<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Mon, May 1, 2017 at 6:56 PM, Wei Mi <span dir="ltr"><<a href="mailto:wmi@google.com" target="_blank">wmi@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">> This also needs compile time and performance measurements.<br>
<br>
</span>I used spec2006 to do compile time testing. The compile time had a<br>
small increase.<br>
<br>
400.perlbench       0.54%<br>
401.bzip2              1.62%<br>
403.gcc                 0.79%<br>
429.mcf                 1.15%<br>
433.milc                 0.24%<br>
445.gobmk             0.71%<br>
456.hmmer            0.26%<br>
458.sjeng              -0.53%<br>
462.libquantum     0.22%<br>
464.h264ref          -0.18%<br>
470.lbm                 0.86%<br>
482.sphinx3           -1.01%<br>
998.specrand        -2.35%<br>
999.specrand        -3.45%<br>
444.namd              0.46%<br>
447.dealII              1.65%<br>
450.soplex             0.92%<br>
453.povray             0.88%<br>
471.omnetpp          1.77%<br>
473.astar                1.23%<br>
483.xalancbmk       1.33%<br>
<br>
However, the compile time increase is not caused by phiTranslate call,<br>
but caused by some data structure setting like numberingExpression map<br>
(I removed the phiTranslate call and the compile time increase is<br>
almost the same). I think that is because GVN is already a big compile<br>
time consumer (top3 according to<br>
<a href="http://lists.llvm.org/pipermail/llvm-dev/2016-November/107110.html" rel="noreferrer" target="_blank">http://lists.llvm.org/<wbr>pipermail/llvm-dev/2016-<wbr>November/107110.html</a>),<br>
and even small increase in the existing cost will be reflected in the<br>
total compile time.<br>
<br>
Considering this is a temporary solution, is this compile time<br>
increase acceptable?<br></blockquote><div><br></div><div>Are you willing to commit to solving it?</div><div><br></div><div>I'm generally going to be okay with increasing compile time a bit if the same people are willing to commit to solving that, but less so if they just want someone else to solve it later :)</div><div><br></div></div></div></div>