commit d02d5edfa15ba6c04a9986a8a362a945cb38ac31
author Michele Di Giorgio <michele.digiorgio@arm.com> Fri Jan 22 09:47:04 2021 +0000
committer Michele Di Giorgio <michele.digiorgio@arm.com> Fri Jun 18 10:33:48 2021 +0000
tree ced4f49691d6c7038e347a8709b315bff59c64cf
parent b014c27ba6db9840e4a72519760d51a87a2af7e7

Integrate improved CPU depthwise convolution kernels

* Replace assembly kernels for depthwise convolution with more optimized
  ones.
* Add int8 assembly kernels.
* Fix implicit padding on optimized kernels

Resolves: COMPMID-3867, COMPMID-4361

Change-Id: I0b0867e05f61be4f368f62190d55e14d0ab3ebf2
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5622
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

SConscript

diff --git a/SConscript b/SConscript
index 94ba6d4..3e834e3 100644
--- a/SConscript
+++ b/SConscript
@@ -1,4 +1,4 @@
-# Copyright (c) 2016, 2017 Arm Limited.
+# Copyright (c) 2016-2021 Arm Limited.
 #
 # SPDX-License-Identifier: MIT
 #
@@ -252,12 +252,8 @@
 if env['neon']:
     core_files += Glob('src/core/NEON/*.cpp')
     core_files += Glob('src/core/NEON/kernels/*.cpp')
-    core_files += Glob('src/core/NEON/kernels/assembly/*.cpp')
 
     core_files += Glob('src/core/NEON/kernels/arm_gemm/*.cpp')
-    core_files += Glob('src/core/NEON/kernels/arm_conv/*.cpp')
-    core_files += Glob('src/core/NEON/kernels/arm_conv/pooling/*.cpp')
-    core_files += Glob('src/core/NEON/kernels/arm_conv/pooling/kernels/cpp_*/*.cpp')
 
     # build winograd/depthwise sources for either v7a / v8a
     core_files += Glob('src/core/NEON/kernels/convolution/*/*.cpp')
@@ -275,11 +271,22 @@
         core_files += Glob('src/core/NEON/kernels/arm_gemm/kernels/a32_*/*.cpp')
 
     if env['estate'] == '64':
+        core_files += Glob('src/core/NEON/kernels/assembly/*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/kernels/cpp_*/*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/interleaves/8b_mla.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/pooling/*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/pooling/kernels/cpp_*/*.cpp')
+
         core_files += Glob('src/core/NEON/kernels/arm_gemm/kernels/a64_*/*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/interleaves/a64_*.cpp')
+        core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_*/*.cpp')
         core_files += Glob('src/core/NEON/kernels/arm_conv/pooling/kernels/a64_*/*.cpp')
         if "sve" in env['arch'] or env['fat_binary']:
             core_files_sve += filelist['cpu']['core']['sve']['all']
             core_files_sve += Glob('src/core/NEON/kernels/arm_gemm/kernels/sve_*/*.cpp')
+            core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/interleaves/sve_*.cpp')
+            core_files += Glob('src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_*/*.cpp')
             core_files_sve += Glob('src/core/NEON/kernels/arm_conv/pooling/kernels/sve_*/*.cpp')
 
     if any(i in env['data_layout_support'] for i in ['all', 'nchw']):