From a8aef2916379402e241d9f2c5e0faf3f99c860f7 Mon Sep 17 00:00:00 2001 From: Gian Marco Iodice Date: Mon, 14 May 2018 14:21:39 +0100 Subject: COMPMID-792 - Re-enabled Winograd on NEON in all graph examples. Since now the input transform can be multi-threaded, I re-ebaled Winograd in all graph examples Change-Id: I39ef78243bb47fdae135e18dcae2102af0675b3b Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/131048 Reviewed-by: Anthony Barbier Tested-by: Jenkins --- tests/validation/CL/ConvolutionLayer.cpp | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'tests/validation/CL/ConvolutionLayer.cpp') diff --git a/tests/validation/CL/ConvolutionLayer.cpp b/tests/validation/CL/ConvolutionLayer.cpp index 935a6ebefa..0c40953524 100644 --- a/tests/validation/CL/ConvolutionLayer.cpp +++ b/tests/validation/CL/ConvolutionLayer.cpp @@ -77,22 +77,25 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z 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(23U, 27U, 31U, 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), - TensorInfo(TensorShape(17U, 31U, 2U), 1, DataType::F32, 0), + TensorInfo(TensorShape(17U, 31U, 32U), 1, DataType::F32, 0), TensorInfo(TensorShape(17U, 31U, 2U), 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, 31U, 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), - TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 0), + TensorInfo(TensorShape(5U, 5U, 32U, 19U), 1, DataType::F32, 0), TensorInfo(TensorShape(5U, 5U, 2U, 19U), 1, DataType::F32, 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), + TensorInfo(TensorShape(21U, 25U, 21U, 4U), 1, DataType::F32, 0), TensorInfo(TensorShape(11U, 25U, 21U), 1, DataType::F32, 0), TensorInfo(TensorShape(11U, 12U, 16U, 4U), 1, DataType::F32, 0), TensorInfo(TensorShape(17U, 31U, 19U), 1, DataType::F32, 0), @@ -101,6 +104,7 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z framework::dataset::make("ConvInfo", { PadStrideInfo(1, 2, 1, 1), PadStrideInfo(1, 2, 1, 1), PadStrideInfo(1, 1, 0, 0), + PadStrideInfo(1, 1, 0, 0), PadStrideInfo(2, 1, 0, 0), PadStrideInfo(3, 2, 1, 0), PadStrideInfo(1, 1, 2, 2), @@ -109,6 +113,7 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z framework::dataset::make("GpuTarget", { GPUTarget::BIFROST, GPUTarget::MIDGARD, GPUTarget::G71, + GPUTarget::G71, GPUTarget::MIDGARD, GPUTarget::BIFROST, GPUTarget::BIFROST, @@ -122,12 +127,13 @@ DATA_TEST_CASE(ValidateConvolutionMethod, framework::DatasetMode::ALL, zip(zip(z Size2D(1U, 1U), Size2D(1U, 1U), Size2D(1U, 1U), + Size2D(1U, 1U), Size2D(2U, 1U), })), -framework::dataset::make("EnableFastMath", { false, false, false, false, false, true, true })), +framework::dataset::make("EnableFastMath", { false, false, false, false, false, false, true, true })), framework::dataset::make("Expected", { - ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, + ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, ConvolutionMethod::GEMM, ConvolutionMethod::WINOGRAD, ConvolutionMethod::GEMM, })), input_info, weights_info, output_info, conv_info, gpu_target, dilation, enable_fast_math, expected) { -- cgit v1.2.1