[llvm-commits] [PATCH] Phi speculation improvement for BasicAA

[Hal Finkel]

Hello,

As a followup, and partial solution, to PR14351, the attached patch removes some of the special significance of the first incoming phi value in the phi aliasing checking logic in BasicAA. In the context of a loop, the current logic assumes that the first incoming value is the interesting one (meaning that it is the one that comes from outside the loop), but this is often not the case. With this patch, we now test first the incoming value that comes from a block other than the parent of the phi being tested. This is important for, among other things, alias-analysis-based instruction dependency breaking for instruction scheduling.

As noted in the PR, Arnold is currently working on a more general solution. In the mean time, this fix captures several cases I need (and has an associated test case). Please review.

Thanks again,
Hal

-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phi_spec_order.patch
Type: text/x-patch
Size: 7987 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20121116/94ff88e5/attachment.bin>