diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2018-04-13 13:44:10 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:49:37 +0000 |
| commit | e7e96e09ff0d3e47797adf197aff2bc39671788c (patch) | |
| tree | b52ecdd7627bdf51b8b8da9b9553cb900460222f /src/core/NEON/kernels/arm_gemm/gemm_int8.cpp | |
| parent | 1ed1fc6d3b7d8494ce3bbc5f8b46bfde6fc586f9 (diff) | |
| download | ComputeLibrary-e7e96e09ff0d3e47797adf197aff2bc39671788c.tar.gz | |
COMPMID-1054 Update RSH's GEMM to add batch+multi support
Change-Id: Ib9d91b77f1d51976da4449fa1e6eeeffae307353
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/127876
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/gemm_int8.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_int8.cpp | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_int8.cpp b/src/core/NEON/kernels/arm_gemm/gemm_int8.cpp index 856d407cfa..f13406284c 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_int8.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_int8.cpp @@ -35,16 +35,17 @@ namespace arm_gemm { template <> UniqueGemmCommon<int8_t, int32_t> gemm<int8_t, int32_t>(const CPUInfo &ci, const unsigned int M, const unsigned int N, const unsigned int K, + const unsigned int nbatches, const unsigned int nmulti, const bool trA, const bool trB, const int32_t alpha, const int32_t beta, const int maxthreads, const bool pretransposed_hint) { if(ci.has_dotprod()) { // Dot product supporting CPUs. This family has a special version for A55r1. - return UniqueGemmCommon<int8_t, int32_t>(new GemmInterleaved<gemm_s8_12x8, int8_t, int32_t>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); + return UniqueGemmCommon<int8_t, int32_t>(new GemmInterleaved<gemm_s8_12x8, int8_t, int32_t>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); } - return UniqueGemmCommon<int8_t, int32_t>(new GemmInterleaved<gemm_s8_4x4, int8_t, int32_t>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); + return UniqueGemmCommon<int8_t, int32_t>(new GemmInterleaved<gemm_s8_4x4, int8_t, int32_t>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); // TODO: There's a better approach for A53, but it doesn't work // well on heterogeneous systems as the required data formats |