onednn 3.0 threadripper

AMD Ryzen Threadripper 3990X 64-Core testing with a Gigabyte TRX40 AORUS PRO WIFI (F4p BIOS) and AMD Radeon RX 5700 8GB on Ubuntu 22.10 via the Phoronix Test Suite.

HTML result view exported from: https://openbenchmarking.org/result/2212207-PTS-ONEDNN3042.

onednn 3.0 threadripperProcessorMotherboardChipsetMemoryDiskGraphicsAudioMonitorNetworkOSKernelDesktopDisplay ServerOpenGLVulkanCompilerFile-SystemScreen ResolutionabccAMD Ryzen Threadripper 3990X 64-Core @ 2.90GHz (64 Cores / 128 Threads)Gigabyte TRX40 AORUS PRO WIFI (F4p BIOS)AMD Starship/Matisse128GBSamsung SSD 970 EVO Plus 500GBAMD Radeon RX 5700 8GB (1750/875MHz)AMD Navi 10 HDMI AudioDELL P2415QIntel I211 + Intel Wi-Fi 6 AX200Ubuntu 22.106.1.0-rc8-phx-mglru (x86_64)GNOME Shell 43.0X Server 1.21.1.4 + Wayland4.6 Mesa 22.2.1 (LLVM 15.0.2 DRM 3.49)1.3.224GCC 12.2.0 + LLVM 15.0.2ext43840x2160OpenBenchmarking.orgKernel Details- Transparent Huge Pages: madviseCompiler Details- --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-12-U8K4Qv/gcc-12-12.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-12-U8K4Qv/gcc-12-12.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v Processor Details- Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0x8301055Security Details- itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Mitigation of untrained return thunk; SMT enabled with STIBP protection + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional STIBP: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

onednn 3.0 threadripperonednn: IP Shapes 1D - f32 - CPUonednn: IP Shapes 3D - f32 - CPUonednn: IP Shapes 1D - u8s8f32 - CPUonednn: IP Shapes 3D - u8s8f32 - CPUonednn: Convolution Batch Shapes Auto - f32 - CPUonednn: Deconvolution Batch shapes_1d - f32 - CPUonednn: Deconvolution Batch shapes_3d - f32 - CPUonednn: Convolution Batch Shapes Auto - u8s8f32 - CPUonednn: Deconvolution Batch shapes_1d - u8s8f32 - CPUonednn: Deconvolution Batch shapes_3d - u8s8f32 - CPUonednn: Recurrent Neural Network Training - f32 - CPUonednn: Recurrent Neural Network Inference - f32 - CPUonednn: Recurrent Neural Network Training - u8s8f32 - CPUonednn: Recurrent Neural Network Inference - u8s8f32 - CPUonednn: Matrix Multiply Batch Shapes Transformer - f32 - CPUonednn: Recurrent Neural Network Training - bf16bf16bf16 - CPUonednn: Recurrent Neural Network Inference - bf16bf16bf16 - CPUonednn: Matrix Multiply Batch Shapes Transformer - u8s8f32 - CPUabcc3.925265.521212.445561.349621.039839.822692.089006.482811.729260.9959895197.151291.225182.241281.828.643835273.921358.9312.76272.033997.54122.620541.367471.108059.889782.095616.580941.748840.9755955182.41291.115208.021324.2910.28065109.341314.1610.61971.650817.68972.420891.42830.94129110.11482.072446.673251.60670.9578025246.91297.125236.441322.5710.41485269.951268.4713.8974OpenBenchmarking.org

oneDNN

Harness: IP Shapes 1D - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: IP Shapes 1D - Data Type: f32 - Engine: CPUabcc0.88321.76642.64963.53284.416SE +/- 1.76988, N = 123.925262.033991.65081MIN: 1.42MIN: 1.68MIN: 1.431. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: IP Shapes 3D - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: IP Shapes 3D - Data Type: f32 - Engine: CPUabcc246810SE +/- 0.02020, N = 35.521217.541207.68970MIN: 5.21MIN: 7.39MIN: 7.561. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: IP Shapes 1D - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: IP Shapes 1D - Data Type: u8s8f32 - Engine: CPUabcc0.58961.17921.76882.35842.948SE +/- 0.05461, N = 152.445562.620542.42089MIN: 1.4MIN: 2.09MIN: 1.861. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPUabcc0.32140.64280.96421.28561.607SE +/- 0.01605, N = 121.349621.367471.42830MIN: 1.05MIN: 1.1MIN: 1.161. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPUabcc0.24930.49860.74790.99721.2465SE +/- 0.014383, N = 31.0398301.1080500.941291MIN: 0.95MIN: 1.04MIN: 0.881. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Deconvolution Batch shapes_1d - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Deconvolution Batch shapes_1d - Data Type: f32 - Engine: CPUabcc3691215SE +/- 0.08002, N = 39.822699.8897810.11480MIN: 8.11MIN: 8.3MIN: 8.291. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPUabcc0.47150.9431.41451.8862.3575SE +/- 0.00438, N = 32.089002.095612.07244MIN: 2.03MIN: 2.04MIN: 2.031. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Convolution Batch Shapes Auto - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Convolution Batch Shapes Auto - Data Type: u8s8f32 - Engine: CPUabcc246810SE +/- 0.00429, N = 36.482816.580946.67325MIN: 6.36MIN: 6.49MIN: 6.561. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPUabcc0.39350.7871.18051.5741.9675SE +/- 0.01830, N = 151.729261.748841.60670MIN: 1.48MIN: 1.52MIN: 1.461. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPUabcc0.22410.44820.67230.89641.1205SE +/- 0.002014, N = 30.9959890.9755950.957802MIN: 0.93MIN: 0.92MIN: 0.921. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPUabcc11002200330044005500SE +/- 55.54, N = 55197.155182.405246.90MIN: 4918.58MIN: 5129.76MIN: 5190.591. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPUabcc30060090012001500SE +/- 17.88, N = 31291.221291.111297.12MIN: 1243.36MIN: 1261.39MIN: 1258.041. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPUabcc11002200330044005500SE +/- 43.75, N = 35182.245208.025236.44MIN: 5039.08MIN: 5148.07MIN: 5173.951. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPUabcc30060090012001500SE +/- 14.13, N = 41281.821324.291322.57MIN: 1216.85MIN: 1292.79MIN: 1292.361. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPUabcc3691215SE +/- 0.48813, N = 158.6438310.2806010.41480MIN: 5.39MIN: 9.92MIN: 10.111. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPUabcc11002200330044005500SE +/- 13.14, N = 35273.925109.345269.95MIN: 5193.2MIN: 5051.41MIN: 5209.891. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPUabcc30060090012001500SE +/- 76.68, N = 151358.931314.161268.47MIN: 1193.45MIN: 1256.02MIN: 1237.011. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

oneDNN

Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPU

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 3.0Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPUabcc48121620SE +/- 0.37, N = 1212.7610.6213.90MIN: 9.89MIN: 9.93MIN: 13.321. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -pie -ldl -lpthread

Phoronix Test Suite v10.8.5