<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <p><br>
    </p>
    <div class="moz-cite-prefix">On 10/03/2017 05:26 AM, Ristow, Warren
      wrote:<br>
    </div>
    <blockquote
cite="mid:1B965C6BCD3780449482E91EE11D7638C4716E5B@USCULXMSG05.am.sony.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
      <meta name="Generator" content="Microsoft Word 12 (filtered
        medium)">
      <style><!--
/* Font Definitions */
@font-face
        {font-family:"Cambria Math";
        panose-1:2 4 5 3 5 4 6 3 2 4;}
@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;}
@font-face
        {font-family:Consolas;
        panose-1:2 11 6 9 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
        {margin:0in;
        margin-bottom:.0001pt;
        font-size:12.0pt;
        font-family:"Times New Roman","serif";
        color:black;}
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;}
p
        {mso-style-priority:99;
        mso-margin-top-alt:auto;
        margin-right:0in;
        mso-margin-bottom-alt:auto;
        margin-left:0in;
        font-size:12.0pt;
        font-family:"Times New Roman","serif";
        color:black;}
pre
        {mso-style-priority:99;
        mso-style-link:"HTML Preformatted Char";
        margin:0in;
        margin-bottom:.0001pt;
        font-size:10.0pt;
        font-family:"Courier New";
        color:black;}
span.HTMLPreformattedChar
        {mso-style-name:"HTML Preformatted Char";
        mso-style-priority:99;
        mso-style-link:"HTML Preformatted";
        font-family:Consolas;
        color:black;}
span.EmailStyle20
        {mso-style-type:personal-reply;
        font-family:"Calibri","sans-serif";
        color:#44546A;}
.MsoChpDefault
        {mso-style-type:export-only;
        font-size:10.0pt;}
@page WordSection1
        {size:8.5in 11.0in;
        margin:1.0in 1.0in 1.0in 1.0in;}
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">
        <div>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A">>
              I agree that using SubclassOptionalData is going to be
              problematic when we run out of bits. ...<o:p></o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p> </o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A">I
              don't have a good view of the big-picture here (in terms
              of the cost of size and speed of the metadata approach, vs
              a member of Instruction, vs something else).<o:p></o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p> </o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A">We
              could tackle this problem now, or defer it hoping that
              we're not going to want to add more flags for finer
              granularity control of fast-math transformations in the
              future.  It seems the general sense is that we should
              tackle it now.<o:p></o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A">Sanjay
              and Hal, is that what you're view is?</span></p>
        </div>
      </div>
    </blockquote>
    <br>
    I think that it might be a good idea, but I don't have an opinion on
    ordering. They should be separate patches if possible (and, if it is
    not possible because we've run out of bits, then obviously the
    decision has been made for us).<br>
    <br>
     -Hal<br>
    <br>
    <blockquote
cite="mid:1B965C6BCD3780449482E91EE11D7638C4716E5B@USCULXMSG05.am.sony.com"
      type="cite">
      <div class="WordSection1">
        <div>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p></o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p> </o:p></span></p>
          <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A">-Warren</span><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p></o:p></span></p>
        </div>
        <p class="MsoNormal"><span
style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#44546A"><o:p> </o:p></span></p>
        <div>
          <div style="border:none;border-top:solid #B5C4DF
            1.0pt;padding:3.0pt 0in 0in 0in">
            <p class="MsoNormal"><b><span
style="font-size:10.0pt;font-family:"Tahoma","sans-serif";color:windowtext">From:</span></b><span
style="font-size:10.0pt;font-family:"Tahoma","sans-serif";color:windowtext">
                Hal Finkel [<a class="moz-txt-link-freetext" href="mailto:hfinkel@anl.gov">mailto:hfinkel@anl.gov</a>]
                <br>
                <b>Sent:</b> Tuesday, October 3, 2017 1:49 AM<br>
                <b>To:</b> Sanjay Patel; Ristow, Warren<br>
                <b>Cc:</b> <a class="moz-txt-link-abbreviated" href="mailto:llvm-dev@lists.llvm.org">llvm-dev@lists.llvm.org</a><br>
                <b>Subject:</b> Re: [llvm-dev] Trouble when suppressing
                a portion of fast-math-transformations<o:p></o:p></span></p>
          </div>
        </div>
        <p class="MsoNormal"><o:p> </o:p></p>
        <p><o:p> </o:p></p>
        <div>
          <p class="MsoNormal">On 10/01/2017 06:05 PM, Sanjay Patel
            wrote:<o:p></o:p></p>
        </div>
        <blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
          <div>
            <div>
              <p class="MsoNormal">Are we confident that we just need
                those 7 bits to represent all of the relaxed FP states
                that we need/want to support?
                <o:p></o:p></p>
            </div>
            <div>
              <p class="MsoNormal"><o:p> </o:p></p>
            </div>
            <div>
              <p class="MsoNormal" style="margin-bottom:12.0pt">I'm
                asking because FMF in IR is currently mapped onto the
                SubclassOptionalData of Value...and we have exactly 7
                bits there. :)<o:p></o:p></p>
            </div>
            <p class="MsoNormal">If we're redoing the definitions, I'm
              wondering if we can share the struct with the backend's
              SDNodeFlags, but that already has one extra bit for vector
              reduction. Should we give up on SubclassOptionalData for
              FMF? We have a MD_fpmath enum value for metadata, so we
              could move things over there?<o:p></o:p></p>
          </div>
        </blockquote>
        <p class="MsoNormal"><br>
          I agree that using SubclassOptionalData is going to be
          problematic when we run out of bits. As I recall, the reason
          that we didn't use metadata in the first place was because
          metadata is (generically) expensive. This case is very much
          like the case of debug info: in some modes, we add the
          debugging metadata to nearly every instruction. We use
          metadata for debug locations, syntactically, but we actually
          have a DebugLoc in each instruction that's used for the
          underlying representation. Here we'd have a similar problem:
          in some modes, we'd add this metadata to a large subset of all
          instructions. That could measurably slow down the optimizer.
          We may need to find some other place to put the data (e.g., an
          actual member variable of Instruction or more tail-allocated
          data in places)<br>
          <br>
           -Hal<br>
          <br>
          <br>
          <o:p></o:p></p>
        <div>
          <div>
            <p class="MsoNormal"><o:p> </o:p></p>
          </div>
        </div>
        <div>
          <p class="MsoNormal"><o:p> </o:p></p>
          <div>
            <p class="MsoNormal">On Fri, Sep 29, 2017 at 8:16 PM,
              Ristow, Warren via llvm-dev <<a moz-do-not-send="true"
                href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>>
              wrote:<o:p></o:p></p>
            <p class="MsoNormal">Hi Hal,<br>
              <br>
              >> 4. To fix this, I think that additional
              fast-math-flags are likely<br>
              >> needed in the IR.  Instead of the following set:<br>
              >><br>
              >> 'nnan' + 'ninf' + 'nsz' + 'arcp' + 'contract'<br>
              >><br>
              >> something like this:<br>
              >><br>
              >> 'reassoc' + 'libm' + 'nnan' + 'ninf' + 'nsz' +
              'arcp' + 'contract'<br>
              >><br>
              >> would be more useful.  Related to this, the
              current 'fast' flag which acts<br>
              >> as an umbrella (enabling 'nnan' + 'ninf' + 'nsz'
              + 'arcp' + 'contract') may<br>
              >> not be needed.  A discussion on this point was
              raised last November on the<br>
              >> mailing list:<br>
              >><br>
              >> <a moz-do-not-send="true"
href="http://lists.llvm.org/pipermail/llvm-dev/2016-November/107104.html"
                target="_blank">
http://lists.llvm.org/pipermail/llvm-dev/2016-November/107104.html</a><br>
              ><br>
              > I agree. I'm happy to help review the patches. It
              will be best to have<br>
              > only the finer-grained flags where there's no "fast"
              flag that implies<br>
              > all of the others.<br>
              <br>
              Thanks for the quick response, and for the willingness to
              review.  I won't let<br>
              this languish so long, like the post from last November.<br>
              <br>
              Happy to hear that you feel it's best not to have the
              umbrella "fast" flag.<br>
              <br>
              Thanks again,<o:p></o:p></p>
            <div>
              <div>
                <p class="MsoNormal">-Warren<br>
                  _______________________________________________<br>
                  LLVM Developers mailing list<br>
                  <a moz-do-not-send="true"
                    href="mailto:llvm-dev@lists.llvm.org">llvm-dev@lists.llvm.org</a><br>
                  <a moz-do-not-send="true"
                    href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev"
                    target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><o:p></o:p></p>
              </div>
            </div>
          </div>
          <p class="MsoNormal"><o:p> </o:p></p>
        </div>
        <p class="MsoNormal"><br>
          <br>
          <o:p></o:p></p>
        <pre>-- <o:p></o:p></pre>
        <pre>Hal Finkel<o:p></o:p></pre>
        <pre>Lead, Compiler Technology and Programming Languages<o:p></o:p></pre>
        <pre>Leadership Computing Facility<o:p></o:p></pre>
        <pre>Argonne National Laboratory<o:p></o:p></pre>
      </div>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory</pre>
  </body>
</html>