• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc < %s -march=x86 -mcpu=core2 | FileCheck %s
2
3; test vector shifts converted to proper SSE2 vector shifts when the shift
4; amounts are the same when using a shuffle splat.
5
6define void @shift1a(<2 x i64> %val, <2 x i64>* %dst, <2 x i64> %sh) nounwind {
7entry:
8; CHECK-LABEL: shift1a:
9; CHECK: psllq
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
13  ret void
14}
15
16; shift1b can't use a packed shift but can shift lanes separately and shuffle back together
17define void @shift1b(<2 x i64> %val, <2 x i64>* %dst, <2 x i64> %sh) nounwind {
18entry:
19; CHECK-LABEL: shift1b:
20; CHECK:       pshufd {{.*#+}} xmm2 = xmm1[2,3,0,1]
21; CHECK-NEXT:  movdqa %xmm0, %xmm3
22; CHECK-NEXT:  psllq  %xmm2, %xmm3
23; CHECK-NEXT:  movq   {{.*#+}} xmm1 = xmm1[0],zero
24; CHECK-NEXT:  psllq  %xmm1, %xmm0
25; CHECK-NEXT:  movsd  {{.*#+}} xmm3 = xmm0[0],xmm3[1]
26  %shamt = shufflevector <2 x i64> %sh, <2 x i64> undef, <2 x i32> <i32 0, i32 1>
27  %shl = shl <2 x i64> %val, %shamt
28  store <2 x i64> %shl, <2 x i64>* %dst
29  ret void
30}
31
32define void @shift2a(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
33entry:
34; CHECK-LABEL: shift2a:
35; CHECK: pslld
36  %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
37  %shl = shl <4 x i32> %val, %shamt
38  store <4 x i32> %shl, <4 x i32>* %dst
39  ret void
40}
41
42define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
43entry:
44; CHECK-LABEL: shift2b:
45; CHECK: pslld
46  %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 1, i32 1>
47  %shl = shl <4 x i32> %val, %shamt
48  store <4 x i32> %shl, <4 x i32>* %dst
49  ret void
50}
51
52define void @shift2c(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
53entry:
54; CHECK-LABEL: shift2c:
55; CHECK: pslld
56  %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
57  %shl = shl <4 x i32> %val, %shamt
58  store <4 x i32> %shl, <4 x i32>* %dst
59  ret void
60}
61
62define void @shift3a(<8 x i16> %val, <8 x i16>* %dst, <8 x i16> %amt) nounwind {
63entry:
64; CHECK-LABEL: shift3a:
65; CHECK: pextrw $6
66; CHECK: psllw
67  %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>
68  %shl = shl <8 x i16> %val, %shamt
69  store <8 x i16> %shl, <8 x i16>* %dst
70  ret void
71}
72
73define void @shift3b(<8 x i16> %val, <8 x i16>* %dst, i16 %amt) nounwind {
74entry:
75; CHECK-LABEL: shift3b:
76; CHECK: movzwl
77; CHECK: psllw
78  %0 = insertelement <8 x i16> undef, i16 %amt, i32 0
79  %1 = insertelement <8 x i16> %0, i16 %amt, i32 1
80  %2 = insertelement <8 x i16> %1, i16 %amt, i32 2
81  %3 = insertelement <8 x i16> %2, i16 %amt, i32 3
82  %4 = insertelement <8 x i16> %3, i16 %amt, i32 4
83  %5 = insertelement <8 x i16> %4, i16 %amt, i32 5
84  %6 = insertelement <8 x i16> %5, i16 %amt, i32 6
85  %7 = insertelement <8 x i16> %6, i16 %amt, i32 7
86  %shl = shl <8 x i16> %val, %7
87  store <8 x i16> %shl, <8 x i16>* %dst
88  ret void
89}
90
91