[PATCH] D101290: [LV] Try to sink and hoist inside candidate loops for vectorization.

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

As far as i can tell, this doesn't reintroduce the problem that was fixed by disabling early hoisting in simplifycfg, so LG.



================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:9989
+  bool Changed = false;
+  DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Eager);
+  for (BasicBlock *BB : make_early_inc_range(L->blocks())) {
----------------
FWIW these sink/hoist helpers are fine with lazy updates.


================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10006-10016
+      if (hoistThenElseCodeToIf(BI, TTI, false, &DTU)) {
+        Changed = true;
+        SmallPtrSet<BasicBlock *, 2> NewSuccs(succ_begin(BB), succ_end(BB));
+        if (!NewSuccs.count(OldThen)) {
+          LI.removeBlock(OldThen);
+          DeleteDeadBlock(OldThen);
+        }
----------------
I think it would be fine to sink that into `hoistThenElseCodeToIf()`.


================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:10025-10031
+      if (sinkCommonCodeFromPredecessors(BB, &DTU)) {
+        // sinkCommonCodeFromPredecessors may introduce a new split block in
+        // the loop. Add it.
+        for (auto *Pred : predecessors(BB)) {
+          if (!L->contains(Pred))
+            L->addBasicBlockToLoop(Pred, LI);
+        }
----------------
I think it would be fine to sink that into `sinkCommonCodeFromPredecessors()`.



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101290/new/

https://reviews.llvm.org/D101290