<div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Jul 29, 2013 at 4:38 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><br><div class="gmail_quote">
<div class="im">On Mon, Jul 29, 2013 at 4:24 PM, Hal Finkel <span dir="ltr"><<a href="mailto:hfinkel@anl.gov" target="_blank">hfinkel@anl.gov</a>></span> wrote:<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">

Out of curiosity, has anyone tried to optimize the pass ordering in some (quasi-)automated way? Naively, a genetic algorithm seems like a perfect fit for this.<br></blockquote><div><br></div></div><div>This is the closest I've seen: <a href="http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/" target="_blank">http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/</a> </div>

<div><br></div><div>However, it deals with a "toy" example. Doing something similar over an entire benchmark suite would be interesting (and it may find non-obvious, highly-profitable interactions between passes that we aren't currently exploiting).</div>
<span class=""><font color="#888888">
<div><br></div></font></span></div></div></div></blockquote><div><br></div><div>One more I saw recently: <a href="http://gcc.gnu.org/wiki/cauldron2013?action=AttachFile&do=get&target=machine_guided_energy_energy_efficient_compilation.pdf">http://gcc.gnu.org/wiki/cauldron2013?action=AttachFile&do=get&target=machine_guided_energy_energy_efficient_compilation.pdf</a> </div>
<div>(it deals with LLVM and GCC, but focuses on "energy efficiency" and not raw performance).</div></div></div></div>