core: dispatch matmul
authorAlexander Alekhin <alexander.alekhin@intel.com>
Thu, 14 Feb 2019 13:00:17 +0000 (16:00 +0300)
committerAlexander Alekhin <alexander.alekhin@intel.com>
Mon, 18 Feb 2019 11:36:46 +0000 (14:36 +0300)
commitcd66f6e3db101bdcf1406ec3604592662c6aebef
treeacf8dc963f1d87c3411551b871e15ea8c2094749
parentfbde57dba838e65bcfced90710035c700f671f9b
core: dispatch matmul

- gemm: keep baseline only (lapack is 10x+ faster, lets reduce binary size)
- transform / distTransform
- scaleAdd (32f/64f only)
- Mahalanobis: keep baseline only (no perf tests)
- mulTransposed: keep baseline only (no perf tests)
- dot
modules/core/CMakeLists.txt
modules/core/include/opencv2/core/cv_cpu_dispatch.h
modules/core/src/matmul.dispatch.cpp
modules/core/src/matmul.simd.hpp