[PATCH] D79175: [ARM][MVE] Tail-Predication: use @llvm.get.active.lane.mask to get the BTC

[Sjoerd Meijer via Phabricator via llvm-commits]

SjoerdMeijer updated this revision to Diff 267898.
SjoerdMeijer added a comment.

I am not entire done yet, but this adds `IsSafeActiveMask`, which performs checks on the induction variable and backedge-taken count that are arguments to @llvm.get.active.lane.mask, and tests have been added for this to `test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-const.ll`. I have also now updated all other tests to the new situation, i.e. manually added the @llvm.get.active.lane.mask instead of the icmp.

The approach for the overflow check is to use SCEV and query if the loop entry is protected by a conditional BTC + 1 < 0. In other words, if the scalar trip count overflows and becomes negative, we shouldn't enter the loop and create a tripcount expression BTC + 1 as that won't be valid. As I said, not entirely done yet, but wanted to check this after our overflow discussion while I fix up these things:

- the vctp can be cloned in the exit block, and looks like I am missing that at the moment.
- I am always creating a new num.elements = BTC + 1 expression, but it looks like that value might exist and I can reuse, hopefully reducing some of the codegen changes that we see in some of the tests.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D79175/new/

https://reviews.llvm.org/D79175

Files:
  llvm/lib/Target/ARM/MVETailPredication.cpp
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/basic-tail-pred.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/clear-maskedinsts.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/cond-vector-reduce-mve-codegen.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/extending-loads.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/fast-fp-loops.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/mve-tail-data-types.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/nested.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-const.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-widen.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-reduce.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/vector-arith-codegen.ll
  llvm/test/CodeGen/Thumb2/LowOverheadLoops/vector-reduce-mve-tail.ll
  llvm/test/CodeGen/Thumb2/mve-fma-loops.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D79175.267898.patch
Type: text/x-patch
Size: 161076 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200602/a0f9de15/attachment-0001.bin>