BRL-CAD/mkl-dnn
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
7,276 Commits 3 Branches 36 Tags
oidn
Commit Graph
97 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Attila Afra f53274c9fe merged 2.2.4 2021-06-28 18:51:36 +03:00
Attila Afra 59e3577cc3 merged 2.2.3 2021-06-28 18:32:17 +03:00
Attila Afra eb3e967005 fixed compile error for some older macOS versions 2020-10-13 01:21:56 +03:00
Attila Afra 603620ba45 Merge remote-tracking branch 'origin/oidn_v1.6.1' into oidn 2020-09-10 00:28:36 +03:00
Attila Afra d4b6cff247 Merge remote-tracking branch 'upstream/rls-v1.6' into oidn_v1.6.1 2020-09-10 00:17:10 +03:00
Attila Afra c2e39639f7 enabled AVX-512 direct conv, disabled Winograd 2020-09-09 03:16:17 +03:00
Attila Afra 1bb88ebf60 Merge remote-tracking branch 'origin/rls-v1.6' into oidn_v1.6.1 
# Conflicts:
#	.github/pull_request_template.md
#	CMakeLists.txt
#	CONTRIBUTING.md
#	README.binary.in
#	README.md
#	cmake/MKL.cmake
#	cmake/Threading.cmake
#	cmake/options.cmake
#	doc/advanced/primitive_cache.md
#	doc/build/link.md
#	doc/performance_considerations/dispatcher_control.md
#	doc/performance_considerations/profilers.md
#	doc/performance_considerations/verbose.md
#	doc/primitives/convolution.md
#	doc/primitives/eltwise.md
#	doc/primitives/layer_normalization.md
#	doc/programming_model/attributes_post_ops.md
#	doc/programming_model/attributes_quantization.md
#	doc/programming_model/data_types.md
#	doc/usage_models/inference_int8.md
#	examples/cnn_inference_int8.cpp
#	examples/cnn_training_bf16.cpp
#	examples/memory_format_propagation.cpp
#	examples/performance_profiling.cpp
#	include/dnnl.h
#	src/common/eltwise_pd.hpp
#	src/common/engine.hpp
#	src/common/math_utils.hpp
#	src/common/primitive_hashing.hpp
#	src/common/verbose.cpp
#	src/cpu/CMakeLists.txt
#	src/cpu/cpu_concat.cpp
#	src/cpu/cpu_convolution_list.cpp
#	src/cpu/cpu_pooling_list.cpp
#	src/cpu/cpu_resampling_list.cpp
#	src/cpu/cpu_sum.cpp
#	src/cpu/eltwise/cpu_eltwise_list.cpp
#	src/cpu/eltwise/jit_uni_eltwise.hpp
#	src/cpu/eltwise/ref_eltwise.cpp
#	src/cpu/eltwise/ref_eltwise.hpp
#	src/cpu/gemm/gemm.cpp
#	src/cpu/gemm/gemm_pack.cpp
#	src/cpu/gemm/gemm_utils.hpp
#	src/cpu/gemm_bf16_convolution.cpp
#	src/cpu/gemm_bf16_inner_product.cpp
#	src/cpu/gemm_convolution.cpp
#	src/cpu/gemm_convolution.hpp
#	src/cpu/gemm_inner_product.cpp
#	src/cpu/gemm_inner_product_utils.cpp
#	src/cpu/gemm_inner_product_utils.hpp
#	src/cpu/gemm_x8s8s32x_convolution.hpp
#	src/cpu/gemm_x8s8s32x_inner_product.cpp
#	src/cpu/jit_uni_eltwise.hpp
#	src/cpu/jit_utils/jitprofiling/README.md
#	src/cpu/matmul/gemm_bf16_matmul.cpp
#	src/cpu/matmul/gemm_f32_matmul.cpp
#	src/cpu/matmul/gemm_x8s8s32x_matmul.cpp
#	src/cpu/matmul/ref_matmul.hpp
#	src/cpu/ref_binary.hpp
#	src/cpu/ref_convolution.hpp
#	src/cpu/rnn/rnn_reorders.hpp
#	src/cpu/rnn/rnn_utils.hpp
#	src/cpu/x64/gemm/gemm_driver.hpp
#	src/cpu/x64/gemm_bf16_convolution.hpp
#	src/cpu/x64/jit_avx2_1x1_conv_kernel_f32.hpp
#	src/cpu/x64/jit_avx2_conv_kernel_f32.hpp
#	src/cpu/x64/jit_avx2_x8s8s32x_1x1_conv_kernel.hpp
#	src/cpu/x64/jit_avx2_x8s8s32x_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_common_1x1_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_common_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_bf16_1x1_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_bf16_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_bf16_dw_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_x8s8s32x_1x1_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_x8s8s32x_conv_kernel.hpp
#	src/cpu/x64/jit_avx512_core_x8s8s32x_deconvolution.hpp
#	src/cpu/x64/jit_sse41_1x1_conv_kernel_f32.hpp
#	src/cpu/x64/jit_sse41_conv_kernel_f32.hpp
#	src/cpu/x64/jit_uni_binary.cpp
#	src/cpu/x64/jit_uni_dw_conv_kernel_f32.hpp
#	src/cpu/x64/jit_uni_dw_conv_kernel_utils.hpp
#	src/cpu/x64/jit_uni_eltwise.hpp
#	src/cpu/x64/jit_uni_i8i8_binary.cpp
#	src/cpu/x64/jit_uni_softmax.cpp
#	src/cpu/x64/rnn/jit_uni_rnn_common_postgemm.hpp
#	src/gpu/ocl/ocl_post_ops.h
#	tests/benchdnn/README.md
#	tests/benchdnn/binary/binary.cpp
#	tests/benchdnn/bnorm/bnorm.cpp
#	tests/benchdnn/common.cpp
#	tests/benchdnn/concat/concat.cpp
#	tests/benchdnn/conv/conv.cpp
#	tests/benchdnn/conv/conv_dw_fusion.cpp
#	tests/benchdnn/conv/deconv.cpp
#	tests/benchdnn/dnnl_common.cpp
#	tests/benchdnn/doc/knobs_attr.md
#	tests/benchdnn/eltwise/eltwise.cpp
#	tests/benchdnn/inputs/binary/shapes_common
#	tests/benchdnn/ip/ip.cpp
#	tests/benchdnn/lnorm/lnorm.cpp
#	tests/benchdnn/lrn/lrn.cpp
#	tests/benchdnn/matmul/matmul.cpp
#	tests/benchdnn/pool/pool.cpp
#	tests/benchdnn/reorder/reorder.cpp
#	tests/benchdnn/resampling/resampling.cpp
#	tests/benchdnn/rnn/rnn.cpp
#	tests/benchdnn/shuffle/shuffle.cpp
#	tests/benchdnn/softmax/softmax.cpp
#	tests/benchdnn/sum/sum.cpp
#	tests/gtests/test_gemm_common.hpp
 2020-09-06 00:13:18 +03:00
Attila Afra bcceee6838 fixed incorrect detection of AVX-512 on macOS 2020-08-26 01:07:17 +03:00
Attila Afra 9730e5bf70 disabled KNL convolution kernels 2020-06-11 00:47:39 +03:00
Attila Afra 873e42b368 Merge branch 'master' into oidn 
# Conflicts:
#	CMakeLists.txt
#	examples/performance_profiling.cpp
#	src/common/memory_tracking.hpp
#	src/cpu/binary/jit_uni_binary.cpp
#	src/cpu/cpu_convolution_list.cpp
#	src/cpu/eltwise/ref_eltwise.cpp
#	src/cpu/jit_avx2_x8s8s32x_conv_kernel.cpp
#	src/cpu/jit_avx2_x8s8s32x_convolution.cpp
#	src/cpu/jit_avx512_common_conv_kernel.cpp
#	src/cpu/jit_avx512_core_bf16_conv_kernel.cpp
#	src/cpu/jit_avx512_core_x8s8s32x_conv_kernel.cpp
#	src/cpu/rnn/cell_common.cpp
#	src/cpu/rnn/cell_gru.cpp
#	src/cpu/rnn/cell_gru_lbr.cpp
#	src/cpu/rnn/ref_rnn.hpp
#	tests/benchdnn/CMakeLists.txt
#	tests/benchdnn/inputs/binary/shapes_common
#	tests/benchdnn/inputs/binary/test_binary_bfloat16
#	tests/benchdnn/inputs/conv/harness_conv_regression_general
#	tests/benchdnn/inputs/conv/test_conv_bfloat16
#	tests/benchdnn/inputs/resampling/test_resampling_all
#	tests/benchdnn/resampling/resampling.cpp
#	tests/benchdnn/rnn/bench_rnn.cpp
#	tests/benchdnn/rnn/rnn.cpp
#	tests/gtests/test_global_scratchpad.cpp
#	tests/gtests/test_rnn_forward.cpp
 2020-05-10 01:37:28 +03:00
Attila Afra 8158955c35 Merge commit '8e96ef49488c65e0738c552cec5c0563ab92c1af' into oidn 2020-03-21 17:49:23 +02:00
Attila Afra eaf9b5216c Merge remote-tracking branch 'upstream/rls-v1.2' into oidn 2020-03-07 23:49:13 +02:00
Attila Afra 331629476f enabled more reorders 2020-01-28 20:11:06 +02:00
Attila Afra d946bcc9fa enabled resampling 2020-01-28 14:02:52 +02:00
Attila Afra cbd8cd0dc5 re-enabled wino_reorder 2020-01-26 16:15:56 +02:00
Attila Afra 2bbcd88ac8 Merge tag 'v1.2-rc' into oidn 
# Conflicts:
#	src/CMakeLists.txt
#	src/common/memory.cpp
#	src/common/scratchpad.cpp
#	src/cpu/cpu_engine.cpp
#	src/cpu/cpu_memory_storage.hpp
#	src/cpu/cpu_reorder.cpp
#	src/cpu/gemm_x8s8s32x_convolution.cpp
#	src/cpu/jit_avx512_common_convolution_winograd.cpp
#	src/cpu/jit_avx512_common_convolution_winograd.hpp
#	src/cpu/jit_avx512_core_f32_wino_conv_4x3.cpp
#	src/cpu/wino_reorder.hpp
#	src/cpu/xbyak/xbyak.h
#	src/cpu/xbyak/xbyak_util.h
#	src/ocl/ocl_memory_storage.cpp
#	src/ocl/ocl_memory_storage.hpp
#	src/ocl/ocl_utils.hpp
#	tests/benchdnn/inputs/conv/harness_conv_regression_general
#	tests/benchdnn/inputs/conv/shapes_gemm
 2020-01-26 15:34:19 +02:00
Attila Afra 04dfb3d2f7 cpu: winograd: add TBB support for forward convolutions 2019-11-15 11:56:43 -08:00
Attila Afra caf815d7c8 cpu: wino reorder: replace parallel_nd with for_nd to improve performance 
Using parallel_nd this way (many small parallel loops) actually hurts performance and it's much better to simply use sequential loops. This improves initialization speed by more than 4x for Open Image Denoise on a 2 x Xeon 8180 machine.
 2019-11-15 11:56:43 -08:00
Attila Afra 9ab68b1b33 cpu, ocl: detect and properly handle memory storage allocation and OCL runtime errors 
The library checks for small memory allocation errors but it does
*not* report out of memory errors when allocating the *actual* memory
storage. On the CPU these allocations are not checked at all and the
whole application crashes, and on the GPU exit() is called, which also
kills the application. This is a major issue because out of memory
errors usually occur for large allocations, which are not handled
properly (or at all). Also, exit() should never be called inside the
library, and assert() should not be used for checking whether an
allocation was successful. This commit fixes all these issues.
 2019-10-28 21:24:02 -07:00
Attila Afra c36a6567a2 reverted all CMake changes 2019-10-21 23:36:09 +03:00
Attila Afra a467a7c499 minor change 2019-10-21 23:35:58 +03:00
Attila Afra 4373397776 CMake cleanup and fixes 2019-10-15 16:04:59 +03:00
Attila Afra 9c15f2e74b fixed bug in TBB threading (reverted to original) 2019-10-11 15:11:49 +03:00
Attila Afra ad5893a811 minor cleanup 2019-10-11 13:12:55 +03:00
Attila Afra 12d00a2fd2 ocl: return out of memory or other error status instead of calling exit() 2019-10-11 02:14:32 +03:00
Attila Afra df34aa8f57 minor merge fix 2019-10-09 01:32:33 +03:00
Attila Afra 63ffa8162a Merge remote-tracking branch 'upstream/rls-v1.1' into oidn 2019-10-08 18:54:48 +03:00
Attila Afra f8bff5ada1 Merge branch 'master' into oidn 
# Conflicts:
#	cmake/OpenMP.cmake
#	cmake/TBB.cmake
#	cmake/lnx/TBBConfig.cmake
#	cmake/mac/TBBConfig.cmake
#	cmake/options.cmake
#	cmake/platform.cmake
#	cmake/win/TBBConfig.cmake
#	doc/ex_int8_simplenet.md
#	include/mkldnn.h
#	include/mkldnn_types.h
#	src/CMakeLists.txt
#	src/common/dnnl_thread_parallel_nd.hpp
#	src/common/mkldnn_thread.hpp
#	src/cpu/cpu_concat.cpp
#	src/cpu/cpu_engine.cpp
#	src/cpu/cpu_isa_traits.hpp
#	src/cpu/cpu_memory.hpp
#	src/cpu/cpu_reorder.cpp
#	src/cpu/cpu_sum.cpp
#	src/cpu/jit_avx512_common_convolution_winograd.cpp
#	src/cpu/jit_avx512_common_convolution_winograd.hpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.cpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.hpp
#	src/cpu/simple_reorder.hpp
#	src/cpu/wino_reorder.hpp
#	tests/other/subproject/main.c
 2019-10-08 17:34:47 +03:00
Attila Afra 096bda1ca2 enabled NDEBUG in Release builds for all compilers 2019-07-16 14:03:02 +03:00
Attila Afra 6241ac736f fixed TBB linking issue 2019-05-09 18:22:07 +03:00
Attila Afra f804bd98c9 fixed Windows link error 2019-05-09 04:04:28 +03:00
Attila Afra d82ace6f70 minor formatting fix 2019-05-08 16:01:56 +03:00
Attila Afra 01eddd74dd never call exit() when an internal error occurs 2019-05-08 15:58:22 +03:00
Attila Afra 308ec2b17f return error status if a memory allocation fails 2019-03-26 13:30:56 +02:00
Attila Afra 7ef75ed70e CMake cleanup 2019-03-24 21:57:56 +02:00
Attila Afra cf16a4b567 Merge remote-tracking branch 'origin/master' into oidn 
# Conflicts:
#	CMakeLists.txt
#	cmake/options.cmake
#	cmake/profiling.cmake
#	src/CMakeLists.txt
#	src/cpu/cpu_engine.cpp
 2019-03-24 21:53:59 +02:00
Attila Afra 130f0ce235 Merge commit '002d23b36d96a9150acf57df6a8cfcb89a7debc1' into oidn 
# Conflicts:
#	CMakeLists.txt
#	README.md
#	cmake/OpenMP.cmake
#	cmake/SDL.cmake
#	cmake/TBB.cmake
#	cmake/options.cmake
#	cmake/platform.cmake
#	cmake/profiling.cmake
#	cmake/utils.cmake
#	doc/mainpage.md
#	src/CMakeLists.txt
#	src/cpu/cpu_concat.cpp
#	src/cpu/cpu_engine.cpp
#	src/cpu/cpu_memory.cpp
#	src/cpu/cpu_reorder.cpp
#	src/cpu/cpu_sum.cpp
#	src/cpu/gemm_convolution_utils.cpp
#	src/cpu/gemm_x8s8s32x_convolution.hpp
#	src/cpu/jit_avx512_common_convolution_winograd.hpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.hpp
#	src/cpu/simple_concat.cpp
#	src/cpu/simple_reorder.hpp
#	src/cpu/xbyak/xbyak_util.h
#	tests/gtests/test_softmax_forward.cpp
 2019-03-21 01:47:25 +02:00
Attila Afra 4e64c7586d TBB CMake cleanup 2019-03-21 00:19:49 +02:00
Attila Afra 267cc475b5 CMake fix for TBB 2019-03-20 01:21:11 +02:00
Attila Afra 2d609bc8b3 cleanups 2019-03-18 01:32:21 +02:00
Attila Afra 63cee68420 Merge remote-tracking branch 'upstream/rls-v0.18' into oidn 
# Conflicts:
#	src/CMakeLists.txt
 2019-03-18 01:03:48 +02:00
Attila Afra 3fdcd90904 replaced hardcoded TBB for 2019-03-18 00:52:57 +02:00
Attila Afra 8876941fcf fixed wrong parallel_nd call 2019-03-18 00:39:50 +02:00
Attila Afra cd61768e13 CMake fix for Windows static library 2019-03-18 00:37:31 +02:00
Attila Afra de7b35023d Merge commit '863ff6e7042cec7d2e29897fe9f0872e0888b0fc' into oidn 
# Conflicts:
#	CMakeLists.txt
#	README.md
#	cmake/Doxygen.cmake
#	cmake/SDL.cmake
#	cmake/TBB.cmake
#	cmake/platform.cmake
#	cmake/profiling.cmake
#	scripts/prepare_mkl.bat
#	scripts/prepare_mkl.sh
#	src/CMakeLists.txt
#	src/common/rnn_pd.hpp
#	src/common/scratchpad.cpp
#	src/cpu/cpu_engine.cpp
#	src/cpu/cpu_reorder.cpp
#	src/cpu/gemm/f32/jit_avx512_common_gemm_f32.cpp
#	src/cpu/gemm/f32/jit_avx_gemm_f32.cpp
#	src/cpu/gemm/f32/ref_gemm_f32.cpp
#	src/cpu/gemm/gemm.cpp
#	src/cpu/gemm/jit_avx512_common_gemm_f32.hpp
#	src/cpu/gemm/jit_avx_gemm_f32.hpp
#	src/cpu/jit_avx512_common_convolution.hpp
#	src/cpu/jit_avx512_common_convolution_winograd.cpp
#	src/cpu/jit_avx512_common_convolution_winograd.hpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.cpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.hpp
#	src/cpu/ref_rnn.cpp
#	src/cpu/ref_rnn.hpp
#	src/cpu/simple_concat.hpp
#	src/cpu/wino_reorder.hpp
#	src/cpu/xbyak/xbyak_util.h
#	tests/benchdnn/rnn/ref_rnn.cpp
#	tests/gtests/test_gemm_common.hpp
 2019-03-18 00:31:30 +02:00
Attila Afra 2856104855 Merge remote-tracking branch 'upstream/rls-v0.17' into oidn 2019-03-17 22:27:56 +02:00
Attila Afra 9507eb93f6 fixed link error for Win+ICC+Debug, added MSVC optimization flags 2019-02-02 22:21:14 +02:00
Attila Afra ae6028e7e7 fixed bug that prevented some compiler flags to be passed to MSVC (e.g. parallel build) 2019-02-02 01:52:31 +02:00
Attila Afra 78ba6e1032 fixed wrong detection of MSVC, which caused some flags to be not set 2019-02-02 01:32:02 +02:00
Attila Afra 0f233d6887 fixed compile error on some macOS systems 2019-02-01 19:01:55 +02:00
Attila Afra cdbcddd53b Merge remote-tracking branch 'origin/oidn' into oidn 2019-02-01 12:41:32 +02:00
Attila Afra 07658e7cd3 lowered the ISA requirement from SSE4.2 to SSE4.1 2019-02-01 12:40:20 +02:00
Attila Afra ad659f0339 fixed wrong rpath on macOS 2019-01-30 16:36:00 +02:00
Attila Afra 3da8669dff improved TBB detection on Windows 2019-01-19 02:28:00 +02:00
Attila Afra db512e89dc fixed bug in TBB detection on Windows 2019-01-19 02:05:18 +02:00
Attila Afra c9ac721a8a removed unused ARCH_OPT_FLAGS option 2019-01-17 01:12:13 +02:00
Attila Afra 4eb83f4893 fixes for ICC on Windows 2018-12-18 22:40:44 +02:00
Attila Afra 7ca14b33ce fix for commercial TBB on Windows 2018-12-18 19:00:02 +02:00
Attila Afra 765eca671d do wino_reorder serially because parallel is much slower (~4x on 2 x Xeon 8180) due to its large overhead 2018-12-14 15:56:32 +02:00
Attila Afra 63883aa229 slightly improved TBB performance 2018-12-14 15:49:03 +02:00
Attila Afra 5eca3411fa disabled unused code 2018-12-14 01:47:16 +02:00
Attila Afra 52131c7f03 implemented KNL support 2018-12-14 01:41:18 +02:00
Attila Afra ea9855b333 CMake ISA flags cleanup 2018-12-14 00:47:52 +02:00
Attila Afra 5750e01dd6 CMake cleanups 2018-12-13 19:06:01 +02:00
Attila Afra 205750484b reverted some CMake changes 2018-12-13 13:01:58 +02:00
Attila Afra 2854ff4437 TBB cleanup 2018-12-13 01:05:39 +02:00
Attila Afra c6bfcd0554 TBB performance improvement 2018-12-12 23:46:54 +02:00
Attila Afra 29ba11ae67 disabled unused primitives 2018-12-12 23:23:31 +02:00
Attila Afra a513cfa26a CMake cleanups 2018-12-11 19:42:56 +02:00
Attila Afra 241d6121bd added support for older macOS versions 2018-12-11 19:31:01 +02:00
Attila Afra c7d1db87fe compile for SSE4.2 with Clang too 2018-12-11 18:25:44 +02:00
Attila Afra 8979faa638 Merge commit 'a7c5f53832acabade6e5086e72c960adedb3c38a' (0.17.1) into oidn 
# Conflicts:
#	CMakeLists.txt
#	cmake/MKL.cmake
#	cmake/OpenMP.cmake
#	cmake/TBB.cmake
#	cmake/platform.cmake
#	src/CMakeLists.txt
#	src/common/mkldnn_thread.hpp
#	src/cpu/cpu_engine.cpp
#	src/cpu/cpu_reorder.cpp
#	src/cpu/jit_avx2_convolution.cpp
#	src/cpu/jit_avx2_convolution.hpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.cpp
#	src/cpu/jit_avx512_core_fp32_wino_conv_4x3.hpp
#	src/cpu/jit_avx512_core_u8s8s32x_wino_convolution.cpp
#	src/cpu/jit_sse42_convolution.cpp
#	src/cpu/jit_uni_pooling.cpp
 2018-12-11 15:41:40 +02:00
Attila Afra d8e2527cd4 fixed linking error on macOS 2018-12-07 18:10:36 +02:00
Attila Afra 37ce5425ab fixed link error on Windows in debug mode 2018-11-25 18:20:18 +02:00
Attila Afra 77f87ee922 fixed Windows compile error 2018-11-23 01:00:21 +02:00
Attila Afra 799bfd0838 fixed memory usage regression caused by the previous fix 2018-08-04 01:20:42 +03:00
Attila Afra 1598c1face fixed crash when a primitive was initialized and executed on different threads 2018-08-03 20:17:05 +03:00
Attila Afra 97890978b1 re-enabled -Wno-unknown-pragmas 2018-08-01 19:13:05 +03:00
Attila Afra b346c669c9 re-enabled -Wall, but -Werror is still disabled 2018-08-01 19:11:25 +03:00
Attila Afra 9b344eed17 disabled -Wall 2018-08-01 17:48:37 +03:00
Attila Afra 145ce834dc fixed TBB bug in AVX2 and SSE4.2 convolution 2018-07-18 17:29:57 +03:00
Attila Afra 8b6ed3e3c4 cmake cleanups 2018-07-17 19:51:56 +03:00
Attila Afra 57e626bf02 compile non-JIT code for SSE4.2 only 2018-07-17 19:29:44 +03:00
Attila Afra 0fb1a5cbcf use TBB for SSE4.2 convolution 2018-07-17 18:32:12 +03:00
Attila Afra 6b8b69463b use TBB for AVX2 convolution 2018-07-17 16:51:45 +03:00
Attila Afra 0246a72a36 re-enabled OpenMP SIMD 2018-07-17 00:40:01 +03:00
Attila Afra ae04e7c055 removed regular MKL support 2018-07-17 00:33:09 +03:00
Attila Afra 4cb10834ce disabled unused code 2018-07-16 12:57:24 +03:00
Attila Afra c069ddec8e switched to static library 2018-07-14 23:42:25 +03:00
Attila Afra 0e086c49ec fixed version 2018-07-14 23:24:57 +03:00
Attila Afra d9b006b311 cleanups 2018-07-14 15:07:54 +03:00
Attila Afra 2f69fd6039 Merge remote-tracking branch 'origin/master' into atafra/tbb 
# Conflicts:
#	src/cpu/jit_avx512_core_convolution_winograd.cpp
 2018-07-14 13:42:21 +03:00
Attila Afra 6b1a8fa7c0 do not link OpenMP lib 2018-07-13 00:44:57 +03:00
Attila Afra 9d5d020931 minor fix 2018-07-13 00:29:19 +03:00
Attila Afra aeb153f885 always use static partitioning 2018-07-12 19:39:40 +03:00
Attila Afra d6f568fde2 removed OpenMP from cmake 2018-07-12 17:20:02 +03:00
Attila Afra 1ec36aa6ff added TBB to cmake 2018-07-12 16:10:35 +03:00