LLVM Clang 3.2 Loop Vectorizer

Intel Core i7-3960X testing of the automatic loop vectorizer in LLVM 3.2 with the Clang compiler. Benchmarking by Michael Larabel for a future article on phoronix.com.

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 1210264-RA-LLVMLOOPV66
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:

Bioinformatics 2 Tests
C/C++ Compiler Tests 5 Tests
CPU Massive 5 Tests
Creator Workloads 3 Tests
HPC - High Performance Computing 2 Tests
Multi-Core 4 Tests
Renderers 2 Tests
Scientific Computing 2 Tests
Server CPU Tests 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
Default
October 26 2012
 
Loop Vectorization
October 26 2012
 
Invert Hiding All Results Option
 
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):


LLVM Clang 3.2 Loop VectorizerOpenBenchmarking.orgPhoronix Test SuiteIntel Core i7-3960X @ 3.30GHz (12 Cores)Intel DX79SIIntel Xeon E5/Core8192MB64GB OCZ VERTEXAMD Radeon HD 4650 512MBRealtek ALC892DELL S2409WIntel 82579LM Gigabit ConnectionUbuntu 12.103.5.0-17-generic (x86_64)Unity 6.8.0X Server 1.13.0radeon 6.99.992.1 Mesa 9.0 Gallium 0.4Clang 3.2 (SVN 166775) + LLVM 3.2svnext41920x1080ProcessorMotherboardChipsetMemoryDiskGraphicsAudioMonitorNetworkOSKernelDesktopDisplay ServerDisplay DriverOpenGLCompilerFile-SystemScreen ResolutionLLVM Clang 3.2 Loop Vectorizer BenchmarksSystem Logs- Optimized build; Built Oct 26 2012 (10:29:50); Default target: x86_64-unknown-linux-gnu; Host CPU: corei7-avx - Scaling Governor: ondemand- Compiz was running on this system.

Loop Vectorization vs. Default ComparisonPhoronix Test SuiteBaseline+30%+30%+60%+60%+90%+90%+120%+120%120%12.3%10.7%6%2.6%2.6%L.A.TBlurTotal TimeResizingT.B.T.P.S3.7%P.P.SG.I.R.1.SGraphicsMagickGraphicsMagickC-RayGraphicsMagickPostgreSQL pgbenchHimeno BenchmarkSmallptLoop VectorizationDefault

LLVM Clang 3.2 Loop Vectorizergraphics-magick: Blurgraphics-magick: Sharpengraphics-magick: Resizinggraphics-magick: HWB Color Spacegraphics-magick: Local Adaptive Thresholdinghimeno: Poisson Pressure Solverpgbench: TPC-B Transactions Per Secondhmmer: Pfam Database Searchc-ray: Total Timesmallpt: Global Illumination Renderer; 100 SamplesLoop VectorizationDefault733184122201554.66336.6515.3723.15157823189123441595.35324.7715.3520.92153OpenBenchmarking.org

GraphicsMagick

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: BlurLoop VectorizationDefault20406080100SE +/- 0.00, N = 3SE +/- 0.00, N = 37382-mllvm -vectorize -lpng121. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: BlurLoop VectorizationDefault1632486480Min: 73 / Avg: 73 / Max: 73Min: 82 / Avg: 82 / Max: 821. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: SharpenLoop VectorizationDefault714212835SE +/- 0.00, N = 3SE +/- 0.00, N = 33131-mllvm -vectorize -lpng121. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: SharpenLoop VectorizationDefault714212835Min: 31 / Avg: 31 / Max: 31Min: 31 / Avg: 31 / Max: 311. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: ResizingLoop VectorizationDefault20406080100SE +/- 0.00, N = 3SE +/- 0.00, N = 38489-mllvm -vectorize -lpng121. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: ResizingLoop VectorizationDefault20406080100Min: 84 / Avg: 84 / Max: 84Min: 89 / Avg: 89 / Max: 891. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: HWB Color SpaceLoop VectorizationDefault306090120150SE +/- 0.33, N = 3SE +/- 0.00, N = 3122123-mllvm -vectorize -lpng121. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: HWB Color SpaceLoop VectorizationDefault20406080100Min: 121 / Avg: 121.67 / Max: 122Min: 123 / Avg: 123 / Max: 1231. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread

OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: Local Adaptive ThresholdingLoop VectorizationDefault1020304050SE +/- 0.00, N = 3SE +/- 0.00, N = 32044-mllvm -vectorize -lpng121. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread
OpenBenchmarking.orgIterations Per Minute, More Is BetterGraphicsMagick 1.3.16Operation: Local Adaptive ThresholdingLoop VectorizationDefault918273645Min: 20 / Avg: 20 / Max: 20Min: 44 / Avg: 44 / Max: 441. (CC) gcc options: -O3 -march=native -pthread -lXext -lX11 -lz -lm -lpthread

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 SolverLoop VectorizationDefault30060090012001500SE +/- 6.98, N = 3SE +/- 3.05, N = 31554.661595.35-mllvm -vectorize1. (CC) gcc options: -O3 -march=native
OpenBenchmarking.orgMFLOPS, More Is BetterHimeno Benchmark 3.0Poisson Pressure SolverLoop VectorizationDefault30060090012001500Min: 1542.56 / Avg: 1554.66 / Max: 1566.74Min: 1589.3 / Avg: 1595.35 / Max: 1599.031. (CC) gcc options: -O3 -march=native

PostgreSQL pgbench

OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 8.4.11TPC-B Transactions Per SecondLoop VectorizationDefault70140210280350SE +/- 3.97, N = 3SE +/- 0.52, N = 3336.65324.77-mllvm -vectorize1. (CC) gcc options: -O3 -march=native -fno-strict-aliasing -fwrapv -lpgport -lpq -lcrypt -ldl -lm
OpenBenchmarking.orgTPS, More Is BetterPostgreSQL pgbench 8.4.11TPC-B Transactions Per SecondLoop VectorizationDefault60120180240300Min: 332.05 / Avg: 336.65 / Max: 344.55Min: 324.06 / Avg: 324.77 / Max: 325.791. (CC) gcc options: -O3 -march=native -fno-strict-aliasing -fwrapv -lpgport -lpq -lcrypt -ldl -lm

Timed HMMer Search

This test searches through the Pfam database of profile hidden markov models. The search finds the domain structure of Drosophila Sevenless protein. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed HMMer Search 2.3.2Pfam Database SearchLoop VectorizationDefault48121620SE +/- 0.02, N = 3SE +/- 0.05, N = 315.3715.35-mllvm -vectorize1. (CC) gcc options: -O3 -march=native -pthread -lhmmer -lsquid -lm
OpenBenchmarking.orgSeconds, Fewer Is BetterTimed HMMer Search 2.3.2Pfam Database SearchLoop VectorizationDefault48121620Min: 15.35 / Avg: 15.37 / Max: 15.41Min: 15.29 / Avg: 15.35 / Max: 15.461. (CC) gcc options: -O3 -march=native -pthread -lhmmer -lsquid -lm

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 TimeLoop VectorizationDefault612182430SE +/- 0.04, N = 3SE +/- 0.03, N = 323.1520.92-mllvm -vectorize1. (CC) gcc options: -lm -lpthread -O3 -march=native
OpenBenchmarking.orgSeconds, Fewer Is BetterC-Ray 1.1Total TimeLoop VectorizationDefault510152025Min: 23.11 / Avg: 23.15 / Max: 23.23Min: 20.89 / Avg: 20.92 / Max: 20.971. (CC) gcc options: -lm -lpthread -O3 -march=native

Smallpt

Smallpt is a C++ global illumination renderer written in less than 100 lines of code. Global illumination is done via unbiased Monte Carlo path tracing and there is multi-threading support via the OpenMP library. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterSmallpt 1.0Global Illumination Renderer; 100 SamplesLoop VectorizationDefault306090120150SE +/- 0.33, N = 3SE +/- 0.33, N = 3157153-mllvm -vectorize1. (CXX) g++ options: -fopenmp -O3 -march=native
OpenBenchmarking.orgSeconds, Fewer Is BetterSmallpt 1.0Global Illumination Renderer; 100 SamplesLoop VectorizationDefault306090120150Min: 157 / Avg: 157.33 / Max: 158Min: 152 / Avg: 152.67 / Max: 1531. (CXX) g++ options: -fopenmp -O3 -march=native