<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Hello, sorry to revisit this old thread but progress seems to
      have stalled on my review again. It's been nearly six months since
      I submitted it and I didn't really think it would be such a hard
      change to get through. Is there a point when we could consider the
      original reviewer no longer to be interested and ask someone else
      to step in?</p>
    <p>What I've done is add a new (static) function addExtension to
      TargetPassConfig to give out of tree code a way to manipulate
      MachineFunctions. I did get some good feedback and made some code
      changes but I haven't been getting any responses for a while now
      and it's hard for me to know what to do at this point. Would
      someone else be able to take a look, please?</p>
    <p>The review is at <a class="moz-txt-link-freetext" href="https://reviews.llvm.org/D98591">https://reviews.llvm.org/D98591</a> and would
      probably need to be rebased again if we want to go ahead with it.<br>
    </p>
    <p>Regards,<br>
      Raoul.<br>
    </p>
    <div class="moz-cite-prefix"><br>
    </div>
    <div class="moz-cite-prefix">On 17/06/2021 08:52, Raoul Gough wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:0b076dc7-2797-58ae-d5ec-0371c800a059@gmail.com">
      <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
      <div class="moz-cite-prefix">Thanks for looking at this. Just so
        you're aware I see that @ychen has just posted some further
        comments on the review which I'll be addressing shortly.</div>
      <div class="moz-cite-prefix"><br>
      </div>
      <div class="moz-cite-prefix">The TargetPassConfig::insertPass
        (after) function would work in theory except that the
        TargetPassConfig objects only have transitory lifetimes within,
        e.g., LLVMTargetMachine::addPassesToEmitFile. I couldn't find
        any existing way to access the TargetPassConfig objects from
        out-of-tree code in order to manipulate them. The solution with
        TargetPassConfig::<span class="nf">addExtension</span><span
          class="p"> is to register callback(s) in a global which get
          passed the TargetPassConfig object during execution of
          TargetPassConfig::addMachinePasses.</span></div>
      <div class="moz-cite-prefix"><span class="p"><br>
        </span></div>
      <div class="moz-cite-prefix"><br>
      </div>
      <div class="moz-cite-prefix">On 17/06/2021 00:31, Craig Topper
        wrote:<br>
      </div>
      <blockquote type="cite"
cite="mid:CAF7ks-OR+PWuhZOiHfmPXNpGspeoyG11eZALYiCJt__TuPEbRQ@mail.gmail.com">
        <meta http-equiv="content-type" content="text/html;
          charset=UTF-8">
        <div dir="ltr">I just got back from vacation so I haven't had
          much time to look at your patch yet. My first question is
          going to be, is this something that can already be done with
          the existing TargetPassConfig::insertPassAfter mechanism? I
          don't know the history of that mechanism just that it exists.
          <div><br>
          </div>
          <div>
            <div>
              <div dir="ltr" class="gmail_signature"
                data-smartmail="gmail_signature">~Craig</div>
            </div>
            <br>
          </div>
        </div>
        <br>
        <div class="gmail_quote">
          <div dir="ltr" class="gmail_attr">On Wed, Jun 16, 2021 at
            12:59 PM Raoul Gough <<a
              href="mailto:github.drti@gmail.com" moz-do-not-send="true">github.drti@gmail.com</a>>
            wrote:<br>
          </div>
          <blockquote class="gmail_quote" style="margin:0px 0px 0px
            0.8ex;border-left:1px solid
            rgb(204,204,204);padding-left:1ex">
            <div>
              <div>Hi Craig, I see you listed as the code owner for the
                x86 backend and I was wondering if you'd have time to
                comment on my review at <a
                  href="https://reviews.llvm.org/D98591" target="_blank"
                  moz-do-not-send="true">https://reviews.llvm.org/D98591</a>
                ? This adds an extension mechanism to <span>TargetPassConfig
                  to allow out-of-tree code to provide machine passes
                  for an existing target.<br>
                </span></div>
              <div><br>
              </div>
              <div>Arthur has raised the question of whether it's
                appropriate to have such an extension mechanism so it
                might help if I explain my motivating example. I'm
                working on a way to do runtime inlining of functions
                referenced via pointers, including C++ virtual function
                calls. This works using ahead-of-time compiled code that
                has been annotated to derive call-tree information at
                runtime and potentially recompile parts of the code
                using bitcode embedded as data in the executable. To
                build the call tree I install a context pointer in a
                callee-saved register at the relevant call sites and
                retrieve it at function entry, both of which I do via a
                machine pass that runs just before the register
                allocator. There are actually several other
                complications but that's the starting point for wanting
                to inject a custom machine pass for an existing target.
                What do you think?</div>
              <div><br>
              </div>
              <div>There are some more details on my github page <a
                  href="https://github.com/drti/drti" target="_blank"
                  moz-do-not-send="true">https://github.com/drti/drti</a>
                if you're interested.<br>
              </div>
              <div><br>
              </div>
              <div>Regards,</div>
              <div>Raoul.<br>
              </div>
              <div><br>
              </div>
              <div>On 15/06/2021 23:22, Arthur Eubanks wrote:<br>
              </div>
              <blockquote type="cite">
                <div dir="ltr">I'm not super familiar with the codegen
                  pipeline, perhaps some backend people can take a look?
                  <div><br>
                  </div>
                  <div>A question is, do we want extensions like in the
                    middle-end optimization pipeline? I don't see why
                    not, this mirrors the existing extensions for the
                    optimization pipeline. But it'd be good to get some
                    more opinions from people more knowledgeable.</div>
                </div>
                <br>
                <div class="gmail_quote">
                  <div dir="ltr" class="gmail_attr">On Mon, Jun 14, 2021
                    at 8:04 AM David Blaikie <<a
                      href="mailto:dblaikie@gmail.com" target="_blank"
                      moz-do-not-send="true">dblaikie@gmail.com</a>>
                    wrote:<br>
                  </div>
                  <blockquote class="gmail_quote" style="margin:0px 0px
                    0px 0.8ex;border-left:1px solid
                    rgb(204,204,204);padding-left:1ex">
                    <div dir="ltr">Arthur - any idea who might be good
                      to review this?</div>
                    <br>
                    <div class="gmail_quote">
                      <div dir="ltr" class="gmail_attr">On Mon, Jun 7,
                        2021 at 6:20 AM Raoul Gough via llvm-dev <<a
                          href="mailto:llvm-dev@lists.llvm.org"
                          target="_blank" moz-do-not-send="true">llvm-dev@lists.llvm.org</a>>
                        wrote:<br>
                      </div>
                      <blockquote class="gmail_quote" style="margin:0px
                        0px 0px 0.8ex;border-left:1px solid
                        rgb(204,204,204);padding-left:1ex">Hello
                        llvm-dev, I have a code contribution for the
                        legacy pass manager <br>
                        and I'm having trouble finding anyone to finish
                        reviewing it. I've made <br>
                        some improvements based on earlier feedback from
                        @ychen on Phabricator <br>
                        but since then nobody has given final approval.
                        Can anyone help with <br>
                        this or point me in the right direction?<br>
                        <br>
                        I couldn't see anyone listed in the
                        CODE_OWNERS.txt for this area of the <br>
                        code base. From what I understand the new pass
                        manager still doesn't <br>
                        handle the target-specific code, so I think the
                        code is in the right <br>
                        place for now at least. I'd be happy to move it
                        to the new pass manager <br>
                        if that's currently possible, of course.<br>
                        <br>
                        This is related to my (out of tree) runtime
                        inlining project and adds <br>
                        minimal support for target-level extensions in a
                        similar way to <br>
                        addGlobalExtension in the legacy IR
                        PassManagerBuilder... <br>
                        <a href="https://reviews.llvm.org/D98591"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://reviews.llvm.org/D98591</a><br>
                        <br>
                        Regards,<br>
                        Raoul Gough.<br>
                        <br>
                        _______________________________________________<br>
                        LLVM Developers mailing list<br>
                        <a href="mailto:llvm-dev@lists.llvm.org"
                          target="_blank" moz-do-not-send="true">llvm-dev@lists.llvm.org</a><br>
                        <a
                          href="https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev"
                          rel="noreferrer" target="_blank"
                          moz-do-not-send="true">https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><br>
                      </blockquote>
                    </div>
                  </blockquote>
                </div>
              </blockquote>
              <p><br>
              </p>
            </div>
          </blockquote>
        </div>
      </blockquote>
      <p><br>
      </p>
    </blockquote>
    <p><br>
    </p>
  </body>
</html>