<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <br>
    <div class="moz-cite-prefix">On 11/05/2014 03:08 PM, Arnaud A. de
      Grandmaison wrote:<br>
    </div>
    <blockquote cite="mid:000f01cff94d$59df5400$0d9dfc00$@arm.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
      <meta name="Generator" content="Microsoft Word 14 (filtered
        medium)">
      <style><!--
/* Font Definitions */
@font-face
        {font-family:Calibri;
        panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
        {font-family:Tahoma;
        panose-1:2 11 6 4 3 5 4 4 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
        {margin:0cm;
        margin-bottom:.0001pt;
        font-size:12.0pt;
        font-family:"Times New Roman","serif";}
a:link, span.MsoHyperlink
        {mso-style-priority:99;
        color:blue;
        text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
        {mso-style-priority:99;
        color:purple;
        text-decoration:underline;}
span.EmailStyle17
        {mso-style-type:personal-reply;
        font-family:"Calibri","sans-serif";
        color:#1F497D;}
.MsoChpDefault
        {mso-style-type:export-only;
        font-family:"Calibri","sans-serif";
        mso-fareast-language:EN-US;}
@page WordSection1
        {size:612.0pt 792.0pt;
        margin:72.0pt 72.0pt 72.0pt 72.0pt;}
div.WordSection1
        {page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
      <div class="WordSection1">
        <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
        <p class="MsoNormal" style="margin-left:36.0pt"><b><span
style="font-size:10.0pt;font-family:"Tahoma","sans-serif""
              lang="EN-US">From:</span></b><span
style="font-size:10.0pt;font-family:"Tahoma","sans-serif""
            lang="EN-US"> Reid Kleckner [<a class="moz-txt-link-freetext" href="mailto:rnk@google.com">mailto:rnk@google.com</a>] <br>
            <b>Sent:</b> 05 November 2014 22:49<br>
            <b>To:</b> Philip Reames<br>
            <b>Cc:</b> Arnaud De Grandmaison; LLVM Developers Mailing
            List<br>
            <b>Subject:</b> Re: [LLVMdev] lifetime.start/end
            clarification<o:p></o:p></span></p>
        <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
        <div>
          <div>
            <div>
              <p class="MsoNormal" style="margin-left:36.0pt">On Wed,
                Nov 5, 2014 at 11:17 AM, Philip Reames <<a
                  moz-do-not-send="true"
                  href="mailto:listmail@philipreames.com"
                  target="_blank">listmail@philipreames.com</a>>
                wrote:<o:p></o:p></p>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
                <div>
                  <p class="MsoNormal" style="margin-left:36.0pt">On
                    11/05/2014 10:54 AM, Reid Kleckner wrote:<o:p></o:p></p>
                </div>
                <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
                  <div>
                    <div>
                      <div>
                        <blockquote style="border:none;border-left:solid
                          #CCCCCC 1.0pt;padding:0cm 0cm 0cm
                          6.0pt;margin-left:4.8pt;margin-right:0cm">
                          <div>
                            <p class="MsoNormal"
                              style="margin-left:36.0pt">This seems fine
                              to me.  The optimizer can (soundly)
                              conclude that %p is dead after the
                              "lifetime.end" (for the two instructions),
                              and dead before the "lifetime.start" (for
                              the *single* instruction in that basic
                              block, *not* for the previous BB).  This
                              seems like the proper result for this
                              example, am I missing something?<o:p></o:p></p>
                          </div>
                        </blockquote>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt"><o:p> </o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">What if I put
                            that in a loop, unroll it once, and prove
                            that the lifetime.start is unreachable? We
                            would end up with IR like:<o:p></o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt"><o:p> </o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">loop:<o:p></o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">  ... use %p<o:p></o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">  call void
                            @lifetime.end( %p )<o:p></o:p></p>
                        </div>
                        <div>
                          <div>
                            <p class="MsoNormal"
                              style="margin-left:36.0pt">  ... use %p<o:p></o:p></p>
                          </div>
                          <div>
                            <p class="MsoNormal"
                              style="margin-left:36.0pt">  call void
                              @lifetime.end( %p )<o:p></o:p></p>
                          </div>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">  br i1 %c, label
                            %loop, label %exit<o:p></o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt"><o:p> </o:p></p>
                        </div>
                        <div>
                          <p class="MsoNormal"
                            style="margin-left:36.0pt">Are the second
                            uses of %p uses of dead memory?<o:p></o:p></p>
                        </div>
                      </div>
                    </div>
                  </div>
                </blockquote>
                <p class="MsoNormal" style="margin-left:36.0pt">It's
                  hard to discuss this without being specific about the
                  starting IR and transforms.  My general response is
                  that either a) such a transform wouldn't be valid or
                  b) the behaviour of the original program was
                  undefined.  <o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">The
                  starting IR would be something that jumps into the
                  middle of a lifetime region, like the example that
                  Arnaud gave. This was assuming the current state of
                  the world where we haven't added a second lifetime
                  start call prior to the goto branch.<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">Start
                  with something like:<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">void
                  f(int x) {<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">  while
                  (x) {<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">    goto
                  skip_start;<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">    {<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">     
                  int y; // lifetime.start<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">skip_start:<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">      y
                  = g();<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">      x
                  -= y;<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">      //
                  lifetime.end<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">    }<br>
                    }<br>
                  }<o:p></o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt"><o:p> </o:p></p>
              </div>
              <div>
                <p class="MsoNormal" style="margin-left:36.0pt">The
                  block containing the lifetime start of y is
                  unreachable and can be deleted.<o:p></o:p></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">On
                    this example, I think the critical edge to the
                    skip_start label need to get a lifetime.start
                    inserted (btw, we already insert there a gep to the
                    alloca to solve a dominator issue), and this makes
                    sense because such an edge also contains (in
                    essence) the trivial constructors of the scope where
                    the label is.</span></p>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    Well said.  This is exactly the point I've been trying to make.  <br>
    <blockquote cite="mid:000f01cff94d$59df5400$0d9dfc00$@arm.com"
      type="cite">
      <div class="WordSection1">
        <div>
          <div>
            <div>
              <div>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">This
                    looks to me as a hint that when they are used with
                    an alloca, lifetime.start/end really have to be
                    paired, and that all possible uses of the alloca
                    should be covered by a [start,end] :<o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">
                    - In the above testcase, when the path from
                    lifetime.end is walked-up the cfg, it reaches the
                    alloca, and this is weird.<o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">
                    - In my testcase, when the paths from lifetime.start
                    and lifetime.end are walked-up the cfg, it becomes
                    inconsistent in the entry block: one path will tell
                    it’s dead, the other it’s alive.<o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">This
                    puts the burden onto clang to generate consistent
                    marker info, which sounds reasonable after all.</span></p>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    My analysis is slightly different, but the result is the same.  I
    don't believe that a dynamic trace of {end, use, end} should be well
    defined.  I'm suggesting we extend the current semantics to make the
    trace {end, start, use, end} meaningful.  It's the frontends job to
    ensure that no dynamic traces of the first type are possible if the
    second is meant.  <br>
    <blockquote cite="mid:000f01cff94d$59df5400$0d9dfc00$@arm.com"
      type="cite">
      <div class="WordSection1">
        <div>
          <div>
            <div>
              <div>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I
                    can probably add some checks to enforce this in
                    llvm, or at least detect such cases: this will
                    definitely help me to spot candidates for the
                    miscompilation I observe, which by the way is not
                    caused by unnamed temporaries, but by standard
                    VarDecl + lifetime markers. I do not know yet where
                    though. It means that we have a lurking bug today
                    in-tree, and changing the size threshold for marker
                    insertion is enough to trigger it.</span></p>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    I agree with this analysis.  <br>
    <blockquote cite="mid:000f01cff94d$59df5400$0d9dfc00$@arm.com"
      type="cite">
      <div class="WordSection1">
        <div>
          <div>
            <div>
              <div>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,<o:p></o:p></span></p>
                <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud<o:p></o:p></span></p>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>