<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <br>
    <br>
    <div class="moz-cite-prefix">On 04/18/2016 09:52 AM, Mehdi Amini via
      llvm-dev wrote:<br>
    </div>
    <blockquote
      cite="mid:67633203-CC62-49EE-BA7A-F66DE2E96CF3@apple.com"
      type="cite">
      <div>
        <blockquote type="cite" class="">
          <div class="">
            <div class="">
              <div class="" style="word-wrap:break-word">
                <div class=""><br class="">
                </div>
                <div class="">Initially I came across the problem with
                  my recent change which added an overloaded type to the
                  masked load/store intrinsics (<a
                    moz-do-not-send="true"
                    href="http://reviews.llvm.org/D17270" class=""><a class="moz-txt-link-freetext" href="http://reviews.llvm.org/D17270">http://reviews.llvm.org/D17270</a></a>).
                  The discrepancy between the name and the signature
                  triggers auto-upgrade bit from my patch converting an
                  incorrect mangling to the correct one. But later after
                  remapping of isomorphic types when we return to the
                  original type name this “updated" intrinsic name
                  become invalid.</div>
              </div>
            </div>
          </div>
        </blockquote>
        <div><br class="">
        </div>
        <div>In the same way, I'd try to avoid "autoupgrading" in this
          case? I'm puzzled by the fact that round-tripping to bitcode
          within the same version of LLVM triggers an "upgrade".</div>
      </div>
    </blockquote>
    This does seem questionable.  It's doesn't avoid the root issue -
    which is that we're not renaming intrinsics when renaming types -
    but changing this would seem reasonable to me.  It sounds like we
    might be relying on the upgrade functionality to paper over problems
    with type renaming.  (And possibly other things...?)<br>
    <br>
    Philip<br>
  </body>
</html>