aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/cl_kernels/gemm.cl
AgeCommit message (Collapse)Author
2018-11-02COMPMID-1032 - Fixing bug in CLGEMM when is_interleaved_transposed=trueGian Marco Iodice
The bug concerned the collapse of the window in CLGEMMMatrixMultiplyKernel Change-Id: I5043bf37b72eeb615ebe7fb3f2c8e72d006bf341 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/126262 Tested-by: Jenkins <bsgcomp@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-11-02COMPMID-935 - Implementing Convolution with Winograd on OpenCL (part 4)Gian Marco Iodice
Implemented Winograd Output Transform (2x2,3x3) on OpenCL Implemented CLWinogradConvolutionLayer on OpenCL Change-Id: I6a113fc5f052ca07f878d2b800d2ab003f84af65 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/125148 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Jenkins <bsgcomp@arm.com>
2018-11-02COMPMID-935 - Implementing Convolution with Winograd on OpenCL (Part 1)Gian Marco
This patch enables GEMM to execute multiple batches in parallel https://confluence.arm.com/display/MLENG/Winograd%3A+batched+GEMM Change-Id: I66222db041dd35e82af11fbb262fd1ebd3ca4b2f Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/120866 Tested-by: Jenkins <bsgcomp@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-11-02COMPMID-882 - Optimizing GEMMLowp on OpenCL reshaping matricesGian Marco
This new optimization allows to achieve 36.3 % of MAC utilisation on Mate 9 @ 1GHz. The performance have been reported here https://confluence.arm.com/display/MLENG/GEMMLowp+performance%3A+ACL+18.02 Change-Id: I71b6a217068763dfdc11bbf3574ee0eb94f93679 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/118531 Reviewed-by: Anthony Barbier <anthony.barbier@arm.com> Tested-by: Jenkins <bsgcomp@arm.com>
2018-11-02COMPMID-748 - Integrating optimized SGEMM for bifrostGian Marco
This patch introduces a new GEMM capable to improve the mac utilisation of 10% compared to the GEMM without reshape. However this implementation is not faster in all cases as we need to take into account the time for reshaping the matrices. For this reason an heuristic solution to select the optimal GEMM to use has been added to the function. More information about the heuristic implementation can be found at COMPMID-852. With this new patch, GoogleNet, MobileNet, VGG16 and SqueezeNet can improved the performance of 1.5x. More information about the performance uplift can be found here: https://confluence.arm.com/display/MLENG/GEMM+FP32+performance%3A+ACL+18.02 Change-Id: I024563c06b9aed02a211a974e452bae5c233b04c Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/117140 Reviewed-by: Pablo Tello <pablo.tello@arm.com> Tested-by: Jenkins <bsgcomp@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-11-02COMPMID-697 - Rework GEMMLowp interface on OpenCLGian Marco
Reworked the interface of GemmLowp in order to make easy the integration in Android NN - Added support for different output stage - Added validation for both matrix multiplication and output stage - Added bounded relu support in the output stage - Added in32_t bias support - Added optimized path for vector by matrix case This rework is required for: - Convolution quantized - Fully connected quantized Change-Id: I512283d406099cf8c614dd89d0a97ed411143afc Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/110625 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
2018-11-02COMPMID-661: Optimize FC layer with 2 new Bifrost kernels and LWS tuning (#33)Anton Lokhmotov
Change-Id: Ie56ac88dff5ff339572cec562e8cd62dc7f0aa8b Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/109805 Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-11-02Allow running without cl_khr_fp16Matthew Bentham
Change-Id: I0f5396c8f32acc28914e2ff9fe953f977a3077b9 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/93405 Reviewed-by: Anthony Barbier <anthony.barbier@arm.com> Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
2018-11-02COMPMID-640: FullyConnectedLayer failures on both NEON/CLGeorgios Pinitas
Change-Id: Idd830cff054114123229c189e423b753b8064146 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/92623 Reviewed-by: Robert Hughes <robert.hughes@arm.com> Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-11-02COMPMID-477 - Optimized batched case in CLConvolutionLayerGian Marco Iodice
Change-Id: I4ef18f49f1da0cb816aaa0762466b940792c15ed Reviewed-on: http://mpd-gerrit.cambridge.arm.com/84162 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-09-17COMPMID-415: Use half_float library for F16Moritz Pflanzer
3RDPARTY_UPDATE Change-Id: Iee572e18d5b1df71300d738cc8690f49d7203d5c Reviewed-on: http://mpd-gerrit.cambridge.arm.com/81353 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
2018-09-17COMPMID-420, COMPMID-414 - Port CLConvolutionLayer and CLFullyConnectedLayer ↵Gian Marco Iodice
to use 8 bit fixed point Change-Id: I1cb1b4d7711ad7b569ee691e13a5df1b3430292b Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79565 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2018-09-17COMPMID-417 - Fixed bug in gemm_interleave_16bit and gemm_interleave_32_bit ↵Gian Marco Iodice
due to the non non representable numbers in half and float Change-Id: I1590699e788a2cd3ed6e72de4afff75e6346f6d8 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79800 Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
2018-09-17COMPMID-434 - Port CLGEMM to support 16 bit fixed pointGian Marco Iodice
Change-Id: I30aef3c7ecd1ee740c2a7f2ce65a63c7dcd66e49 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79630 Reviewed-by: Anthony Barbier <anthony.barbier@arm.com> Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
2018-09-17COMPMID-418 Add check and fix comments after preprocessor conditionsAnthony Barbier
Change-Id: I1353fd652ee180e3931e58b4ce13d651a48c7e2c Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79567 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com>
2018-09-17COMPMID-411 - Port CLGEMM to support 8 bit fixed pointGian Marco Iodice
Change-Id: I6c8bd69ae9715e4d83d128b2162fc15aa5561afb Reviewed-on: http://mpd-gerrit.cambridge.arm.com/78804 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com>
2018-09-17COMPMID-414 - Port CLConvolutionLayer to support 8 bit fixed point - ↵Gian Marco Iodice
CLGEMMMatrixAccumulateBiasesKernel Change-Id: Idba13b578dc564b8003ce2fa3392eea2af3ce806 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/78664 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2018-09-17COMPMID-411 - Ported CLGEMMInterleave4x4Kernel and CLGEMMTranspose1xWKernel ↵Gian Marco Iodice
to support 8 bit fixed point Change-Id: If236c9047ed536e808a0ed26e97e1799ca938e03 Reviewed-on: http://mpd-gerrit.cambridge.arm.com/78529 Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com> Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2018-09-17COMPMID-344 Updated doxygenAnthony Barbier
Change-Id: I32f7b84daa560e460b77216add529c8fa8b327ae
generated by cgit v1.2.1 (git 2.23.0) at 2024-07-30 03:39:27 +0000