diff options
author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2018-10-18 10:21:02 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:55:45 +0000 |
commit | 4b90865ab985d571f70c60583cdfb8c7a65f1670 (patch) | |
tree | f116a4ffef5f5e823689dd00c1e5c9d987f3d295 /tests/validation/NEON/GEMMLowp.cpp | |
parent | c55beee7ef70fa08a5d217619083b288a74fcb27 (diff) | |
download | ComputeLibrary-4b90865ab985d571f70c60583cdfb8c7a65f1670.tar.gz |
COMPMID-1413 - Improve the performance of GEMMLowp with 8 bit dot product on OpenCL
COMPMID-1424 - Add dot product support for CLDepthwise QASYMM8 3x3 NHWC non-unit stride
With this patch we are able to improve the performance of MobileNet v1-qasymm8 by 37 %
Tried to use the dot product instruction in CLDepthwise QASYMM8 3x3 NHWC non-unit stride
but I have not seen any benefit (maybe because we have few arithemtic operation and we
do not have more load instructions). However Depthwise convolution has been improved by
30%
Change-Id: Id768a99c2e53a04276707e427af5d0ec93419ada
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/155082
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'tests/validation/NEON/GEMMLowp.cpp')
-rw-r--r-- | tests/validation/NEON/GEMMLowp.cpp | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/tests/validation/NEON/GEMMLowp.cpp b/tests/validation/NEON/GEMMLowp.cpp index 9eba3c85c1..1458c9fdc3 100644 --- a/tests/validation/NEON/GEMMLowp.cpp +++ b/tests/validation/NEON/GEMMLowp.cpp @@ -95,7 +95,7 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::c // Create and configure function NEGEMMLowpMatrixMultiplyCore gemmlowp_mm; - gemmlowp_mm.configure(&a, &b, &c); + gemmlowp_mm.configure(&a, &b, nullptr, &c); } // *INDENT-OFF* @@ -125,6 +125,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip( // Lock tensors Status status = NEGEMMLowpMatrixMultiplyCore::validate(&a_info.clone()->set_is_resizable(false), &b_info.clone()->set_is_resizable(false), + nullptr, &output_info.clone()->set_is_resizable(false)); ARM_COMPUTE_EXPECT(bool(status) == expected, framework::LogLevel::ERRORS); } |