Another entry about shuffles.

llvm-svn: 27245
This commit is contained in:
Evan Cheng 2006-03-29 03:03:46 +00:00
parent da59b0d2a8
commit 3a1c4e75de
1 changed files with 6 additions and 0 deletions

View File

@ -667,3 +667,9 @@ directly %esp[0] if there are no other uses.
Use movhps to update upper 64-bits of a v4sf value. Also movlps on lower half
of a v4sf value.
//===---------------------------------------------------------------------===//
Better codegen for vector_shuffles like this { x, 0, 0, 0 } or { x, 0, x, 0}.
Perhaps use pxor / xorp* to clear a XMM register first?