[PATCH] Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g

[Yunzhong Gao]

  From: Manman Ren [manman.ren at gmail.com]
  Sent: Wednesday, November 13, 2013 1:49 PM

  > Hi Yunzhong,
  > A few nits:
  > In the file name "inst_limits.ll", we prefer "-" instead of "_".
  > +      // Debug intrinsics don't (and can't) cause dependences.
  > --> dependencies
  > I agree with David that you should only need one dbg.value intrinsic to show the problem.

  Many thanks for looking at my patch!
  I fixed the file name and the typo as you suggested.

  From: David Blaikie [dblaikie at gmail.com]
  Sent: Wednesday, November 13, 2013 1:26 PM

  > I assume you only need one dbg.value intrinsic to demonstrate
  > the problem, right?
  > And you only really need it in the inside limit case - in the
  > sense that the presence of that one intrinsic shouldn't tip you
  > over the limit.

  You are right about both points. I need only one dbg.value intrinsic
  in the inside-limit test case. Thanks for catching these!

  - Gao.

http://llvm-reviews.chandlerc.com/D2141

CHANGE SINCE LAST DIFF
  http://llvm-reviews.chandlerc.com/D2141?vs=5516&id=5528#toc

Files:
  lib/Analysis/MemoryDependenceAnalysis.cpp
  test/Transforms/DeadStoreElimination/inst-limits.ll
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D2141.4.patch
Type: text/x-patch
Size: 10347 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20131113/d7c25128/attachment.bin>