<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 12/23/2014 04:55 PM, Chandler
      Carruth wrote:<br>
    </div>
    <blockquote
cite="mid:CAGCO0KjWTJB2+rJYA3HDENSPpp3fCpo6GEvFMVbYnfZAUbFR8g@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div class="gmail_extra"><br>
          <div class="gmail_quote">On Tue, Dec 23, 2014 at 4:51 PM,
            Sanjoy Das <span dir="ltr"><<a moz-do-not-send="true"
                href="mailto:sanjoy@playingwithpointers.com"
                target="_blank">sanjoy@playingwithpointers.com</a>></span>
            wrote:<br>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex"><span
                class="">> Perhaps it would help to explain it
                slightly differently: There are<br>
                > optimizations which we can only perform by
                *removing* any no-wrap flags. But<br>
                > we don't know whether that optimization or the
                no-wrap flags is more<br>
                > important, and so we have to make a choice, and the
                choice is to preserve<br>
                > the flags at the expense of the transformation.
                Thus, adding flags can<br>
                <br>
              </span>Is this choice always in favor of retaining the
              flags?  Or is it a<br>
              per-transform decision?</blockquote>
          </div>
          <br>
          per-transform, but at least recently I've been seeing a trend
          towards preserving flags.</div>
      </div>
      <br>
    </blockquote>
    Chandler, thanks for reframing this.   This suddenly makes a lot
    more sense to me.  :)<br>
    <br>
    This is starting to seem like an canonicalization issue.   If we
    canonicalized to the form with flags, and deferred flag stripping
    until later in the optimization pipeline, would that get us a better
    practical result?  <br>
    <br>
    Philip
  </body>
</html>