From a23b4686a091a7960a4b336d0fe53f15db4ae538 Mon Sep 17 00:00:00 2001 From: Jakub Sujak Date: Thu, 5 Oct 2023 10:20:59 +0100 Subject: 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 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10443 Tested-by: Arm Jenkins Reviewed-by: Viet-Hoa Do Comments-Addressed: Arm Jenkins Benchmark: Arm Jenkins --- tests/validation/fixtures/TransposeFixture.h | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'tests/validation/fixtures') diff --git a/tests/validation/fixtures/TransposeFixture.h b/tests/validation/fixtures/TransposeFixture.h index 92eb9af0c1..212c76cc9a 100644 --- a/tests/validation/fixtures/TransposeFixture.h +++ b/tests/validation/fixtures/TransposeFixture.h @@ -21,8 +21,8 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#ifndef ARM_COMPUTE_TEST_TRANSPOSE_FIXTURE -#define ARM_COMPUTE_TEST_TRANSPOSE_FIXTURE +#ifndef ACL_TESTS_VALIDATION_FIXTURES_TRANSPOSEFIXTURE_H +#define ACL_TESTS_VALIDATION_FIXTURES_TRANSPOSEFIXTURE_H #include "arm_compute/core/TensorShape.h" #include "arm_compute/core/Types.h" @@ -32,7 +32,7 @@ #include "tests/IAccessor.h" #include "tests/framework/Asserts.h" #include "tests/framework/Fixture.h" -#include "tests/validation/reference/Transpose.h" +#include "tests/validation/reference/Permute.h" namespace arm_compute { @@ -97,7 +97,7 @@ protected: // Fill reference fill(src); - return reference::transpose(src); + return reference::permute(src, PermutationVector(1U, 0U)); } TensorType _target{}; @@ -106,4 +106,4 @@ protected: } // namespace validation } // namespace test } // namespace arm_compute -#endif /* ARM_COMPUTE_TEST_TRANSPOSE_FIXTURE */ +#endif // ACL_TESTS_VALIDATION_FIXTURES_TRANSPOSEFIXTURE_H -- cgit v1.2.1