commit 32741725ac6e6c9658c51ed8585d314a1707ee8c
author Jakub Sujak <jakub.sujak@arm.com> Fri Nov 25 16:43:18 2022 +0000
committer Jakub Sujak <jakub.sujak@arm.com> Tue Dec 13 09:58:36 2022 +0000
tree a32c508ade253975a892387f7d5eba5e3f02e650
parent 5945070e4d73962f6d1e0b9eff2ab7488ee5c787

Add CLAMP operator to Dynamic Fusion interface

Add the CLAMP activation function for GPU backend with generic activation Component and TemplateWriter modules.

CLAMP is internally implemented as LU_BOUNDED_RELU activation function with the alpha and beta variables swapped.

We do NOT consider in-place computation cases in this patch.

* CLAMP operator for GPU backend.
* Activation Component and TemplateWriter for CL backend.
* TemplateWriter generates tiled kernel code.
* Supported data types: F16, F32.
* Validation tests for CLAMP operation.

Resolves: COMPMID-5519

Change-Id: Ieb097d6b1e6a7ed2b882518e88314454efb402f6
Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8762
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

Android.bp

diff --git a/Android.bp b/Android.bp
index 77ca595..f4d94f9 100644
--- a/Android.bp
+++ b/Android.bp
@@ -589,6 +589,7 @@
         "src/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.cpp",
         "src/dynamic_fusion/sketch/OperatorAttributes.cpp",
         "src/dynamic_fusion/sketch/attributes/CastAttributes.cpp",
+        "src/dynamic_fusion/sketch/attributes/ClampAttributes.cpp",
         "src/dynamic_fusion/sketch/attributes/DepthwiseConv2dAttributes.cpp",
         "src/dynamic_fusion/sketch/gpu/GpuKernelArgument.cpp",
         "src/dynamic_fusion/sketch/gpu/GpuKernelComponentGraph.cpp",
@@ -598,6 +599,7 @@
         "src/dynamic_fusion/sketch/gpu/GpuOperatorGroup.cpp",
         "src/dynamic_fusion/sketch/gpu/GpuWorkloadContext.cpp",
         "src/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.cpp",
+        "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentActivation.cpp",
         "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentCast.cpp",
         "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentDepthwiseConv2d.cpp",
         "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentDirectConv2d.cpp",
@@ -605,10 +607,12 @@
         "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentStore.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuAdd.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuCast.cpp",
+        "src/dynamic_fusion/sketch/gpu/operators/GpuClamp.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuConv2d.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuDepthwiseConv2d.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/internal/GpuElementwiseBinaryCommon.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/GpuKernelVariableTable.cpp",
+        "src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateActivation.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateCast.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateDepthwiseConv2d.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateDirectConv2d.cpp",