelerator-cshift \
--disable-unified \
CXX=nvcc \
LDFLAGS="-cudart shared " \
CXXFLAGS="-ccbin mpicxx -gencode arch=compute_80,code=sm_80 -std=c++14 -cudart shared"
CXXFLAGS="-ccbin mpicxx -gencode arch=compute_80,code=sm_80 -std=c++17 -cudart shared -lcublas"

10
systems/Booster/sourceme.sh
View File

@ -1,5 +1,5 @@
module load GCC/9.3.0
module load GMP/6.2.0
module load MPFR/4.1.0
module load OpenMPI/4.1.0rc1
module load CUDA/11.3
module load GCC
module load GMP
module load MPFR
module load OpenMPI
module load CUDA