From 7359a87040c7a901619de21701f540dd5a9a960c Mon Sep 17 00:00:00 2001 From: Jakub Sujak Date: Thu, 5 Jan 2023 14:24:13 +0000 Subject: Add Multiplication operator (FP only) to Dynamic Fusion Interface Note: we use a separate test fixture for Multiplication op instead of reusing ElementwiseBinaryFixture to avoid exposing the internal enum ElementwiseOp to the public utils/TypePrinters.h as required by the data test case macros to print the test data. We also do not consider modifying the enum ArithmeticOp in the standard interface to include MUL without an implementation. Future work should consider refactoring this test fixture into the ElementwiseBinaryFixture to reduce the total number of fixtures/code duplication. Resolves: COMPMID-5779 Change-Id: I84207658ce0407095b028fca0ab7bfa2950255ec Signed-off-by: Jakub Sujak Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9013 Reviewed-by: Viet-Hoa Do Reviewed-by: SiCong Li Reviewed-by: Gunes Bayir Tested-by: Arm Jenkins Comments-Addressed: Arm Jenkins Benchmark: Arm Jenkins --- tests/validation/dynamic_fusion/gpu/cl/Add.cpp | 157 ++++++++++------- tests/validation/dynamic_fusion/gpu/cl/Mul.cpp | 223 +++++++++++++++++++++++++ 2 files changed, 323 insertions(+), 57 deletions(-) create mode 100644 tests/validation/dynamic_fusion/gpu/cl/Mul.cpp (limited to 'tests/validation/dynamic_fusion') diff --git a/tests/validation/dynamic_fusion/gpu/cl/Add.cpp b/tests/validation/dynamic_fusion/gpu/cl/Add.cpp index 0385407ad2..52ba0520ad 100644 --- a/tests/validation/dynamic_fusion/gpu/cl/Add.cpp +++ b/tests/validation/dynamic_fusion/gpu/cl/Add.cpp @@ -34,7 +34,6 @@ #include "tests/datasets/DynamicFusionDataset.h" #include "tests/datasets/ShapeDatasets.h" #include "tests/validation/fixtures/dynamic_fusion/gpu/cl/ElementwiseBinaryFixture.h" -#include "tests/validation/reference/ElementwiseOperations.h" namespace arm_compute { @@ -42,6 +41,14 @@ namespace test { namespace validation { +/* Synced with tests/validation/CL/ArithmeticAddition.cpp from the standard interface. + * + * Difference | Why the difference + * No quantized tests | Not supported yet + * No in place tests | Not supported yet + * No activation tests | Not needed in dynamic fusion interface + * + */ TEST_SUITE(CL) TEST_SUITE(DYNAMIC_FUSION) TEST_SUITE(ADD) @@ -49,29 +56,33 @@ TEST_SUITE(ADD) // *INDENT-OFF* // clang-format off DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip( - framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + framework::dataset::make("LhsInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Invalid data type combination TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16), // S16 is valid data type for Add TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S32), // S32 is valid data type for Add TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Mismatching shapes TensorInfo(TensorShape(32U, 1U, 1U), 1, DataType::F32), // Broadcasting allowed for lhs + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8), // Unsupported data type QASYMM8 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8_SIGNED), // Unsupported data type QASYMM8 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), TensorInfo(TensorShape(15U, 23U, 3U), 1, DataType::F32), // Broadcast Y dimension is not allowed TensorInfo(TensorShape( 3U, 8U, 9U), 1, DataType::S16), // Broadcast Z dimension is not allowed TensorInfo(TensorShape(32U, 13U, 2U, 2), 1, DataType::F32), // Batching is allowed }), - framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + framework::dataset::make("RhsInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16), TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16), TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S32), TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32), TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8), // Unsupported data type QASYMM8 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8_SIGNED), // Unsupported data type QASYMM8 TensorInfo(TensorShape(32U, 1U, 1U), 1, DataType::F32), // Broadcasting allowed for rhs TensorInfo(TensorShape(15U, 1U, 3U), 1, DataType::F32), TensorInfo(TensorShape( 3U, 8U, 1U), 1, DataType::S16), TensorInfo(TensorShape(32U, 13U, 2U, 2), 1, DataType::F32), })), - framework::dataset::make("Expected", { true, false, true, true, false, true, true, false, false, true})), + framework::dataset::make("Expected", { true, false, true, true, false, true, false, false, true, false, false, true})), input1_info, input2_info, expected) { // Create a new workload sketch @@ -79,7 +90,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip( auto gpu_ctx = GpuWorkloadContext{ &cl_compile_ctx }; GpuWorkloadSketch sketch{ &gpu_ctx }; - // Fuse Elementwise Add + // Validate Elementwise Add auto lhs_info = sketch.create_tensor_info(input1_info); auto rhs_info = sketch.create_tensor_info(input2_info); @@ -89,59 +100,73 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip( // clang-format on // *INDENT-ON* -RelativeTolerance tolerance_f32(0.01f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */ -RelativeTolerance tolerance_f16(half_float::half(0.1)); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F16 */ -constexpr float tolerance_num = 0.01f; /**< Tolerance number */ +constexpr AbsoluteTolerance tolerance_f16(0.0001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F16 */ +constexpr AbsoluteTolerance tolerance_f32(0.0001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */ +constexpr float tolerance_num = 0.0001f; /**< Tolerance number */ template -using DynamicFusionAddOpFixture = DynamicFusionGpuElementwiseBinaryOneOpValidationFixture; +using DynamicFusionCLAddFixture = DynamicFusionGpuElementwiseBinaryOneOpValidationFixture; template -using DynamicFusionAddOpBroadcastFixture = DynamicFusionGpuElementwiseBinaryBroadcastOneOpValidationFixture; +using DynamicFusionCLAddBroadcastFixture = DynamicFusionGpuElementwiseBinaryBroadcastOneOpValidationFixture; template -using DynamicFusionGpuFuseTwoAddOpsFixture = DynamicFusionGpuElementwiseBinaryTwoOpsValidationFixture; +using DynamicFusionCLAddTwoOpsFixture = DynamicFusionGpuElementwiseBinaryTwoOpsValidationFixture; TEST_SUITE(FP32) -FIXTURE_DATA_TEST_CASE(RunSmallOneOp, DynamicFusionAddOpFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine( - framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::SmallShapesNoBatches()), - framework::dataset::make("DataType", { DataType::F32 })), - framework::dataset::make("InPlace", { false, true }))) +FIXTURE_DATA_TEST_CASE(RunSmallOneOp, + DynamicFusionCLAddFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::SmallShapes()), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32); } -FIXTURE_DATA_TEST_CASE(RunLargeOneOp, DynamicFusionAddOpFixture, framework::DatasetMode::NIGHTLY, combine(combine(combine( - framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::LargeShapesNoBatches()), - framework::dataset::make("DataType", { DataType::F32 })), - framework::dataset::make("InPlace", { false, true }))) +FIXTURE_DATA_TEST_CASE(RunLargeOneOp, + DynamicFusionCLAddFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::LargeShapes()), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32); } -FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, DynamicFusionAddOpBroadcastFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::TemporaryLimitedSmallShapesBroadcast()), - framework::dataset::make("DataType", { DataType::F32 })), - framework::dataset::make("InPlace", { false, true }))) +FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, + DynamicFusionCLAddBroadcastFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::TemporaryLimitedSmallShapesBroadcast()), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32); } -FIXTURE_DATA_TEST_CASE(RunLargeBroadcastOneOp, DynamicFusionAddOpBroadcastFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::TemporaryLimitedLargeShapesBroadcast()), - framework::dataset::make("DataType", { DataType::F32 })), - framework::dataset::make("InPlace", { false, true }))) +FIXTURE_DATA_TEST_CASE(RunLargeBroadcastOneOp, + DynamicFusionCLAddBroadcastFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::TemporaryLimitedLargeShapesBroadcast()), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32); } -FIXTURE_DATA_TEST_CASE(RunSmallTwoOps, DynamicFusionGpuFuseTwoAddOpsFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::DynamicFusionElementwiseBinaryTwoOpsSmallShapes()), - framework::dataset::make("DataType", { DataType::F32 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunSmallTwoOps, + DynamicFusionCLAddTwoOpsFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::DynamicFusionElementwiseBinaryTwoOpsSmallShapes()), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false })), + framework::dataset::make("FuseTwoOps", { true }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32); @@ -149,19 +174,25 @@ FIXTURE_DATA_TEST_CASE(RunSmallTwoOps, DynamicFusionGpuFuseTwoAddOpsFixture, framework::DatasetMode::ALL, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::SmallShapesNoBatches()), - framework::dataset::make("DataType", { DataType::F16 })), - framework::dataset::make("InPlace", { false, true }))) +FIXTURE_DATA_TEST_CASE(RunSmallOneOp, + DynamicFusionCLAddFixture, + framework::DatasetMode::ALL, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::SmallShapes()), + framework::dataset::make("DataType", { DataType::F16 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32, tolerance_num); } -FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, DynamicFusionAddOpBroadcastFixture, framework::DatasetMode::ALL, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::TemporaryLimitedSmallShapesBroadcast()), - framework::dataset::make("DataType", { DataType::F16 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, + DynamicFusionCLAddBroadcastFixture, + framework::DatasetMode::ALL, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::TemporaryLimitedSmallShapesBroadcast()), + framework::dataset::make("DataType", { DataType::F16 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference, tolerance_f32, tolerance_num); @@ -170,10 +201,13 @@ FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, DynamicFusionAddOpBroadcastFixtur TEST_SUITE_END() // FP16 TEST_SUITE(S32) -FIXTURE_DATA_TEST_CASE(RunSmall, DynamicFusionAddOpFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::SmallShapesNoBatches()), - framework::dataset::make("DataType", { DataType::S32 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunSmall, + DynamicFusionCLAddFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::SmallShapes()), + framework::dataset::make("DataType", { DataType::S32 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference); @@ -181,18 +215,24 @@ FIXTURE_DATA_TEST_CASE(RunSmall, DynamicFusionAddOpFixture, framework:: TEST_SUITE_END() // S32 TEST_SUITE(S16) -FIXTURE_DATA_TEST_CASE(RunSmall, DynamicFusionAddOpFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::SmallShapesNoBatches()), - framework::dataset::make("DataType", { DataType::S16 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunSmall, + DynamicFusionCLAddFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::SmallShapes()), + framework::dataset::make("DataType", { DataType::S16 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference); } -FIXTURE_DATA_TEST_CASE(RunLarge, DynamicFusionAddOpFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::LargeShapesNoBatches()), - framework::dataset::make("DataType", { DataType::S16 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunLarge, + DynamicFusionCLAddFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::LargeShapes()), + framework::dataset::make("DataType", { DataType::S16 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference); @@ -200,10 +240,13 @@ FIXTURE_DATA_TEST_CASE(RunLarge, DynamicFusionAddOpFixture, framework:: TEST_SUITE_END() // S16 TEST_SUITE(U8) -FIXTURE_DATA_TEST_CASE(RunSmall, DynamicFusionAddOpFixture, framework::DatasetMode::PRECOMMIT, combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), - datasets::SmallShapesNoBatches()), - framework::dataset::make("DataType", { DataType::U8 })), - framework::dataset::make("InPlace", { false }))) +FIXTURE_DATA_TEST_CASE(RunSmall, + DynamicFusionCLAddFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(framework::dataset::make("ElementwiseOp", { ArithmeticOperation::ADD }), + datasets::SmallShapes()), + framework::dataset::make("DataType", { DataType::U8 })), + framework::dataset::make("InPlace", { false }))) { // Validate output validate(CLAccessor(_target), _reference); diff --git a/tests/validation/dynamic_fusion/gpu/cl/Mul.cpp b/tests/validation/dynamic_fusion/gpu/cl/Mul.cpp new file mode 100644 index 0000000000..a9e8f9c15f --- /dev/null +++ b/tests/validation/dynamic_fusion/gpu/cl/Mul.cpp @@ -0,0 +1,223 @@ +/* + * Copyright (c) 2023 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h" +#include "arm_compute/dynamic_fusion/sketch/gpu/operators/GpuMul.h" + +#include "tests/CL/CLAccessor.h" +#include "tests/framework/Fixture.h" +#include "tests/framework/Macros.h" +#include "tests/framework/datasets/Datasets.h" +#include "tests/validation/Validation.h" + +#include "tests/datasets/DynamicFusionDataset.h" +#include "tests/datasets/ShapeDatasets.h" +#include "tests/validation/fixtures/dynamic_fusion/operators/MulFixture.h" + +namespace arm_compute +{ +namespace test +{ +namespace validation +{ +/* Synced with tests/validation/CL/PixelwiseMultiplication.cpp from the standard interface. + * + * Difference | Why the difference + * No integer tests | Not supported yet + * No quantized tests | Not supported yet + * No convert policy tests | Not needed as convert policy is ignored by floating types + * No scale tests | Not supported yet + * No rounding modes tests | Not supported yet + * No in place tests | Not supported yet + * No activation tests | Not needed in dynamic fusion interface + * + */ +namespace +{ +constexpr AbsoluteTolerance tolerance_f16(0.0001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F16 */ +constexpr AbsoluteTolerance tolerance_f32(0.0001f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */ +} // namespace +TEST_SUITE(CL) +TEST_SUITE(DYNAMIC_FUSION) +TEST_SUITE(MUL) + +// *INDENT-OFF* +// clang-format off +DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip( + framework::dataset::make("LhsInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Invalid data type combination + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8), // Unsupported data type U8 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S8), // Unsupported data type S8 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16), // Unsupported data type S16 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S32), // Unsupported data type S32 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8), // Unsupported data type QASYMM8 + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8_SIGNED), // Unsupported data type QASYMM8_SIGNED + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Mismatching shapes + TensorInfo(TensorShape(32U, 1U, 1U), 1, DataType::F32), // Broadcasting allowed for lhs + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(15U, 23U, 3U), 1, DataType::F32), // Broadcast Y dimension is not allowed + TensorInfo(TensorShape( 3U, 8U, 9U), 1, DataType::F32), // Broadcast Z dimension is not allowed + TensorInfo(TensorShape(32U, 13U, 2U, 2), 1, DataType::F32), // Batching is allowed + }), + framework::dataset::make("RhsInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S8), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S32), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8_SIGNED), + TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 1U, 1U), 1, DataType::F32), // Broadcasting allowed for rhs + TensorInfo(TensorShape(15U, 1U, 3U), 1, DataType::F32), + TensorInfo(TensorShape( 3U, 8U, 1U), 1, DataType::F32), + TensorInfo(TensorShape(32U, 13U, 2U, 2), 1, DataType::F32), + })), + framework::dataset::make("Expected", { true, true, false, false, false, false, false, false, false, false, true, true, false, false, true })), + input1_info, input2_info, expected) +{ + // Create a new workload sketch + auto cl_compile_ctx = CLKernelLibrary::get().get_compile_context(); + auto gpu_ctx = GpuWorkloadContext{ &cl_compile_ctx }; + GpuWorkloadSketch sketch{ &gpu_ctx }; + + // Validate Elementwise Mul + auto lhs_info = sketch.create_tensor_info(input1_info); + auto rhs_info = sketch.create_tensor_info(input2_info); + + bool res = bool(GpuMul::validate_op(sketch, &lhs_info, &rhs_info)); + ARM_COMPUTE_EXPECT(res == expected, framework::LogLevel::ERRORS); +} +// clang-format on +// *INDENT-ON* + +template +using DynamicFusionCLMulFixture = DynamicFusionMulOneOpValidationFixture; +template +using DynamicFusionCLMulBroadcastFixture = DynamicFusionMulBroadcastValidationFixture; +template +using DynamicFusionCLMulTwoOpsFixture = DynamicFusionMulTwoOpsValidationFixture; + +TEST_SUITE(F16) +FIXTURE_DATA_TEST_CASE(RunSmallOneOp, + DynamicFusionCLMulFixture, + framework::DatasetMode::ALL, + combine(combine(datasets::SmallShapes(), + framework::dataset::make("DataType", { DataType::F16 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f16); +} + +FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, + DynamicFusionCLMulBroadcastFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(datasets::TemporaryLimitedSmallShapesBroadcast(), + framework::dataset::make("DataType", { DataType::F16 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f16); +} + +FIXTURE_DATA_TEST_CASE(RunLargeBroadcastOneOp, + DynamicFusionCLMulBroadcastFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(datasets::TemporaryLimitedLargeShapesBroadcast(), + framework::dataset::make("DataType", { DataType::F16 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f16); +} +TEST_SUITE_END() // F16 + +TEST_SUITE(F32) +FIXTURE_DATA_TEST_CASE(RunSmallOneOp, + DynamicFusionCLMulFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(datasets::SmallShapes(), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f32); +} + +FIXTURE_DATA_TEST_CASE(RunLargeOneOp, + DynamicFusionCLMulFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(datasets::LargeShapes(), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f32); +} + +FIXTURE_DATA_TEST_CASE(RunSmallBroadcastOneOp, + DynamicFusionCLMulBroadcastFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(datasets::TemporaryLimitedSmallShapesBroadcast(), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f32); +} + +FIXTURE_DATA_TEST_CASE(RunLargeBroadcastOneOp, + DynamicFusionCLMulBroadcastFixture, + framework::DatasetMode::NIGHTLY, + combine(combine(datasets::TemporaryLimitedLargeShapesBroadcast(), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f32); +} + +FIXTURE_DATA_TEST_CASE(RunSmallTwoOps, + DynamicFusionCLMulTwoOpsFixture, + framework::DatasetMode::PRECOMMIT, + combine(combine(combine(datasets::DynamicFusionElementwiseBinaryTwoOpsSmallShapes(), + framework::dataset::make("DataType", { DataType::F32 })), + framework::dataset::make("InPlace", { false })), + framework::dataset::make("FuseTwoOps", { true }))) +{ + // Validate output + validate(CLAccessor(_target), _reference, tolerance_f32); +} +TEST_SUITE_END() // F32 + +TEST_SUITE_END() // MUL +TEST_SUITE_END() // DYNAMIC_FUSION +TEST_SUITE_END() // CL +} // namespace validation +} // namespace test +} // namespace arm_compute -- cgit v1.2.1