[PATCH] D132208: [LoopIntWrapPredication] Loop Integer Wrapping Predication Pass
Sergei Kachkov via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Dec 8 01:17:42 PST 2022
kachkov98 added a comment.
Compile-time results on llvm-test-suite + SPEC2006 + SPEC2017:
Program loop-int-wrap-predication.NumPredicatedChains
test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2029.00
test-suite :: External/SPEC/CFP2006/447.dealII/447.dealII.test 250.00
test-suite :: External/SPEC/CINT2017rate/502.gcc_r/502.gcc_r.test 24.00
test-suite :: External/SPEC/CINT2017speed/602.gcc_s/602.gcc_s.test 24.00
test-suite :: SingleSource/UnitTests/matrix-types-spec.test 17.00
test-suite :: External/SPEC/CINT2006/403.gcc/403.gcc.test 12.00
test-suite :: MultiSource/Applications/JM/lencod/lencod.test 8.00
test-suite :: External/SPEC/CINT2017rate/523.xalancbmk_r/523.xalancbmk_r.test 7.00
test-suite :: External/SPEC/CINT2017speed/623.xalancbmk_s/623.xalancbmk_s.test 7.00
test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 7.00
test-suite :: External/SPEC/CINT2006/483.xalancbmk/483.xalancbmk.test 7.00
test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test 7.00
test-suite :: MultiSource/Applications/ClamAV/clamscan.test 6.00
test-suite :: SingleSource/UnitTests/Vectorizer/runtime-checks.test 6.00
test-suite :: MultiSource/Benchmarks/Prolangs-C/agrep/agrep.test 6.00
test-suite :: MultiSource/Benchmarks/FreeBench/fourinarow/fourinarow.test 5.00
test-suite :: MultiSource/Benchmarks/MiBench/consumer-lame/consumer-lame.test 5.00
test-suite :: External/SPEC/CINT2006/464.h264ref/464.h264ref.test 4.00
test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C++/CLAMR/CLAMR.test 4.00
test-suite :: MultiSource/Benchmarks/PAQ8p/paq8p.test 4.00
test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 3.00
test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 3.00
test-suite :: MultiSource/Benchmarks/Rodinia/backprop/backprop.test 2.00
test-suite :: External/SPEC/CINT2017speed/600.perlbench_s/600.perlbench_s.test 2.00
test-suite :: SingleSource/Benchmarks/Misc-C++/oopack_v1p8.test 2.00
test-suite :: External/SPEC/CINT2017rate/500.perlbench_r/500.perlbench_r.test 2.00
test-suite :: MultiSource/Benchmarks/mediabench/mpeg2/mpeg2dec/mpeg2decode.test 2.00
test-suite :: External/SPEC/CFP2006/453.povray/453.povray.test 1.00
test-suite :: External/SPEC/CINT2017speed/657.xz_s/657.xz_s.test 1.00
test-suite :: MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4.test 1.00
test-suite :: External/SPEC/CINT2017rate/557.xz_r/557.xz_r.test 1.00
test-suite :: MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg.test 1.00
test-suite :: External/SPEC/CINT2006/462.libquantum/462.libquantum.test 1.00
test-suite :: MultiSource/Benchmarks/mediabench/jpeg/jpeg-6a/cjpeg.test 1.00
test-suite :: External/SPEC/CFP2017rate/511.povray_r/511.povray_r.test 1.00
Big number of cases in 510.parest_r is due to template instantiations of almost the same code (sparse_matrix class methods). In this benchmark (on ref data), I've observed 1.7 % improvement on T-head RVB-ICE RISC-V board with XuanTie C910 core (before: 3 090 706 ms, after: 3 038 376 ms). Results on SPEC2006:
Benchmark baseline, ms optimized, ms diff
400.perlbench 1 493 603,64 1 497 198,00 100,24%
401.bzip2 1 905 760,43 1 903 672,13 99,89%
403.gcc 1 495 701,76 1 492 812,78 99,81%
429.mcf 2 009 662,24 2 018 930,79 100,46%
445.gobmk 1 502 378,31 1 495 364,06 99,53%
456.hmmer 1 900 147,45 1 860 074,98 97,89%
458.sjeng 1 837 567,39 1 839 056,42 100,08%
462.libquantum 1 171 120,66 1 156 115,71 98,72%
464.h264ref 2 386 369,44 2 359 021,70 98,85%
471.omnetpp 1 688 319,54 1 685 899,41 99,86%
473.astar 1 670 467,01 1 661 233,42 99,45%
483.xalancbmk 1 450 736,57 1 463 321,58 100,87%
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D132208/new/
https://reviews.llvm.org/D132208
More information about the llvm-commits
mailing list