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 /src/core/CL/CLHelpers.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 'src/core/CL/CLHelpers.cpp')
| -rw-r--r-- | src/core/CL/CLHelpers.cpp | 7 |
1 files changed, 6 insertions, 1 deletions
diff --git a/src/core/CL/CLHelpers.cpp b/src/core/CL/CLHelpers.cpp index 5c435ddc22..0947d58973 100644 --- a/src/core/CL/CLHelpers.cpp +++ b/src/core/CL/CLHelpers.cpp @@ -144,7 +144,12 @@ bool fp16_supported(const cl::Device &device) bool dot8_supported(const cl::Device &device) { - return device_supports_extension(device, "cl_arm_integer_dot_product_int8"); + std::string device_name = device.getInfo<CL_DEVICE_NAME>(); + const GPUTarget gpu_target = get_target_from_name(device_name); + + // SW_WORKAROUND: Workaround for DDK revision r14p0.to enable cl_arm_integer_dot_product_int8 + std::set<GPUTarget> sw_workaround_issue = {GPUTarget::G76}; + return (device_supports_extension(device, "cl_arm_integer_dot_product_int8") || sw_workaround_issue.count(gpu_target) != 0); } bool dot8_acc_supported(const cl::Device &device) |