<div dir="ltr">No pointer analysis will show any benefits until it is on by default and tuning starts :)<br><br><div>or everything is tuned and then it's turned on, whichever way you want to do it.</div><div><br><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Mar 22, 2016 at 2:33 PM, Jia Chen <span dir="ltr"><<a href="mailto:jchen@cs.utexas.edu" target="_blank">jchen@cs.utexas.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    It's something that I am certainly interested in and qualified to
    do.<br>
    <br>
    However, the way I read Daniel's reply in this thread is: "LLVM, in
    its current form, is unlikely to benefit from a more precise aa". He
    did mention that cfl-aa is "more understandable and maintainable",
    and is "fast enough", but nothing is said about the benefits. There
    was some discussions I could find back in 2014, which basically says
    that cfl-aa offers no significant improvement in performance. Things
    may got greatly improved since then, yet it is not clear to me what
    the current situation is.<br>
    <br>
    With the benefits being unclear a GSoC proposal on this topic may
    not look motivated enough.<div><div class="h5"><br>
    <br>
    <div>On 03/22/2016 01:55 PM, Philip Reames
      wrote:<br>
    </div>
    <blockquote type="cite">
      
      It's found more and more like "get CFL-AA turned on by default"
      might be a viable GSoC project for the right student.  It would
      require someone with existing knowledge of AA and a willingness to
      debug nasty problems, but it sounds like there's definitely
      interest in the community in seeing this happen.<br>
      <br>
      If the student finished early (unlikely), they could start on
      SCEV-AA as well.  <br>
      <br>
      Philip<br>
      <br>
      <div>On 03/21/2016 01:10 PM, George
        Burgess IV via llvm-dev wrote:<br>
      </div>
      <blockquote type="cite">
        <div dir="ltr">
          <div><span style="color:rgb(0,0,0);font-family:arial,helvetica,sans-serif;font-size:13.3333px">As

              of late-August 2015, putting CFL-AA behind BasicAA caused
              miscompiles when trying to bootstrap Clang/LLVM, yeah. It
              didn't seem that there were many new errors (I think it
              caused ~10 tests to fail, where fail = either segv or
              produce the wrong output), but it did end up breaking
              things. I don't recall if standalone CFL-AA causes
              miscompiles, but I highly doubt the breakages I observed
              were BasicAA's fault.</span></div>
          <div><span style="color:rgb(0,0,0);font-family:arial,helvetica,sans-serif;font-size:13.3333px"><br>
            </span></div>
          <div><font color="#000000" face="arial, helvetica, sans-serif"><span style="font-size:13.3333px">WRT speed, `time make -j14`
                on my box (6c/12t) didn't show a meaningful increase in
                compile time when CFL-AA gets enabled (read: it got lost
                in the noise). So, I agree that it's probably fast
                enough at the moment; if we want to enhance it, we
                should focus on making it bootstrap clang+LLVM/making it
                more accurate.</span></font></div>
          <div class="gmail_extra"><br>
            <div class="gmail_quote">On Mon, Mar 21, 2016 at 12:26 PM,
              Hal Finkel <span dir="ltr"><<a href="mailto:hfinkel@anl.gov" target="_blank">hfinkel@anl.gov</a>></span>
              wrote:<br>
              <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
                <div>
                  <div style="font-family:arial,helvetica,sans-serif;font-size:10pt;color:#000000"><br>
                    <hr>
                    <blockquote style="border-left:2px solid rgb(16,16,255);margin-left:5px;padding-left:5px;color:rgb(0,0,0);font-weight:normal;font-style:normal;text-decoration:none;font-family:Helvetica,Arial,sans-serif;font-size:12pt"><b>From:

                      </b>"Daniel Berlin via llvm-dev" <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank"></a><a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>><br>
                      <b>To: </b>"Renato Golin" <<a href="mailto:renato.golin@linaro.org" target="_blank"></a><a href="mailto:renato.golin@linaro.org" target="_blank">renato.golin@linaro.org</a>>,

                      "George Burgess IV" <<a href="mailto:george.burgess.iv@gmail.com" target="_blank">george.burgess.iv@gmail.com</a>><br>
                      <b>Cc: </b>"llvm-dev" <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank"></a><a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>>,
                      "Jia Chen" <<a href="mailto:jchen@cs.utexas.edu" target="_blank">jchen@cs.utexas.edu</a>><br>
                      <b>Sent: </b>Monday, March 21, 2016 2:07:44 PM<br>
                      <b>Subject: </b>Re: [llvm-dev] Existing studies
                      on the benefits of pointer analysis<span><br>
                        <br>
                        <div dir="ltr"><br>
                          <div class="gmail_extra"><br>
                            <div class="gmail_quote">On Mon, Mar 21,
                              2016 at 12:05 PM, Renato Golin <span dir="ltr"><<a href="mailto:renato.golin@linaro.org" target="_blank">renato.golin@linaro.org</a>></span>
                              wrote:<br>
                              <blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><span>On

                                  21 March 2016 at 18:59, Daniel Berlin
                                  <<a href="mailto:dberlin@dberlin.org" target="_blank">dberlin@dberlin.org</a>>

                                  wrote:<br>
                                  > Which is why i've never mentioned
                                  it or used it in the community ;)<br>
                                  <br>
                                </span>Makes sense. :)<br>
                                <span><br>
                                  <br>
                                  > I would rather see someone spend
                                  their time getting SCEV-AA on by
                                  default or<br>
                                  > CFL-AA on by default than doing
                                  another evaluation.<br>
                                  <br>
                                </span>But those may not be simple
                                enough for a GSOC, that's why I
                                mentioned it.<br>
                                <br>
                              </blockquote>
                              <div><br>
                              </div>
                              <div>CFL-AA should just be fixing
                                performance regressions, and maybe a
                                little bug fixing, which is hopefully
                                easy enough.  It's already fast enough
                                as a pass.</div>
                            </div>
                          </div>
                        </div>
                      </span></blockquote>
                    <br>
                    My understanding from George is that there are
                    self-hosting miscompiles if you disable all AA
                    except for CFL-AA. This is what is preventing us
                    from enabling it by default. George, is that right?<br>
                    <br>
                     -Hal<br>
                    <blockquote style="border-left:2px solid rgb(16,16,255);margin-left:5px;padding-left:5px;color:rgb(0,0,0);font-weight:normal;font-style:normal;text-decoration:none;font-family:Helvetica,Arial,sans-serif;font-size:12pt"><span>
                        <div dir="ltr">
                          <div class="gmail_extra">
                            <div class="gmail_quote">
                              <div><br>
                              </div>
                              <div>SCEV-AA would be harder (must make
                                SCEV-AA faster).</div>
                              <div><br>
                              </div>
                              <blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> The
                                analysis could not only get us a birds
                                view of the problem ahead,<br>
                                but also introduce new developers to AA,
                                which would make their future<br>
                                work on SCEV-AA or CFL-AA easier. Kind
                                of a teaching tool to get more<br>
                                AA-savvy people.<br>
                              </blockquote>
                              <div><br>
                              </div>
                              <div>Sure.  </div>
                              <blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> <br>
                                cheers,<br>
                                --renato<br>
                              </blockquote>
                            </div>
                            <br>
                          </div>
                        </div>
                        <br>
                      </span><span>_______________________________________________<br>
                        LLVM Developers mailing list<br>
                        <a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
                        <a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><br>
                      </span></blockquote>
                    <span><font color="#888888"><br>
                        <br>
                        <br>
                        -- <br>
                        <div><span name="x"></span>Hal Finkel<br>
                          Assistant Computational Scientist<br>
                          Leadership Computing Facility<br>
                          Argonne National Laboratory<span name="x"></span><br>
                        </div>
                      </font></span></div>
                </div>
              </blockquote>
            </div>
            <br>
          </div>
        </div>
        <br>
        <fieldset></fieldset>
        <br>
        <pre>_______________________________________________
LLVM Developers mailing list
<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a>
</pre>
      </blockquote>
      <br>
    </blockquote></div></div><span class="">
    -- <br>
    Best Regards,<br>
    <br>
    --<br>
    Jia Chen<br>
  </span></div>

</blockquote></div><br></div>