mkl-dnn-O3native AMD Ryzen 7 2700 Eight-Core testing with a ASRock X470 Master SLI (P3.70 BIOS) and Zotac NVIDIA GeForce GTX 550 Ti 1GB on Mageia 9 via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/2103181-HA-MKLDNNO3N46 .
mkl-dnn-O3native Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server Display Driver OpenGL Compiler File-System Screen Resolution mga9-O3native AMD Ryzen 7 2700 Eight-Core @ 4.05GHz (8 Cores / 16 Threads) ASRock X470 Master SLI (P3.70 BIOS) AMD 17h 16GB Samsung SSD 970 EVO 250GB + 525GB Crucial CT525MX3 + 1000GB Seagate ST31000528AS + 512GB INTEL SSDSCKKW51 Zotac NVIDIA GeForce GTX 550 Ti 1GB NVIDIA GF116 HD Audio DELL U2412M Intel I211 Mageia 9 5.11.6-desktop-4.mga9 (x86_64) KDE Plasma 5.21.2 X Server 1.20.10 NVIDIA 390.141 4.6.0 GCC 10.2.1 20210313 + Clang 12.0.0 + LLVM 12.0.0 ext4 1920x1200 OpenBenchmarking.org - Transparent Huge Pages: madvise - CXXFLAGS="-O3 -mtune=native -march=native" CFLAGS="-O3 -mtune=native -march=native" - --build=x86_64-mageia-linux-gnu --disable-libffi --disable-libssp --disable-libunwind-exceptions --disable-werror --enable-__cxa_atexit --enable-checking=release --enable-clocale=gnu --enable-languages=c,c++,ada,fortran,objc,obj-c++,d --enable-long-long --enable-lto --enable-shared --enable-ssp --enable-threads=posix --host=x86_64-mageia-linux-gnu --mandir=/usr/share/man --with-cpu=generic --with-gcc-major-version-only --with-isl --with-linker-hash-style=gnu --with-python-dir=/lib/python3.8/site-packages --with-slibdir=/lib64 - Scaling Governor: acpi-cpufreq conservative (Boost: Enabled) - CPU Microcode: 0x800820d - itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Full AMD retpoline IBPB: conditional STIBP: disabled RSB filling + srbds: Not affected + tsx_async_abort: Not affected
mkl-dnn-O3native mkl-dnn: IP Batch 1D - f32 mkl-dnn: IP Batch All - f32 mkl-dnn: Deconvolution Batch deconv_1d - f32 mkl-dnn: Deconvolution Batch deconv_3d - f32 mkl-dnn: Recurrent Neural Network Training - f32 mkl-dnn: Recurrent Neural Network Inference - f32 mga9-O3native 7.53791 109.942 8.39442 12.1666 498.701 49.4729 OpenBenchmarking.org
oneDNN MKL-DNN Harness: IP Batch 1D - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: IP Batch 1D - Data Type: f32 mga9-O3native 2 4 6 8 10 SE +/- 0.01963, N = 3 7.53791 MIN: 6.83 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
oneDNN MKL-DNN Harness: IP Batch All - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: IP Batch All - Data Type: f32 mga9-O3native 20 40 60 80 100 SE +/- 0.17, N = 3 109.94 MIN: 103.94 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
oneDNN MKL-DNN Harness: Deconvolution Batch deconv_1d - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: Deconvolution Batch deconv_1d - Data Type: f32 mga9-O3native 2 4 6 8 10 SE +/- 0.04881, N = 3 8.39442 MIN: 7.64 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
oneDNN MKL-DNN Harness: Deconvolution Batch deconv_3d - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: Deconvolution Batch deconv_3d - Data Type: f32 mga9-O3native 3 6 9 12 15 SE +/- 0.10, N = 3 12.17 MIN: 11.84 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
oneDNN MKL-DNN Harness: Recurrent Neural Network Training - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: Recurrent Neural Network Training - Data Type: f32 mga9-O3native 110 220 330 440 550 SE +/- 3.51, N = 3 498.70 MIN: 464.59 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
oneDNN MKL-DNN Harness: Recurrent Neural Network Inference - Data Type: f32 OpenBenchmarking.org ms, Fewer Is Better oneDNN MKL-DNN 1.3 Harness: Recurrent Neural Network Inference - Data Type: f32 mga9-O3native 11 22 33 44 55 SE +/- 0.06, N = 3 49.47 MIN: 43.54 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
Phoronix Test Suite v10.8.4