| blob | 2c53769e069dfdcfce8c1ca8cad1b81c0d1c918c |
1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
3 ; RUN: -mtriple=powerpc64le-unknown-unknown < %s | FileCheck %s
4 %struct.poly2 = type { [11 x i64] }
6 ; Function Attrs: nofree norecurse nounwind
7 define dso_local void @poly2_lshift1(%struct.poly2* nocapture %p) local_unnamed_addr #0 {
8 ; CHECK-LABEL: poly2_lshift1:
9 ; CHECK: # %bb.0: # %entry
10 ; CHECK-NEXT: li r4, 72
11 ; CHECK-NEXT: addis r5, r2, .LCPI0_0@toc@ha
12 ; CHECK-NEXT: addis r6, r2, .LCPI0_1@toc@ha
13 ; CHECK-NEXT: ld r7, 64(r3)
14 ; CHECK-NEXT: ld r8, 16(r3)
15 ; CHECK-NEXT: ld r10, 24(r3)
16 ; CHECK-NEXT: ld r11, 32(r3)
17 ; CHECK-NEXT: lxvd2x vs0, r3, r4
18 ; CHECK-NEXT: addi r5, r5, .LCPI0_0@toc@l
19 ; CHECK-NEXT: addi r6, r6, .LCPI0_1@toc@l
20 ; CHECK-NEXT: ld r12, 56(r3)
21 ; CHECK-NEXT: lxvd2x vs1, 0, r5
22 ; CHECK-NEXT: mtfprd f2, r7
23 ; CHECK-NEXT: ld r5, 0(r3)
24 ; CHECK-NEXT: xxswapd v2, vs0
25 ; CHECK-NEXT: lxvd2x vs0, 0, r6
26 ; CHECK-NEXT: ld r6, 8(r3)
27 ; CHECK-NEXT: rotldi r9, r5, 1
28 ; CHECK-NEXT: sldi r5, r5, 1
29 ; CHECK-NEXT: xxswapd v3, vs1
30 ; CHECK-NEXT: std r5, 0(r3)
31 ; CHECK-NEXT: rotldi r5, r10, 1
32 ; CHECK-NEXT: rldimi r9, r6, 1, 0
33 ; CHECK-NEXT: rotldi r6, r6, 1
34 ; CHECK-NEXT: xxpermdi v4, v2, vs2, 2
35 ; CHECK-NEXT: xxswapd v5, vs0
36 ; CHECK-NEXT: rldimi r6, r8, 1, 0
37 ; CHECK-NEXT: rotldi r8, r8, 1
38 ; CHECK-NEXT: std r9, 8(r3)
39 ; CHECK-NEXT: ld r9, 40(r3)
40 ; CHECK-NEXT: rldimi r8, r10, 1, 0
41 ; CHECK-NEXT: rldimi r5, r11, 1, 0
42 ; CHECK-NEXT: std r6, 16(r3)
43 ; CHECK-NEXT: rotldi r10, r11, 1
44 ; CHECK-NEXT: ld r11, 48(r3)
45 ; CHECK-NEXT: std r5, 32(r3)
46 ; CHECK-NEXT: rotldi r6, r12, 1
47 ; CHECK-NEXT: vsrd v3, v4, v3
48 ; CHECK-NEXT: rldimi r10, r9, 1, 0
49 ; CHECK-NEXT: rotldi r9, r9, 1
50 ; CHECK-NEXT: std r8, 24(r3)
51 ; CHECK-NEXT: vsld v2, v2, v5
52 ; CHECK-NEXT: rotldi r5, r11, 1
53 ; CHECK-NEXT: rldimi r9, r11, 1, 0
54 ; CHECK-NEXT: std r10, 40(r3)
55 ; CHECK-NEXT: rldimi r5, r12, 1, 0
56 ; CHECK-NEXT: rldimi r6, r7, 1, 0
57 ; CHECK-NEXT: std r9, 48(r3)
58 ; CHECK-NEXT: xxlor vs0, v2, v3
59 ; CHECK-NEXT: std r5, 56(r3)
60 ; CHECK-NEXT: std r6, 64(r3)
61 ; CHECK-NEXT: xxswapd vs0, vs0
62 ; CHECK-NEXT: stxvd2x vs0, r3, r4
63 ; CHECK-NEXT: blr
64 entry:
65 %arrayidx = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 0
66 %0 = load i64, i64* %arrayidx, align 8
67 %shl = shl i64 %0, 1
68 store i64 %shl, i64* %arrayidx, align 8
69 %arrayidx.1 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 1
70 %1 = load i64, i64* %arrayidx.1, align 8
71 %or.1 = call i64 @llvm.fshl.i64(i64 %1, i64 %0, i64 1)
72 store i64 %or.1, i64* %arrayidx.1, align 8
73 %arrayidx.2 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 2
74 %2 = load i64, i64* %arrayidx.2, align 8
75 %or.2 = call i64 @llvm.fshl.i64(i64 %2, i64 %1, i64 1)
76 store i64 %or.2, i64* %arrayidx.2, align 8
77 %arrayidx.3 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 3
78 %3 = load i64, i64* %arrayidx.3, align 8
79 %or.3 = call i64 @llvm.fshl.i64(i64 %3, i64 %2, i64 1)
80 store i64 %or.3, i64* %arrayidx.3, align 8
81 %arrayidx.4 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 4
82 %4 = load i64, i64* %arrayidx.4, align 8
83 %or.4 = call i64 @llvm.fshl.i64(i64 %4, i64 %3, i64 1)
84 store i64 %or.4, i64* %arrayidx.4, align 8
85 %arrayidx.5 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 5
86 %5 = load i64, i64* %arrayidx.5, align 8
87 %or.5 = call i64 @llvm.fshl.i64(i64 %5, i64 %4, i64 1)
88 store i64 %or.5, i64* %arrayidx.5, align 8
89 %arrayidx.6 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 6
90 %6 = load i64, i64* %arrayidx.6, align 8
91 %or.6 = call i64 @llvm.fshl.i64(i64 %6, i64 %5, i64 1)
92 store i64 %or.6, i64* %arrayidx.6, align 8
93 %arrayidx.7 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 7
94 %7 = load i64, i64* %arrayidx.7, align 8
95 %or.7 = call i64 @llvm.fshl.i64(i64 %7, i64 %6, i64 1)
96 store i64 %or.7, i64* %arrayidx.7, align 8
97 %arrayidx.8 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 8
98 %8 = load i64, i64* %arrayidx.8, align 8
99 %or.8 = call i64 @llvm.fshl.i64(i64 %8, i64 %7, i64 1)
100 store i64 %or.8, i64* %arrayidx.8, align 8
101 %arrayidx.9 = getelementptr inbounds %struct.poly2, %struct.poly2* %p, i64 0, i32 0, i64 9
102 %9 = bitcast i64* %arrayidx.9 to <2 x i64>*
103 %10 = load <2 x i64>, <2 x i64>* %9, align 8
104 %11 = insertelement <2 x i64> undef, i64 %8, i32 0
105 %12 = shufflevector <2 x i64> %11, <2 x i64> %10, <2 x i32> <i32 0, i32 2>
106 %13 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %10, <2 x i64> %12, <2 x i64> <i64 1, i64 1>)
107 %14 = bitcast i64* %arrayidx.9 to <2 x i64>*
108 store <2 x i64> %13, <2 x i64>* %14, align 8
109 ret void
110 }
112 ; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
113 declare i64 @llvm.fshl.i64(i64, i64, i64) #1
115 ; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
116 declare <2 x i64> @llvm.fshl.v2i64(<2 x i64>, <2 x i64>, <2 x i64>) #1