<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div><blockquote type="cite" class=""><div class="">On May 18, 2017, at 3:30 PM, Dehao Chen via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" class="">llvm-dev@lists.llvm.org</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class="">Hi,<div class=""><br class=""></div><div class="">I'm proposing to make vectorizer-maximize-bandwidth on by default for loop vectorizer because it should generally help performance.</div><div class=""><br class=""></div><div class="">I've tested the performance impact on Intel sandybridge machine with speccpu benchmarks:</div><div class=""><br class=""></div><div class=""><div class=""> Benchmark Base:Reference (1) </div><div class="">-------------------------------------------------------</div><div class="">spec/2006/fp/C++/444.namd 26.84 -0.31%</div><div class="">spec/2006/fp/C++/447.dealII 46.19 +0.89%</div><div class="">spec/2006/fp/C++/450.soplex 42.92 -0.44%</div><div class="">spec/2006/fp/C++/453.povray 38.57 -2.25%</div><div class="">spec/2006/fp/C/433.milc 24.54 -0.76%</div><div class="">spec/2006/fp/C/470.lbm 41.08 +0.26%</div><div class="">spec/2006/fp/C/482.sphinx3 47.58 -0.99%</div><div class="">spec/2006/int/C++/471.omnetpp 22.06 +1.87%</div><div class="">spec/2006/int/C++/473.astar 22.65 -0.12%</div><div class="">spec/2006/int/C++/483.xalancbmk 33.69 +4.97%</div><div class="">spec/2006/int/C/400.perlbench 33.43 +1.70%</div><div class="">spec/2006/int/C/401.bzip2 23.02 -0.19%</div><div class="">spec/2006/int/C/403.gcc 32.57 -0.43%</div><div class="">spec/2006/int/C/429.mcf 40.35 +0.27%</div><div class="">spec/2006/int/C/445.gobmk 26.96 +0.06%</div><div class="">spec/2006/int/C/456.hmmer 24.4 +0.19%</div><div class="">spec/2006/int/C/458.sjeng 27.91 -0.08%</div><div class="">spec/2006/int/C/462.libquantum 57.47 -0.20%</div><div class="">spec/2006/int/C/464.h264ref 46.52 +1.35%</div><div class=""><br class=""></div><div class="">geometric mean +0.29%</div><div class=""><br class=""></div><div class=""> Scores are benchmark specific.</div></div><div class=""><br class=""></div><div class="">We do have regression on 453.povray, but it's due to secondary effects as all hot functions are the same. I've also tested the code size impact, it does not change for tested speccpu benchmarks.</div></div></div></blockquote><div><br class=""></div><div>Can you please describe the config for the runs (optimization level, PGO/no-PGO, etc).</div><div><br class=""></div><div>It would be good to provide analysis for the changes >1%. I.e. we need to make sure that the improvements are not noise either ;).</div><div><br class=""></div><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class=""><br class=""></div><div class="">I've prepared <a href="https://reviews.llvm.org/D33341" rel="noreferrer" target="_blank" style="font-size:12.8px" class="">https://reviews.llvm.org/<wbr class="">D33341</a> to do this.</div><div class=""><br class=""></div><div class="">I really appreciate if the community can help test the performance impact of this change on other architectures so that we can decide if this should go target-dependent.</div></div></div></blockquote><div><br class=""></div><div>I will run it on Cyclone/AArch64 next week.</div><div><br class=""></div><div>Adam</div><br class=""><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class=""><br class=""></div><div class="">Any comments/concerns?</div><div class=""><br class=""></div><div class="">Thanks,</div><div class="">Dehao</div></div>
_______________________________________________<br class="">LLVM Developers mailing list<br class=""><a href="mailto:llvm-dev@lists.llvm.org" class="">llvm-dev@lists.llvm.org</a><br class="">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev<br class=""></div></blockquote></div><br class=""></body></html>