portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-14 13:57:07 +01:00
Code Releases Activity

USQCD benchmark

Browse Source
This commit is contained in:
Peter Boyle
2024-03-01 00:05:04 -05:00
parent 04ca065281
commit c805f86343
4 changed files with 78 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
Grid/algorithms/blas/BatchedBlas.cc Normal file
View File

@ -0,0 +1,34 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: BatchedBlas.h
Copyright (C) 2023
Author: Peter Boyle <pboyle@bnl.gov>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/GridCore.h>
#include <Grid/algorithms/blas/BatchedBlas.h>
NAMESPACE_BEGIN(Grid);
gridblasHandle_t GridBLAS::gridblasHandle;
int GridBLAS::gridblasInit;
NAMESPACE_END(Grid);

37
Grid/algorithms/blas/BatchedBlas.h
View File

@ -615,9 +615,10 @@ public:
deviceVector<ComplexD> beta_p(1);
acceleratorCopyToDevice((void *)&alpha,(void *)&alpha_p[0],sizeof(ComplexD));
acceleratorCopyToDevice((void *)&beta ,(void *)&beta_p[0],sizeof(ComplexD));
std::cout << "blasZgemmStridedBatched mnk "<<m<<","<<n<<","<<k<<" count "<<batchCount<<std::endl;
std::cout << "blasZgemmStridedBatched ld "<<lda<<","<<ldb<<","<<ldc<<std::endl;
std::cout << "blasZgemmStridedBatched sd "<<sda<<","<<sdb<<","<<sdc<<std::endl;
// std::cout << "blasZgemmStridedBatched mnk "<<m<<","<<n<<","<<k<<" count "<<batchCount<<std::endl;
// std::cout << "blasZgemmStridedBatched ld "<<lda<<","<<ldb<<","<<ldc<<std::endl;
// std::cout << "blasZgemmStridedBatched sd "<<sda<<","<<sdb<<","<<sdc<<std::endl;
#ifdef GRID_HIP
auto err = hipblasZgemmStridedBatched(gridblasHandle,
HIPBLAS_OP_N,
@ -672,21 +673,23 @@ public:
ComplexD alpha(1.0);
ComplexD beta (1.0);
RealD flops = 8.0*M*N*K*BATCH;
for(int i=0;i<10;i++){
RealD t0 = usecond();
gemmStridedBatched(M,N,K,
alpha,
&A[0], // m x k
&B[0], // k x n
beta,
&C[0], // m x n
BATCH);
synchronise();
RealD t1 = usecond();
RealD bytes = 1.0*sizeof(ComplexD)*(M*N*2+N*K+M*K)*BATCH;
flops = flops/(t1-t0)/1.e3;
int ncall=10;
RealD t0 = usecond();
for(int i=0;i<ncall;i++){
gemmStridedBatched(M,N,K,
alpha,
&A[0], // m x k
&B[0], // k x n
beta,
&C[0], // m x n
BATCH);
}
return flops;
synchronise();
RealD t1 = usecond();
RealD bytes = 1.0*sizeof(ComplexD)*(M*N*2+N*K+M*K)*BATCH;
flops = 8.0*M*N*K*BATCH*ncall;
flops = flops/(t1-t0)/1.e3;
return flops; // Returns gigaflops
}