<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Mar 30, 2016 at 6:17 PM, Rui Ueyama <span dir="ltr"><<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><span class="">On Wed, Mar 30, 2016 at 5:34 PM, Sean Silva <span dir="ltr"><<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span>On Wed, Mar 30, 2016 at 4:25 PM, Rui Ueyama <span dir="ltr"><<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><span><div>On Wed, Mar 30, 2016 at 4:20 PM, Sean Silva <span dir="ltr"><<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>></span> wrote:<br></div></span><div class="gmail_extra"><div class="gmail_quote"><span><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">I believe the relocation stuff that Rafael is currently working on will make this a non-issue (it will make relocation application much friendlier for the CPU).</div></blockquote><div><br></div></span><div>I don't think Rafael's patch would make this a non-issue. He's making scanRelocs to create data, which would reduce the number of calls to the virtual functions, but it wouldn't be reduced to zero.</div><span><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div>However, even in the current scheme, since the target is fixed, all the indirect call sites should be monomorphic and so there shouldn't be much branch-prediction cost (certainly nothing that would cause 1.8% performance delta for the entire link).</div></div></blockquote><div><br></div></span><div>Agreed. We could template functions that call TargetInfo's member functions for each target to eliminate the virtual function calls.</div></div></div></div></blockquote><div><br></div></span><div>From what has been presented I would not conclude that virtual calls are actually the problem (or a problem at all). A root-cause analysis is necessary. As <span style="font-size:12.8px">r263227 shows, the relocation application loop is very sensitive to small changes.</span></div><div><span style="font-size:12.8px"><br></span></div><div><span style="font-size:12.8px">One quick thing you may also want to try as a sanity check is inserting nops in different places in the function. I suspect you'll find that the performance swings (both speedups and slowdowns) from doing that are similar in magnitude to what you are seeing. You may also want to try editing the indirect call instruction to a direct call without otherwise modifying the binary; if that reproduces the 1.8% speedup then it will be convincing.</span></div></div></div></div></blockquote><div><br></div></span><div>Honestly I was somewhat skeptical about what you wrote here, but I observed 0.4% *slowdown* when I used gcc to compile it, so looks like I was wrong. It is possible that devirtualization might have been effective for clang-generated code, but it is more likely that that was a result of some performance deviation caused by some other factor.</div><div><br></div><div>The relocation handling loop is really a tight loop and therefore sensitive to small changes. How can we optimize this? Maybe PGO?</div></div></div></div></blockquote><div><br></div><div>Rafael's change will fix it. That is why he is doing it in the first place :)</div><div>(The idea came when we were crunching the numbers for "LLD performance w.r.t. local symbols (and --build-id)" and looked at r263227. I suggested that this looked like it was because this loop is getting long enough to run-out the CPU's reorder buffer waiting on the cache misses, preventing it from seeing the memory accesses of the next iteration and thus failing to pipeline the memory accesses across iterations. Small changes in scheduling, instruction count, etc. will tickle this and cause large performance changes. The solution is to make the relocation application loop tighter. Especially by separating some of the stuff that we currently have inside one huge loop to be in separate loops, but also by making the loops tighter.)</div><div><br></div><div>-- Sean Silva</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><span class=""><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><div><span style="font-size:12.8px">If you haven't read it, I think you would enjoy this paper: <a href="http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/37077.pdf" target="_blank">http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/37077.pdf</a></span></div><span><font color="#888888"><div><br></div><div>-- Sean Silva</div></font></span><span><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><span><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr"><div>Notice that 1.8% is smaller than the performance variation from <span style="font-size:12.8px">r263227 which is a very innocuous-looking change but caused </span><span style="font-size:12.8px">~2-3% slowdown for ScyllaDB (see the thread "LLD performance w.r.t. local symbols (and --build-id)").</span><br></div><div><div><br></div><div>-- Sean Silva</div></div></div><div class="gmail_extra"><br><div class="gmail_quote"><div><div>On Wed, Mar 30, 2016 at 3:39 PM, Rui Ueyama via llvm-dev <span dir="ltr"><<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>></span> wrote:<br></div></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div><div><div dir="ltr">I was wandering how much is the overhead of virtual function calls of TargetInfo member functions. TargetInfo handles platform-specific details, and we have target-specific subclasses of that class. The subclasses override functions defined in TargetInfo.<div><br><div>The TargetInfo member functions are called multiple times for each relocation. So the cost of virtual function calls may be non-neglible. That is a motication to do the following test.<div><br></div><div>As a test, I removed all TargetInfo subclasses except for x86-64, move all code from X86_64TargetInfo to TargetInfo, and remove `virtual` from TargetInfo.</div><div><br></div><div>The original LLD links itself (with debug info) in 7.499 seconds. The de-virtualized version did the same thing in 7.364 seconds. So it can improve it by 1.8%.</div><div><br></div><div>I'm just pointing out that there's room there to improve performance, and I'm not suggesting we do something for this right now. We probably shouldn't do anything for this because the current code is pretty straightforward. But I'd expect that we will eventually want do something for this in future.</div></div></div></div>
<br></div></div>_______________________________________________<br>
LLVM Developers mailing list<br>
<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><br>
<br></blockquote></div><br></div>
</blockquote></span></div><br></div></div>
</blockquote></span></div><br></div></div>
</blockquote></span></div><br></div></div>
</blockquote></div><br></div></div>