mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-21 12:02:58 +02:00
a825efbf95
This patch adds vectorization support for uniform constant i64 arithmetic shift right operators. Differential Revision: http://reviews.llvm.org/D9645 llvm-svn: 241514
14 lines
396 B
LLVM
14 lines
396 B
LLVM
; RUN: llc < %s -march=x86 -mattr=+sse4.2 | FileCheck %s
|
|
; CHECK: psllq $48, %xmm0
|
|
; CHECK: psrad $16, %xmm0
|
|
; CHECK: pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
|
|
|
|
; sign extension v2i16 to v2i32
|
|
|
|
define void @convert(<2 x i32>* %dst.addr, <2 x i16> %src) nounwind {
|
|
entry:
|
|
%signext = sext <2 x i16> %src to <2 x i32> ; <<12 x i8>> [#uses=1]
|
|
store <2 x i32> %signext, <2 x i32>* %dst.addr
|
|
ret void
|
|
}
|