mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-13 23:18:58 +00:00
6970be03d1
This patch adds vectorization support for uniform constant i64 arithmetic shift right operators. Differential Revision: http://reviews.llvm.org/D9645 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241514 91177308-0d34-0410-b5e6-96231b3b80d8
67 lines
1.9 KiB
LLVM
67 lines
1.9 KiB
LLVM
; RUN: llc < %s -march=x86 -mattr=+sse2 | FileCheck %s
|
|
|
|
; test vector shifts converted to proper SSE2 vector shifts when the shift
|
|
; amounts are the same.
|
|
|
|
; Note that x86 does have ashr
|
|
|
|
define void @shift1a(<2 x i64> %val, <2 x i64>* %dst) nounwind {
|
|
entry:
|
|
; CHECK-LABEL: shift1a:
|
|
; CHECK: psrad $31
|
|
%ashr = ashr <2 x i64> %val, < i64 32, i64 32 >
|
|
store <2 x i64> %ashr, <2 x i64>* %dst
|
|
ret void
|
|
}
|
|
|
|
define void @shift2a(<4 x i32> %val, <4 x i32>* %dst) nounwind {
|
|
entry:
|
|
; CHECK-LABEL: shift2a:
|
|
; CHECK: psrad $5
|
|
%ashr = ashr <4 x i32> %val, < i32 5, i32 5, i32 5, i32 5 >
|
|
store <4 x i32> %ashr, <4 x i32>* %dst
|
|
ret void
|
|
}
|
|
|
|
define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, i32 %amt) nounwind {
|
|
entry:
|
|
; CHECK-LABEL: shift2b:
|
|
; CHECK: movd
|
|
; CHECK: psrad
|
|
%0 = insertelement <4 x i32> undef, i32 %amt, i32 0
|
|
%1 = insertelement <4 x i32> %0, i32 %amt, i32 1
|
|
%2 = insertelement <4 x i32> %1, i32 %amt, i32 2
|
|
%3 = insertelement <4 x i32> %2, i32 %amt, i32 3
|
|
%ashr = ashr <4 x i32> %val, %3
|
|
store <4 x i32> %ashr, <4 x i32>* %dst
|
|
ret void
|
|
}
|
|
|
|
define void @shift3a(<8 x i16> %val, <8 x i16>* %dst) nounwind {
|
|
entry:
|
|
; CHECK-LABEL: shift3a:
|
|
; CHECK: psraw $5
|
|
%ashr = ashr <8 x i16> %val, < i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5 >
|
|
store <8 x i16> %ashr, <8 x i16>* %dst
|
|
ret void
|
|
}
|
|
|
|
define void @shift3b(<8 x i16> %val, <8 x i16>* %dst, i16 %amt) nounwind {
|
|
entry:
|
|
; CHECK-LABEL: shift3b:
|
|
; CHECK: movzwl
|
|
; CHECK: movd
|
|
; CHECK: psraw
|
|
%0 = insertelement <8 x i16> undef, i16 %amt, i32 0
|
|
%1 = insertelement <8 x i16> %0, i16 %amt, i32 1
|
|
%2 = insertelement <8 x i16> %1, i16 %amt, i32 2
|
|
%3 = insertelement <8 x i16> %2, i16 %amt, i32 3
|
|
%4 = insertelement <8 x i16> %3, i16 %amt, i32 4
|
|
%5 = insertelement <8 x i16> %4, i16 %amt, i32 5
|
|
%6 = insertelement <8 x i16> %5, i16 %amt, i32 6
|
|
%7 = insertelement <8 x i16> %6, i16 %amt, i32 7
|
|
%ashr = ashr <8 x i16> %val, %7
|
|
store <8 x i16> %ashr, <8 x i16>* %dst
|
|
ret void
|
|
}
|