portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-25 21:25:56 +01:00
Code Releases Activity

Gemm batched fix

Browse Source
This commit is contained in:
Peter Boyle 2024-08-27 19:24:05 +00:00
parent a465ecece9
commit 4e2a6d87c4
3 changed files with 2 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
BLAS_benchmark/BatchBlasBench.cc
View File

@ -12,9 +12,6 @@
#include <iostream>
#include <sys/time.h>
#undef GRID_SYCL
#define GRID_HIP
#undef GRID_CUDA
#ifdef GRID_HIP
#include <hipblas/hipblas.h>
@ -586,36 +583,6 @@ public:
beta,
Cmn);
}
void gemmBatched(int m,int n, int k,
RealD alpha,
deviceVector<RealD*> &Amk, // pointer list to matrices
deviceVector<RealD*> &Bkn,
RealD beta,
deviceVector<RealD*> &Cmn)
{
gemmBatched(GridBLAS_OP_N,GridBLAS_OP_N,
m,n,k,
alpha,
Amk,
Bkn,
beta,
Cmn);
}
void gemmBatched(int m,int n, int k,
RealF alpha,
deviceVector<RealF*> &Amk, // pointer list to matrices
deviceVector<RealF*> &Bkn,
RealF beta,
deviceVector<RealF*> &Cmn)
{
gemmBatched(GridBLAS_OP_N,GridBLAS_OP_N,
m,n,k,
alpha,
Amk,
Bkn,
beta,
Cmn);
}
void gemmBatched(GridBLASOperation_t OpA,
GridBLASOperation_t OpB,

2
BLAS_benchmark/compile-command
View File

@ -1,2 +1,2 @@
mpicxx -qmkl=parallel -fsycl BatchBlasBench.cc -o BatchBlasBench
mpicxx -qmkl=parallel -fsycl BatchBlasBench.cc -o BatchBlasBench -DGRID_SYCL

2
BLAS_benchmark/compile-command-frontier
View File

@ -1,5 +1,5 @@
CXX=hipcc
MPICXX=mpicxx
CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -I${MPICH_DIR}/include -L/lib64 -I/opt/cray/pe/mpich/8.1.28/ofi/gnu/12.3/include"
CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -I${MPICH_DIR}/include -L/lib64 -I/opt/cray/pe/mpich/8.1.28/ofi/gnu/12.3/include -DGRID_HIP"
LDFLAGS="-L/lib64 -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 -lhipblas -lrocblas -lmpi_gnu_123"
hipcc $CXXFLAGS $LDFLAGS BatchBlasBench.cc -o BatchBlasBench