<html>
  <head>
    <meta content="text/html; charset=windows-1252"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <br>
    <div class="moz-cite-prefix">On 11/20/2014 12:07 PM, Mark Heffernan
      wrote:<br>
    </div>
    <blockquote
cite="mid:CAOi2v+RoUAvUvATS9EnVsQOQq3MQqtkOX0RqYx=GWXg7wwb7aA@mail.gmail.com"
      type="cite">
      <div dir="ltr">Rolling back this CL which just improves analysis
        precision (trip count computation specifically) is probably not
        the right approach.</div>
    </blockquote>
    Strongly, strongly agreed with this point.  <br>
    <blockquote
cite="mid:CAOi2v+RoUAvUvATS9EnVsQOQq3MQqtkOX0RqYx=GWXg7wwb7aA@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div><br>
          </div>
          <div>Mark</div>
        </div>
      </div>
      <div class="gmail_extra"><br>
        <div class="gmail_quote">On Thu, Nov 20, 2014 at 11:13 AM, Hans
          Wennborg <span dir="ltr"><<a moz-do-not-send="true"
              href="mailto:hans@chromium.org" target="_blank">hans@chromium.org</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">Sorry for
            reviving an old thread.<br>
            <span class=""><br>
              On Fri, Oct 10, 2014 at 10:39 AM, Mark Heffernan <<a
                moz-do-not-send="true" href="mailto:meheff@google.com">meheff@google.com</a>>
              wrote:<br>
              > Author: meheff<br>
              > Date: Fri Oct 10 12:39:11 2014<br>
              > New Revision: 219517<br>
              ><br>
              > URL: <a moz-do-not-send="true"
                href="http://llvm.org/viewvc/llvm-project?rev=219517&view=rev"
                target="_blank">http://llvm.org/viewvc/llvm-project?rev=219517&view=rev</a><br>
              > Log:<br>
              > This patch de-pessimizes the calculation of loop trip
              counts in<br>
              > ScalarEvolution in the presence of multiple exits.
              Previously all<br>
              > loops exits had to have identical counts for a loop
              trip count to be<br>
              > considered computable. This pessimization was
              implemented by calling<br>
              > getBackedgeTakenCount(L) rather than getExitCount(L,
              ExitingBlock)<br>
              > inside of
              ScalarEvolution::getSmallConstantTripCount() (see the
              FIXME<br>
              > in the comments of that function). The pessimization
              was added to fix<br>
              > a corner case involving undefined behavior
              (pr/16130). This patch more<br>
              > precisely handles the undefined behavior case
              allowing the pessimization<br>
              > to be removed.<br>
              <br>
            </span>This caused a 50 k regression in Chromium's binary
            size due to more<br>
            unrolling. I suspect it's something we'll have to take, but
            was<br>
            wondering if you have any numbers showing performance
            benefit here?<br>
            <br>
            Thanks,<br>
            Hans<br>
          </blockquote>
        </div>
        <br>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
llvm-commits mailing list
<a class="moz-txt-link-abbreviated" href="mailto:llvm-commits@cs.uiuc.edu">llvm-commits@cs.uiuc.edu</a>
<a class="moz-txt-link-freetext" href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>