[llvm-commits] [PATCH] Defer some shl transformations from InstCombine to DAGCombine

[Jakob Stoklund Olesen]

On Apr 18, 2012, at 4:28 PM, Chandler Carruth <chandlerc at google.com> wrote:

> Because it is a phase-ordering kind of problem, I worry a lot about regressions. Specifically, if we ever get some other pass that nukes your "silly function", will we stop testing that this works?
> 
> I wonder, could we add a direct test that SCEV can undrestand this pattern when InstCombine runs first? That would seem more targeted.

Committed as r155136, including test/Transforms/PhaseOrdering/scev.ll which runs all the standard passes followed by scalar evolution. It is used to test that certain expressions are not obfuscated by the optimizers.

/jakob

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20120419/0356416f/attachment.html>