<div dir="ltr">Thanks. For future reference, do these benchmarking tests live some place where I can run them myself in the future?</div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Jun 30, 2016 at 2:15 AM, Eric Fiselier <span dir="ltr"><<a href="mailto:eric@efcs.ca" target="_blank">eric@efcs.ca</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">EricWF accepted this revision.<br>
EricWF added a comment.<br>
This revision is now accepted and ready to land.<br>
<br>
LGTM. I benchmarked the change against different key types and:<br>
<br>
1. The change doesn't have a large detrimental impact when the key equality is as expensive as hash equality. I benchmarked std::unordered_set<int>.find(...) at 27ns and 29ns before and after the change for a load factor >= 3.5, and 15ns vs 17 ns when the load factor is less than one.<br>
<br>
2. The change has a large positive impact when the load factor is > 1 and where key equality is more expensive than hash equality. For strings of size 1024 that only differed in the last characters I noticed a change of 880ns to 650ns. for a load factor >= 3.5.<br>
<br>
3. This change has a slight positive inpact when the load factor is < 1. For the same string inputs (mentioned above) I saw timings of 661ns and 623ns before and after.<br>
<div class="HOEnZb"><div class="h5"><br>
<br>
<br>
<br>
Repository:<br>
rL LLVM<br>
<br>
<a href="http://reviews.llvm.org/D21510" rel="noreferrer" target="_blank">http://reviews.llvm.org/D21510</a><br>
<br>
<br>
<br>
</div></div></blockquote></div><br></div>