commit 4bfc70e31766587c951204c93a127a486e007d0c
author Gunes Bayir <gunes.bayir@arm.com> Fri Dec 10 16:17:56 2021 +0000
committer Gunes Bayir <gunes.bayir@arm.com> Wed Jul 13 14:47:44 2022 +0000
tree 198b2150c43b14c571c100b8dfa0d3aaa4c968d0
parent 29cab36ddd73c174bf6b2de453663aa49c1cc576

Add Gemm MMUL Reshaped Only Rhs Support for FP32/FP16

This patch introduces a GEMM routine that is optimized for Arm(R) Mali(TM)-G715 and Arm(R) Mali(TM)-G615

Resolves: COMPMID-5216
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: I2e5d7806f5904347185bb3e250f73d73d6669dba
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7914
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

src/gpu/cl/ClKernelLibrary.cpp

diff --git a/src/gpu/cl/ClKernelLibrary.cpp b/src/gpu/cl/ClKernelLibrary.cpp
index 1bf7f2b..52661d6 100644
--- a/src/gpu/cl/ClKernelLibrary.cpp
+++ b/src/gpu/cl/ClKernelLibrary.cpp
@@ -272,6 +272,8 @@
     { "gemm_mv", "common/gemv.cl" },
     { "gemm_mv_quantized", "common/gemv.cl" },
     { "gemm_mm_native", "common/gemm.cl" },
+    { "gemm_mm_reshaped_only_rhs_nt_mmul", "common/gemm_reshaped_only_rhs_mmul.cl" },
+    { "gemm_mm_reshaped_only_rhs_nt_mmul_texture", "common/gemm_reshaped_only_rhs_mmul.cl" },
     { "gemm_mm_native_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_native.cl" },
     { "gemm_mm_reshaped_lhs_nt_rhs_t", "common/gemm.cl" },
     { "gemm_mm_reshaped_lhs_nt_rhs_t_texture", "common/gemm.cl" },
@@ -584,6 +586,10 @@
 #include "./cl_kernels/common/gemm.clembed"
     },
     {
+        "common/gemm_reshaped_only_rhs_mmul.cl",
+#include "./cl_kernels/common/gemm_reshaped_only_rhs_mmul.clembed"
+    },
+    {
         "common/gemm_utils.cl",
 #include "./cl_kernels/common/gemm_utils.clembed"
     },