aboutsummaryrefslogtreecommitdiff
path: root/tests/validation/CL
diff options
context:
space:
mode:
authorGian Marco Iodice <gianmarco.iodice@arm.com>2018-04-11 15:59:10 +0100
committerAnthony Barbier <anthony.barbier@arm.com>2018-11-02 16:49:37 +0000
commite52a3000d2c13bc1b66ca66b3d12b6b836982394 (patch)
tree70e8ef5ba216762604f84228805aac9bd65747b6 /tests/validation/CL
parentdd03870b63784abe499761da2b26b209b33f2db2 (diff)
downloadComputeLibrary-e52a3000d2c13bc1b66ca66b3d12b6b836982394.tar.gz
COMPMID-1026 - Add support for 4x4 output tile in CLWinogradConvolutionLayer
The performance achieved can be found at the following confluence page: https://confluence.arm.com/display/MLENG/GEMM-based+convolution+vs+Winograd-based+convolution+on+OpenCL Change-Id: I4b690cfdd4eb4ff0cd17b14fdd49ccaa1d1dc85c Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/127729 Tested-by: Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'tests/validation/CL')
-rw-r--r--tests/validation/CL/ConvolutionLayer.cpp37
-rw-r--r--tests/validation/CL/DilatedConvolutionLayer.cpp37
-rw-r--r--tests/validation/CL/Winograd.cpp14
3 files changed, 41 insertions, 47 deletions
diff --git a/tests/validation/CL/ConvolutionLayer.cpp b/tests/validation/CL/ConvolutionLayer.cpp
index b50bb94bbb..52ad417cd0 100644
--- a/tests/validation/CL/ConvolutionLayer.cpp
+++ b/tests/validation/CL/ConvolutionLayer.cpp
@@ -72,25 +72,19 @@ const auto ActivationFunctionsDataset = framework::dataset::make("ActivationInfo
TEST_SUITE(CL)
TEST_SUITE(ConvolutionLayer)
-DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(
- framework::dataset::make("InputInfo", { TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(23U, 27U, 5U, 4U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 2U, 1U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(33U, 27U, 7U, 4U), 1, DataType::F32, 0)
- }),
- framework::dataset::make("WeightsInfo", { TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(5U, 5U, 7U, 16U), 1, DataType::F16, 0)
- })),
- framework::dataset::make("BiasesInfo", { TensorInfo(TensorShape(19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(16U), 1, DataType::F32, 0)
- })),
+DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(
+ framework::dataset::make("InputInfo", { TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(23U, 27U, 5U, 4U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 2U, 1U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(33U, 27U, 7U, 4U), 1, DataType::F32, 0)
+ }),
+ framework::dataset::make("WeightsInfo", { TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(5U, 5U, 7U, 16U), 1, DataType::F16, 0)
+ })),
framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(15U, 15U, 19U), 1, DataType::F32, 0),
TensorInfo(TensorShape(15U, 15U, 19U), 1, DataType::F32, 0),
TensorInfo(TensorShape(21U, 25U, 21U, 4U), 1, DataType::F32, 0),
@@ -110,12 +104,11 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z
GPUTarget::BIFROST
})),
- framework::dataset::make("Expected", { ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM })),
- input_info, weights_info, biases_info, output_info, conv_info, gpu_target, expected)
+ framework::dataset::make("Expected", { ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM })),
+ input_info, weights_info, output_info, conv_info, gpu_target, expected)
{
ConvolutionMethod is_valid = CLConvolutionLayer::get_convolution_method(&input_info.clone()->set_is_resizable(false),
&weights_info.clone()->set_is_resizable(false),
- &biases_info.clone()->set_is_resizable(false),
&output_info.clone()->set_is_resizable(false), conv_info, WeightsInfo(), ActivationLayerInfo(), gpu_target);
ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS);
}
diff --git a/tests/validation/CL/DilatedConvolutionLayer.cpp b/tests/validation/CL/DilatedConvolutionLayer.cpp
index 532d04b431..e6a765bbe1 100644
--- a/tests/validation/CL/DilatedConvolutionLayer.cpp
+++ b/tests/validation/CL/DilatedConvolutionLayer.cpp
@@ -63,25 +63,19 @@ const auto CNNDataTypes = framework::dataset::make("DataType",
TEST_SUITE(CL)
TEST_SUITE(DilatedConvolutionLayer)
-DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zip(
- framework::dataset::make("InputInfo", { TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(23U, 27U, 5U, 4U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 2U, 1U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(33U, 27U, 7U, 4U), 1, DataType::F32, 0)
- }),
- framework::dataset::make("WeightsInfo", { TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(5U, 5U, 7U, 16U), 1, DataType::F16, 0)
- })),
- framework::dataset::make("BiasesInfo", { TensorInfo(TensorShape(19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(19U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(21U), 1, DataType::F32, 0),
- TensorInfo(TensorShape(16U), 1, DataType::F32, 0)
- })),
+DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(
+ framework::dataset::make("InputInfo", { TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(23U, 27U, 5U, 4U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 2U, 1U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(33U, 27U, 7U, 4U), 1, DataType::F32, 0)
+ }),
+ framework::dataset::make("WeightsInfo", { TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(3U, 3U, 5U, 21U), 1, DataType::F32, 0),
+ TensorInfo(TensorShape(5U, 5U, 7U, 16U), 1, DataType::F16, 0)
+ })),
framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(15U, 15U, 19U), 1, DataType::F32, 0),
TensorInfo(TensorShape(15U, 15U, 19U), 1, DataType::F32, 0),
TensorInfo(TensorShape(21U, 25U, 21U, 4U), 1, DataType::F32, 0),
@@ -107,12 +101,11 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z
Size2D(3U, 3U)
})),
- framework::dataset::make("Expected", { ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM })),
- input_info, weights_info, biases_info, output_info, conv_info, gpu_target, dilation, expected)
+ framework::dataset::make("Expected", { ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM })),
+ input_info, weights_info, output_info, conv_info, gpu_target, dilation, expected)
{
ConvolutionMethod is_valid = CLConvolutionLayer::get_convolution_method(&input_info.clone()->set_is_resizable(false),
&weights_info.clone()->set_is_resizable(false),
- &biases_info.clone()->set_is_resizable(false),
&output_info.clone()->set_is_resizable(false), conv_info, WeightsInfo(), ActivationLayerInfo(), gpu_target, dilation);
ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS);
}
diff --git a/tests/validation/CL/Winograd.cpp b/tests/validation/CL/Winograd.cpp
index 4bfe4c8f42..6e673a5f96 100644
--- a/tests/validation/CL/Winograd.cpp
+++ b/tests/validation/CL/Winograd.cpp
@@ -236,7 +236,9 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
TensorInfo(TensorShape(512U, 49U, 16U, 5U), 1, DataType::F32), // Valid
TensorInfo(TensorShape(13U, 108U, 16U, 4U), 1, DataType::F32), // Padding needed
TensorInfo(TensorShape(7U, 20U, 16U, 7U), 1, DataType::F32), // Valid
- TensorInfo(TensorShape(7U, 20U, 16U, 7U), 1, DataType::F32) // Wrong WinogradInfo
+ TensorInfo(TensorShape(7U, 20U, 16U, 7U), 1, DataType::F32), // Wrong WinogradInfo
+ TensorInfo(TensorShape(7U, 256U, 36U, 3U), 1, DataType::F32), // Valid
+ TensorInfo(TensorShape(7U, 256U, 16U, 3U), 1, DataType::F32) // Wrong number of batches
}),
framework::dataset::make("BiasInfo", {
TensorInfo(TensorShape(512U), 1, DataType::F16),
@@ -245,6 +247,8 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
TensorInfo(TensorShape(512U), 1, DataType::F32),
TensorInfo(TensorShape(13U), 1, DataType::F32),
TensorInfo(TensorShape(7U), 1, DataType::F32),
+ TensorInfo(TensorShape(7U), 1, DataType::F32),
+ TensorInfo(TensorShape(7U), 1, DataType::F32),
TensorInfo(TensorShape(7U), 1, DataType::F32)
})),
framework::dataset::make("OutputInfo", {
@@ -254,7 +258,9 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
TensorInfo(TensorShape(14U, 14U, 512U, 5U), 1, DataType::F32),
TensorInfo(TensorShape(17U, 23U, 13U, 4U), 1, DataType::F32),
TensorInfo(TensorShape(8U, 10U, 7U, 7U), 1, DataType::F32),
- TensorInfo(TensorShape(7U, 9U, 7U, 7U), 1, DataType::F32)
+ TensorInfo(TensorShape(7U, 9U, 7U, 7U), 1, DataType::F32),
+ TensorInfo(TensorShape(64U, 64U, 7U, 3U), 1, DataType::F32),
+ TensorInfo(TensorShape(64U, 64U, 7U, 3U), 1, DataType::F32)
})),
framework::dataset::make("WinogradInfo", {
WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(14U, 14U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW),
@@ -264,8 +270,10 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(17U, 23U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW),
WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(8U, 10U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW),
WinogradInfo(Size2D(2U, 3U), Size2D(3U, 3U), Size2D(8U, 10U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW),
+ WinogradInfo(Size2D(4U, 4U), Size2D(3U, 3U), Size2D(64U, 64U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW),
+ WinogradInfo(Size2D(4U, 4U), Size2D(3U, 3U), Size2D(64U, 64U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW)
})),
- framework::dataset::make("Expected", { false, false, false, true, false, true, false })),
+ framework::dataset::make("Expected", { false, false, false, true, false, true, false, true, false })),
input_info, bias_info, output_info, winograd_info, expected)
{
ARM_COMPUTE_EXPECT(bool(CLWinogradOutputTransformKernel::validate(&input_info.clone()->set_is_resizable(false), &bias_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), winograd_info)) == expected, framework::LogLevel::ERRORS);
generated by cgit v1.2.1 (git 2.23.0) at 2024-06-13 06:46:01 +0000