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_fp16.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_fp16.cpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp index 484892dc81..d1180b13cb 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp @@ -38,6 +38,7 @@ namespace arm_gemm { template <> UniqueGemmCommon<__fp16, __fp16> gemm(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 __fp16 alpha, const __fp16 beta, const int maxthreads, const bool pretransposed_hint) { @@ -56,15 +57,15 @@ UniqueGemmCommon<__fp16, __fp16> gemm(const CPUInfo &ci, const unsigned int M, c // If FP16 is supported, use it. if(use_fp16) { - return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<hgemm_24x8, __fp16, __fp16>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); + return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<hgemm_24x8, __fp16, __fp16>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); } #endif // Fallback to using the blocked SGEMM kernel. - return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_12x8, __fp16, __fp16>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); + return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_12x8, __fp16, __fp16>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); #else // For AArch32, only support the SGEMM route for now. - return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_8x6, __fp16, __fp16>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); + return UniqueGemmCommon<__fp16, __fp16>(new GemmInterleaved<sgemm_8x6, __fp16, __fp16>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint)); #endif } |