<div dir="ltr">The table is this.<div><font face="monospace, monospace"><br></font></div><div><div><font face="monospace, monospace"> # of cores Before After</font></div><div><font face="monospace, monospace"> 1 13.462 15.795 +17.33%</font></div><div><font face="monospace, monospace"> 2 9.766 10.046 +2.86%</font></div><div><font face="monospace, monospace"> 4 7.697 7.228 -6.09%</font></div><div><font face="monospace, monospace"> 8 6.888 5.672 -17.65%</font></div><div><font face="monospace, monospace"> 12 7.073 5.848 -17.31%</font></div><div><font face="monospace, monospace"> 16 7.066 5.746 -18.68%</font></div><div><font face="monospace, monospace"> 20 6.846 5.482 -19.92%</font></div></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Dec 5, 2016 at 11:21 PM, Rui Ueyama via Phabricator <span dir="ltr"><<a href="mailto:reviews@reviews.llvm.org" target="_blank">reviews@reviews.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">ruiu updated this revision to Diff 80381.<br>
ruiu added a comment.<br>
<br>
I was about to commit <a href="https://reviews.llvm.org/D27155" rel="noreferrer" target="_blank">https://reviews.llvm.org/<wbr>D27155</a>, but it turned out that the single-core<br>
performance of that change is too poor to check-in. So I want to submit<br>
this patch instead.<br>
<br>
This doesn't scale that much compared to <a href="https://reviews.llvm.org/D27155" rel="noreferrer" target="_blank">https://reviews.llvm.org/<wbr>D27155</a>, but its single-core<br>
performance doesn't such as well. Here is clang link time in seconds.<br>
As you can see, it's performance is competitive if you have at least<br>
two cores. However, it doesn't scale more than 8 cores. (I guess the<br>
small improvements over 8 cores comes from other parts of the linker.)<br>
<br>
1. of cores Before After 1 13.462 15.795 +17.33% 2 9.766 10.046 +2.86% 4 7.697 7.228 -6.09% 8 6.888 5.672 -17.65% 12 7.073 5.848 -17.31% 16 7.066 5.746 -18.68% 20 6.846 5.482 -19.92%<br>
<br>
Linking clang-scale programs only with single core is very painful<br>
anyways, so I think this performance characteristics is okay.<br>
<span class=""><br>
<br>
<a href="https://reviews.llvm.org/D27152" rel="noreferrer" target="_blank">https://reviews.llvm.org/<wbr>D27152</a><br>
<br>
Files:<br>
ELF/InputSection.h<br>
ELF/OutputSections.cpp<br>
ELF/OutputSections.h<br>
</span> test/ELF/comment-gc.s<br>
test/ELF/compressed-debug-<wbr>input.s<br>
test/ELF/debug-gc.s<br>
test/ELF/gc-sections-merge.s<br>
test/ELF/merge-string-align.s<br>
test/ELF/merge.s<br>
<br>
</blockquote></div><br></div>