[PATCH] D16100: Fix for two constant propagation problems in GVN with assume intrinsic instruction

[Yuanrui Zhang via llvm-commits]

Ray updated this revision to Diff 44903.
Ray added a comment.

I actually agree with Daniel's comments, especially about the proposal to rewrite propagateEquality. I would see my fix as a necessary co-implementation with Piotr's earlier patch http://reviews.llvm.org/D12170, since my changes only happen in that particular commit.

Probably we can submit this "necessary" fix as well as the test cases (for "assume") first, and then think of restructure the propagateEquality.

I update this new patch to fix my previously wrong comments pointed out by Daniel. Thanks for that:)


http://reviews.llvm.org/D16100

Files:
  include/llvm/Transforms/Utils/Local.h
  lib/Transforms/Scalar/GVN.cpp
  lib/Transforms/Utils/Local.cpp
  test/Transforms/GVN/assume-equal.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D16100.44903.patch
Type: text/x-patch
Size: 4297 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160114/35af5f8c/attachment.bin>