ARM vs Thumb GCC comparison

ARMv7 codegen impact of ARM/Thumb mode in GCC 4.9, 7.1, 8.0

HTML result view exported from: https://openbenchmarking.org/result/1705048-RI-MERGE366949.

ARM vs Thumb GCC comparisonProcessorMotherboardMemoryDiskGraphicsOSKernelDisplay ServerCompilerFile-SystemScreen ResolutionGCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #1GCC 4.9 thumb #1GCC 4.9 arm #1GCC 4.9 thumb #2GCC 4.9 arm #2GCC 7.1 thumb #2GCC 7.1 arm #2GCC 8.0 thumb #2GCC 8.0 arm #2ARMv7 rev 1 @ 1.73GHz (4 Cores)ODROIDC915MB64GB 00000ARM MaliUbuntu 14.043.10.80-20-odrobian (armv7l)X Server 1.15.1GCC 7.0.1 20170422 + Clang 3.6.0-2ubuntu1~trusty1 + LLVM 3.6.0ext41280x1024GCC 8.0.0 20170501 + Clang 3.6.0-2ubuntu1~trusty1 + LLVM 3.6.0GCC 4.9.4 + Clang 3.6.0-2ubuntu1~trusty1 + LLVM 3.6.0GCC 7.0.1 20170422 + Clang 3.6.0-2ubuntu1~trusty1 + LLVM 3.6.0GCC 8.0.0 20170501 + Clang 3.6.0-2ubuntu1~trusty1 + LLVM 3.6.0OpenBenchmarking.orgCompiler Details- GCC 7.1 thumb #1: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 7.1 arm #1: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 8.0 thumb #1: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 8.0 arm #1: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 4.9 thumb #1: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-objc-gc --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb -v - GCC 4.9 arm #1: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-objc-gc --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb -v - GCC 4.9 thumb #2: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-objc-gc --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb -v - GCC 4.9 arm #2: --build=arm-linux-gnueabihf --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-objc-gc --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-float=hard --with-fpu=vfpv3-d16 --with-mode=thumb -v - GCC 7.1 thumb #2: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 7.1 arm #2: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 8.0 thumb #2: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v - GCC 8.0 arm #2: --build=arm-linux-gnueabihf --disable-bootstrap --disable-browser-plugin --disable-libitm --disable-libquadmath --disable-libstdcxx-dual-abi --disable-sjlj-exceptions --disable-werror --enable-checking=release --enable-clocale=gnu --enable-gnu-unique-object --enable-languages=c,c++,fortran --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-multilib --enable-nls --enable-plugin --enable-shared --enable-threads=posix --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf --with-arch-directory=arm --with-arch=armv7-a --with-default-libstdcxx-abi=gcc4-compatible --with-float=hard --with-fpu=vfpv3 --with-mode=arm -v Processor Details- Scaling Governor: meson_cpufreq performance

ARM vs Thumb GCC comparisonttsiod-renderer: Phong Rendering With Soft-Shadow Mappinghimeno: Poisson Pressure Solverc-ray: Total Timeprimesieve: 1e12 Prime Number Generationstockfish: Total Timetachyon: Total Timescimark2: Compositescimark2: Monte Carloscimark2: Fast Fourier Transformscimark2: Sparse Matrix Multiplyscimark2: Dense LU Matrix Factorizationscimark2: Jacobi Successive Over-Relaxationtscp: AI Chess Performancesmallpt: Global Illumination Renderer; 100 Samplesn-queens: Elapsed TimeGCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #1GCC 4.9 thumb #1GCC 4.9 arm #1GCC 4.9 thumb #2GCC 4.9 arm #2GCC 7.1 thumb #2GCC 7.1 arm #2GCC 8.0 thumb #2GCC 8.0 arm #215.6288.14244.361025.3543893109.82102.6758.1122.8898.19115.42218.72186603285195.8515.8679.09245.331027.1651235108.99104.6467.7022.2998.18115.53219.48189365284201.2416.5284.18243.381028.3743547109.73104.0558.1421.0799.02116.86225.17185241285195.8315.8362.49243.761028.0252205109.52103.4667.1922.4597.00114.10216.56190842285201.3015.5691.27265.69877.1541922110.57100.3056.9521.3298.24124.60200.40179364284185.4115.7981.69266.741029.4353864108.8799.9764.5622.6497.15122.69192.80187058284227.37OpenBenchmarking.org

TTSIOD 3D Renderer

Phong Rendering With Soft-Shadow Mapping

OpenBenchmarking.orgFPS, More Is BetterTTSIOD 3D Renderer 2.3aPhong Rendering With Soft-Shadow MappingGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #148121620SE +/- 0.08, N = 3SE +/- 0.02, N = 2SE +/- 0.02, N = 3SE +/- 0.02, N = 315.5615.7915.6215.8616.5215.83-marm-mthumb-marm-mthumb-marm1. (CXX) g++ options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -ffast-math -mtune=native -flto -lSDL -lstdc++

Himeno Benchmark

Poisson Pressure Solver

OpenBenchmarking.orgMFLOPS, More Is BetterHimeno Benchmark 3.0Poisson Pressure SolverGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #120406080100SE +/- 1.21, N = 3SE +/- 0.96, N = 3SE +/- 0.23, N = 3SE +/- 1.02, N = 3SE +/- 1.59, N = 3SE +/- 1.31, N = 691.2781.6988.1479.0984.1862.49-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -O3 -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers

C-Ray

Total Time

OpenBenchmarking.orgSeconds, Fewer Is BetterC-Ray 1.1Total TimeGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #160120180240300SE +/- 1.04, N = 3SE +/- 1.92, N = 3SE +/- 0.30, N = 3SE +/- 1.31, N = 3SE +/- 0.19, N = 3SE +/- 0.22, N = 3265.69266.74244.36245.33243.38243.76-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -lm -lpthread -O3 -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers

Primesieve

1e12 Prime Number Generation

OpenBenchmarking.orgSeconds, Fewer Is BetterPrimesieve 5.4.21e12 Prime Number GenerationGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #12004006008001000SE +/- 3.08, N = 3SE +/- 10.97, N = 3SE +/- 5.98, N = 3SE +/- 0.31, N = 3SE +/- 1.95, N = 3SE +/- 5.44, N = 3877.151029.431025.351027.161028.371028.02-mthumb-marm-mthumb-marm-mthumb-marm1. (CXX) g++ options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -fopenmp

Stockfish

Total Time

OpenBenchmarking.orgms, Fewer Is BetterStockfish 2014-11-26Total TimeGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #112K24K36K48K60KSE +/- 122.93, N = 3SE +/- 60.67, N = 3SE +/- 663.13, N = 6SE +/- 58.33, N = 3SE +/- 710.15, N = 3SE +/- 221.70, N = 3419225386443893512354354752205-mthumb-marm-mthumb-marm-mthumb-marm1. (CXX) g++ options: -lpthread -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -fno-exceptions -fno-rtti -ansi -pedantic -O3 -flto

Tachyon

Total Time

OpenBenchmarking.orgSeconds, Fewer Is BetterTachyon 0.98.9Total TimeGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #120406080100SE +/- 0.36, N = 3SE +/- 0.13, N = 3SE +/- 0.45, N = 3SE +/- 0.39, N = 3SE +/- 0.36, N = 3SE +/- 0.28, N = 3110.57108.87109.82108.99109.73109.52

SciMark

Computational Test: Composite

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: CompositeGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #120406080100SE +/- 1.12, N = 4SE +/- 1.17, N = 4SE +/- 1.09, N = 4SE +/- 1.08, N = 4SE +/- 1.07, N = 4SE +/- 1.35, N = 4100.3099.97102.67104.64104.05103.46-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

SciMark

Computational Test: Monte Carlo

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: Monte CarloGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #11530456075SE +/- 0.47, N = 4SE +/- 0.56, N = 4SE +/- 0.43, N = 4SE +/- 0.49, N = 4SE +/- 0.44, N = 4SE +/- 0.60, N = 456.9564.5658.1167.7058.1467.19-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

SciMark

Computational Test: Fast Fourier Transform

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: Fast Fourier TransformGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #1510152025SE +/- 0.32, N = 4SE +/- 0.24, N = 4SE +/- 0.18, N = 4SE +/- 0.17, N = 4SE +/- 0.17, N = 4SE +/- 0.21, N = 421.3222.6422.8822.2921.0722.45-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

SciMark

Computational Test: Sparse Matrix Multiply

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: Sparse Matrix MultiplyGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #120406080100SE +/- 1.21, N = 4SE +/- 1.31, N = 4SE +/- 1.12, N = 4SE +/- 1.13, N = 4SE +/- 1.20, N = 4SE +/- 1.37, N = 498.2497.1598.1998.1899.0297.00-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

SciMark

Computational Test: Dense LU Matrix Factorization

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: Dense LU Matrix FactorizationGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #1306090120150SE +/- 1.39, N = 4SE +/- 1.56, N = 4SE +/- 1.27, N = 4SE +/- 1.25, N = 4SE +/- 1.44, N = 4SE +/- 1.55, N = 4124.60122.69115.42115.53116.86114.10-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

SciMark

Computational Test: Jacobi Successive Over-Relaxation

OpenBenchmarking.orgMflops, More Is BetterSciMark 2.0Computational Test: Jacobi Successive Over-RelaxationGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #150100150200250SE +/- 2.23, N = 4SE +/- 2.19, N = 4SE +/- 2.47, N = 4SE +/- 2.37, N = 4SE +/- 2.33, N = 4SE +/- 3.03, N = 4200.40192.80218.72219.48225.17216.56-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -lm

TSCP

AI Chess Performance

OpenBenchmarking.orgNodes Per Second, More Is BetterTSCP 1.81AI Chess PerformanceGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #140K80K120K160K200KSE +/- 1328.11, N = 5SE +/- 1704.24, N = 5SE +/- 1317.11, N = 5SE +/- 1793.38, N = 5SE +/- 1601.83, N = 5SE +/- 1539.07, N = 5179364187058186603189365185241190842-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -O3 -march=native

Smallpt

Global Illumination Renderer; 100 Samples

OpenBenchmarking.orgSeconds, Fewer Is BetterSmallpt 1.0Global Illumination Renderer; 100 SamplesGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #160120180240300SE +/- 0.58, N = 3SE +/- 0.33, N = 3SE +/- 1.53, N = 3284284285284285285-mthumb-marm-mthumb-marm-mthumb-marm1. (CXX) g++ options: -fopenmp -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers

N-Queens

Elapsed Time

OpenBenchmarking.orgSeconds, Fewer Is BetterN-Queens 1.0Elapsed TimeGCC 4.9 thumb #1GCC 4.9 arm #1GCC 7.1 thumb #1GCC 7.1 arm #1GCC 8.0 thumb #1GCC 8.0 arm #150100150200250SE +/- 0.01, N = 3SE +/- 0.01, N = 3SE +/- 0.01, N = 3SE +/- 0.02, N = 3SE +/- 0.02, N = 3SE +/- 0.06, N = 3185.41227.37195.85201.24195.83201.30-mthumb-marm-mthumb-marm-mthumb-marm1. (CC) gcc options: -static -fopenmp -mcpu=cortex-a5 -Ofast -fomit-frame-pointer -fipa-pta -mfpu=neon-vfpv4 -ftree-vectorize -frename-registers -O3 -march=native

Phoronix Test Suite v10.8.4