potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 15:23:05 +04:00
Code Packages Releases Wiki Activity

Merge pull request #20203 from JoeHowse:clMath-patches

Browse Source 
Fix dynamic loading of clBLAS and clFFT (formerly, clAmdBlas and clAmdFft)

* Fix dynamic loading of clBLAS and clFFT

* Update filenames and function names for clBLAS (formerly, clAmdBlas)

* Update filenames and function names for clFFT (formerly, clAmdFft)

* Uncomment teardown of clFFT; tear down clFFT in same way as clBLAS

* Fix generators for clBLAS and clFFT headers

* Update generators to parse recent clBLAS and clFFT library headers

* Update generators to be compatible with Python 3

* Re-generate OpenCV's clBLAS and clFFT headers

* Update function calls to match names in newly generated headers

* Disable (and comment on) teardown code for clBLAS and clFFT

* Renaming *clamd* files

* Renaming *clamdblas* files to *clblas*

* Renaming *clamdfft* files to *clfft*

* Update generator for CL headers

* Update generator to be compatible with Python 3
This commit is contained in:
JoeHowse
2021-06-07 17:24:27 -03:00
committed by GitHub
parent 286ec92967
commit 34183237ce
28 changed files with 2562 additions and 2880 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/core/src/matmul.dispatch.cpp
+26 -26
View File
@@ -43,7 +43,7 @@
#include "precomp.hpp"
#include "opencl_kernels_core.hpp"
#include "opencv2/core/opencl/runtime/opencl_clamdblas.hpp"
#include "opencv2/core/opencl/runtime/opencl_clblas.hpp"
#include "opencv2/core/opencl/runtime/opencl_core.hpp"
#include "intel_gpu_gemm.inl.hpp"
@@ -106,47 +106,47 @@ static bool ocl_gemm_amdblas( InputArray matA, InputArray matB, double alpha,
int offa = (int)A.offset / esz, offb = (int)B.offset / esz, offc = (int)D.offset / esz;
cl_command_queue clq = (cl_command_queue)ocl::Queue::getDefault().ptr();
clAmdBlasTranspose transA = atrans ? clAmdBlasTrans : clAmdBlasNoTrans;
clAmdBlasTranspose transB = btrans ? clAmdBlasTrans : clAmdBlasNoTrans;
clAmdBlasOrder order = clAmdBlasRowMajor;
clAmdBlasStatus status = clAmdBlasSuccess;
clblasTranspose transA = atrans ? clblasTrans : clblasNoTrans;
clblasTranspose transB = btrans ? clblasTrans : clblasNoTrans;
clblasOrder order = clblasRowMajor;
clblasStatus status = clblasSuccess;
if (type == CV_32FC1)
status = clAmdBlasSgemmEx(order, transA, transB, M, N, K,
(cl_float)alpha, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
(cl_float)beta, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
status = clblasSgemm(order, transA, transB, M, N, K,
(cl_float)alpha, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
(cl_float)beta, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
else if (type == CV_64FC1)
status = clAmdBlasDgemmEx(order, transA, transB, M, N, K,
alpha, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
status = clblasDgemm(order, transA, transB, M, N, K,
alpha, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
else if (type == CV_32FC2)
{
cl_float2 alpha_2 = { { (cl_float)alpha, 0 } };
cl_float2 beta_2 = { { (cl_float)beta, 0 } };
status = clAmdBlasCgemmEx(order, transA, transB, M, N, K,
alpha_2, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta_2, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
status = clblasCgemm(order, transA, transB, M, N, K,
alpha_2, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta_2, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
}
else if (type == CV_64FC2)
{
cl_double2 alpha_2 = { { alpha, 0 } };
cl_double2 beta_2 = { { beta, 0 } };
status = clAmdBlasZgemmEx(order, transA, transB, M, N, K,
alpha_2, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta_2, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
status = clblasZgemm(order, transA, transB, M, N, K,
alpha_2, (const cl_mem)A.handle(ACCESS_READ), offa, lda,
(const cl_mem)B.handle(ACCESS_READ), offb, ldb,
beta_2, (cl_mem)D.handle(ACCESS_RW), offc, ldc,
1, &clq, 0, NULL, NULL);
}
else
CV_Error(Error::StsUnsupportedFormat, "");
return status == clAmdBlasSuccess;
return status == clblasSuccess;
}
#endif