From 146138378c1587b7297d245b7177641315f6180b Mon Sep 17 00:00:00 2001 From: Georgios Pinitas Date: Fri, 1 Mar 2019 19:07:11 +0000 Subject: COMPMID-1995: Update RSH GEMM assembly kernels. -Updates u8/s8 hybrid dot product kernels to work for any N and any K >=16. -Adds hybrid FP32 kernels with generic and A55 variants. -Adds SVE native kernels for fp16/u8/s8. Change-Id: Ifc0eaba9e3c8ea5bb19d334e870e1b39e4e7e728 Signed-off-by: Georgios Pinitas Reviewed-on: https://review.mlplatform.org/c/863 Tested-by: Arm Jenkins Reviewed-by: Michele Di Giorgio Comments-Addressed: Arm Jenkins Reviewed-by: Michalis Spyrou --- .../core/NEON/kernels/assembly/arm_gemm.hpp | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'arm_compute/core/NEON/kernels/assembly/arm_gemm.hpp') diff --git a/arm_compute/core/NEON/kernels/assembly/arm_gemm.hpp b/arm_compute/core/NEON/kernels/assembly/arm_gemm.hpp index 26c1f3df89..1b511ba79a 100644 --- a/arm_compute/core/NEON/kernels/assembly/arm_gemm.hpp +++ b/arm_compute/core/NEON/kernels/assembly/arm_gemm.hpp @@ -42,13 +42,13 @@ enum class GemmMethod GEMM_INTERLEAVED }; - struct KernelDescription { - GemmMethod method = GemmMethod::DEFAULT; - std::string name = ""; + GemmMethod method = GemmMethod::DEFAULT; + std::string name = ""; + bool is_default = false; - KernelDescription(GemmMethod m, std::string n) : method(m), name(n) { } + KernelDescription(GemmMethod m, std::string n, bool d=false) : method(m), name(n), is_default(d) { } KernelDescription() { } }; @@ -166,16 +166,16 @@ UniqueGemmCommon gemm(const CPUInfo &ci, } template -std::vector get_compatible_kernels(const GemmArgs &args); +std::vector get_compatible_kernels(const GemmArgs &args); template -std::vector get_compatible_kernels(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 Tret alpha, const Tret beta, - const int maxthreads, const bool pretransposed_hint) +std::vector get_compatible_kernels(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 Tret alpha, const Tret beta, + const int maxthreads, const bool pretransposed_hint, GemmConfig *cfg=nullptr) { - GemmArgs args(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint); + GemmArgs args(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint, cfg); return get_compatible_kernels(args); } -- cgit v1.2.1