[PATCH] D30211: [LV] Merge floating point and integer induction widening code

[Matthew Simpson via Phabricator via llvm-commits]

mssimpso updated this revision to Diff 89392.
mssimpso marked 4 inline comments as done.
mssimpso added a comment.

Addressed comments from Michael and Elena. Thanks!

- Changed the constant helper to use signed values.
- Simplified assert in widenIntOrFpInduction.
- Used the induction opcode from the induction descriptor for the "AddOp" in buildScalarSteps and createVectorIntOrFPInductionPHI. I corrected one of the test cases I incorrectly updated the first time around. It now correctly uses fsub instead of fadd for the induction update.
- Rebased new test case.


https://reviews.llvm.org/D30211

Files:
  lib/Transforms/Vectorize/LoopVectorize.cpp
  test/Transforms/LoopVectorize/float-induction.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D30211.89392.patch
Type: text/x-patch
Size: 30189 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170222/e9c7833d/attachment.bin>