[PATCH] D57221: [LoopSimplifyCFG] Change logic of dead loops removal to avoid hitting asserts

[Mikael Holmén via Phabricator via llvm-commits]

uabelho added inline comments.


================
Comment at: lib/Transforms/Scalar/LoopSimplifyCFG.cpp:408-409
+        Loop *DL = LI.getLoopFor(BB);
+        if (DL->getParentLoop()) {
+          for (auto *PL = DL->getParentLoop(); PL; PL = PL->getParentLoop())
+            for (auto *BB : DL->getBlocks())
----------------
mkazantsev wrote:
> uabelho wrote:
> > mkazantsev wrote:
> > > uabelho wrote:
> > > > Perhaps replace
> > > > 
> > > >         if (DL->getParentLoop()) {
> > > >           for (auto *PL = DL->getParentLoop(); PL; PL = PL->getParentLoop())
> > > > 
> > > > with
> > > > 
> > > >         while (auto *PL = DL->getParentLoop()) {
> > > > 
> > > > ?
> > > And loop infinitely? :) We need to change DL somehow, otherwise PL is always the same loop.
> > Doesn't
> >  DL->getParentLoop()->removeChildLoop(DL);
> > disconnect DL from PL, so in the next loop round DL doesn't have the old PL as parent anymore?
> The line `DL->getParentLoop()->removeChildLoop(DL);` is executed once *after* this loop. You seem to be misreading the code. :)
Right. When I tried what I suggested I did

```
        while (auto *PL = DL->getParentLoop()) {
          for (auto *BB : DL->getBlocks())
            PL->removeBlockFromLoop(BB);
          DL->getParentLoop()->removeChildLoop(DL);
          LI.addTopLevelLoop(DL);
        }
```
with that all the testcases (including the reproducer in this patch) passes but perhaps it's stupid to call LI.addTopLevelLoop(DL) several times.
But nvm, I just thought the

```
if (DL->getParentLoop()) {
  for (auto *PL = DL->getParentLoop(); PL; PL = PL->getParentLoop())
```
was cumbersome and could be simplified, I didn't mean to make a mess :)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57221/new/

https://reviews.llvm.org/D57221