[all-commits] [llvm/llvm-project] 764223: [SLP]Support reordered buildvector nodes for bette...

Alexey Bataev via All-commits all-commits at lists.llvm.org
Wed Nov 6 07:51:38 PST 2024 

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 76422385c3081475ed1bf0e23aa2f3913e66c5b8
      https://github.com/llvm/llvm-project/commit/76422385c3081475ed1bf0e23aa2f3913e66c5b8
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2024-11-06 (Wed, 06 Nov 2024)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll

  Log Message:
  -----------
  [SLP]Support reordered buildvector nodes for better clustering

Patch adds reordering of the buildvector nodes for better clustering of
the compatible operations and future vectorization. Includes basic cost
estimation and if the transformation is not profitable - reverts it.

AVX512, -O3+LTO
Metric: size..text

Program                                                                          size..text
                                                                                       results     results0    diff
                        test-suite :: External/SPEC/CINT2006/401.bzip2/401.bzip2.test    74565.00    75701.00  1.5%
                test-suite :: External/SPEC/CINT2017rate/541.leela_r/541.leela_r.test    75773.00    76397.00  0.8%
               test-suite :: External/SPEC/CINT2017speed/641.leela_s/641.leela_s.test    75773.00    76397.00  0.8%
               test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2014462.00  2024494.00  0.5%
                         test-suite :: MultiSource/Applications/JM/ldecod/ldecod.test   395219.00   396979.00  0.4%
                         test-suite :: MultiSource/Applications/JM/lencod/lencod.test   857795.00   859667.00  0.2%
                    test-suite :: External/SPEC/CINT2006/464.h264ref/464.h264ref.test   800472.00   802440.00  0.2%
                       test-suite :: External/SPEC/CFP2006/447.dealII/447.dealII.test   590699.00   591403.00  0.1%
        test-suite :: MultiSource/Benchmarks/MiBench/consumer-lame/consumer-lame.test   203006.00   203102.00  0.0%
            test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/miniGMG/miniGMG.test    42408.00    42424.00  0.0%
            test-suite ::  External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12451575.00  12451927.00  0.0%
            test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1396480.00  1396448.00 -0.0%
             test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1396480.00  1396448.00 -0.0%
                        test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test  1047708.00  1047580.00 -0.0%
        test-suite :: MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg.test   111344.00   111328.00 -0.0%
                test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test  1087660.00  1087500.00 -0.0%
       test-suite :: MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc.test   280664.00   280616.00 -0.0%
                          test-suite :: MultiSource/Applications/sqlite3/sqlite3.test   502646.00   502006.00 -0.1%
                      test-suite :: MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4.test  1033135.00  1031567.00 -0.2%
        test-suite :: External/SPEC/CINT2017rate/500.perlbench_r/500.perlbench_r.test  2070917.00  2065845.00 -0.2%
       test-suite :: External/SPEC/CINT2017speed/600.perlbench_s/600.perlbench_s.test  2070917.00  2065845.00 -0.2%
                        test-suite :: External/SPEC/CINT2006/473.astar/473.astar.test    33893.00    33797.00 -0.3%
          test-suite :: MultiSource/Benchmarks/MiBench/telecomm-gsm/telecomm-gsm.test    39677.00    39549.00 -0.3%
                 test-suite :: MultiSource/Benchmarks/mediabench/gsm/toast/toast.test    39674.00    39546.00 -0.3%
test-suite :: MultiSource/Benchmarks/MiBench/security-blowfish/security-blowfish.test    11560.00    11512.00 -0.4%
                 test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   653867.00   649275.00 -0.7%
                  test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   653867.00   649275.00 -0.7%

CINT2006/401.bzip2 - extra code vectorized
CINT2017rate/541.leela_r
CINT2017speed/641.leela_s - function
_ZN9FastBoard25get_pattern3_augment_specEiib not inlined anymore, better
vectorization
CFP2017rate/510.parest_r - better vectorization
JM/ldecod - better vectorization
JM/lencod - same
CINT2006/464.h264ref - extra code vectorized
CFP2006/447.dealII - extra vector code
MiBench/consumer-lame - vectorized 2 loops previously scalar
DOE-ProxyApps-C/miniGMG - small changes
Benchmarks/7zip - extra code vectorized, better vectorization
CFP2017rate/526.blender_r - extra vectorization
CFP2017speed/638.imagick_s
CFP2017rate/538.imagick_r - extra vectorization
MiBench/consumer-jpeg - extra vectorization
CINT2006/400.perlbench - extra vectorization
Prolangs-C/TimberWolfMC - small variations
Applications/sqlite3 - extra function vectorized and inlined
Benchmarks/tramp3d-v4 - extra code vectorized
CINT2017rate/500.perlbench_r
CINT2017speed/600.perlbench_s - extra code vectorized, function digcpy gets
vectorized and inlined
CINT2006/473.astar - extra code vectorized
MiBench/telecomm-gsm - extra code vectorized, better vector code
mediabench/gsm - same
MiBench/security-blowfish - extra code vectorized
CINT2017speed/625.x264_s
CINT2017rate/525.x264_r - sub4x4_dct function vectorized and gets
inlined

RISCV-V, SiFive-p670, O3+LTO

CFP2017rate/510.parest_r - extra vectorization
CFP2017rate/526.blender_r - extra vectorization
MiBench/consumer-lame - extra vectorized code

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/114284



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications

More information about the All-commits mailing list