[PATCH] D75746: [LoopVectorizer] Simplify branch in the remainder loop for trivial cases

[Danila Malyutin via Phabricator via llvm-commits]

danilaml marked an inline comment as done.
danilaml added a comment.

Updated revision with additional checks and rebased.

I'm not sure that `llvm.assume` can be reliably made to work (and be simpler, than just eliminating back edge).

Going with `{0, +1, %TripCount}` might be more beneficial in the end (it should also be trivial to call vectorize/unroll recursively in the remainder loop in such case, if I understood things correctly).



================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:3107
+  // If VFxUF is 2 and vector loop is not skipped then remainder executes once.
+  if (VF * UF == 2 && !areSafetyChecksAdded()) {
+    if (BasicBlock *Latch = OrigLoop->getLoopLatch())
----------------
Ayal wrote:
> danilaml wrote:
> > I think this check is enough unless there are other cases in which "remainder loop has `N % (VF*UF)` iterations doesn't hold.
> Note that the trip count of the remainder loop may be **equal** to VF*UF, when loop `requiresScalarEpilogue()`; so in the above special case of VF*UF==2 remainder loop may iterate once **or twice**.
> 
> Note that `emitMinimumIterationCountCheck()` also takes care of the case where adding 1 to the backedge-taken count overflows, leading to an incorrect trip count of zero; here too "remainder" loop iterates (much) more than once.
Thanks. I knew I might've missed something. This makes me more skeptical about potential llvm.assume solution.
Am I understanding your note correctly, that adding requiresScalarEpilogue check is enough?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75746/new/

https://reviews.llvm.org/D75746