1
; RUN: llc < %s -march=x86 -mattr=+sse2 -disable-mmx | FileCheck %s
3
; test vector shifts converted to proper SSE2 vector shifts when the shift
4
; amounts are the same when using a shuffle splat.
6
define void @shift1a(<2 x i64> %val, <2 x i64>* %dst, <2 x i64> %sh) nounwind {
10
%shamt = shufflevector <2 x i64> %sh, <2 x i64> undef, <2 x i32> <i32 0, i32 0>
11
%shl = shl <2 x i64> %val, %shamt
12
store <2 x i64> %shl, <2 x i64>* %dst
16
; shift1b can't use a packed shift
17
define void @shift1b(<2 x i64> %val, <2 x i64>* %dst, <2 x i64> %sh) nounwind {
21
%shamt = shufflevector <2 x i64> %sh, <2 x i64> undef, <2 x i32> <i32 0, i32 1>
22
%shl = shl <2 x i64> %val, %shamt
23
store <2 x i64> %shl, <2 x i64>* %dst
27
define void @shift2a(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
31
%shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
32
%shl = shl <4 x i32> %val, %shamt
33
store <4 x i32> %shl, <4 x i32>* %dst
37
define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
41
%shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 1, i32 1>
42
%shl = shl <4 x i32> %val, %shamt
43
store <4 x i32> %shl, <4 x i32>* %dst
47
define void @shift2c(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
51
%shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
52
%shl = shl <4 x i32> %val, %shamt
53
store <4 x i32> %shl, <4 x i32>* %dst
57
define void @shift3a(<8 x i16> %val, <8 x i16>* %dst, <8 x i16> %amt) nounwind {
62
%shamt = shufflevector <8 x i16> %amt, <8 x i16> undef, <8 x i32> <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
63
%shl = shl <8 x i16> %val, %shamt
64
store <8 x i16> %shl, <8 x i16>* %dst
68
define void @shift3b(<8 x i16> %val, <8 x i16>* %dst, i16 %amt) nounwind {
73
%0 = insertelement <8 x i16> undef, i16 %amt, i32 0
74
%1 = insertelement <8 x i16> %0, i16 %amt, i32 1
75
%2 = insertelement <8 x i16> %0, i16 %amt, i32 2
76
%3 = insertelement <8 x i16> %0, i16 %amt, i32 3
77
%4 = insertelement <8 x i16> %0, i16 %amt, i32 4
78
%5 = insertelement <8 x i16> %0, i16 %amt, i32 5
79
%6 = insertelement <8 x i16> %0, i16 %amt, i32 6
80
%7 = insertelement <8 x i16> %0, i16 %amt, i32 7
81
%shl = shl <8 x i16> %val, %7
82
store <8 x i16> %shl, <8 x i16>* %dst