IVGCVSW-1897 : Android support for pluggable backends

Change-Id: I9143485f542d58307a604248200a4ecf839ff39a

diff --git a/Android.mk b/Android.mk
index cc1076f01ba59aa8fab17bed7205744e5c089cb2..4d0664b1cfd711e46a75a25e39c53d70b485df29 100644 (file)
--- a/Android.mk
+++ b/Android.mk
@@ -14,6 +14,11 @@ ARMNN_MAIN_HEADER_PATH := $(LOCAL_PATH)/src
 ARMNN_SOURCE_HEADER_PATH := $(LOCAL_PATH)/src/armnn
 ARMNN_SOURCE_UTILS_HEADER_PATH := $(LOCAL_PATH)/src/armnnUtils
 
+# find the backend.mk files in the backend source folders
+ARMNN_BACKEND_MAKEFILE_LOCAL_PATHS := $(wildcard $(LOCAL_PATH)/src/backends/*/backend.mk)
+ARMNN_BACKEND_MAKEFILE_PATHS := $(subst $(LOCAL_PATH),,$(ARMNN_BACKEND_MAKEFILE_LOCAL_PATHS))
+ARMNN_BACKEND_MAKEFILE_DIRS := $(subst /backend.mk,,$(ARMNN_BACKEND_MAKEFILE_PATHS))
+
 ##############
 # libarmnn.a #
 ##############
@@ -24,8 +29,20 @@ LOCAL_MODULE_TAGS := eng optional
 LOCAL_ARM_MODE := arm
 LOCAL_PROPRIETARY_MODULE := true
 
-# Mark source files as dependent on Android.mk
-LOCAL_ADDITIONAL_DEPENDENCIES := $(LOCAL_PATH)/Android.mk
+# placeholder to hold all backend source files
+ARMNN_BACKEND_SOURCES :=
+
+#
+# iterate through the backend include paths, include them into the current makefile
+# and append the sources held by the BACKEND_SOURCES variable to the ARMNN_BACKEND_SOURCES list
+#
+$(foreach mkPath,$(ARMNN_BACKEND_MAKEFILE_DIRS),\
+   $(eval include $(LOCAL_PATH)/$(mkPath)/backend.mk)\
+   $(eval ARMNN_BACKEND_SOURCES := $(ARMNN_BACKEND_SOURCES) $(patsubst %,$(mkPath)/%,$(BACKEND_SOURCES))))
+
+# Mark source files as dependent on Android.mk and backend makefiles
+LOCAL_ADDITIONAL_DEPENDENCIES := $(LOCAL_PATH)/Android.mk \
+                                 $(ARMNN_BACKEND_MAKEFILE_LOCAL_PATHS)
 
 LOCAL_EXPORT_C_INCLUDES := \
         $(ARMNN_MAIN_HEADER_PATH) \
@@ -41,131 +58,26 @@ LOCAL_C_INCLUDES := \
        $(ARMNN_SOURCE_UTILS_HEADER_PATH)
 
 LOCAL_SRC_FILES := \
+        $(ARMNN_BACKEND_SOURCES) \
         src/armnnUtils/DotSerializer.cpp \
         src/armnnUtils/FloatingPointConverter.cpp \
         src/armnnUtils/Logging.cpp \
         src/armnnUtils/Permute.cpp \
         src/backends/ArmComputeTensorUtils.cpp \
-        src/backends/ClWorkloads/ClActivationFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClActivationUint8Workload.cpp \
-        src/backends/ClWorkloads/ClAdditionWorkload.cpp \
-        src/backends/ClWorkloads/ClSubtractionWorkload.cpp \
-        src/backends/ClWorkloads/ClBaseConstantWorkload.cpp \
-        src/backends/ClWorkloads/ClBatchNormalizationFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClConstantFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClConstantUint8Workload.cpp \
-        src/backends/ClWorkloads/ClConvertFp16ToFp32Workload.cpp \
-        src/backends/ClWorkloads/ClConvertFp32ToFp16Workload.cpp \
-        src/backends/ClWorkloads/ClConvolution2dBaseWorkload.cpp \
-        src/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClConvolution2dUint8Workload.cpp \
-        src/backends/ClWorkloads/ClDepthwiseConvolutionBaseWorkload.cpp \
-        src/backends/ClWorkloads/ClDepthwiseConvolutionFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClDepthwiseConvolutionUint8Workload.cpp \
-        src/backends/ClWorkloads/ClDivisionFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClFloorFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClFullyConnectedWorkload.cpp \
-        src/backends/ClWorkloads/ClL2NormalizationFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClLstmFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClMergerFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClMergerUint8Workload.cpp \
-        src/backends/ClWorkloads/ClMultiplicationFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClNormalizationFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClPermuteWorkload.cpp \
-        src/backends/ClWorkloads/ClPooling2dBaseWorkload.cpp \
-        src/backends/ClWorkloads/ClPooling2dFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClPooling2dUint8Workload.cpp \
-        src/backends/ClWorkloads/ClReshapeFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClReshapeUint8Workload.cpp \
-        src/backends/ClWorkloads/ClResizeBilinearFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClSoftmaxBaseWorkload.cpp \
-        src/backends/ClWorkloads/ClSoftmaxFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp \
-        src/backends/ClWorkloads/ClSplitterFloatWorkload.cpp \
-        src/backends/ClWorkloads/ClSplitterUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonActivationFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonActivationUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonAdditionFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonBatchNormalizationFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonConstantFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonConstantUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonConvertFp16ToFp32Workload.cpp \
-        src/backends/NeonWorkloads/NeonConvertFp32ToFp16Workload.cpp \
-        src/backends/NeonWorkloads/NeonConvolution2dBaseWorkload.cpp \
-        src/backends/NeonWorkloads/NeonConvolution2dFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonConvolution2dUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonDepthwiseConvolutionBaseWorkload.cpp \
-        src/backends/NeonWorkloads/NeonDepthwiseConvolutionFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonDepthwiseConvolutionUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonFloorFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonFullyConnectedFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonL2NormalizationFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonLstmFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonMergerFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonMergerUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonMultiplicationFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonNormalizationFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonPermuteWorkload.cpp \
-        src/backends/NeonWorkloads/NeonPooling2dBaseWorkload.cpp \
-        src/backends/NeonWorkloads/NeonPooling2dFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonPooling2dUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonReshapeFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonReshapeUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonSoftmaxBaseWorkload.cpp \
-        src/backends/NeonWorkloads/NeonSoftmaxFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonSoftmaxUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonSplitterFloatWorkload.cpp \
-        src/backends/NeonWorkloads/NeonSplitterUint8Workload.cpp \
-        src/backends/NeonWorkloads/NeonSubtractionFloatWorkload.cpp \
         src/backends/ClWorkloadFactory.cpp \
         src/backends/ClContextControl.cpp \
         src/backends/CpuTensorHandle.cpp \
         src/backends/RefWorkloadFactory.cpp \
-        src/backends/RefWorkloads/RefMergerUint8Workload.cpp \
-        src/backends/RefWorkloads/RefResizeBilinearUint8Workload.cpp \
-        src/backends/RefWorkloads/FullyConnected.cpp \
-        src/backends/RefWorkloads/RefFullyConnectedFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefSoftmaxFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefActivationFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefBatchNormalizationUint8Workload.cpp \
-        src/backends/RefWorkloads/RefBaseConstantWorkload.cpp \
-        src/backends/RefWorkloads/RefResizeBilinearFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefBatchNormalizationFloat32Workload.cpp \
-        src/backends/RefWorkloads/Broadcast.cpp \
-        src/backends/RefWorkloads/ArithmeticFunction.cpp \
-        src/backends/RefWorkloads/RefArithmeticWorkload.cpp \
-        src/backends/RefWorkloads/RefFakeQuantizationFloat32Workload.cpp \
-        src/backends/RefWorkloads/ResizeBilinear.cpp \
-        src/backends/RefWorkloads/RefSoftmaxUint8Workload.cpp \
-        src/backends/RefWorkloads/RefDepthwiseConvolution2dFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefPooling2dUint8Workload.cpp \
-        src/backends/RefWorkloads/RefFloorFloat32Workload.cpp \
-        src/backends/RefWorkloads/ConvImpl.cpp \
-        src/backends/RefWorkloads/Activation.cpp \
-        src/backends/RefWorkloads/RefReshapeUint8Workload.cpp \
-        src/backends/RefWorkloads/RefL2NormalizationFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefLstmFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefConvolution2dFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefConvolution2dUint8Workload.cpp \
-        src/backends/RefWorkloads/RefSplitterFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefActivationUint8Workload.cpp \
-        src/backends/RefWorkloads/RefSplitterUint8Workload.cpp \
-        src/backends/RefWorkloads/RefPooling2dFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefReshapeFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefNormalizationFloat32Workload.cpp \
-        src/backends/RefWorkloads/Softmax.cpp \
-        src/backends/RefWorkloads/RefDepthwiseConvolution2dUint8Workload.cpp \
-        src/backends/RefWorkloads/RefConstantUint8Workload.cpp \
-        src/backends/RefWorkloads/RefConstantFloat32Workload.cpp \
-        src/backends/RefWorkloads/Pooling2d.cpp \
-        src/backends/RefWorkloads/RefMergerFloat32Workload.cpp \
-        src/backends/RefWorkloads/RefFullyConnectedUint8Workload.cpp \
-        src/backends/RefWorkloads/RefPermuteWorkload.cpp \
-        src/backends/RefWorkloads/RefConvertFp16ToFp32Workload.cpp \
-        src/backends/RefWorkloads/RefConvertFp32ToFp16Workload.cpp \
         src/backends/MemCopyWorkload.cpp \
         src/backends/WorkloadData.cpp \
         src/backends/WorkloadFactory.cpp \
+        src/backends/OutputHandler.cpp \
+        src/backends/StringMapping.cpp \
+        src/backends/RefLayerSupport.cpp \
+        src/backends/ClLayerSupport.cpp \
+        src/backends/NeonLayerSupport.cpp \
+        src/backends/NeonWorkloadUtils.cpp \
+        src/backends/NeonWorkloadFactory.cpp \
         src/armnn/layers/ActivationLayer.cpp \
         src/armnn/layers/AdditionLayer.cpp \
         src/armnn/layers/ArithmeticBaseLayer.cpp \
@@ -207,7 +119,6 @@ LOCAL_SRC_FILES := \
         src/armnn/NeonInterceptorScheduler.cpp \
         src/armnn/NeonTimer.cpp \
         src/armnn/Network.cpp \
-        src/backends/OutputHandler.cpp \
         src/armnn/OpenClTimer.cpp \
         src/armnn/WallClockTimer.cpp \
         src/armnn/ProfilingEvent.cpp \
@@ -217,12 +128,6 @@ LOCAL_SRC_FILES := \
         src/armnn/Utils.cpp \
         src/armnn/LayerSupport.cpp \
         src/armnn/Observable.cpp \
-        src/backends/StringMapping.cpp \
-        src/backends/RefLayerSupport.cpp \
-        src/backends/ClLayerSupport.cpp \
-        src/backends/NeonLayerSupport.cpp \
-        src/backends/NeonWorkloadUtils.cpp \
-        src/backends/NeonWorkloadFactory.cpp \
         src/armnn/memory/BaseMemoryManager.cpp \
         src/armnn/memory/BlobLifetimeManager.cpp \
         src/armnn/memory/BlobMemoryPool.cpp \

diff --git a/src/backends/ClWorkloads/backend.mk b/src/backends/ClWorkloads/backend.mk
new file mode 100644 (file)
index 0000000..1239408
--- /dev/null
+++ b/src/backends/ClWorkloads/backend.mk
@@ -0,0 +1,44 @@
+#
+# Copyright © 2017 ARM Ltd. All rights reserved.
+# SPDX-License-Identifier: MIT
+#
+
+BACKEND_SOURCES := \
+        ClActivationFloatWorkload.cpp \
+        ClActivationUint8Workload.cpp \
+        ClAdditionWorkload.cpp \
+        ClSubtractionWorkload.cpp \
+        ClBaseConstantWorkload.cpp \
+        ClBatchNormalizationFloatWorkload.cpp \
+        ClConstantFloatWorkload.cpp \
+        ClConstantUint8Workload.cpp \
+        ClConvertFp16ToFp32Workload.cpp \
+        ClConvertFp32ToFp16Workload.cpp \
+        ClConvolution2dBaseWorkload.cpp \
+        ClConvolution2dFloatWorkload.cpp \
+        ClConvolution2dUint8Workload.cpp \
+        ClDepthwiseConvolutionBaseWorkload.cpp \
+        ClDepthwiseConvolutionFloatWorkload.cpp \
+        ClDepthwiseConvolutionUint8Workload.cpp \
+        ClDivisionFloatWorkload.cpp \
+        ClFloorFloatWorkload.cpp \
+        ClFullyConnectedWorkload.cpp \
+        ClL2NormalizationFloatWorkload.cpp \
+        ClLstmFloatWorkload.cpp \
+        ClMergerFloatWorkload.cpp \
+        ClMergerUint8Workload.cpp \
+        ClMultiplicationFloatWorkload.cpp \
+        ClNormalizationFloatWorkload.cpp \
+        ClPermuteWorkload.cpp \
+        ClPooling2dBaseWorkload.cpp \
+        ClPooling2dFloatWorkload.cpp \
+        ClPooling2dUint8Workload.cpp \
+        ClReshapeFloatWorkload.cpp \
+        ClReshapeUint8Workload.cpp \
+        ClResizeBilinearFloatWorkload.cpp \
+        ClSoftmaxBaseWorkload.cpp \
+        ClSoftmaxFloatWorkload.cpp \
+        ClSoftmaxUint8Workload.cpp \
+        ClSplitterFloatWorkload.cpp \
+        ClSplitterUint8Workload.cpp
+

diff --git a/src/backends/NeonWorkloads/backend.mk b/src/backends/NeonWorkloads/backend.mk
new file mode 100644 (file)
index 0000000..06525c9
--- /dev/null
+++ b/src/backends/NeonWorkloads/backend.mk
@@ -0,0 +1,41 @@
+#
+# Copyright © 2017 ARM Ltd. All rights reserved.
+# SPDX-License-Identifier: MIT
+#
+
+BACKEND_SOURCES := \
+        NeonActivationFloatWorkload.cpp \
+        NeonActivationUint8Workload.cpp \
+        NeonAdditionFloatWorkload.cpp \
+        NeonBatchNormalizationFloatWorkload.cpp \
+        NeonConstantFloatWorkload.cpp \
+        NeonConstantUint8Workload.cpp \
+        NeonConvertFp16ToFp32Workload.cpp \
+        NeonConvertFp32ToFp16Workload.cpp \
+        NeonConvolution2dBaseWorkload.cpp \
+        NeonConvolution2dFloatWorkload.cpp \
+        NeonConvolution2dUint8Workload.cpp \
+        NeonDepthwiseConvolutionBaseWorkload.cpp \
+        NeonDepthwiseConvolutionFloatWorkload.cpp \
+        NeonDepthwiseConvolutionUint8Workload.cpp \
+        NeonFloorFloatWorkload.cpp \
+        NeonFullyConnectedFloatWorkload.cpp \
+        NeonL2NormalizationFloatWorkload.cpp \
+        NeonLstmFloatWorkload.cpp \
+        NeonMergerFloatWorkload.cpp \
+        NeonMergerUint8Workload.cpp \
+        NeonMultiplicationFloatWorkload.cpp \
+        NeonNormalizationFloatWorkload.cpp \
+        NeonPermuteWorkload.cpp \
+        NeonPooling2dBaseWorkload.cpp \
+        NeonPooling2dFloatWorkload.cpp \
+        NeonPooling2dUint8Workload.cpp \
+        NeonReshapeFloatWorkload.cpp \
+        NeonReshapeUint8Workload.cpp \
+        NeonSoftmaxBaseWorkload.cpp \
+        NeonSoftmaxFloatWorkload.cpp \
+        NeonSoftmaxUint8Workload.cpp \
+        NeonSplitterFloatWorkload.cpp \
+        NeonSplitterUint8Workload.cpp \
+        NeonSubtractionFloatWorkload.cpp
+

diff --git a/src/backends/RefWorkloads/backend.mk b/src/backends/RefWorkloads/backend.mk
new file mode 100644 (file)
index 0000000..8cde8c9
--- /dev/null
+++ b/src/backends/RefWorkloads/backend.mk
@@ -0,0 +1,49 @@
+#
+# Copyright © 2017 ARM Ltd. All rights reserved.
+# SPDX-License-Identifier: MIT
+#
+
+BACKEND_SOURCES := \
+        FullyConnected.cpp \
+        RefMergerUint8Workload.cpp \
+        RefResizeBilinearUint8Workload.cpp \
+        RefFullyConnectedFloat32Workload.cpp \
+        RefSoftmaxFloat32Workload.cpp \
+        RefActivationFloat32Workload.cpp \
+        RefBatchNormalizationUint8Workload.cpp \
+        RefBaseConstantWorkload.cpp \
+        RefResizeBilinearFloat32Workload.cpp \
+        RefBatchNormalizationFloat32Workload.cpp \
+        Broadcast.cpp \
+        ArithmeticFunction.cpp \
+        RefArithmeticWorkload.cpp \
+        RefFakeQuantizationFloat32Workload.cpp \
+        ResizeBilinear.cpp \
+        RefSoftmaxUint8Workload.cpp \
+        RefDepthwiseConvolution2dFloat32Workload.cpp \
+        RefPooling2dUint8Workload.cpp \
+        RefFloorFloat32Workload.cpp \
+        ConvImpl.cpp \
+        Activation.cpp \
+        RefReshapeUint8Workload.cpp \
+        RefL2NormalizationFloat32Workload.cpp \
+        RefLstmFloat32Workload.cpp \
+        RefConvolution2dFloat32Workload.cpp \
+        RefConvolution2dUint8Workload.cpp \
+        RefSplitterFloat32Workload.cpp \
+        RefActivationUint8Workload.cpp \
+        RefSplitterUint8Workload.cpp \
+        RefPooling2dFloat32Workload.cpp \
+        RefReshapeFloat32Workload.cpp \
+        RefNormalizationFloat32Workload.cpp \
+        Softmax.cpp \
+        RefDepthwiseConvolution2dUint8Workload.cpp \
+        RefConstantUint8Workload.cpp \
+        RefConstantFloat32Workload.cpp \
+        Pooling2d.cpp \
+        RefMergerFloat32Workload.cpp \
+        RefFullyConnectedUint8Workload.cpp \
+        RefPermuteWorkload.cpp \
+        RefConvertFp16ToFp32Workload.cpp \
+        RefConvertFp32ToFp16Workload.cpp
+