<div dir="ltr">In particular:<br>"<div>I'm not so sure that it's sufficient. Suppose, for instance, that I wanted to insert a MemoryDef between 1 and 2 in the below AccessList:<div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"> 0 = MemoryDef(liveOnEntry)<br class="gmail-m_-868984512930983029gmail_msg"> 1 = MemoryDef(0)<br class="gmail-m_-868984512930983029gmail_msg"></span></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"> 2 = MemoryDef(1)<br class="gmail-m_-868984512930983029gmail_msg"> 3 = MemoryUse(1)<br class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div><div class="gmail-m_-868984512930983029gmail_msg">Invoking createMemoryAccess followed by RAUW of 1's uses with 4 would result in:</div><div class="gmail-m_-868984512930983029gmail_msg">....</div><div class="gmail-m_-868984512930983029gmail_msg">"</div><div class="gmail-m_-868984512930983029gmail_msg"><br></div><div class="gmail-m_-868984512930983029gmail_msg">Which is why this is not the way one does this, anywhere.</div><div class="gmail-m_-868984512930983029gmail_msg">If you wish to *replace* an existing access, the normal way is:<br><div class="gmail-m_-868984512930983029gmail_msg"><br></div><div class="gmail-m_-868984512930983029gmail_msg"> MemoryAccess *Def = OldMemAcc->getDefiningAccess();<br></div><div class="gmail-m_-868984512930983029gmail_msg"> NewMemAcc =</div><div class="gmail-m_-868984512930983029gmail_msg"> MSSA->createMemoryAccessAfter(Repl, Def, OldMemAccess->getMemoryInst());</div><div class="gmail-m_-868984512930983029gmail_msg"> OldMemAcc->replaceAllUsesWith(NewMemAcc);</div><div class="gmail-m_-868984512930983029gmail_msg"><br></div><div class="gmail-m_-868984512930983029gmail_msg">It doesn't make any sense to create a replacing access with the thing you are going to replace as the definition.</div><div class="gmail-m_-868984512930983029gmail_msg"><br></div><div class="gmail-m_-868984512930983029gmail_msg">The same thing, btw, would happen in normal LLVM SSA IR if you used the thing you are replacing in the definition of the replacement.<br><br></div><div class="gmail-m_-868984512930983029gmail_msg">%c = add %a, %b</div><div class="gmail-m_-868984512930983029gmail_msg">%e = add %c, %d</div><div class="gmail-m_-868984512930983029gmail_msg"><br></div><div class="gmail-m_-868984512930983029gmail_msg">If you wish to replace the first, you can't do this:</div><div class="gmail-m_-868984512930983029gmail_msg"><br><div class="gmail-m_-868984512930983029gmail_msg">%c = add %a, %b</div><div class="gmail-m_-868984512930983029gmail_msg">%tmp = add %a, %c</div><div class="gmail-m_-868984512930983029gmail_msg">%e = add %c, %d</div><div><br></div><div>because calling %c->replaceAllUsesWith %tmp</div><div>will result in:<br><br></div><div><div class="gmail-m_-868984512930983029gmail_msg">%tmp = add %a, %tmp</div><div class="gmail-m_-868984512930983029gmail_msg">%e = add %c, %d</div></div><div><br></div><div><br></div><div>IE the behavior you are citing is completely consistent with existing LLVM IR. If you define a replacement partially in terms of the thing it is replacing, it breaks.</div><div><br></div></div><div class="gmail-m_-868984512930983029gmail_msg"><br></div></div><div class="gmail-m_-868984512930983029gmail_msg"><br></div></div></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Sun, Oct 30, 2016 at 1:56 PM, Daniel Berlin <span dir="ltr"><<a href="mailto:dberlin@dberlin.org" target="_blank">dberlin@dberlin.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><span>For starters, createDefinedAccess is private, so nobody can use it.</span><div>Second, the API calls that use it explicitly take a defining access to avoid the problem you cite. </div><div><br></div><div>Can you produce a case where normal usage of the update API actually does the wrong thing?</div><div><br></div><div>I'm entirely unsure what actual usage problem you are trying to solve, so I can't help explain how to do it.<div><div class="gmail-h5"><br><br><div class="gmail_quote"><div dir="ltr">On Sun, Oct 30, 2016, 1:17 PM Bryant Wong <<a href="mailto:314472@gmail.com" target="_blank">314472@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" class="gmail-m_-868984512930983029gmail_msg">Thanks for clearing up my misunderstanding regarding MemoryUses.<br class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg">On Sun, Oct 30, 2016 at 2:04 PM, Daniel Berlin <span dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><<a href="mailto:dberlin@dberlin.org" class="gmail-m_-868984512930983029gmail_msg" target="_blank">dberlin@dberlin.org</a>></span> wrote:<br class="gmail-m_-868984512930983029gmail_msg"><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg">On Sun, Oct 30, 2016 at 3:03 AM, bryant <span dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><<a href="mailto:3.14472+reviews.llvm.org@gmail.com" class="gmail-m_-868984512930983029gmail_msg" target="_blank">3.14472+reviews.llvm.org@<wbr>gmail.com</a>></span> wrote:<br class="gmail-m_-868984512930983029gmail_msg"><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">bryant created this revision.<br class="gmail-m_-868984512930983029gmail_msg">
bryant added reviewers: dberlin, george.burgess.iv.<br class="gmail-m_-868984512930983029gmail_msg">
bryant added a subscriber: llvm-commits.<br class="gmail-m_-868984512930983029gmail_msg">
bryant set the repository for this revision to rL LLVM.<br class="gmail-m_-868984512930983029gmail_msg">
<br class="gmail-m_-868984512930983029gmail_msg">
An invariant of AccessLists is that the defining access of a Use or Def is the nearest preceding Def node.</blockquote></span><div class="gmail-m_-868984512930983029gmail_msg">False</div><div class="gmail-m_-868984512930983029gmail_msg">This is correct for def's but not for uses.</div><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div><div class="gmail-m_-868984512930983029gmail_msg"> </div><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> For instance, within a BasicBlock:<br class="gmail-m_-868984512930983029gmail_msg">
<br class="gmail-m_-868984512930983029gmail_msg">
0 = MemoryDef(liveOnEntry)<br class="gmail-m_-868984512930983029gmail_msg">
1 = MemoryDef(0)<br class="gmail-m_-868984512930983029gmail_msg">
2 = MemoryUse(n)<br class="gmail-m_-868984512930983029gmail_msg">
3 = MemoryDef(m)<br class="gmail-m_-868984512930983029gmail_msg">
<br class="gmail-m_-868984512930983029gmail_msg">
1 is the nearest parent Def of 2 and 3, and m and n must be 1.<br class="gmail-m_-868984512930983029gmail_msg"></blockquote><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div></span><div class="gmail-m_-868984512930983029gmail_msg">Given the above, this is incorrect.</div><div class="gmail-m_-868984512930983029gmail_msg">n may be 0</div><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div><div class="gmail-m_-868984512930983029gmail_msg">IE the following is perfectly legal</div><div class="gmail-m_-868984512930983029gmail_msg"><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"> 0 = MemoryDef(liveOnEntry)<br class="gmail-m_-868984512930983029gmail_msg"> 1 = MemoryDef(0)<br class="gmail-m_-868984512930983029gmail_msg"></span> 2 = MemoryUse(0)<br class="gmail-m_-868984512930983029gmail_msg"> 3 = MemoryDef(1)<br class="gmail-m_-868984512930983029gmail_msg"></div><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"> </div><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br class="gmail-m_-868984512930983029gmail_msg">
This patch simplifies the interfaces of createMemoryAccessBefore/<wbr>After, and augments them to maintain this invariant.<br class="gmail-m_-868984512930983029gmail_msg">
<br class="gmail-m_-868984512930983029gmail_msg">
Additionally, when inserting a new Def after an existing Def, there is currently no (clean) way to update the users of the old Def to use the new Def. </blockquote><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div></span><div class="gmail-m_-868984512930983029gmail_msg">RAUW works exactly to do this case.</div><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"> </div></span></div></div></div></blockquote><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg">I'm not so sure that it's sufficient. Suppose, for instance, that I wanted to insert a MemoryDef between 1 and 2 in the below AccessList:</div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"> 0 = MemoryDef(liveOnEntry)<br class="gmail-m_-868984512930983029gmail_msg"> 1 = MemoryDef(0)<br class="gmail-m_-868984512930983029gmail_msg"></span></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"> 2 = MemoryDef(1)<br class="gmail-m_-868984512930983029gmail_msg"> 3 = MemoryUse(1)<br class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div><div class="gmail-m_-868984512930983029gmail_msg">Invoking createMemoryAccess followed by RAUW of 1's uses with 4 would result in:<br class="gmail-m_-868984512930983029gmail_msg"></div><div class="gmail-m_-868984512930983029gmail_msg"></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"><br></div></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg">So RAUW needs to happen before setting 4's defining access to 1, but this isn't possible with the current createDefiningAccess. What other options are there? RAUW with a bogus Def, create, then re-RAUW to the newly created Def? I feel like I'm missing something obvious, so clarification would be much appreciated.<br class="gmail-m_-868984512930983029gmail_msg"></div></div></div></div><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><div class="gmail-m_-868984512930983029gmail_msg"> </div><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr" class="gmail-m_-868984512930983029gmail_msg"><div class="gmail_extra gmail-m_-868984512930983029gmail_msg"><div class="gmail_quote gmail-m_-868984512930983029gmail_msg"><span class="gmail-m_-868984512930983029m_-6466548616601174784m_7998958474538251886gmail- gmail-m_-868984512930983029gmail_msg"><blockquote class="gmail_quote gmail-m_-868984512930983029gmail_msg" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">So createDefiningAccess now permits the option of updating the users.<br class="gmail-m_-868984512930983029gmail_msg"></blockquote><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div></span><div class="gmail-m_-868984512930983029gmail_msg">Please, no.</div><div class="gmail-m_-868984512930983029gmail_msg"><br class="gmail-m_-868984512930983029gmail_msg"></div></div></div></div>
</blockquote></div></div></div></blockquote></div></div></div></div></blockquote></div><br></div></div></div>