GCC 11 vs. LLVM Clang 12 Benchmarks On Xeon Ice Lake

Xeon Platinum 8380 compiler benchmarks by Michael Larabel looking at GCC 11 against LLVM Clang 12 for some initial holiday weekend tests...

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 2105299-IB-COMPILERT91
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

Limit displaying results to tests within:

Audio Encoding 4 Tests
Bioinformatics 2 Tests
C/C++ Compiler Tests 16 Tests
CPU Massive 16 Tests
Creator Workloads 18 Tests
Cryptography 3 Tests
Encoding 9 Tests
HPC - High Performance Computing 8 Tests
Imaging 4 Tests
Common Kernel Benchmarks 2 Tests
Machine Learning 4 Tests
Multi-Core 13 Tests
NVIDIA GPU Compute 2 Tests
Renderers 2 Tests
Scientific Computing 3 Tests
Server 2 Tests
Server CPU Tests 10 Tests
Single-Threaded 5 Tests
Video Encoding 5 Tests
Common Workstation Benchmarks 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
View Logs
Performance Per
Dollar
Date
Run
  Test
  Duration
GCC 11.1
May 28 2021
  7 Hours, 55 Minutes
Clang 12.0
May 28 2021
  6 Hours, 38 Minutes
Invert Hiding All Results Option
  7 Hours, 16 Minutes
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):

GCC 11.1

Processor: 2 x Intel Xeon Platinum 8380 @ 3.40GHz (80 Cores / 160 Threads), Motherboard: Intel M50CYP2SB2U (SE5C6200.86B.0022.D08.2103221623 BIOS), Chipset: Intel Device 0998, Memory: 16 x 32 GB DDR4-3200MT/s Hynix HMA84GR7CJR4N-XN, Disk: 800GB INTEL SSDPF21Q800GB, Graphics: ASPEED, Network: 2 x Intel X710 for 10GBASE-T + 2 x Intel E810-C for QSFP

OS: Fedora 34, Kernel: 5.12.6-300.fc34.x86_64 (x86_64), Compiler: GCC 11.1.1 20210428, File-System: xfs, Screen Resolution: 1024x768

Kernel Notes: Transparent Huge Pages: madvise
Environment Notes: CXXFLAGS="-O3 -march=native -flto" CFLAGS="-O3 -march=native -flto"
Compiler Notes: --build=x86_64-redhat-linux --disable-libunwind-exceptions --enable-__cxa_atexit --enable-bootstrap --enable-cet --enable-checking=release --enable-gnu-indirect-function --enable-gnu-unique-object --enable-initfini-array --enable-languages=c,c++,fortran,objc,obj-c++,ada,go,d,lto --enable-multilib --enable-offload-targets=nvptx-none --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-arch_32=i686 --with-gcc-major-version-only --with-linker-hash-style=gnu --with-tune=generic --without-cuda-driver
Processor Notes: Scaling Governor: intel_pstate performance - CPU Microcode: 0xd000270
Python Notes: Python 3.9.5
Security Notes: SELinux + 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 Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

Clang 12.0

OS: Fedora 34, Kernel: 5.12.6-300.fc34.x86_64 (x86_64), Compiler: Clang 12.0.0, File-System: xfs, Screen Resolution: 1024x768

Kernel Notes: Transparent Huge Pages: madvise
Environment Notes: CXXFLAGS="-O3 -march=native -flto" CFLAGS="-O3 -march=native -flto"
Processor Notes: Scaling Governor: intel_pstate performance - CPU Microcode: 0xd000270
Python Notes: Python 3.9.5
Security Notes: SELinux + 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 Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

GCC 11 vs. LLVM Clang 12 Benchmarks On Xeon Ice LakeOpenBenchmarking.orgPhoronix Test Suite2 x Intel Xeon Platinum 8380 @ 3.40GHz (80 Cores / 160 Threads)Intel M50CYP2SB2U (SE5C6200.86B.0022.D08.2103221623 BIOS)Intel Device 099816 x 32 GB DDR4-3200MT/s Hynix HMA84GR7CJR4N-XN800GB INTEL SSDPF21Q800GBASPEED2 x Intel X710 for 10GBASE-T + 2 x Intel E810-C for QSFPFedora 345.12.6-300.fc34.x86_64 (x86_64)GCC 11.1.1 20210428Clang 12.0.0xfs1024x768ProcessorMotherboardChipsetMemoryDiskGraphicsNetworkOSKernelCompilersFile-SystemScreen ResolutionGCC 11 Vs. LLVM Clang 12 Benchmarks On Xeon Ice Lake PerformanceSystem Logs- Transparent Huge Pages: madvise- CXXFLAGS="-O3 -march=native -flto" CFLAGS="-O3 -march=native -flto"- GCC 11.1: --build=x86_64-redhat-linux --disable-libunwind-exceptions --enable-__cxa_atexit --enable-bootstrap --enable-cet --enable-checking=release --enable-gnu-indirect-function --enable-gnu-unique-object --enable-initfini-array --enable-languages=c,c++,fortran,objc,obj-c++,ada,go,d,lto --enable-multilib --enable-offload-targets=nvptx-none --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-arch_32=i686 --with-gcc-major-version-only --with-linker-hash-style=gnu --with-tune=generic --without-cuda-driver - Scaling Governor: intel_pstate performance - CPU Microcode: 0xd000270- Python 3.9.5- SELinux + 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 Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

GCC 11.1 vs. Clang 12.0 ComparisonPhoronix Test SuiteBaseline+64.2%+64.2%+128.4%+128.4%+192.6%+192.6%256.6%158.4%126.5%98%82.6%80%79.3%78.8%61.7%61.6%54.7%42.4%38.9%36.5%35.8%35.3%27.2%26.9%26.3%25.6%22.6%17.9%16.9%16.7%16.4%15.9%15.8%15.5%15.2%14.3%11.2%8.5%8.2%8.1%8%7.8%7.5%6.9%6.3%6.1%6%5.7%5.6%5.3%4.7%4.5%4.5%4.1%3.9%3.9%3.7%3.5%3.5%3.4%3%2.8%2.6%2.5%2.4%2%CPU - regnety_400mCPU - blazefaceCPU-v3-v3 - mobilenet-v3CPU-v2-v2 - mobilenet-v2Total Time - 4.1.R.P.P97%M.M.B.S.T - u8s8f32 - CPUD.B.s - u8s8f32 - CPUCPU - mnasnetCPU - shufflenet-v2CPU - efficientnet-b0ResizingM.M.B.S.T - f32 - CPUR.4.b.P54.1%CPU - MobileNet v246.5%MediumIP Shapes 1D - f32 - CPUCPU - mobilenetThoroughIP Shapes 3D - u8s8f32 - CPUR.N.N.I - f32 - CPUCPU - googlenetR.N.N.I - u8s8f32 - CPUR.N.N.I - bf16bf16bf16 - CPU8, Long Mode - Compression Speed25.4%IP Shapes 1D - u8s8f32 - CPU1000 Stack21.5%CoreMark Size 666 - I.P.S18.4%CPU - squeezenet_ssdExhaustiveSharpen16.8%M.M.B.S.T - bf16bf16bf16 - CPUR.N.N.T - f32 - CPUR.N.N.T - bf16bf16bf16 - CPU160 - 256 - 57R.N.N.T - u8s8f32 - CPUEnhanced15.2%IP Shapes 3D - f32 - CPUD.B.s - bf16bf16bf16 - CPUQ.1.C.E.512.7%P.P.S11.8%WAV To Opus Encode11.6%WAV To MP311.3%D.B.s - u8s8f32 - CPU3000 Fall11.1%10.9%136 Ragdolls9.6%CPU - yolov4-tiny9.3%1000 Convex9.2%Prim Trimesh9.2%Convex Trimesh8.9%IP Shapes 1D - bf16bf16bf16 - CPUBosphorus 4KQ.1.H.CUnkeyed AlgorithmsCPU - vgg167.9%Q.9.C.E.7C.B.S.A - u8s8f32 - CPUQ.7.C.E.7CPU - SqueezeNet v1.16.8%2048 x 2048 - Total Time6.5%IP Shapes 3D - bf16bf16bf16 - CPUDefault1 - Bosphorus 1080pBosphorus 4K - Very Fast7 - Bosphorus 1080p8 - Compression SpeedPreset 4 - Bosphorus 1080p100 - 250 - Read Only - Average LatencyBosphorus 1080p - Ultra FastQ.1.L4.1%Bosphorus 1080p - Very Fast100 - 250 - Read Only19, Long Mode - Compression Speed3.7%D.T100 - 250 - Read Write100 - 250 - Read Write - Average LatencyV.Q.O - Bosphorus 1080p3.5%C.B.S.A - f32 - CPUQ.1.L.H.C3.1%Q.1.L.C3%Bosphorus 4K - Ultra FastPreset 4 - Bosphorus 4K19 - Compression Speed2.8%P.P.AP.S.O - Bosphorus 1080p2.5%CPU - resnet18T.T.S.SVMAF Optimized - Bosphorus 1080p2%Preset 8 - Bosphorus 4KNCNNNCNNNCNNNCNNC-RayoneDNNoneDNNNCNNNCNNNCNNGraphicsMagickoneDNNOpenSSLTNNASTC EncoderoneDNNNCNNASTC EncoderoneDNNoneDNNNCNNoneDNNoneDNNZstd CompressiononeDNNBullet Physics EngineCoremarkNCNNASTC EncoderGraphicsMagickoneDNNoneDNNoneDNNLiquid-DSPoneDNNGraphicsMagickoneDNNoneDNNWebP2 Image EncodeHimeno BenchmarkOpus Codec EncodingLAME MP3 EncodingoneDNNBullet Physics EngineKripkeBullet Physics EngineNCNNBullet Physics EngineBullet Physics EngineBullet Physics EngineoneDNNx265WebP Image EncodeCrypto++NCNNWebP2 Image EncodeoneDNNWebP2 Image EncodeTNNAOBenchoneDNNWebP2 Image EncodeSVT-HEVCKvazaarSVT-HEVCZstd CompressionSVT-AV1PostgreSQL pgbenchGcrypt LibraryKvazaarWebP Image EncodeKvazaarPostgreSQL pgbenchZstd Compressionlibjpeg-turbo tjbenchPostgreSQL pgbenchPostgreSQL pgbenchSVT-VP9oneDNNWebP Image EncodeWebP2 Image EncodeKvazaarSVT-AV1Zstd CompressionTimed MrBayes AnalysisSVT-VP9NCNNeSpeak-NG Speech EngineSVT-VP9SVT-AV1GCC 11.1Clang 12.0

GCC 11 vs. LLVM Clang 12 Benchmarks On Xeon Ice Lakencnn: CPU - blazefacencnn: CPU-v3-v3 - mobilenet-v3ncnn: CPU-v2-v2 - mobilenet-v2c-ray: Total Time - 4K, 16 Rays Per Pixelonednn: Matrix Multiply Batch Shapes Transformer - u8s8f32 - CPUonednn: Deconvolution Batch shapes_1d - u8s8f32 - CPUonednn: Matrix Multiply Batch Shapes Transformer - f32 - CPUopenssl: RSA 4096-bit Performancetnn: CPU - MobileNet v2astcenc: Mediumonednn: IP Shapes 1D - f32 - CPUastcenc: Thoroughonednn: IP Shapes 3D - u8s8f32 - CPUonednn: Recurrent Neural Network Inference - f32 - CPUonednn: Recurrent Neural Network Inference - u8s8f32 - CPUonednn: Recurrent Neural Network Inference - bf16bf16bf16 - CPUcompress-zstd: 8, Long Mode - Compression Speedonednn: IP Shapes 1D - u8s8f32 - CPUbullet: 1000 Stackcoremark: CoreMark Size 666 - Iterations Per Secondastcenc: Exhaustivegraphics-magick: Sharpenonednn: Matrix Multiply Batch Shapes Transformer - bf16bf16bf16 - CPUonednn: Recurrent Neural Network Training - f32 - CPUonednn: Recurrent Neural Network Training - bf16bf16bf16 - CPUliquid-dsp: 160 - 256 - 57onednn: Recurrent Neural Network Training - u8s8f32 - CPUgraphics-magick: Enhancedonednn: IP Shapes 3D - f32 - CPUonednn: Deconvolution Batch shapes_1d - bf16bf16bf16 - CPUwebp2: Quality 100, Compression Effort 5himeno: Poisson Pressure Solverencode-opus: WAV To Opus Encodeencode-mp3: WAV To MP3onednn: Deconvolution Batch shapes_3d - u8s8f32 - CPUbullet: 3000 Fallkripke: bullet: 136 Ragdollsncnn: CPU - yolov4-tinybullet: 1000 Convexbullet: Prim Trimeshbullet: Convex Trimeshonednn: IP Shapes 1D - bf16bf16bf16 - CPUx265: Bosphorus 4Kwebp: Quality 100, Highest Compressioncryptopp: Unkeyed Algorithmswebp2: Quality 95, Compression Effort 7onednn: Convolution Batch Shapes Auto - u8s8f32 - CPUwebp2: Quality 75, Compression Effort 7tnn: CPU - SqueezeNet v1.1aobench: 2048 x 2048 - Total Timeonednn: IP Shapes 3D - bf16bf16bf16 - CPUwebp2: Defaultsvt-hevc: 1 - Bosphorus 1080pkvazaar: Bosphorus 4K - Very Fastsvt-hevc: 7 - Bosphorus 1080pcompress-zstd: 8 - Compression Speedsvt-av1: Preset 4 - Bosphorus 1080pgcrypt: kvazaar: Bosphorus 1080p - Ultra Fastwebp: Quality 100, Losslesskvazaar: Bosphorus 1080p - Very Fastcompress-zstd: 19, Long Mode - Compression Speedtjbench: Decompression Throughputpgbench: 100 - 250 - Read Writepgbench: 100 - 250 - Read Write - Average Latencysvt-vp9: Visual Quality Optimized - Bosphorus 1080ponednn: Convolution Batch Shapes Auto - f32 - CPUwebp: Quality 100, Lossless, Highest Compressionwebp2: Quality 100, Lossless Compressionkvazaar: Bosphorus 4K - Ultra Fastsvt-av1: Preset 4 - Bosphorus 4Kcompress-zstd: 19 - Compression Speedmrbayes: Primate Phylogeny Analysissvt-vp9: PSNR/SSIM Optimized - Bosphorus 1080pespeak: Text-To-Speech Synthesissvt-vp9: VMAF Optimized - Bosphorus 1080psvt-av1: Preset 8 - Bosphorus 4Kwebp: Quality 100compress-zstd: 19 - Decompression Speedcompress-zstd: 19, Long Mode - Decompression Speedliquid-dsp: 1 - 256 - 57webp: Defaultsvt-av1: Preset 8 - Bosphorus 1080pprimesieve: 1e12 Prime Number Generationonednn: Convolution Batch Shapes Auto - bf16bf16bf16 - CPUcompress-zstd: 8 - Decompression Speedcompress-zstd: 8, Long Mode - Decompression Speedencode-flac: WAV To FLACx265: Bosphorus 1080ponednn: Deconvolution Batch shapes_3d - f32 - CPUcaffe: AlexNet - CPU - 200caffe: GoogleNet - CPU - 200svt-hevc: 10 - Bosphorus 1080pencode-wavpack: WAV To WavPackonednn: Deconvolution Batch shapes_3d - bf16bf16bf16 - CPUdaphne: OpenMP - Euclidean Clusterdaphne: OpenMP - Points2Imagedaphne: OpenMP - NDT Mappinggraphics-magick: Rotategmpbench: Total Timencnn: CPU - regnety_400mncnn: CPU - squeezenet_ssdncnn: CPU - resnet18ncnn: CPU - vgg16ncnn: CPU - googlenetncnn: CPU - efficientnet-b0ncnn: CPU - mnasnetncnn: CPU - shufflenet-v2ncnn: CPU - mobilenetpgbench: 100 - 250 - Read Only - Average Latencypgbench: 100 - 250 - Read Onlygraphics-magick: ResizingGCC 11.1Clang 12.06.159.569.807.7940.2198990.3616910.24755417804.2376.7416.42700.9238769.42190.439708448.247445.263444.2281040.61.237074.4517652522898.56822216.38858980.601530686.190677.7083182866667686.04513151.402683.278295.7654651.5279088.7688.6190.1945873.8730601776136002.55330323.124.340600.862001.0542733.0025226.928.026359.900917196.4860.969972106.658377.43033.8811.813512.64439.5942.31336.372611.08.996265.172176.6619.473159.7147.9174.229617894252.797393.171.4593040.912389.07046.584.21383.8142.197477.9030.511476.8555.1902.6452537.32670.8609853331.638167.3883.7802.092462959.33168.49.38276.880.840383298291662408609.5617.3603.571271013.9914507.8061035371046.607453871.694.3821.1811.1025.3419.4612.489.4310.5519.400.2779074013802.384.224.9515.3520.1204060.2009170.16001711555.8552.0284.51200.6653556.93800.325027352.474352.588353.576830.01.008835.4105112130829.93175314.01337690.515284589.283584.5353686466667594.10911411.217262.867666.4954161.6535149.7889.5910.1750604.3018231601556752.79832225.284.7410880.9409031.1483672.7668129.137.422388.572346182.2900.90221399.777402.91436.0751.705672.49241.9644.72355.182748.19.418253.856183.9720.267166.0046.2180.596292925762.702379.931.4116742.184400.61747.964.33381.5138.549466.0629.806467.5756.2732.6922495.22632.3618403331.616169.6063.8302.066292996.73204.69.47477.240.842984297554663282608.9517.3433.5728874526.4717.9610.8327.3415.347.725.265.9014.210.265943043614OpenBenchmarking.org

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: blazefaceClang 12.0GCC 11.1246810SE +/- 0.02, N = 13SE +/- 0.05, N = 122.386.15-lomp - MIN: 2.14 / MAX: 10.57-lgomp - MIN: 5.69 / MAX: 22.061. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: blazefaceClang 12.0GCC 11.1246810Min: 2.29 / Avg: 2.38 / Max: 2.54Min: 5.91 / Avg: 6.15 / Max: 6.481. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU-v3-v3 - Model: mobilenet-v3Clang 12.0GCC 11.13691215SE +/- 0.04, N = 13SE +/- 0.09, N = 124.229.56-lomp - MIN: 3.66 / MAX: 22.54-lgomp - MIN: 8.9 / MAX: 68.781. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU-v3-v3 - Model: mobilenet-v3Clang 12.0GCC 11.13691215Min: 4.02 / Avg: 4.22 / Max: 4.47Min: 9.14 / Avg: 9.56 / Max: 10.231. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU-v2-v2 - Model: mobilenet-v2Clang 12.0GCC 11.13691215SE +/- 0.07, N = 13SE +/- 0.05, N = 124.959.80-lomp - MIN: 4.07 / MAX: 22.91-lgomp - MIN: 9.23 / MAX: 38.711. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU-v2-v2 - Model: mobilenet-v2Clang 12.0GCC 11.13691215Min: 4.68 / Avg: 4.95 / Max: 5.47Min: 9.46 / Avg: 9.8 / Max: 10.051. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

C-Ray

This is a test of C-Ray, a simple raytracer designed to test the floating-point CPU performance. This test is multi-threaded (16 threads per core), will shoot 8 rays per pixel for anti-aliasing, and will generate a 1600 x 1200 image. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterC-Ray 1.1Total Time - 4K, 16 Rays Per PixelGCC 11.1Clang 12.048121620SE +/- 0.006, N = 3SE +/- 0.018, N = 37.79415.3521. (CC) gcc options: -lm -lpthread -O3 -march=native -flto
OpenBenchmarking.orgSeconds, Fewer Is BetterC-Ray 1.1Total Time - 4K, 16 Rays Per PixelGCC 11.1Clang 12.048121620Min: 7.78 / Avg: 7.79 / Max: 7.8Min: 15.32 / Avg: 15.35 / Max: 15.381. (CC) gcc options: -lm -lpthread -O3 -march=native -flto

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.04950.0990.14850.1980.2475SE +/- 0.000357, N = 3SE +/- 0.000536, N = 30.1204060.219899-fopenmp=libomp - MIN: 0.11-fopenmp - MIN: 0.191. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.112345Min: 0.12 / Avg: 0.12 / Max: 0.12Min: 0.22 / Avg: 0.22 / Max: 0.221. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.08140.16280.24420.32560.407SE +/- 0.000390, N = 3SE +/- 0.001289, N = 30.2009170.361691-fopenmp=libomp - MIN: 0.17-fopenmp - MIN: 0.321. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_1d - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.112345Min: 0.2 / Avg: 0.2 / Max: 0.2Min: 0.36 / Avg: 0.36 / Max: 0.361. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPUClang 12.0GCC 11.10.05570.11140.16710.22280.2785SE +/- 0.000879, N = 3SE +/- 0.000532, N = 30.1600170.247554-fopenmp=libomp - MIN: 0.14-fopenmp - MIN: 0.231. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: f32 - Engine: CPUClang 12.0GCC 11.112345Min: 0.16 / Avg: 0.16 / Max: 0.16Min: 0.25 / Avg: 0.25 / Max: 0.251. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenSSL

OpenSSL is an open-source toolkit that implements SSL (Secure Sockets Layer) and TLS (Transport Layer Security) protocols. This test measures the RSA 4096-bit performance of OpenSSL. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSigns Per Second, More Is BetterOpenSSL 1.1.1RSA 4096-bit PerformanceGCC 11.1Clang 12.04K8K12K16K20KSE +/- 44.60, N = 3SE +/- 10.32, N = 317804.211555.8-Qunused-arguments1. (CC) gcc options: -pthread -m64 -O3 -march=native -flto -lssl -lcrypto -ldl
OpenBenchmarking.orgSigns Per Second, More Is BetterOpenSSL 1.1.1RSA 4096-bit PerformanceGCC 11.1Clang 12.03K6K9K12K15KMin: 17715 / Avg: 17804.2 / Max: 17849.9Min: 11535.2 / Avg: 11555.8 / Max: 11567.21. (CC) gcc options: -pthread -m64 -O3 -march=native -flto -lssl -lcrypto -ldl

TNN

TNN is an open-source deep learning reasoning framework developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterTNN 0.2.3Target: CPU - Model: MobileNet v2GCC 11.1Clang 12.0120240360480600SE +/- 2.55, N = 3SE +/- 1.30, N = 3376.74552.03-fopenmp - MIN: 373.07 / MAX: 547.54-fopenmp=libomp - MIN: 544.23 / MAX: 587.281. (CXX) g++ options: -O3 -march=native -flto -pthread -fvisibility=hidden -O2 -rdynamic -ldl
OpenBenchmarking.orgms, Fewer Is BetterTNN 0.2.3Target: CPU - Model: MobileNet v2GCC 11.1Clang 12.0100200300400500Min: 374.15 / Avg: 376.74 / Max: 381.84Min: 550.72 / Avg: 552.03 / Max: 554.631. (CXX) g++ options: -O3 -march=native -flto -pthread -fvisibility=hidden -O2 -rdynamic -ldl

ASTC Encoder

ASTC Encoder (astcenc) is for the Adaptive Scalable Texture Compression (ASTC) format commonly used with OpenGL, OpenGL ES, and Vulkan graphics APIs. This test profile does a coding test of both compression/decompression. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: MediumClang 12.0GCC 11.1246810SE +/- 0.0037, N = 3SE +/- 0.0021, N = 34.51206.42701. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread
OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: MediumClang 12.0GCC 11.13691215Min: 4.51 / Avg: 4.51 / Max: 4.52Min: 6.42 / Avg: 6.43 / Max: 6.431. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: f32 - Engine: CPUClang 12.0GCC 11.10.20790.41580.62370.83161.0395SE +/- 0.000134, N = 3SE +/- 0.001474, N = 30.6653550.923876-fopenmp=libomp - MIN: 0.61-fopenmp - MIN: 0.851. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1246810Min: 0.67 / Avg: 0.67 / Max: 0.67Min: 0.92 / Avg: 0.92 / Max: 0.931. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

ASTC Encoder

ASTC Encoder (astcenc) is for the Adaptive Scalable Texture Compression (ASTC) format commonly used with OpenGL, OpenGL ES, and Vulkan graphics APIs. This test profile does a coding test of both compression/decompression. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: ThoroughClang 12.0GCC 11.13691215SE +/- 0.0107, N = 3SE +/- 0.0219, N = 36.93809.42191. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread
OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: ThoroughClang 12.0GCC 11.13691215Min: 6.92 / Avg: 6.94 / Max: 6.96Min: 9.39 / Avg: 9.42 / Max: 9.461. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.09890.19780.29670.39560.4945SE +/- 0.001298, N = 3SE +/- 0.000912, N = 30.3250270.439708-fopenmp=libomp - MIN: 0.28-fopenmp - MIN: 0.41. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.112345Min: 0.32 / Avg: 0.33 / Max: 0.33Min: 0.44 / Avg: 0.44 / Max: 0.441. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1100200300400500SE +/- 0.21, N = 3SE +/- 3.11, N = 3352.47448.25-fopenmp=libomp - MIN: 333.78-fopenmp - MIN: 427.181. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: f32 - Engine: CPUClang 12.0GCC 11.180160240320400Min: 352.17 / Avg: 352.47 / Max: 352.88Min: 442.26 / Avg: 448.25 / Max: 452.711. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.1100200300400500SE +/- 1.62, N = 3SE +/- 1.07, N = 3352.59445.26-fopenmp=libomp - MIN: 334.39-fopenmp - MIN: 427.381. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.180160240320400Min: 350.86 / Avg: 352.59 / Max: 355.82Min: 443.34 / Avg: 445.26 / Max: 447.061. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1100200300400500SE +/- 2.41, N = 3SE +/- 1.02, N = 3353.58444.23-fopenmp=libomp - MIN: 333.93-fopenmp - MIN: 427.491. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Inference - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.180160240320400Min: 350.76 / Avg: 353.58 / Max: 358.38Min: 442.24 / Avg: 444.23 / Max: 445.631. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8, Long Mode - Compression SpeedGCC 11.1Clang 12.02004006008001000SE +/- 3.20, N = 3SE +/- 9.83, N = 31040.6830.01. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8, Long Mode - Compression SpeedGCC 11.1Clang 12.02004006008001000Min: 1034.2 / Avg: 1040.6 / Max: 1044.1Min: 817.4 / Avg: 830.03 / Max: 849.41. (CC) gcc options: -O3 -march=native -flto -pthread -lz

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.27830.55660.83491.11321.3915SE +/- 0.00107, N = 3SE +/- 0.00988, N = 91.008831.23707-fopenmp=libomp - MIN: 0.69-fopenmp - MIN: 0.881. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.1246810Min: 1.01 / Avg: 1.01 / Max: 1.01Min: 1.17 / Avg: 1.24 / Max: 1.271. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Bullet Physics Engine

This is a benchmark of the Bullet Physics Engine. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 1000 StackGCC 11.1Clang 12.01.21742.43483.65224.86966.087SE +/- 0.003456, N = 3SE +/- 0.002175, N = 34.4517655.410511-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 1000 StackGCC 11.1Clang 12.0246810Min: 4.45 / Avg: 4.45 / Max: 4.46Min: 5.41 / Avg: 5.41 / Max: 5.411. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

Coremark

This is a test of EEMBC CoreMark processor benchmark. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgIterations/Sec, More Is BetterCoremark 1.0CoreMark Size 666 - Iterations Per SecondGCC 11.1Clang 12.0500K1000K1500K2000K2500KSE +/- 1575.92, N = 3SE +/- 4844.43, N = 32522898.572130829.931. (CC) gcc options: -O2 -O3 -march=native -flto -lrt" -lrt
OpenBenchmarking.orgIterations/Sec, More Is BetterCoremark 1.0CoreMark Size 666 - Iterations Per SecondGCC 11.1Clang 12.0400K800K1200K1600K2000KMin: 2521273.24 / Avg: 2522898.57 / Max: 2526049.89Min: 2123517.96 / Avg: 2130829.93 / Max: 2139991.081. (CC) gcc options: -O2 -O3 -march=native -flto -lrt" -lrt

ASTC Encoder

ASTC Encoder (astcenc) is for the Adaptive Scalable Texture Compression (ASTC) format commonly used with OpenGL, OpenGL ES, and Vulkan graphics APIs. This test profile does a coding test of both compression/decompression. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: ExhaustiveClang 12.0GCC 11.148121620SE +/- 0.01, N = 3SE +/- 0.01, N = 314.0116.391. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread
OpenBenchmarking.orgSeconds, Fewer Is BetterASTC Encoder 2.4Preset: ExhaustiveClang 12.0GCC 11.148121620Min: 13.99 / Avg: 14.01 / Max: 14.03Min: 16.38 / Avg: 16.39 / Max: 16.41. (CXX) g++ options: -O3 -march=native -flto -O2 -pthread

GraphicsMagick

This is a test of GraphicsMagick with its OpenMP implementation that performs various imaging tests on a sample 6000x4000 pixel JPEG image. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: SharpenGCC 11.1Clang 12.02004006008001000SE +/- 2.40, N = 3SE +/- 2.73, N = 38987691. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: SharpenGCC 11.1Clang 12.0160320480640800Min: 895 / Avg: 898.33 / Max: 903Min: 765 / Avg: 768.67 / Max: 7741. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.10.13530.27060.40590.54120.6765SE +/- 0.001561, N = 3SE +/- 0.000449, N = 30.5152840.601530-fopenmp=libomp - MIN: 0.48-fopenmp - MIN: 0.561. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Matrix Multiply Batch Shapes Transformer - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1246810Min: 0.51 / Avg: 0.52 / Max: 0.52Min: 0.6 / Avg: 0.6 / Max: 0.61. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1150300450600750SE +/- 2.16, N = 3SE +/- 3.66, N = 3589.28686.19-fopenmp=libomp - MIN: 555.66-fopenmp - MIN: 652.781. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1120240360480600Min: 585.43 / Avg: 589.28 / Max: 592.89Min: 681.38 / Avg: 686.19 / Max: 693.361. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1150300450600750SE +/- 0.97, N = 3SE +/- 2.54, N = 3584.54677.71-fopenmp=libomp - MIN: 556.26-fopenmp - MIN: 648.931. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1120240360480600Min: 582.6 / Avg: 584.54 / Max: 585.55Min: 673.73 / Avg: 677.71 / Max: 682.421. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Liquid-DSP

LiquidSDR's Liquid-DSP is a software-defined radio (SDR) digital signal processing library. This test profile runs a multi-threaded benchmark of this SDR/DSP library focused on embedded platform usage. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgsamples/s, More Is BetterLiquid-DSP 2021.01.31Threads: 160 - Buffer Length: 256 - Filter Length: 57Clang 12.0GCC 11.1800M1600M2400M3200M4000MSE +/- 34322846.29, N = 3SE +/- 22402480.02, N = 3368646666731828666671. (CC) gcc options: -O3 -march=native -flto -pthread -lm -lc -lliquid
OpenBenchmarking.orgsamples/s, More Is BetterLiquid-DSP 2021.01.31Threads: 160 - Buffer Length: 256 - Filter Length: 57Clang 12.0GCC 11.1600M1200M1800M2400M3000MMin: 3648800000 / Avg: 3686466666.67 / Max: 3755000000Min: 3150800000 / Avg: 3182866666.67 / Max: 32260000001. (CC) gcc options: -O3 -march=native -flto -pthread -lm -lc -lliquid

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.1150300450600750SE +/- 6.09, N = 5SE +/- 2.28, N = 3594.11686.05-fopenmp=libomp - MIN: 556.85-fopenmp - MIN: 656.341. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Recurrent Neural Network Training - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.1120240360480600Min: 583.61 / Avg: 594.11 / Max: 615.45Min: 682.18 / Avg: 686.04 / Max: 690.081. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

GraphicsMagick

This is a test of GraphicsMagick with its OpenMP implementation that performs various imaging tests on a sample 6000x4000 pixel JPEG image. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: EnhancedGCC 11.1Clang 12.030060090012001500SE +/- 1.20, N = 3SE +/- 4.10, N = 3131511411. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: EnhancedGCC 11.1Clang 12.02004006008001000Min: 1313 / Avg: 1315.33 / Max: 1317Min: 1133 / Avg: 1140.67 / Max: 11471. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: f32 - Engine: CPUClang 12.0GCC 11.10.31560.63120.94681.26241.578SE +/- 0.00153, N = 3SE +/- 0.00224, N = 31.217261.40268-fopenmp=libomp - MIN: 1.19-fopenmp - MIN: 1.361. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1246810Min: 1.22 / Avg: 1.22 / Max: 1.22Min: 1.4 / Avg: 1.4 / Max: 1.411. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_1d - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.10.73761.47522.21282.95043.688SE +/- 0.00123, N = 3SE +/- 0.00172, N = 32.867663.27829-fopenmp=libomp - MIN: 2.67-fopenmp - MIN: 3.111. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_1d - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1246810Min: 2.87 / Avg: 2.87 / Max: 2.87Min: 3.28 / Avg: 3.28 / Max: 3.281. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

WebP2 Image Encode

This is a test of Google's libwebp2 library with the WebP2 image encode utility and using a sample 6000x4000 pixel JPEG image as the input, similar to the WebP/libwebp test profile. WebP2 is currently experimental and under heavy development as ultimately the successor to WebP. WebP2 supports 10-bit HDR, more efficienct lossy compression, improved lossless compression, animation support, and full multi-threading support compared to WebP. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 100, Compression Effort 5GCC 11.1Clang 12.0246810SE +/- 0.004, N = 3SE +/- 0.013, N = 35.7656.4951. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg
OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 100, Compression Effort 5GCC 11.1Clang 12.03691215Min: 5.76 / Avg: 5.76 / Max: 5.77Min: 6.47 / Avg: 6.5 / Max: 6.511. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg

Himeno Benchmark

The Himeno benchmark is a linear solver of pressure Poisson using a point-Jacobi method. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMFLOPS, More Is BetterHimeno Benchmark 3.0Poisson Pressure SolverGCC 11.1Clang 12.010002000300040005000SE +/- 1.58, N = 3SE +/- 0.79, N = 34651.534161.651. (CC) gcc options: -O3 -march=native -flto -mavx2
OpenBenchmarking.orgMFLOPS, More Is BetterHimeno Benchmark 3.0Poisson Pressure SolverGCC 11.1Clang 12.08001600240032004000Min: 4649.46 / Avg: 4651.53 / Max: 4654.64Min: 4160.1 / Avg: 4161.65 / Max: 4162.641. (CC) gcc options: -O3 -march=native -flto -mavx2

Opus Codec Encoding

Opus is an open audio codec. Opus is a lossy audio compression format designed primarily for interactive real-time applications over the Internet. This test uses Opus-Tools and measures the time required to encode a WAV file to Opus. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterOpus Codec Encoding 1.3.1WAV To Opus EncodeGCC 11.1Clang 12.03691215SE +/- 0.003, N = 5SE +/- 0.008, N = 58.7689.788-fvisibility=hidden1. (CXX) g++ options: -O3 -march=native -flto -logg -lm
OpenBenchmarking.orgSeconds, Fewer Is BetterOpus Codec Encoding 1.3.1WAV To Opus EncodeGCC 11.1Clang 12.03691215Min: 8.76 / Avg: 8.77 / Max: 8.77Min: 9.76 / Avg: 9.79 / Max: 9.81. (CXX) g++ options: -O3 -march=native -flto -logg -lm

LAME MP3 Encoding

LAME is an MP3 encoder licensed under the LGPL. This test measures the time required to encode a WAV file to MP3 format. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterLAME MP3 Encoding 3.100WAV To MP3GCC 11.1Clang 12.03691215SE +/- 0.004, N = 3SE +/- 0.003, N = 38.6199.591-ffast-math -funroll-loops -fschedule-insns2 -fbranch-count-reg -fforce-addr1. (CC) gcc options: -O3 -pipe -march=native -flto -lncurses -lm
OpenBenchmarking.orgSeconds, Fewer Is BetterLAME MP3 Encoding 3.100WAV To MP3GCC 11.1Clang 12.03691215Min: 8.61 / Avg: 8.62 / Max: 8.62Min: 9.59 / Avg: 9.59 / Max: 9.61. (CC) gcc options: -O3 -pipe -march=native -flto -lncurses -lm

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.04380.08760.13140.17520.219SE +/- 0.000222, N = 3SE +/- 0.000672, N = 30.1750600.194587-fopenmp=libomp - MIN: 0.16-fopenmp - MIN: 0.181. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.112345Min: 0.17 / Avg: 0.18 / Max: 0.18Min: 0.19 / Avg: 0.19 / Max: 0.21. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Bullet Physics Engine

This is a benchmark of the Bullet Physics Engine. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 3000 FallGCC 11.1Clang 12.00.96791.93582.90373.87164.8395SE +/- 0.007493, N = 3SE +/- 0.005089, N = 33.8730604.301823-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 3000 FallGCC 11.1Clang 12.0246810Min: 3.86 / Avg: 3.87 / Max: 3.89Min: 4.3 / Avg: 4.3 / Max: 4.311. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

Kripke

Kripke is a simple, scalable, 3D Sn deterministic particle transport code. Its primary purpose is to research how data layout, programming paradigms and architectures effect the implementation and performance of Sn transport. Kripke is developed by LLNL. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgThroughput FoM, More Is BetterKripke 1.2.4GCC 11.1Clang 12.040M80M120M160M200MSE +/- 1982388.01, N = 5SE +/- 1966921.21, N = 4177613600160155675-fopenmp-fopenmp=libomp1. (CXX) g++ options: -O3 -march=native -flto -O2
OpenBenchmarking.orgThroughput FoM, More Is BetterKripke 1.2.4GCC 11.1Clang 12.030M60M90M120M150MMin: 170975200 / Avg: 177613600 / Max: 182800400Min: 154910600 / Avg: 160155675 / Max: 1637469001. (CXX) g++ options: -O3 -march=native -flto -O2

Bullet Physics Engine

This is a benchmark of the Bullet Physics Engine. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 136 RagdollsGCC 11.1Clang 12.00.62961.25921.88882.51843.148SE +/- 0.004472, N = 3SE +/- 0.000181, N = 32.5533032.798322-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 136 RagdollsGCC 11.1Clang 12.0246810Min: 2.55 / Avg: 2.55 / Max: 2.56Min: 2.8 / Avg: 2.8 / Max: 2.81. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: yolov4-tinyGCC 11.1Clang 12.0612182430SE +/- 0.18, N = 12SE +/- 0.25, N = 1323.1225.28-lgomp - MIN: 21.63 / MAX: 253.04-lomp - MIN: 23.29 / MAX: 148.871. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: yolov4-tinyGCC 11.1Clang 12.0612182430Min: 22.27 / Avg: 23.12 / Max: 23.97Min: 24.24 / Avg: 25.28 / Max: 27.711. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

Bullet Physics Engine

This is a benchmark of the Bullet Physics Engine. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 1000 ConvexGCC 11.1Clang 12.01.06672.13343.20014.26685.3335SE +/- 0.003221, N = 3SE +/- 0.003944, N = 34.3406004.741088-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: 1000 ConvexGCC 11.1Clang 12.0246810Min: 4.34 / Avg: 4.34 / Max: 4.35Min: 4.73 / Avg: 4.74 / Max: 4.751. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: Prim TrimeshGCC 11.1Clang 12.00.21170.42340.63510.84681.0585SE +/- 0.002011, N = 3SE +/- 0.000250, N = 30.8620000.940903-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: Prim TrimeshGCC 11.1Clang 12.0246810Min: 0.86 / Avg: 0.86 / Max: 0.87Min: 0.94 / Avg: 0.94 / Max: 0.941. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: Convex TrimeshGCC 11.1Clang 12.00.25840.51680.77521.03361.292SE +/- 0.001326, N = 3SE +/- 0.001342, N = 31.0542731.148367-lglut -lGL -lGLU1. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterBullet Physics Engine 2.81Test: Convex TrimeshGCC 11.1Clang 12.0246810Min: 1.05 / Avg: 1.05 / Max: 1.06Min: 1.15 / Avg: 1.15 / Max: 1.151. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.10.67561.35122.02682.70243.378SE +/- 0.00228, N = 3SE +/- 0.00315, N = 32.766813.00252-fopenmp=libomp - MIN: 2.62-fopenmp - MIN: 2.871. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 1D - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1246810Min: 2.76 / Avg: 2.77 / Max: 2.77Min: 3 / Avg: 3 / Max: 3.011. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

x265

This is a simple test of the x265 encoder run on the CPU with 1080p and 4K options for H.265 video encode performance with x265. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betterx265 3.4Video Input: Bosphorus 4KClang 12.0GCC 11.1714212835SE +/- 0.16, N = 3SE +/- 0.27, N = 329.1326.921. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread -lrt -ldl
OpenBenchmarking.orgFrames Per Second, More Is Betterx265 3.4Video Input: Bosphorus 4KClang 12.0GCC 11.1612182430Min: 28.84 / Avg: 29.13 / Max: 29.37Min: 26.61 / Avg: 26.92 / Max: 27.451. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread -lrt -ldl

WebP Image Encode

This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, Highest CompressionClang 12.0GCC 11.1246810SE +/- 0.007, N = 3SE +/- 0.007, N = 37.4228.0261. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg
OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, Highest CompressionClang 12.0GCC 11.13691215Min: 7.41 / Avg: 7.42 / Max: 7.44Min: 8.01 / Avg: 8.03 / Max: 8.031. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg

Crypto++

Crypto++ is a C++ class library of cryptographic algorithms. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMiB/second, More Is BetterCrypto++ 8.2Test: Unkeyed AlgorithmsClang 12.0GCC 11.180160240320400SE +/- 0.06, N = 3SE +/- 0.06, N = 3388.57359.901. (CXX) g++ options: -O3 -march=native -flto -fPIC -pthread -pipe
OpenBenchmarking.orgMiB/second, More Is BetterCrypto++ 8.2Test: Unkeyed AlgorithmsClang 12.0GCC 11.170140210280350Min: 388.5 / Avg: 388.57 / Max: 388.68Min: 359.81 / Avg: 359.9 / Max: 360.021. (CXX) g++ options: -O3 -march=native -flto -fPIC -pthread -pipe

WebP2 Image Encode

This is a test of Google's libwebp2 library with the WebP2 image encode utility and using a sample 6000x4000 pixel JPEG image as the input, similar to the WebP/libwebp test profile. WebP2 is currently experimental and under heavy development as ultimately the successor to WebP. WebP2 supports 10-bit HDR, more efficienct lossy compression, improved lossless compression, animation support, and full multi-threading support compared to WebP. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 95, Compression Effort 7Clang 12.0GCC 11.14080120160200SE +/- 0.01, N = 3SE +/- 0.01, N = 3182.29196.491. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg
OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 95, Compression Effort 7Clang 12.0GCC 11.14080120160200Min: 182.27 / Avg: 182.29 / Max: 182.32Min: 196.48 / Avg: 196.49 / Max: 196.51. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.10.21820.43640.65460.87281.091SE +/- 0.002259, N = 3SE +/- 0.007538, N = 30.9022130.969972-fopenmp=libomp - MIN: 0.85-fopenmp - MIN: 0.891. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: u8s8f32 - Engine: CPUClang 12.0GCC 11.1246810Min: 0.9 / Avg: 0.9 / Max: 0.91Min: 0.96 / Avg: 0.97 / Max: 0.981. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

WebP2 Image Encode

This is a test of Google's libwebp2 library with the WebP2 image encode utility and using a sample 6000x4000 pixel JPEG image as the input, similar to the WebP/libwebp test profile. WebP2 is currently experimental and under heavy development as ultimately the successor to WebP. WebP2 supports 10-bit HDR, more efficienct lossy compression, improved lossless compression, animation support, and full multi-threading support compared to WebP. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 75, Compression Effort 7Clang 12.0GCC 11.120406080100SE +/- 0.04, N = 3SE +/- 0.01, N = 399.78106.661. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg
OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 75, Compression Effort 7Clang 12.0GCC 11.120406080100Min: 99.7 / Avg: 99.78 / Max: 99.83Min: 106.64 / Avg: 106.66 / Max: 106.691. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg

TNN

TNN is an open-source deep learning reasoning framework developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterTNN 0.2.3Target: CPU - Model: SqueezeNet v1.1GCC 11.1Clang 12.090180270360450SE +/- 0.01, N = 3SE +/- 0.02, N = 3377.43402.91-fopenmp - MIN: 377.31 / MAX: 377.64-fopenmp=libomp - MIN: 402.36 / MAX: 405.481. (CXX) g++ options: -O3 -march=native -flto -pthread -fvisibility=hidden -O2 -rdynamic -ldl
OpenBenchmarking.orgms, Fewer Is BetterTNN 0.2.3Target: CPU - Model: SqueezeNet v1.1GCC 11.1Clang 12.070140210280350Min: 377.41 / Avg: 377.43 / Max: 377.45Min: 402.87 / Avg: 402.91 / Max: 402.951. (CXX) g++ options: -O3 -march=native -flto -pthread -fvisibility=hidden -O2 -rdynamic -ldl

AOBench

AOBench is a lightweight ambient occlusion renderer, written in C. The test profile is using a size of 2048 x 2048. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterAOBenchSize: 2048 x 2048 - Total TimeGCC 11.1Clang 12.0816243240SE +/- 0.02, N = 3SE +/- 0.01, N = 333.8836.081. (CC) gcc options: -lm -O3 -march=native -flto
OpenBenchmarking.orgSeconds, Fewer Is BetterAOBenchSize: 2048 x 2048 - Total TimeGCC 11.1Clang 12.0816243240Min: 33.84 / Avg: 33.88 / Max: 33.92Min: 36.06 / Avg: 36.08 / Max: 36.081. (CC) gcc options: -lm -O3 -march=native -flto

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.10.4080.8161.2241.6322.04SE +/- 0.00299, N = 3SE +/- 0.00150, N = 31.705671.81351-fopenmp=libomp - MIN: 1.54-fopenmp - MIN: 1.681. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: IP Shapes 3D - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1246810Min: 1.7 / Avg: 1.71 / Max: 1.71Min: 1.81 / Avg: 1.81 / Max: 1.821. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

WebP2 Image Encode

This is a test of Google's libwebp2 library with the WebP2 image encode utility and using a sample 6000x4000 pixel JPEG image as the input, similar to the WebP/libwebp test profile. WebP2 is currently experimental and under heavy development as ultimately the successor to WebP. WebP2 supports 10-bit HDR, more efficienct lossy compression, improved lossless compression, animation support, and full multi-threading support compared to WebP. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: DefaultClang 12.0GCC 11.10.59491.18981.78472.37962.9745SE +/- 0.009, N = 3SE +/- 0.035, N = 32.4922.6441. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg
OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: DefaultClang 12.0GCC 11.1246810Min: 2.48 / Avg: 2.49 / Max: 2.51Min: 2.57 / Avg: 2.64 / Max: 2.681. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg

SVT-HEVC

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-HEVC CPU-based multi-threaded video encoder for the HEVC / H.265 video format with a sample 1080p YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 1 - Input: Bosphorus 1080pClang 12.0GCC 11.11020304050SE +/- 0.21, N = 3SE +/- 0.28, N = 341.9639.591. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 1 - Input: Bosphorus 1080pClang 12.0GCC 11.1918273645Min: 41.68 / Avg: 41.96 / Max: 42.38Min: 39.14 / Avg: 39.59 / Max: 40.111. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt

Kvazaar

This is a test of Kvazaar as a CPU-based H.265 video encoder written in the C programming language and optimized in Assembly. Kvazaar is the winner of the 2016 ACM Open-Source Software Competition and developed at the Ultra Video Group, Tampere University, Finland. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 4K - Video Preset: Very FastClang 12.0GCC 11.11020304050SE +/- 0.28, N = 3SE +/- 0.35, N = 344.7242.31-lpthread1. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 4K - Video Preset: Very FastClang 12.0GCC 11.1918273645Min: 44.39 / Avg: 44.72 / Max: 45.27Min: 41.76 / Avg: 42.31 / Max: 42.951. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt

SVT-HEVC

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-HEVC CPU-based multi-threaded video encoder for the HEVC / H.265 video format with a sample 1080p YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 7 - Input: Bosphorus 1080pClang 12.0GCC 11.180160240320400SE +/- 1.29, N = 3SE +/- 2.94, N = 3355.18336.371. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 7 - Input: Bosphorus 1080pClang 12.0GCC 11.160120180240300Min: 353.15 / Avg: 355.18 / Max: 357.57Min: 331.31 / Avg: 336.37 / Max: 341.491. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8 - Compression SpeedClang 12.0GCC 11.16001200180024003000SE +/- 30.31, N = 5SE +/- 30.03, N = 152748.12611.01. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8 - Compression SpeedClang 12.0GCC 11.15001000150020002500Min: 2696.9 / Avg: 2748.14 / Max: 2859.2Min: 2453.6 / Avg: 2610.96 / Max: 2835.81. (CC) gcc options: -O3 -march=native -flto -pthread -lz

SVT-AV1

This is a benchmark of the SVT-AV1 open-source video encoder/decoder. SVT-AV1 was originally developed by Intel as part of their Open Visual Cloud / Scalable Video Technology (SVT). Development of SVT-AV1 has since moved to the Alliance for Open Media as part of upstream AV1 development. SVT-AV1 is a CPU-based multi-threaded video encoder for the AV1 video format with a sample YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 4 - Input: Bosphorus 1080pClang 12.0GCC 11.13691215SE +/- 0.082, N = 3SE +/- 0.043, N = 39.4188.9961. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 4 - Input: Bosphorus 1080pClang 12.0GCC 11.13691215Min: 9.26 / Avg: 9.42 / Max: 9.54Min: 8.94 / Avg: 9 / Max: 9.081. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie

Gcrypt Library

Libgcrypt is a general purpose cryptographic library developed as part of the GnuPG project. This is a benchmark of libgcrypt's integrated benchmark and is measuring the time to run the benchmark command with a cipher/mac/hash repetition count set for 50 times as simple, high level look at the overall crypto performance of the system under test. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterGcrypt Library 1.9Clang 12.0GCC 11.160120180240300SE +/- 0.78, N = 3SE +/- 0.95, N = 3253.86265.171. (CC) gcc options: -O3 -march=native -flto -fvisibility=hidden -lgpg-error
OpenBenchmarking.orgSeconds, Fewer Is BetterGcrypt Library 1.9Clang 12.0GCC 11.150100150200250Min: 252.37 / Avg: 253.86 / Max: 254.99Min: 263.7 / Avg: 265.17 / Max: 266.951. (CC) gcc options: -O3 -march=native -flto -fvisibility=hidden -lgpg-error

Kvazaar

This is a test of Kvazaar as a CPU-based H.265 video encoder written in the C programming language and optimized in Assembly. Kvazaar is the winner of the 2016 ACM Open-Source Software Competition and developed at the Ultra Video Group, Tampere University, Finland. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 1080p - Video Preset: Ultra FastClang 12.0GCC 11.14080120160200SE +/- 1.71, N = 3SE +/- 1.31, N = 3183.97176.66-lpthread1. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 1080p - Video Preset: Ultra FastClang 12.0GCC 11.1306090120150Min: 180.59 / Avg: 183.97 / Max: 186.04Min: 174.17 / Avg: 176.66 / Max: 178.61. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt

WebP Image Encode

This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, LosslessGCC 11.1Clang 12.0510152025SE +/- 0.01, N = 3SE +/- 0.00, N = 319.4720.271. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg
OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, LosslessGCC 11.1Clang 12.0510152025Min: 19.46 / Avg: 19.47 / Max: 19.49Min: 20.26 / Avg: 20.27 / Max: 20.271. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg

Kvazaar

This is a test of Kvazaar as a CPU-based H.265 video encoder written in the C programming language and optimized in Assembly. Kvazaar is the winner of the 2016 ACM Open-Source Software Competition and developed at the Ultra Video Group, Tampere University, Finland. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 1080p - Video Preset: Very FastClang 12.0GCC 11.14080120160200SE +/- 0.77, N = 3SE +/- 0.42, N = 3166.00159.71-lpthread1. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 1080p - Video Preset: Very FastClang 12.0GCC 11.1306090120150Min: 164.93 / Avg: 166 / Max: 167.5Min: 159 / Avg: 159.71 / Max: 160.461. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19, Long Mode - Compression SpeedGCC 11.1Clang 12.01122334455SE +/- 0.48, N = 15SE +/- 0.37, N = 1547.946.21. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19, Long Mode - Compression SpeedGCC 11.1Clang 12.01020304050Min: 43.3 / Avg: 47.91 / Max: 49.8Min: 42.8 / Avg: 46.25 / Max: 481. (CC) gcc options: -O3 -march=native -flto -pthread -lz

libjpeg-turbo tjbench

tjbench is a JPEG decompression/compression benchmark that is part of libjpeg-turbo, a JPEG image codec library optimized for SIMD instructions on modern CPU architectures. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMegapixels/sec, More Is Betterlibjpeg-turbo tjbench 2.1.0Test: Decompression ThroughputClang 12.0GCC 11.14080120160200SE +/- 0.47, N = 3SE +/- 1.34, N = 3180.60174.231. (CC) gcc options: -O3 -march=native -flto -rdynamic -lm
OpenBenchmarking.orgMegapixels/sec, More Is Betterlibjpeg-turbo tjbench 2.1.0Test: Decompression ThroughputClang 12.0GCC 11.1306090120150Min: 179.69 / Avg: 180.6 / Max: 181.28Min: 171.55 / Avg: 174.23 / Max: 175.81. (CC) gcc options: -O3 -march=native -flto -rdynamic -lm

PostgreSQL pgbench

This is a benchmark of PostgreSQL using pgbench for facilitating the database benchmarks. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read WriteClang 12.0GCC 11.120K40K60K80K100KSE +/- 239.57, N = 3SE +/- 106.75, N = 392576894251. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm
OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read WriteClang 12.0GCC 11.116K32K48K64K80KMin: 92316.77 / Avg: 92576.05 / Max: 93054.64Min: 89238.17 / Avg: 89425.44 / Max: 89607.871. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm

OpenBenchmarking.orgms, Fewer Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read Write - Average LatencyClang 12.0GCC 11.10.62931.25861.88792.51723.1465SE +/- 0.007, N = 3SE +/- 0.003, N = 32.7022.7971. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm
OpenBenchmarking.orgms, Fewer Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read Write - Average LatencyClang 12.0GCC 11.1246810Min: 2.69 / Avg: 2.7 / Max: 2.71Min: 2.79 / Avg: 2.8 / Max: 2.81. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm

SVT-VP9

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-VP9 CPU-based multi-threaded video encoder for the VP9 video format with a sample YUV input video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: Visual Quality Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.090180270360450SE +/- 0.84, N = 3SE +/- 1.57, N = 3393.17379.931. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: Visual Quality Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.070140210280350Min: 391.49 / Avg: 393.17 / Max: 394.11Min: 377.41 / Avg: 379.93 / Max: 382.821. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPUClang 12.0GCC 11.10.32830.65660.98491.31321.6415SE +/- 0.00418, N = 3SE +/- 0.00729, N = 31.411671.45930-fopenmp=libomp - MIN: 1.26-fopenmp - MIN: 1.281. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: f32 - Engine: CPUClang 12.0GCC 11.1246810Min: 1.4 / Avg: 1.41 / Max: 1.42Min: 1.45 / Avg: 1.46 / Max: 1.471. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

WebP Image Encode

This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, Lossless, Highest CompressionGCC 11.1Clang 12.01020304050SE +/- 0.05, N = 3SE +/- 0.02, N = 340.9142.181. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg
OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100, Lossless, Highest CompressionGCC 11.1Clang 12.0918273645Min: 40.85 / Avg: 40.91 / Max: 41.01Min: 42.15 / Avg: 42.18 / Max: 42.211. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg

WebP2 Image Encode

This is a test of Google's libwebp2 library with the WebP2 image encode utility and using a sample 6000x4000 pixel JPEG image as the input, similar to the WebP/libwebp test profile. WebP2 is currently experimental and under heavy development as ultimately the successor to WebP. WebP2 supports 10-bit HDR, more efficienct lossy compression, improved lossless compression, animation support, and full multi-threading support compared to WebP. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 100, Lossless CompressionGCC 11.1Clang 12.090180270360450SE +/- 0.07, N = 3SE +/- 0.16, N = 3389.07400.621. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg
OpenBenchmarking.orgSeconds, Fewer Is BetterWebP2 Image Encode 20210126Encode Settings: Quality 100, Lossless CompressionGCC 11.1Clang 12.070140210280350Min: 388.96 / Avg: 389.07 / Max: 389.19Min: 400.36 / Avg: 400.62 / Max: 400.911. (CXX) g++ options: -O3 -march=native -flto -fno-rtti -O2 -rdynamic -lpthread -ljpeg

Kvazaar

This is a test of Kvazaar as a CPU-based H.265 video encoder written in the C programming language and optimized in Assembly. Kvazaar is the winner of the 2016 ACM Open-Source Software Competition and developed at the Ultra Video Group, Tampere University, Finland. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 4K - Video Preset: Ultra FastClang 12.0GCC 11.11122334455SE +/- 0.33, N = 3SE +/- 0.54, N = 347.9646.58-lpthread1. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterKvazaar 2.0Video Input: Bosphorus 4K - Video Preset: Ultra FastClang 12.0GCC 11.11020304050Min: 47.32 / Avg: 47.96 / Max: 48.43Min: 45.55 / Avg: 46.58 / Max: 47.361. (CC) gcc options: -pthread -ftree-vectorize -fvisibility=hidden -O3 -march=native -flto -lm -lrt

SVT-AV1

This is a benchmark of the SVT-AV1 open-source video encoder/decoder. SVT-AV1 was originally developed by Intel as part of their Open Visual Cloud / Scalable Video Technology (SVT). Development of SVT-AV1 has since moved to the Alliance for Open Media as part of upstream AV1 development. SVT-AV1 is a CPU-based multi-threaded video encoder for the AV1 video format with a sample YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 4 - Input: Bosphorus 4KClang 12.0GCC 11.10.97491.94982.92473.89964.8745SE +/- 0.006, N = 3SE +/- 0.029, N = 34.3334.2131. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 4 - Input: Bosphorus 4KClang 12.0GCC 11.1246810Min: 4.32 / Avg: 4.33 / Max: 4.34Min: 4.16 / Avg: 4.21 / Max: 4.261. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19 - Compression SpeedGCC 11.1Clang 12.020406080100SE +/- 0.39, N = 3SE +/- 0.87, N = 583.881.51. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19 - Compression SpeedGCC 11.1Clang 12.01632486480Min: 83 / Avg: 83.77 / Max: 84.3Min: 79.7 / Avg: 81.5 / Max: 84.81. (CC) gcc options: -O3 -march=native -flto -pthread -lz

Timed MrBayes Analysis

This test performs a bayesian analysis of a set of primate genome sequences in order to estimate their phylogeny. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed MrBayes Analysis 3.2.7Primate Phylogeny AnalysisClang 12.0GCC 11.1306090120150SE +/- 0.91, N = 3SE +/- 0.46, N = 3138.55142.20-mabm1. (CC) gcc options: -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msha -maes -mavx -mfma -mavx2 -mavx512f -mavx512cd -mavx512vl -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -mrdrnd -mbmi -mbmi2 -madx -O3 -std=c99 -pedantic -march=native -flto -lm
OpenBenchmarking.orgSeconds, Fewer Is BetterTimed MrBayes Analysis 3.2.7Primate Phylogeny AnalysisClang 12.0GCC 11.1306090120150Min: 136.75 / Avg: 138.55 / Max: 139.59Min: 141.39 / Avg: 142.2 / Max: 142.991. (CC) gcc options: -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msha -maes -mavx -mfma -mavx2 -mavx512f -mavx512cd -mavx512vl -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -mrdrnd -mbmi -mbmi2 -madx -O3 -std=c99 -pedantic -march=native -flto -lm

SVT-VP9

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-VP9 CPU-based multi-threaded video encoder for the VP9 video format with a sample YUV input video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: PSNR/SSIM Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.0100200300400500SE +/- 3.23, N = 3SE +/- 0.60, N = 3477.90466.061. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: PSNR/SSIM Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.080160240320400Min: 472.63 / Avg: 477.9 / Max: 483.76Min: 465.15 / Avg: 466.06 / Max: 467.21. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm

eSpeak-NG Speech Engine

This test times how long it takes the eSpeak speech synthesizer to read Project Gutenberg's The Outline of Science and output to a WAV file. This test profile is now tracking the eSpeak-NG version of eSpeak. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BettereSpeak-NG Speech Engine 20200907Text-To-Speech SynthesisClang 12.0GCC 11.1714212835SE +/- 0.02, N = 4SE +/- 0.09, N = 429.8130.511. (CC) gcc options: -O3 -march=native -flto -std=c99 -lpthread -lm
OpenBenchmarking.orgSeconds, Fewer Is BettereSpeak-NG Speech Engine 20200907Text-To-Speech SynthesisClang 12.0GCC 11.1714212835Min: 29.74 / Avg: 29.81 / Max: 29.84Min: 30.23 / Avg: 30.51 / Max: 30.611. (CC) gcc options: -O3 -march=native -flto -std=c99 -lpthread -lm

SVT-VP9

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-VP9 CPU-based multi-threaded video encoder for the VP9 video format with a sample YUV input video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: VMAF Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.0100200300400500SE +/- 2.96, N = 3SE +/- 3.96, N = 3476.85467.571. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-VP9 0.3Tuning: VMAF Optimized - Input: Bosphorus 1080pGCC 11.1Clang 12.080160240320400Min: 471.68 / Avg: 476.85 / Max: 481.93Min: 461.08 / Avg: 467.57 / Max: 474.751. (CC) gcc options: -O3 -fcommon -march=native -flto -fPIE -fPIC -fvisibility=hidden -O2 -pie -rdynamic -lpthread -lrt -lm

SVT-AV1

This is a benchmark of the SVT-AV1 open-source video encoder/decoder. SVT-AV1 was originally developed by Intel as part of their Open Visual Cloud / Scalable Video Technology (SVT). Development of SVT-AV1 has since moved to the Alliance for Open Media as part of upstream AV1 development. SVT-AV1 is a CPU-based multi-threaded video encoder for the AV1 video format with a sample YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 8 - Input: Bosphorus 4KClang 12.0GCC 11.11326395265SE +/- 0.41, N = 3SE +/- 0.23, N = 356.2755.191. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 8 - Input: Bosphorus 4KClang 12.0GCC 11.11122334455Min: 55.82 / Avg: 56.27 / Max: 57.09Min: 54.74 / Avg: 55.19 / Max: 55.491. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie

WebP Image Encode

This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100GCC 11.1Clang 12.00.60571.21141.81712.42283.0285SE +/- 0.001, N = 3SE +/- 0.003, N = 32.6452.6921. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg
OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: Quality 100GCC 11.1Clang 12.0246810Min: 2.64 / Avg: 2.65 / Max: 2.65Min: 2.69 / Avg: 2.69 / Max: 2.71. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19 - Decompression SpeedGCC 11.1Clang 12.05001000150020002500SE +/- 0.57, N = 3SE +/- 13.91, N = 52537.32495.21. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19 - Decompression SpeedGCC 11.1Clang 12.0400800120016002000Min: 2536.2 / Avg: 2537.3 / Max: 2538.1Min: 2439.7 / Avg: 2495.22 / Max: 2512.41. (CC) gcc options: -O3 -march=native -flto -pthread -lz

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19, Long Mode - Decompression SpeedGCC 11.1Clang 12.06001200180024003000SE +/- 2.07, N = 15SE +/- 1.79, N = 152670.82632.31. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 19, Long Mode - Decompression SpeedGCC 11.1Clang 12.05001000150020002500Min: 2653.4 / Avg: 2670.81 / Max: 2680.6Min: 2614.5 / Avg: 2632.33 / Max: 2642.81. (CC) gcc options: -O3 -march=native -flto -pthread -lz

Liquid-DSP

LiquidSDR's Liquid-DSP is a software-defined radio (SDR) digital signal processing library. This test profile runs a multi-threaded benchmark of this SDR/DSP library focused on embedded platform usage. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgsamples/s, More Is BetterLiquid-DSP 2021.01.31Threads: 1 - Buffer Length: 256 - Filter Length: 57Clang 12.0GCC 11.113M26M39M52M65MSE +/- 5206.83, N = 3SE +/- 49079.30, N = 361840333609853331. (CC) gcc options: -O3 -march=native -flto -pthread -lm -lc -lliquid
OpenBenchmarking.orgsamples/s, More Is BetterLiquid-DSP 2021.01.31Threads: 1 - Buffer Length: 256 - Filter Length: 57Clang 12.0GCC 11.111M22M33M44M55MMin: 61831000 / Avg: 61840333.33 / Max: 61849000Min: 60888000 / Avg: 60985333.33 / Max: 610450001. (CC) gcc options: -O3 -march=native -flto -pthread -lm -lc -lliquid

WebP Image Encode

This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: DefaultClang 12.0GCC 11.10.36860.73721.10581.47441.843SE +/- 0.003, N = 3SE +/- 0.000, N = 31.6161.6381. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg
OpenBenchmarking.orgEncode Time - Seconds, Fewer Is BetterWebP Image Encode 1.1Encode Settings: DefaultClang 12.0GCC 11.1246810Min: 1.61 / Avg: 1.62 / Max: 1.62Min: 1.64 / Avg: 1.64 / Max: 1.641. (CC) gcc options: -fvisibility=hidden -O3 -march=native -flto -pthread -lm -ljpeg

SVT-AV1

This is a benchmark of the SVT-AV1 open-source video encoder/decoder. SVT-AV1 was originally developed by Intel as part of their Open Visual Cloud / Scalable Video Technology (SVT). Development of SVT-AV1 has since moved to the Alliance for Open Media as part of upstream AV1 development. SVT-AV1 is a CPU-based multi-threaded video encoder for the AV1 video format with a sample YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 8 - Input: Bosphorus 1080pClang 12.0GCC 11.14080120160200SE +/- 1.14, N = 3SE +/- 0.34, N = 3169.61167.391. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 0.8.7Encoder Mode: Preset 8 - Input: Bosphorus 1080pClang 12.0GCC 11.1306090120150Min: 167.73 / Avg: 169.61 / Max: 171.66Min: 166.78 / Avg: 167.39 / Max: 167.951. (CXX) g++ options: -O3 -march=native -flto -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq -pie

Primesieve

Primesieve generates prime numbers using a highly optimized sieve of Eratosthenes implementation. Primesieve benchmarks the CPU's L1/L2 cache performance. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterPrimesieve 7.41e12 Prime Number GenerationGCC 11.1Clang 12.00.86181.72362.58543.44724.309SE +/- 0.006, N = 3SE +/- 0.003, N = 33.7803.8301. (CXX) g++ options: -O3 -march=native -flto -O2 -lpthread
OpenBenchmarking.orgSeconds, Fewer Is BetterPrimesieve 7.41e12 Prime Number GenerationGCC 11.1Clang 12.0246810Min: 3.77 / Avg: 3.78 / Max: 3.79Min: 3.83 / Avg: 3.83 / Max: 3.841. (CXX) g++ options: -O3 -march=native -flto -O2 -lpthread

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.10.47080.94161.41241.88322.354SE +/- 0.00039, N = 3SE +/- 0.00148, N = 32.066292.09246-fopenmp=libomp - MIN: 1.99-fopenmp - MIN: 2.041. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Convolution Batch Shapes Auto - Data Type: bf16bf16bf16 - Engine: CPUClang 12.0GCC 11.1246810Min: 2.07 / Avg: 2.07 / Max: 2.07Min: 2.09 / Avg: 2.09 / Max: 2.091. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Zstd Compression

This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8 - Decompression SpeedClang 12.0GCC 11.16001200180024003000SE +/- 7.51, N = 5SE +/- 2.87, N = 152996.72959.31. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8 - Decompression SpeedClang 12.0GCC 11.15001000150020002500Min: 2968.2 / Avg: 2996.7 / Max: 3010.8Min: 2942.1 / Avg: 2959.33 / Max: 2979.11. (CC) gcc options: -O3 -march=native -flto -pthread -lz

OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8, Long Mode - Decompression SpeedClang 12.0GCC 11.17001400210028003500SE +/- 4.90, N = 3SE +/- 6.63, N = 33204.63168.41. (CC) gcc options: -O3 -march=native -flto -pthread -lz
OpenBenchmarking.orgMB/s, More Is BetterZstd Compression 1.5.0Compression Level: 8, Long Mode - Decompression SpeedClang 12.0GCC 11.16001200180024003000Min: 3195.4 / Avg: 3204.6 / Max: 3212.1Min: 3159 / Avg: 3168.4 / Max: 3181.21. (CC) gcc options: -O3 -march=native -flto -pthread -lz

FLAC Audio Encoding

This test times how long it takes to encode a sample WAV file to FLAC format five times. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterFLAC Audio Encoding 1.3.2WAV To FLACGCC 11.1Clang 12.03691215SE +/- 0.004, N = 5SE +/- 0.004, N = 59.3829.474-fvisibility=hidden1. (CXX) g++ options: -O3 -march=native -flto -logg -lm
OpenBenchmarking.orgSeconds, Fewer Is BetterFLAC Audio Encoding 1.3.2WAV To FLACGCC 11.1Clang 12.03691215Min: 9.37 / Avg: 9.38 / Max: 9.39Min: 9.46 / Avg: 9.47 / Max: 9.481. (CXX) g++ options: -O3 -march=native -flto -logg -lm

x265

This is a simple test of the x265 encoder run on the CPU with 1080p and 4K options for H.265 video encode performance with x265. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is Betterx265 3.4Video Input: Bosphorus 1080pClang 12.0GCC 11.120406080100SE +/- 0.27, N = 3SE +/- 0.62, N = 377.2476.881. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread -lrt -ldl
OpenBenchmarking.orgFrames Per Second, More Is Betterx265 3.4Video Input: Bosphorus 1080pClang 12.0GCC 11.11530456075Min: 76.83 / Avg: 77.24 / Max: 77.76Min: 75.87 / Avg: 76.88 / Max: 78.011. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread -lrt -ldl

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPUGCC 11.1Clang 12.00.18970.37940.56910.75880.9485SE +/- 0.001374, N = 3SE +/- 0.000915, N = 30.8403830.842984-fopenmp - MIN: 0.8-fopenmp=libomp - MIN: 0.771. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: f32 - Engine: CPUGCC 11.1Clang 12.0246810Min: 0.84 / Avg: 0.84 / Max: 0.84Min: 0.84 / Avg: 0.84 / Max: 0.841. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Caffe

This is a benchmark of the Caffe deep learning framework and currently supports the AlexNet and Googlenet model and execution on both CPUs and NVIDIA GPUs. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2020-02-13Model: AlexNet - Acceleration: CPU - Iterations: 200Clang 12.0GCC 11.160K120K180K240K300KSE +/- 100.93, N = 3SE +/- 46.89, N = 32975542982911. (CXX) g++ options: -O3 -march=native -flto -fPIC -O2 -rdynamic -lboost_system -lboost_thread -lboost_filesystem -lboost_chrono -lboost_date_time -lboost_atomic -lglog -lgflags -lprotobuf -lpthread -lhdf5_cpp -lhdf5 -lhdf5_hl_cpp -lhdf5_hl -llmdb -lopenblas
OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2020-02-13Model: AlexNet - Acceleration: CPU - Iterations: 200Clang 12.0GCC 11.150K100K150K200K250KMin: 297424 / Avg: 297554.33 / Max: 297753Min: 298198 / Avg: 298291.33 / Max: 2983461. (CXX) g++ options: -O3 -march=native -flto -fPIC -O2 -rdynamic -lboost_system -lboost_thread -lboost_filesystem -lboost_chrono -lboost_date_time -lboost_atomic -lglog -lgflags -lprotobuf -lpthread -lhdf5_cpp -lhdf5 -lhdf5_hl_cpp -lhdf5_hl -llmdb -lopenblas

OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2020-02-13Model: GoogleNet - Acceleration: CPU - Iterations: 200GCC 11.1Clang 12.0140K280K420K560K700KSE +/- 192.49, N = 3SE +/- 154.25, N = 36624086632821. (CXX) g++ options: -O3 -march=native -flto -fPIC -O2 -rdynamic -lboost_system -lboost_thread -lboost_filesystem -lboost_chrono -lboost_date_time -lboost_atomic -lglog -lgflags -lprotobuf -lpthread -lhdf5_cpp -lhdf5 -lhdf5_hl_cpp -lhdf5_hl -llmdb -lopenblas
OpenBenchmarking.orgMilli-Seconds, Fewer Is BetterCaffe 2020-02-13Model: GoogleNet - Acceleration: CPU - Iterations: 200GCC 11.1Clang 12.0110K220K330K440K550KMin: 662160 / Avg: 662408 / Max: 662787Min: 663044 / Avg: 663282 / Max: 6635711. (CXX) g++ options: -O3 -march=native -flto -fPIC -O2 -rdynamic -lboost_system -lboost_thread -lboost_filesystem -lboost_chrono -lboost_date_time -lboost_atomic -lglog -lgflags -lprotobuf -lpthread -lhdf5_cpp -lhdf5 -lhdf5_hl_cpp -lhdf5_hl -llmdb -lopenblas

SVT-HEVC

This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-HEVC CPU-based multi-threaded video encoder for the HEVC / H.265 video format with a sample 1080p YUV video file. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 10 - Input: Bosphorus 1080pGCC 11.1Clang 12.0130260390520650SE +/- 1.44, N = 3SE +/- 2.17, N = 3609.56608.951. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-HEVC 1.5.0Tuning: 10 - Input: Bosphorus 1080pGCC 11.1Clang 12.0110220330440550Min: 607.29 / Avg: 609.56 / Max: 612.24Min: 604.84 / Avg: 608.95 / Max: 612.241. (CC) gcc options: -O3 -march=native -flto -fPIE -fPIC -O2 -pie -rdynamic -lpthread -lrt

WavPack Audio Encoding

This test times how long it takes to encode a sample WAV file to WavPack format with very high quality settings. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterWavPack Audio Encoding 5.3WAV To WavPackClang 12.0GCC 11.148121620SE +/- 0.00, N = 5SE +/- 0.00, N = 517.3417.361. (CXX) g++ options: -O3 -march=native -flto -rdynamic
OpenBenchmarking.orgSeconds, Fewer Is BetterWavPack Audio Encoding 5.3WAV To WavPackClang 12.0GCC 11.148121620Min: 17.34 / Avg: 17.34 / Max: 17.34Min: 17.36 / Avg: 17.36 / Max: 17.361. (CXX) g++ options: -O3 -march=native -flto -rdynamic

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 the Intel oneAPI initiative. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: bf16bf16bf16 - Engine: CPUGCC 11.1Clang 12.00.80391.60782.41173.21564.0195SE +/- 0.00574, N = 3SE +/- 0.00202, N = 33.571273.57288-fopenmp - MIN: 3.5-fopenmp=libomp - MIN: 3.481. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl
OpenBenchmarking.orgms, Fewer Is BetteroneDNN 2.1.2Harness: Deconvolution Batch shapes_3d - Data Type: bf16bf16bf16 - Engine: CPUGCC 11.1Clang 12.0246810Min: 3.56 / Avg: 3.57 / Max: 3.58Min: 3.57 / Avg: 3.57 / Max: 3.581. (CXX) g++ options: -O3 -march=native -flto -std=c++11 -msse4.1 -fPIC -O2 -pie -lpthread -ldl

Darmstadt Automotive Parallel Heterogeneous Suite

DAPHNE is the Darmstadt Automotive Parallel HeterogeNEous Benchmark Suite with OpenCL / CUDA / OpenMP test cases for these automotive benchmarks for evaluating programming models in context to vehicle autonomous driving capabilities. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgTest Cases Per Minute, More Is BetterDarmstadt Automotive Parallel Heterogeneous SuiteBackend: OpenMP - Kernel: Euclidean ClusterGCC 11.12004006008001000SE +/- 9.21, N = 31013.991. (CXX) g++ options: -O3 -std=c++11 -fopenmp

OpenBenchmarking.orgTest Cases Per Minute, More Is BetterDarmstadt Automotive Parallel Heterogeneous SuiteBackend: OpenMP - Kernel: Points2ImageGCC 11.13K6K9K12K15KSE +/- 204.19, N = 314507.811. (CXX) g++ options: -O3 -std=c++11 -fopenmp

OpenBenchmarking.orgTest Cases Per Minute, More Is BetterDarmstadt Automotive Parallel Heterogeneous SuiteBackend: OpenMP - Kernel: NDT MappingGCC 11.12004006008001000SE +/- 2.86, N = 31046.601. (CXX) g++ options: -O3 -std=c++11 -fopenmp

GraphicsMagick

This is a test of GraphicsMagick with its OpenMP implementation that performs various imaging tests on a sample 6000x4000 pixel JPEG image. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: RotateClang 12.0GCC 11.1160320480640800SE +/- 6.08, N = 37457451. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: RotateClang 12.0GCC 11.1130260390520650Min: 735 / Avg: 745 / Max: 7561. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread

GNU GMP GMPbench

GMPbench is a test of the GNU Multiple Precision Arithmetic (GMP) Library. GMPbench is a single-threaded integer benchmark that leverages the GMP library to stress the CPU with widening integer multiplication. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgGMPbench Score, More Is BetterGNU GMP GMPbench 6.2.1Total TimeGCC 11.180016002400320040003871.61. (CC) gcc options: -O3 -march=native -flto -lm

NCNN

NCNN is a high performance neural network inference framework optimized for mobile and other platforms developed by Tencent. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: regnety_400mClang 12.0GCC 11.120406080100SE +/- 1.81, N = 13SE +/- 1.10, N = 1226.4794.38-lomp - MIN: 20.72 / MAX: 164.27-lgomp - MIN: 87.04 / MAX: 668.911. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: regnety_400mClang 12.0GCC 11.120406080100Min: 21.67 / Avg: 26.47 / Max: 41.25Min: 88.16 / Avg: 94.38 / Max: 99.311. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: squeezenet_ssdClang 12.0GCC 11.1510152025SE +/- 0.87, N = 13SE +/- 0.10, N = 1217.9621.18-lomp - MIN: 14.31 / MAX: 155.2-lgomp - MIN: 20.25 / MAX: 100.941. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: squeezenet_ssdClang 12.0GCC 11.1510152025Min: 14.88 / Avg: 17.96 / Max: 22.82Min: 20.8 / Avg: 21.18 / Max: 22.011. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: resnet18Clang 12.0GCC 11.13691215SE +/- 0.49, N = 13SE +/- 0.11, N = 1210.8311.10-lomp - MIN: 8.67 / MAX: 27.15-lgomp - MIN: 10.39 / MAX: 64.831. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: resnet18Clang 12.0GCC 11.13691215Min: 8.92 / Avg: 10.83 / Max: 14.69Min: 10.55 / Avg: 11.1 / Max: 11.951. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: vgg16GCC 11.1Clang 12.0612182430SE +/- 0.17, N = 12SE +/- 0.51, N = 1325.3427.34-lgomp - MIN: 23.97 / MAX: 98.11-lomp - MIN: 21.3 / MAX: 132.891. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: vgg16GCC 11.1Clang 12.0612182430Min: 24.58 / Avg: 25.34 / Max: 26.45Min: 25.54 / Avg: 27.34 / Max: 32.291. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: googlenetClang 12.0GCC 11.1510152025SE +/- 0.78, N = 13SE +/- 0.19, N = 1215.3419.46-lomp - MIN: 12.78 / MAX: 96.28-lgomp - MIN: 17.96 / MAX: 72.731. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: googlenetClang 12.0GCC 11.1510152025Min: 13.46 / Avg: 15.34 / Max: 22.06Min: 18.36 / Avg: 19.46 / Max: 20.751. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: efficientnet-b0Clang 12.0GCC 11.13691215SE +/- 0.72, N = 13SE +/- 0.15, N = 127.7212.48-lomp - MIN: 5.61 / MAX: 59.7-lgomp - MIN: 11.37 / MAX: 205.281. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: efficientnet-b0Clang 12.0GCC 11.148121620Min: 6.28 / Avg: 7.72 / Max: 13.74Min: 11.72 / Avg: 12.48 / Max: 13.71. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: mnasnetClang 12.0GCC 11.13691215SE +/- 0.48, N = 13SE +/- 0.05, N = 115.269.43-lomp - MIN: 3.73 / MAX: 46.61-lgomp - MIN: 8.91 / MAX: 24.891. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: mnasnetClang 12.0GCC 11.13691215Min: 4.38 / Avg: 5.26 / Max: 10.42Min: 9.1 / Avg: 9.43 / Max: 9.661. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: shufflenet-v2Clang 12.0GCC 11.13691215SE +/- 0.37, N = 13SE +/- 0.08, N = 125.9010.55-lomp - MIN: 4.8 / MAX: 33.78-lgomp - MIN: 10 / MAX: 25.641. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: shufflenet-v2Clang 12.0GCC 11.13691215Min: 5 / Avg: 5.9 / Max: 10.25Min: 10.13 / Avg: 10.55 / Max: 11.061. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: mobilenetClang 12.0GCC 11.1510152025SE +/- 0.41, N = 13SE +/- 0.19, N = 1214.2119.40-lomp - MIN: 12.25 / MAX: 44.36-lgomp - MIN: 18 / MAX: 399.431. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread
OpenBenchmarking.orgms, Fewer Is BetterNCNN 20201218Target: CPU - Model: mobilenetClang 12.0GCC 11.1510152025Min: 13.07 / Avg: 14.21 / Max: 18.93Min: 18.83 / Avg: 19.4 / Max: 21.381. (CXX) g++ options: -O3 -march=native -flto -O2 -rdynamic -lpthread

PostgreSQL pgbench

This is a benchmark of PostgreSQL using pgbench for facilitating the database benchmarks. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read Only - Average LatencyClang 12.0GCC 11.10.06230.12460.18690.24920.3115SE +/- 0.002, N = 3SE +/- 0.005, N = 130.2650.2771. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm
OpenBenchmarking.orgms, Fewer Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read Only - Average LatencyClang 12.0GCC 11.112345Min: 0.26 / Avg: 0.27 / Max: 0.27Min: 0.26 / Avg: 0.28 / Max: 0.321. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm

OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read OnlyClang 12.0GCC 11.1200K400K600K800K1000KSE +/- 8511.06, N = 3SE +/- 15989.65, N = 139430439074011. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm
OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 13.0Scaling Factor: 100 - Clients: 250 - Mode: Read OnlyClang 12.0GCC 11.1160K320K480K640K800KMin: 929228.39 / Avg: 943042.95 / Max: 958563.17Min: 791821.06 / Avg: 907401.17 / Max: 980446.691. (CC) gcc options: -fno-strict-aliasing -fwrapv -O3 -march=native -flto -lpgcommon -lpgport -lpq -lpthread -lrt -ldl -lm

GraphicsMagick

This is a test of GraphicsMagick with its OpenMP implementation that performs various imaging tests on a sample 6000x4000 pixel JPEG image. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: ResizingClang 12.0GCC 11.1130260390520650SE +/- 15.04, N = 15SE +/- 4.36, N = 36143801. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.33Operation: ResizingClang 12.0GCC 11.1110220330440550Min: 522 / Avg: 613.87 / Max: 761Min: 373 / Avg: 380 / Max: 3881. (CC) gcc options: -fopenmp -O3 -march=native -flto -pthread -ljpeg -lX11 -lz -lm -lpthread

Geometric Mean Of All Test Results

OpenBenchmarking.orgGeometric Mean, More Is BetterGeometric Mean Of All Test ResultsResult Composite - GCC 11 vs. LLVM Clang 12 Benchmarks On Xeon Ice LakeClang 12.0GCC 11.1122436486055.0150.57

114 Results Shown

NCNN:
  CPU - blazeface
  CPU-v3-v3 - mobilenet-v3
  CPU-v2-v2 - mobilenet-v2
C-Ray
oneDNN:
  Matrix Multiply Batch Shapes Transformer - u8s8f32 - CPU
  Deconvolution Batch shapes_1d - u8s8f32 - CPU
  Matrix Multiply Batch Shapes Transformer - f32 - CPU
OpenSSL
TNN
ASTC Encoder
oneDNN
ASTC Encoder
oneDNN:
  IP Shapes 3D - u8s8f32 - CPU
  Recurrent Neural Network Inference - f32 - CPU
  Recurrent Neural Network Inference - u8s8f32 - CPU
  Recurrent Neural Network Inference - bf16bf16bf16 - CPU
Zstd Compression
oneDNN
Bullet Physics Engine
Coremark
ASTC Encoder
GraphicsMagick
oneDNN:
  Matrix Multiply Batch Shapes Transformer - bf16bf16bf16 - CPU
  Recurrent Neural Network Training - f32 - CPU
  Recurrent Neural Network Training - bf16bf16bf16 - CPU
Liquid-DSP
oneDNN
GraphicsMagick
oneDNN:
  IP Shapes 3D - f32 - CPU
  Deconvolution Batch shapes_1d - bf16bf16bf16 - CPU
WebP2 Image Encode
Himeno Benchmark
Opus Codec Encoding
LAME MP3 Encoding
oneDNN
Bullet Physics Engine
Kripke
Bullet Physics Engine
NCNN
Bullet Physics Engine:
  1000 Convex
  Prim Trimesh
  Convex Trimesh
oneDNN
x265
WebP Image Encode
Crypto++
WebP2 Image Encode
oneDNN
WebP2 Image Encode
TNN
AOBench
oneDNN
WebP2 Image Encode
SVT-HEVC
Kvazaar
SVT-HEVC
Zstd Compression
SVT-AV1
Gcrypt Library
Kvazaar
WebP Image Encode
Kvazaar
Zstd Compression
libjpeg-turbo tjbench
PostgreSQL pgbench:
  100 - 250 - Read Write
  100 - 250 - Read Write - Average Latency
SVT-VP9
oneDNN
WebP Image Encode
WebP2 Image Encode
Kvazaar
SVT-AV1
Zstd Compression
Timed MrBayes Analysis
SVT-VP9
eSpeak-NG Speech Engine
SVT-VP9
SVT-AV1
WebP Image Encode
Zstd Compression:
  19 - Decompression Speed
  19, Long Mode - Decompression Speed
Liquid-DSP
WebP Image Encode
SVT-AV1
Primesieve
oneDNN
Zstd Compression:
  8 - Decompression Speed
  8, Long Mode - Decompression Speed
FLAC Audio Encoding
x265
oneDNN
Caffe:
  AlexNet - CPU - 200
  GoogleNet - CPU - 200
SVT-HEVC
WavPack Audio Encoding
oneDNN
Darmstadt Automotive Parallel Heterogeneous Suite:
  OpenMP - Euclidean Cluster
  OpenMP - Points2Image
  OpenMP - NDT Mapping
GraphicsMagick
GNU GMP GMPbench
NCNN:
  CPU - regnety_400m
  CPU - squeezenet_ssd
  CPU - resnet18
  CPU - vgg16
  CPU - googlenet
  CPU - efficientnet-b0
  CPU - mnasnet
  CPU - shufflenet-v2
  CPU - mobilenet
PostgreSQL pgbench:
  100 - 250 - Read Only - Average Latency
  100 - 250 - Read Only
GraphicsMagick
Geometric Mean Of All Test Results

GCC 11.1

Processor: 2 x Intel Xeon Platinum 8380 @ 3.40GHz (80 Cores / 160 Threads), Motherboard: Intel M50CYP2SB2U (SE5C6200.86B.0022.D08.2103221623 BIOS), Chipset: Intel Device 0998, Memory: 16 x 32 GB DDR4-3200MT/s Hynix HMA84GR7CJR4N-XN, Disk: 800GB INTEL SSDPF21Q800GB, Graphics: ASPEED, Network: 2 x Intel X710 for 10GBASE-T + 2 x Intel E810-C for QSFP

OS: Fedora 34, Kernel: 5.12.6-300.fc34.x86_64 (x86_64), Compiler: GCC 11.1.1 20210428, File-System: xfs, Screen Resolution: 1024x768

Kernel Notes: Transparent Huge Pages: madvise
Environment Notes: CXXFLAGS="-O3 -march=native -flto" CFLAGS="-O3 -march=native -flto"
Compiler Notes: --build=x86_64-redhat-linux --disable-libunwind-exceptions --enable-__cxa_atexit --enable-bootstrap --enable-cet --enable-checking=release --enable-gnu-indirect-function --enable-gnu-unique-object --enable-initfini-array --enable-languages=c,c++,fortran,objc,obj-c++,ada,go,d,lto --enable-multilib --enable-offload-targets=nvptx-none --enable-plugin --enable-shared --enable-threads=posix --mandir=/usr/share/man --with-arch_32=i686 --with-gcc-major-version-only --with-linker-hash-style=gnu --with-tune=generic --without-cuda-driver
Processor Notes: Scaling Governor: intel_pstate performance - CPU Microcode: 0xd000270
Python Notes: Python 3.9.5
Security Notes: SELinux + 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 Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 28 May 2021 11:02 by user .

Clang 12.0

Processor: 2 x Intel Xeon Platinum 8380 @ 3.40GHz (80 Cores / 160 Threads), Motherboard: Intel M50CYP2SB2U (SE5C6200.86B.0022.D08.2103221623 BIOS), Chipset: Intel Device 0998, Memory: 16 x 32 GB DDR4-3200MT/s Hynix HMA84GR7CJR4N-XN, Disk: 800GB INTEL SSDPF21Q800GB, Graphics: ASPEED, Network: 2 x Intel X710 for 10GBASE-T + 2 x Intel E810-C for QSFP

OS: Fedora 34, Kernel: 5.12.6-300.fc34.x86_64 (x86_64), Compiler: Clang 12.0.0, File-System: xfs, Screen Resolution: 1024x768

Kernel Notes: Transparent Huge Pages: madvise
Environment Notes: CXXFLAGS="-O3 -march=native -flto" CFLAGS="-O3 -march=native -flto"
Processor Notes: Scaling Governor: intel_pstate performance - CPU Microcode: 0xd000270
Python Notes: Python 3.9.5
Security Notes: SELinux + 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 Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 28 May 2021 19:40 by user .