forked from OSchip/llvm-project
parent
da59b0d2a8
commit
3a1c4e75de
|
@ -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?
|
||||
|
||||
|
|
Loading…
Reference in New Issue