commit 10e88a73518932abc427e6b12d0267d5f52ac569
author Gian Marco Iodice <gianmarco.iodice@arm.com> Mon Nov 29 12:49:19 2021 +0000
committer Sheri Zhang <sheri.zhang@arm.com> Tue Jan 25 12:48:53 2022 +0000
tree 641a4e5e6d4b540d90923cc105d29b131c848806
parent 639f0f665b4346d287014adc99e14a8497aadf18

Rework gemm_mm_reshaped_only_rhs_ kernels with new macros

- Rework gemm_reshaped_rhs_only with new TILE macros
- Fuse post ops in gemm_reshaped_rhs_only

Resolves COMPMID-4890

Change-Id: I944948ecec6d08deaf3545b80cd3eeac26e44205
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6944
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>

src/gpu/cl/ClKernelLibrary.cpp

diff --git a/src/gpu/cl/ClKernelLibrary.cpp b/src/gpu/cl/ClKernelLibrary.cpp
index 92a9d9c..856d377 100644
--- a/src/gpu/cl/ClKernelLibrary.cpp
+++ b/src/gpu/cl/ClKernelLibrary.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021 Arm Limited.
+ * Copyright (c) 2016-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -281,14 +281,10 @@
     { "gemm_mm_reshaped_lhs_nt_rhs_t_texture_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped.cl" },
     { "gemm_mm_reshaped_lhs_t_rhs_nt_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped.cl" },
     { "gemm_mm_reshaped_lhs_t_rhs_nt_texture_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped.cl" },
-    { "gemm_mm_reshaped_only_rhs_nt", "common/gemm.cl" },
-    { "gemm_mm_reshaped_only_rhs_nt_texture", "common/gemm.cl" },
-    { "gemm_mm_reshaped_only_rhs_t", "common/gemm.cl" },
-    { "gemm_mm_reshaped_only_rhs_t_texture", "common/gemm.cl" },
-    { "gemm_mm_reshaped_only_rhs_nt_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.cl" },
-    { "gemm_mm_reshaped_only_rhs_nt_texture_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.cl" },
-    { "gemm_mm_reshaped_only_rhs_t_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.cl" },
-    { "gemm_mm_reshaped_only_rhs_t_texture_post_act_eltwise_op_act", "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.cl" },
+    { "gemm_mm_reshaped_only_rhs_nt", "common/gemm_reshaped_rhs_only.cl" },
+    { "gemm_mm_reshaped_only_rhs_nt_texture", "common/gemm_reshaped_rhs_only.cl" },
+    { "gemm_mm_reshaped_only_rhs_t", "common/gemm_reshaped_rhs_only.cl" },
+    { "gemm_mm_reshaped_only_rhs_t_texture", "common/gemm_reshaped_rhs_only.cl" },
     { "gemm_lc_vm_f32", "common/gemm.cl" },
     { "gemm_reshape_lhs_matrix_nt", "common/gemm_utils.cl" },
     { "gemm_reshape_lhs_matrix_t", "common/gemm_utils.cl" },
@@ -590,6 +586,10 @@
 #include "./cl_kernels/common/gemm_utils.clembed"
     },
     {
+        "common/gemm_reshaped_rhs_only.cl",
+#include "./cl_kernels/common/gemm_reshaped_rhs_only.clembed"
+    },
+    {
         "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_native.cl",
 #include "./cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_native.clembed"
     },
@@ -598,10 +598,6 @@
 #include "./cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped.clembed"
     },
     {
-        "common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.cl",
-#include "./cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped_only_rhs.clembed"
-    },
-    {
         "common/gemmlowp.cl",
 #include "./cl_kernels/common/gemmlowp.clembed"
     },