tgl onnx onednn

Intel Core i7-1185G7 testing with a Dell 0DXP1F (3.4.0 BIOS) and Intel Xe TGL GT2 3GB on Ubuntu 22.04 via the Phoronix Test Suite.

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 2203319-PTS-TGLONNXO37
Jump To Table - Results

View

Do Not Show Noisy Results
Do Not Show Results With Incomplete Data
Do Not Show Results With Little Change/Spread
List Notable Results
Show Result Confidence Charts

Limit displaying results to tests within:

HPC - High Performance Computing 2 Tests
Machine Learning 2 Tests

Statistics

Show Overall Harmonic Mean(s)
Show Overall Geometric Mean
Show Geometric Means Per-Suite/Category
Show Wins / Losses Counts (Pie Chart)
Normalize Results
Remove Outliers Before Calculating Averages

Graph Settings

Force Line Graphs Where Applicable
Convert To Scalar Where Applicable
Prefer Vertical Bar Graphs

Multi-Way Comparison

Condense Multi-Option Tests Into Single Result Graphs

Table

Show Detailed System Result Table

Run Management

Highlight
Result
Hide
Result
Result
Identifier
Performance Per
Dollar
Date
Run
  Test
  Duration
A
March 30 2022
  37 Minutes
B
March 30 2022
  3 Hours, 42 Minutes
C
March 31 2022
  3 Hours, 47 Minutes
Invert Hiding All Results Option
  2 Hours, 42 Minutes

Only show results where is faster than
Only show results matching title/arguments (delimit multiple options with a comma):
Do not show results matching title/arguments (delimit multiple options with a comma):


tgl onnx onednnOpenBenchmarking.orgPhoronix Test SuiteIntel Core i7-1185G7 @ 4.80GHz (4 Cores / 8 Threads)Dell 0DXP1F (3.4.0 BIOS)Intel Tiger Lake-LP16GBMicron 2300 NVMe 512GBIntel Xe TGL GT2 3GB (1350MHz)Realtek ALC289Intel Wi-Fi 6 AX201Ubuntu 22.045.17.0-051700rc7daily20220309-generic (x86_64)GNOME Shell 41.3X Server + Wayland4.6 Mesa 21.3.51.2.195GCC 11.2.0ext41920x1200ProcessorMotherboardChipsetMemoryDiskGraphicsAudioNetworkOSKernelDesktopDisplay ServerOpenGLVulkanCompilerFile-SystemScreen ResolutionTgl Onnx Onednn BenchmarksSystem Logs- Transparent Huge Pages: madvise- --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-XWYfV6/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-XWYfV6/gcc-11-11.2.0/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-build-config=bootstrap-lto-lean --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 - Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x9a - Thermald 2.4.7 - Python 3.10.2- itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced IBRS plus Retpolines IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

ABCResult OverviewPhoronix Test Suite100%110%121%131%142%ONNX RuntimeONNX RuntimeONNX RuntimeONNX RuntimeoneDNNoneDNNoneDNNONNX RuntimeoneDNNoneDNNONNX RuntimeoneDNNoneDNNoneDNNoneDNNONNX RuntimeoneDNNoneDNNoneDNNoneDNNoneDNNONNX RuntimeoneDNNoneDNNONNX RuntimeoneDNNONNX RuntimeONNX RuntimeoneDNNONNX RuntimeoneDNNoneDNNoneDNNoneDNNoneDNNoneDNNsuper-resolution-10 - CPU - Standardbertsquad-12 - CPU - Standardfcn-resnet101-11 - CPU - Standardfcn-resnet101-11 - CPU - ParallelIP Shapes 1D - f32 - CPUD.B.s - u8s8f32 - CPUIP Shapes 1D - u8s8f32 - CPUArcFace ResNet-100 - CPU - ParallelC.B.S.A - f32 - CPUIP Shapes 1D - bf16bf16bf16 - CPUsuper-resolution-10 - CPU - ParallelIP Shapes 3D - bf16bf16bf16 - CPUC.B.S.A - u8s8f32 - CPUR.N.N.T - f32 - CPUD.B.s - f32 - CPUArcFace ResNet-100 - CPU - StandardD.B.s - f32 - CPUD.B.s - bf16bf16bf16 - CPUM.M.B.S.T - bf16bf16bf16 - CPUD.B.s - u8s8f32 - CPUIP Shapes 3D - f32 - CPUbertsquad-12 - CPU - ParallelIP Shapes 3D - u8s8f32 - CPUD.B.s - bf16bf16bf16 - CPUGPT-2 - CPU - ParallelM.M.B.S.T - u8s8f32 - CPUyolov4 - CPU - Parallelyolov4 - CPU - StandardR.N.N.I - bf16bf16bf16 - CPUGPT-2 - CPU - StandardR.N.N.T - u8s8f32 - CPUC.B.S.A - bf16bf16bf16 - CPUR.N.N.T - bf16bf16bf16 - CPUR.N.N.I - u8s8f32 - CPUM.M.B.S.T - f32 - CPUR.N.N.I - f32 - CPU

tgl onnx onednnonnx: bertsquad-12 - CPU - Parallelonnx: super-resolution-10 - CPU - Standardonnx: fcn-resnet101-11 - CPU - Standardonnx: fcn-resnet101-11 - CPU - Parallelonednn: Recurrent Neural Network Training - f32 - CPUonnx: bertsquad-12 - CPU - Standardonnx: GPT-2 - CPU - Parallelonnx: yolov4 - CPU - Parallelonnx: GPT-2 - CPU - Standardonnx: ArcFace ResNet-100 - CPU - Parallelonnx: ArcFace ResNet-100 - CPU - Standardonnx: yolov4 - CPU - Standardonnx: super-resolution-10 - CPU - Parallelonednn: Recurrent Neural Network Training - u8s8f32 - CPUonednn: Recurrent Neural Network Training - bf16bf16bf16 - CPUonednn: Deconvolution Batch shapes_1d - f32 - CPUonednn: Deconvolution Batch shapes_1d - u8s8f32 - CPUonednn: Recurrent Neural Network Inference - u8s8f32 - CPUonednn: Recurrent Neural Network Inference - bf16bf16bf16 - CPUonednn: Recurrent Neural Network Inference - f32 - CPUonednn: IP Shapes 1D - bf16bf16bf16 - CPUonednn: Deconvolution Batch shapes_1d - bf16bf16bf16 - CPUonednn: IP Shapes 1D - f32 - CPUonednn: IP Shapes 1D - u8s8f32 - CPUonednn: Matrix Multiply Batch Shapes Transformer - bf16bf16bf16 - CPUonednn: Convolution Batch Shapes Auto - u8s8f32 - CPUonednn: Convolution Batch Shapes Auto - f32 - CPUonednn: Deconvolution Batch shapes_3d - f32 - CPUonednn: Deconvolution Batch shapes_3d - u8s8f32 - CPUonednn: Deconvolution Batch shapes_3d - bf16bf16bf16 - CPUonednn: Matrix Multiply Batch Shapes Transformer - f32 - CPUonednn: Matrix Multiply Batch Shapes Transformer - u8s8f32 - CPUonednn: IP Shapes 3D - bf16bf16bf16 - CPUonednn: IP Shapes 3D - f32 - CPUonednn: IP Shapes 3D - u8s8f32 - CPUonednn: Convolution Batch Shapes Auto - bf16bf16bf16 - CPUABC222248840337179.623694466181638855076126218797183.167180.8116.25252.066173703.273702.73708.3124.34858.20556.736381.6083310.80527.853658.75999.995642.4265541.62583.201341.43086.319545.926862.457650.9273227220837327833.653724513182637655477826118887189.337188.5516.14772.467203706.433708.523706.0122.592258.44898.206471.9170610.79268.6829510.1732110.819772.4673638.94673.202911.433716.487285.983012.4432350.9499221175630257177.772784470182637346572126116467195.107189.0615.06202.412353702.643711.783708.6220.849157.26397.332321.6143710.288978.7227210.2447810.58502.5209139.25223.204181.425327.074776.120002.4980950.8909OpenBenchmarking.org

ONNX Runtime

ONNX Runtime is developed by Microsoft and partners as a open-source, cross-platform, high performance machine learning inferencing and training accelerator. This test profile runs the ONNX Runtime with various models available from the ONNX Zoo. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: bertsquad-12 - Device: CPU - Executor: ParallelBAC50100150200250SE +/- 5.99, N = 12SE +/- 4.21, N = 122272222211. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: super-resolution-10 - Device: CPU - Executor: StandardABC5001000150020002500SE +/- 121.56, N = 12SE +/- 78.17, N = 122488220817561. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: fcn-resnet101-11 - Device: CPU - Executor: StandardABC918273645SE +/- 1.89, N = 10SE +/- 2.14, N = 124037301. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: fcn-resnet101-11 - Device: CPU - Executor: ParallelABC816243240SE +/- 0.17, N = 3SE +/- 0.29, N = 123332251. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

oneDNN

This is a test of the Intel oneDNN as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Intel oneDNN was formerly known as DNNL (Deep Neural Network Library) and MKL-DNN before being rebranded as part of Intel oneAPI. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPUCAB2K4K6K8K10KSE +/- 12.61, N = 3SE +/- 462.12, N = 157177.777179.627833.65MIN: 7109.7MIN: 7126.95MIN: 7123.391. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

ONNX Runtime

ONNX Runtime is developed by Microsoft and partners as a open-source, cross-platform, high performance machine learning inferencing and training accelerator. This test profile runs the ONNX Runtime with various models available from the ONNX Zoo. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: bertsquad-12 - Device: CPU - Executor: StandardBAC80160240320400SE +/- 0.17, N = 3SE +/- 19.85, N = 123723692781. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: GPT-2 - Device: CPU - Executor: ParallelBCA10002000300040005000SE +/- 37.78, N = 12SE +/- 27.09, N = 34513447044661. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: yolov4 - Device: CPU - Executor: ParallelCBA4080120160200SE +/- 1.32, N = 3SE +/- 1.64, N = 31821821811. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: GPT-2 - Device: CPU - Executor: StandardABC14002800420056007000SE +/- 1.36, N = 3SE +/- 4.16, N = 36388637663731. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: ArcFace ResNet-100 - Device: CPU - Executor: ParallelBAC120240360480600SE +/- 1.86, N = 3SE +/- 3.82, N = 35545504651. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: ArcFace ResNet-100 - Device: CPU - Executor: StandardBAC2004006008001000SE +/- 0.17, N = 3SE +/- 5.85, N = 37787617211. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: yolov4 - Device: CPU - Executor: StandardACB60120180240300SE +/- 0.17, N = 3SE +/- 0.17, N = 32622612611. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

OpenBenchmarking.orgInferences Per Minute, More Is BetterONNX Runtime 1.11Model: super-resolution-10 - Device: CPU - Executor: ParallelBAC400800120016002000SE +/- 8.61, N = 3SE +/- 3.28, N = 31888187916461. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt

oneDNN

This is a test of the Intel oneDNN as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Intel oneDNN was formerly known as DNNL (Deep Neural Network Library) and MKL-DNN before being rebranded as part of Intel oneAPI. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPUABC15003000450060007500SE +/- 1.53, N = 3SE +/- 5.73, N = 37183.167189.337195.10MIN: 7143.58MIN: 7134.86MIN: 7136.261. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPUABC15003000450060007500SE +/- 2.56, N = 3SE +/- 7.40, N = 37180.817188.557189.06MIN: 7130.5MIN: 7142.75MIN: 7132.271. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_1d - Data Type: f32 - Engine: CPUCBA48121620SE +/- 0.35, N = 15SE +/- 0.38, N = 1515.0616.1516.25MIN: 11.2MIN: 11.65MIN: 13.21. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPUACB0.55511.11021.66532.22042.7755SE +/- 0.05576, N = 15SE +/- 0.04201, N = 152.066172.412352.46720MIN: 1.84MIN: 1.84MIN: 1.921. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPUCAB8001600240032004000SE +/- 8.92, N = 3SE +/- 1.82, N = 33702.643703.273706.43MIN: 3646.09MIN: 3664.66MIN: 3672.11. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPUABC8001600240032004000SE +/- 1.76, N = 3SE +/- 1.69, N = 33702.703708.523711.78MIN: 3669.11MIN: 3671.55MIN: 3673.421. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPUBAC8001600240032004000SE +/- 0.73, N = 3SE +/- 0.71, N = 33706.013708.313708.62MIN: 3668.11MIN: 3667.77MIN: 3673.451. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 1D - Data Type: bf16bf16bf16 - Engine: CPUCBA612182430SE +/- 0.79, N = 15SE +/- 0.77, N = 1520.8522.5924.35MIN: 17.06MIN: 17.03MIN: 23.811. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_1d - Data Type: bf16bf16bf16 - Engine: CPUCAB1326395265SE +/- 0.89, N = 15SE +/- 0.71, N = 357.2658.2158.45MIN: 46.22MIN: 55.58MIN: 55.571. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 1D - Data Type: f32 - Engine: CPUACB246810SE +/- 0.18274, N = 12SE +/- 0.29521, N = 126.736387.332328.20647MIN: 6.11MIN: 5.99MIN: 5.821. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 1D - Data Type: u8s8f32 - Engine: CPUACB0.43130.86261.29391.72522.1565SE +/- 0.02179, N = 3SE +/- 0.02467, N = 151.608331.614371.91706MIN: 1.43MIN: 1.36MIN: 1.391. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Matrix Multiply Batch Shapes Transformer - Data Type: bf16bf16bf16 - Engine: CPUCBA3691215SE +/- 0.23, N = 15SE +/- 0.03, N = 310.2910.7910.81MIN: 7.37MIN: 10.35MIN: 10.41. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Convolution Batch Shapes Auto - Data Type: u8s8f32 - Engine: CPUABC246810SE +/- 0.08839, N = 15SE +/- 0.08702, N = 157.853658.682958.72272MIN: 7.67MIN: 7.61MIN: 7.71. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPUABC3691215SE +/- 0.14982, N = 13SE +/- 0.08939, N = 148.7599010.1732110.24478MIN: 8.17MIN: 8.18MIN: 8.191. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPUACB3691215SE +/- 0.19882, N = 15SE +/- 0.21609, N = 159.9956410.5850010.81977MIN: 9.67MIN: 9.56MIN: 9.581. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPUABC0.56721.13441.70162.26882.836SE +/- 0.02316, N = 15SE +/- 0.03720, N = 152.426552.467362.52091MIN: 2.2MIN: 2.2MIN: 2.21. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Deconvolution Batch shapes_3d - Data Type: bf16bf16bf16 - Engine: CPUBCA918273645SE +/- 1.01, N = 12SE +/- 0.94, N = 1538.9539.2541.63MIN: 35.45MIN: 35.42MIN: 35.451. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPUABC0.72091.44182.16272.88363.6045SE +/- 0.00447, N = 3SE +/- 0.00537, N = 33.201343.202913.20418MIN: 3.11MIN: 3.11MIN: 3.11. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPUCAB0.32260.64520.96781.29041.613SE +/- 0.00427, N = 3SE +/- 0.00013, N = 31.425321.430801.43371MIN: 1.37MIN: 1.38MIN: 1.381. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 3D - Data Type: bf16bf16bf16 - Engine: CPUABC246810SE +/- 0.06647, N = 5SE +/- 0.05421, N = 36.319546.487287.07477MIN: 5.74MIN: 5.73MIN: 5.921. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 3D - Data Type: f32 - Engine: CPUABC246810SE +/- 0.00832, N = 3SE +/- 0.02249, N = 35.926865.983016.12000MIN: 5.65MIN: 5.68MIN: 5.861. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPUBAC0.56211.12421.68632.24842.8105SE +/- 0.00609, N = 3SE +/- 0.01292, N = 32.443232.457602.49809MIN: 2.34MIN: 2.34MIN: 2.311. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.6Harness: Convolution Batch Shapes Auto - Data Type: bf16bf16bf16 - Engine: CPUCAB1122334455SE +/- 0.08, N = 3SE +/- 0.02, N = 350.8950.9350.95MIN: 47.81MIN: 50.57MIN: 50.571. (CXX) g++ options: -O3 -march=native -fopenmp -msse4.1 -fPIC -std=c++11 -pie -ldl

36 Results Shown

ONNX Runtime:
  bertsquad-12 - CPU - Parallel
  super-resolution-10 - CPU - Standard
  fcn-resnet101-11 - CPU - Standard
  fcn-resnet101-11 - CPU - Parallel
oneDNN
ONNX Runtime:
  bertsquad-12 - CPU - Standard
  GPT-2 - CPU - Parallel
  yolov4 - CPU - Parallel
  GPT-2 - CPU - Standard
  ArcFace ResNet-100 - CPU - Parallel
  ArcFace ResNet-100 - CPU - Standard
  yolov4 - CPU - Standard
  super-resolution-10 - CPU - Parallel
oneDNN:
  Recurrent Neural Network Training - u8s8f32 - CPU
  Recurrent Neural Network Training - bf16bf16bf16 - CPU
  Deconvolution Batch shapes_1d - f32 - CPU
  Deconvolution Batch shapes_1d - u8s8f32 - CPU
  Recurrent Neural Network Inference - u8s8f32 - CPU
  Recurrent Neural Network Inference - bf16bf16bf16 - CPU
  Recurrent Neural Network Inference - f32 - CPU
  IP Shapes 1D - bf16bf16bf16 - CPU
  Deconvolution Batch shapes_1d - bf16bf16bf16 - CPU
  IP Shapes 1D - f32 - CPU
  IP Shapes 1D - u8s8f32 - CPU
  Matrix Multiply Batch Shapes Transformer - bf16bf16bf16 - CPU
  Convolution Batch Shapes Auto - u8s8f32 - CPU
  Convolution Batch Shapes Auto - f32 - CPU
  Deconvolution Batch shapes_3d - f32 - CPU
  Deconvolution Batch shapes_3d - u8s8f32 - CPU
  Deconvolution Batch shapes_3d - bf16bf16bf16 - CPU
  Matrix Multiply Batch Shapes Transformer - f32 - CPU
  Matrix Multiply Batch Shapes Transformer - u8s8f32 - CPU
  IP Shapes 3D - bf16bf16bf16 - CPU
  IP Shapes 3D - f32 - CPU
  IP Shapes 3D - u8s8f32 - CPU
  Convolution Batch Shapes Auto - bf16bf16bf16 - CPU