[PATCH][ARM64] Improve ARM64 vector creation

[Louis Gerbarg]

Currently when vectors are of mostly constant floating point values are created on ARM64 we do a lot of unnecessary inserts instead of duping the value across the vector. This patch fixes that, along with adding some patterns to be a bit a bit more efficient when that value is 0.0f.

Louis

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Improve-ARM64-vector-creation.patch
Type: application/octet-stream
Size: 3829 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140417/638f8a3b/attachment.obj>