[PATCH] Improvde code quality of vselect

[Michael Liao]

Hi,

As we canonicalize sext from i1 to select in instcomb, it generates redundant code to prepare all 0s and all 1s if we don't take it into account that, in SSE/AVX, packed comparisons output all 0s or 1s in SIMD lanes. The attached patch adds optimization for those cases and transforms select into AND/OR to reduce BLEND pressure when profitable.

Thanks for review
- Michael

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Add-optimization-to-leverage-all-0s-1s-produced-by-p.patch
Type: text/x-patch
Size: 5233 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130410/fd7f8da7/attachment.bin>