<html><head><meta http-equiv="Content-Type" content="text/html charset=windows-1252"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;">In lto+pgo some (5 out of 12 with usual suspect like perlbench and gcc among them using -flto -Wl,-mllvm,-use-cfl-aa -Wl,-mllvm,-use-cfl-aa-in-codegen) the CINT2006 benchmarks don’t compile. Has the implementation been tested with lto? If not, please stress the implementation more.<div>Do we know reasons for gains? Where did you expect the biggest gains?</div><div>Some of the losses will likely boil down to increased register pressure.</div><div><br></div><div>Looks like the current performance numbers pose a good challenge for gaining new and refreshing insights into our heuristics (and for smoothing out the implementation along the way).</div><div><br></div><div>Cheers</div><div>Gerolf</div><div><br><div><div><br></div><div><br><div><div><div>On Sep 12, 2014, at 1:27 AM, Jiangning Liu <<a href="mailto:liujiangning1@gmail.com">liujiangning1@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div dir="ltr">Hi Hal,<br><br>I run on SPEC2000 on cortex-a57(AArch64), and got the following results,<br><br>(It is to measure run-time reduction, and negative is better performance)<br><br>spec.cpu2000.ref.183_equake 33.77%<br>spec.cpu2000.ref.179_art 13.44%<br>spec.cpu2000.ref.256_bzip2 7.80%<br>spec.cpu2000.ref.186_crafty 3.69%<br>spec.cpu2000.ref.175_vpr 2.96%<br>spec.cpu2000.ref.176_gcc 1.77%<br>spec.cpu2000.ref.252_eon 1.77%<br>spec.cpu2000.ref.254_gap 1.19%<br>spec.cpu2000.ref.197_parser 1.15%<br>spec.cpu2000.ref.253_perlbmk 1.11%<br>spec.cpu2000.ref.300_twolf -1.04%<br><br>So we can see almost all got worse performance.<br><br>The command line option I'm using is "-O3 -std=gnu89 -ffast-math -fslp-vectorize -fvectorize -mcpu=cortex-a57 -mllvm -use-cfl-aa -mllvm -use-cfl-aa-in-codegen"<br><br>I didn't try compile-time, and I think your test on POWER7 native build should already meant something for other hosts. Also I don't have a good benchmark suit for compile time testing. My past experiences showed both llvm-test-suite (single/multiple) and spec benchmark are not good benchmarks for compile time testing.<br><br>Thanks,<br>-Jiangning <div class="gmail_extra"><br><div class="gmail_quote">2014-09-04 1:11 GMT+08:00 Hal Finkel <span dir="ltr"><<a href="mailto:hfinkel@anl.gov" target="_blank">hfinkel@anl.gov</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hello everyone,<br>
<br>
One of Google's summer interns, George Burgess IV, created an implementation of the CFL pointer-aliasing analysis algorithm, and this has now been added to LLVM trunk. Now we should determine whether it is worthwhile adding this to the default optimization pipeline. For ease of testing, I've added the command line option -use-cfl-aa which will cause the CFL analysis to be added to the optimization pipeline. This can be used with the opt program, and also via Clang by passing: -mllvm -use-cfl-aa.<br>
<br>
For the purpose of testing with those targets that make use of aliasing analysis during code generation, there is also a corresponding -use-cfl-aa-in-codegen option.<br>
<br>
Running the test suite on one of our IBM POWER7 systems (comparing -O3 -mcpu=native to -O3 -mcpu=native -mllvm -use-cfl-aa -mllvm -use-cfl-aa-in-codegen [testing without use in code generation were essentially the same]), I see no significant compile-time changes, and the following performance results:<br>
speedup:<br>
MultiSource/Benchmarks/mafft/pairlocalalign: -11.5862% +/- 5.9257%<br>
<br>
slowdown:<br>
MultiSource/Benchmarks/FreeBench/neural/neural: 158.679% +/- 22.3212%<br>
MultiSource/Benchmarks/MiBench/consumer-typeset/consumer-typeset: 0.627176% +/- 0.290698%<br>
MultiSource/Benchmarks/Ptrdist/ks/ks: 57.5457% +/- 21.8869%<br>
<br>
I ran the test suite 20 times in each configuration, using make -j48 each time, so I'll only pick up large changes. I've not yet investigated the cause of the slowdowns (or the speedup), and I really need people to try this on x86, ARM, etc. I appears, however, the better aliasing analysis results might have some negative unintended consequences, and we'll need to look at those closely.<br>
<br>
Please let me know how this fares on your systems!<br>
<br>
Thanks again,<br>
Hal<br>
<span class="HOEnZb"><font color="#888888"><br>
--<br>
Hal Finkel<br>
Assistant Computational Scientist<br>
Leadership Computing Facility<br>
Argonne National Laboratory<br>
_______________________________________________<br>
LLVM Developers mailing list<br>
<a href="mailto:LLVMdev@cs.uiuc.edu">LLVMdev@cs.uiuc.edu</a> <a href="http://llvm.cs.uiuc.edu/" target="_blank">http://llvm.cs.uiuc.edu</a><br>
<a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev</a><br>
</font></span></blockquote></div><br></div></div>
_______________________________________________<br>LLVM Developers mailing list<br><a href="mailto:LLVMdev@cs.uiuc.edu">LLVMdev@cs.uiuc.edu</a> <a href="http://llvm.cs.uiuc.edu">http://llvm.cs.uiuc.edu</a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev</a><br></blockquote></div><br></div></div></div></div></body></html>