diff options
| author | Jakub Sujak <jakub.sujak@arm.com> | 2023-10-05 10:20:59 +0100 |
|---|---|---|
| committer | Jakub Sujak <jakub.sujak@arm.com> | 2023-10-05 20:38:57 +0000 |
| commit | a23b4686a091a7960a4b336d0fe53f15db4ae538 (patch) | |
| tree | 1ad99168638177ccbf4f7c991ac539b5dd270eca /tests/validation/CL | |
| parent | 3831111db26d791cade87fd2d7fe2663e2ceb4a6 (diff) | |
| download | ComputeLibrary-a23b4686a091a7960a4b336d0fe53f15db4ae538.tar.gz | |
Optimize CLTranspose operator
* Transpose higher dimensional tensors (>2D) by collapsing higher
dimensions into the third dimension thus avoiding multiple dispatches
of the CL kernel
* Maximize tile size without register spilling
Resolves: COMPMID-6448
Change-Id: Iac094b8c428bdf319d9c28a8334cb55d58e2d14b
Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10443
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/validation/CL')
| -rw-r--r-- | tests/validation/CL/Transpose.cpp | 25 |
1 files changed, 23 insertions, 2 deletions
diff --git a/tests/validation/CL/Transpose.cpp b/tests/validation/CL/Transpose.cpp index 943534058b..6cf5fe8537 100644 --- a/tests/validation/CL/Transpose.cpp +++ b/tests/validation/CL/Transpose.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2021 Arm Limited. + * Copyright (c) 2017-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -50,12 +50,14 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip( framework::dataset::make("InputInfo", { TensorInfo(TensorShape(21U, 13U), 1, DataType::U16), // Invalid shape TensorInfo(TensorShape(20U, 13U), 1, DataType::U8), // Wrong data type TensorInfo(TensorShape(20U, 16U), 1, DataType::U32), // Valid + TensorInfo(TensorShape(20U, 16U, 3U, 3U), 1, DataType::U16), // Transpose only first two dimensions }), framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(13U, 20U), 1, DataType::U32), TensorInfo(TensorShape(31U, 20U), 1, DataType::U16), TensorInfo(TensorShape(16U, 20U), 1, DataType::U32), + TensorInfo(TensorShape(16U, 20U, 3U, 3U), 1, DataType::U16), })), - framework::dataset::make("Expected", { false, false, true })), + framework::dataset::make("Expected", { false, false, true, true })), a_info, output_info, expected) { // Lock tensors @@ -80,6 +82,16 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLTransposeFixture<uint8_t>, framework::Dataset // Validate output validate(CLAccessor(_target), _reference); } +FIXTURE_DATA_TEST_CASE(RunLargeHighDimensional, + CLTransposeFixture<uint8_t>, + framework::DatasetMode::NIGHTLY, + combine(concat(concat(datasets::Large3DShapes(), datasets::Large4DShapes()), + datasets::Large5dShapes()), + framework::dataset::make("DataType", DataType::U8))) +{ + // Validate output + validate(CLAccessor(_target), _reference); +} TEST_SUITE_END() // U8 TEST_SUITE(U16) @@ -106,6 +118,15 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLTransposeFixture<uint32_t>, framework::Datase // Validate output validate(CLAccessor(_target), _reference); } +FIXTURE_DATA_TEST_CASE(RunSmallHighDimensional, + CLTransposeFixture<uint32_t>, + framework::DatasetMode::PRECOMMIT, + combine(concat(datasets::Small3DShapes(), datasets::Small4DShapes()), + framework::dataset::make("DataType", DataType::U32))) +{ + // Validate output + validate(CLAccessor(_target), _reference); +} FIXTURE_DATA_TEST_CASE(RunLarge, CLTransposeFixture<uint32_t>, framework::DatasetMode::NIGHTLY, combine(concat(datasets::Large1DShapes(), datasets::Large2DShapes()), framework::dataset::make("DataType", DataType::U32))) { |