Looks good! <div>I'll do the actual measurements to see if we missed something after the commit</div><div><br></div><div>--kcc <br><br><div class="gmail_quote">On Mon, Jul 16, 2012 at 1:10 PM, Chandler Carruth <span dir="ltr"><<a href="mailto:chandlerc@gmail.com" target="_blank">chandlerc@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">The patch helps....<div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Jul 16, 2012 at 2:10 AM, Chandler Carruth <span dir="ltr"><<a href="mailto:chandlerc@gmail.com" target="_blank">chandlerc@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">I have no idea if this heuristic is correct, but I suspect it's pretty good...<div><br></div><div>It doesn't really matter what heuristic is used if we insert 'unreachable' -- that should trump all else, but it seemed simpler to just always set a flat heavy bias toward the "nothing wrong" direction. Looking at the asm output, the metadata has the desired effect.</div>
</blockquote></div><br></div>
</div></div></blockquote></div><br></div>