1; RUN: llc -mcpu=pwr9 -O3 -verify-machineinstrs -ppc-vsr-nums-as-vr \
2; RUN:     -mtriple=powerpc64le-unknown-linux-gnu \
3; RUN:     < %s | FileCheck %s
4
5; RUN: llc -mcpu=pwr9 -O3 -verify-machineinstrs -ppc-vsr-nums-as-vr \
6; RUN:     -mtriple=powerpc64-unknown-linux-gnu \
7; RUN:     < %s | FileCheck %s --check-prefix=P9BE
8
9; RUN: llc -mcpu=pwr9 -O3 -verify-machineinstrs -ppc-vsr-nums-as-vr \
10; RUN:     -mtriple=powerpc64-ibm-aix-xcoff -vec-extabi \
11; RUN:     < %s | FileCheck %s --check-prefix=P9BE
12
13define void @test64(i8* nocapture readonly %pix2, i32 signext %i_pix2) {
14; CHECK-LABEL: test64:
15; CHECK-NOT: ldux
16; CHECK-NOT: mtvsrd
17; CHECK: lxsdx [[REG:[0-9]+]]
18; CHECK: vperm {{[0-9]+}}, [[REG]]
19; P9BE-LABEL: test64:
20; P9BE-NOT: ldux
21; P9BE-NOT: mtvsrd
22; P9BE: lxsdx [[REG:[0-9]+]]
23; P9BE: vperm {{[0-9]+}}, {{[0-9]+}}, [[REG]]
24entry:
25  %idx.ext63 = sext i32 %i_pix2 to i64
26  %add.ptr64 = getelementptr inbounds i8, i8* %pix2, i64 %idx.ext63
27  %arrayidx5.1 = getelementptr inbounds i8, i8* %add.ptr64, i64 4
28  %0 = bitcast i8* %add.ptr64 to <4 x i16>*
29  %1 = load <4 x i16>, <4 x i16>* %0, align 1
30  %reorder_shuffle117 = shufflevector <4 x i16> %1, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
31  %2 = zext <4 x i16> %reorder_shuffle117 to <4 x i32>
32  %3 = sub nsw <4 x i32> zeroinitializer, %2
33  %4 = bitcast i8* %arrayidx5.1 to <4 x i16>*
34  %5 = load <4 x i16>, <4 x i16>* %4, align 1
35  %reorder_shuffle115 = shufflevector <4 x i16> %5, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
36  %6 = zext <4 x i16> %reorder_shuffle115 to <4 x i32>
37  %7 = sub nsw <4 x i32> zeroinitializer, %6
38  %8 = shl nsw <4 x i32> %7, <i32 16, i32 16, i32 16, i32 16>
39  %9 = add nsw <4 x i32> %8, %3
40  %10 = sub nsw <4 x i32> %9, zeroinitializer
41  %11 = shufflevector <4 x i32> undef, <4 x i32> %10, <4 x i32> <i32 2, i32 7, i32 0, i32 5>
42  %12 = add nsw <4 x i32> zeroinitializer, %11
43  %13 = shufflevector <4 x i32> %12, <4 x i32> undef, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
44  store <4 x i32> %13, <4 x i32>* undef, align 16
45  ret void
46}
47
48define void @test32(i8* nocapture readonly %pix2, i32 signext %i_pix2) {
49; CHECK-LABEL: test32:
50; CHECK-NOT: lwzux
51; CHECK-NOT: mtvsrwz
52; CHECK: lxsiwzx [[REG:[0-9]+]]
53; CHECK: vperm {{[0-9]+}}, [[REG]]
54; P9BE-LABEL: test32:
55; P9BE-NOT: lwzux
56; P9BE-NOT: mtvsrwz
57; P9BE: lxsiwzx [[REG:[0-9]+]]
58; P9BE: vperm {{[0-9]+}}, {{[0-9]+}}, [[REG]]
59entry:
60  %idx.ext63 = sext i32 %i_pix2 to i64
61  %add.ptr64 = getelementptr inbounds i8, i8* %pix2, i64 %idx.ext63
62  %arrayidx5.1 = getelementptr inbounds i8, i8* %add.ptr64, i64 4
63  %0 = bitcast i8* %add.ptr64 to <4 x i8>*
64  %1 = load <4 x i8>, <4 x i8>* %0, align 1
65  %reorder_shuffle117 = shufflevector <4 x i8> %1, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
66  %2 = zext <4 x i8> %reorder_shuffle117 to <4 x i32>
67  %3 = sub nsw <4 x i32> zeroinitializer, %2
68  %4 = bitcast i8* %arrayidx5.1 to <4 x i8>*
69  %5 = load <4 x i8>, <4 x i8>* %4, align 1
70  %reorder_shuffle115 = shufflevector <4 x i8> %5, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
71  %6 = zext <4 x i8> %reorder_shuffle115 to <4 x i32>
72  %7 = sub nsw <4 x i32> zeroinitializer, %6
73  %8 = shl nsw <4 x i32> %7, <i32 16, i32 16, i32 16, i32 16>
74  %9 = add nsw <4 x i32> %8, %3
75  %10 = sub nsw <4 x i32> %9, zeroinitializer
76  %11 = shufflevector <4 x i32> undef, <4 x i32> %10, <4 x i32> <i32 2, i32 7, i32 0, i32 5>
77  %12 = add nsw <4 x i32> zeroinitializer, %11
78  %13 = shufflevector <4 x i32> %12, <4 x i32> undef, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
79  store <4 x i32> %13, <4 x i32>* undef, align 16
80  ret void
81}
82
83define void @test16(i16* nocapture readonly %sums, i32 signext %delta, i32 signext %thresh) {
84; CHECK-LABEL: test16
85; CHECK-NOT: lhzux
86; CHECK: lxsihzx [[REG:[0-9]+]]
87; CHECK: vmrghh {{[0-9]+}}, {{[0-9]+}}, [[REG]]
88; P9BE-LABEL: test16
89; P9BE-NOT: lhzux
90; P9BE: lxsihzx [[REG:[0-9]+]]
91; P9BE: vperm {{[0-9]+}}, {{[0-9]+}}, [[REG]]
92entry:
93  %idxprom = sext i32 %delta to i64
94  %add14 = add nsw i32 %delta, 8
95  %idxprom15 = sext i32 %add14 to i64
96  br label %for.body
97
98for.body:                                         ; preds = %entry
99  %arrayidx8 = getelementptr inbounds i16, i16* %sums, i64 %idxprom
100  %0 = load i16, i16* %arrayidx8, align 2
101  %arrayidx16 = getelementptr inbounds i16, i16* %sums, i64 %idxprom15
102  %1 = load i16, i16* %arrayidx16, align 2
103  %2 = insertelement <4 x i16> undef, i16 %0, i32 2
104  %3 = insertelement <4 x i16> %2, i16 %1, i32 3
105  %4 = zext <4 x i16> %3 to <4 x i32>
106  %5 = sub nsw <4 x i32> zeroinitializer, %4
107  %6 = sub nsw <4 x i32> zeroinitializer, %5
108  %7 = select <4 x i1> undef, <4 x i32> %6, <4 x i32> %5
109  %bin.rdx = add <4 x i32> %7, zeroinitializer
110  %rdx.shuf54 = shufflevector <4 x i32> %bin.rdx, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
111  %bin.rdx55 = add <4 x i32> %bin.rdx, %rdx.shuf54
112  %8 = extractelement <4 x i32> %bin.rdx55, i32 0
113  %op.extra = add nuw i32 %8, 0
114  %cmp25 = icmp slt i32 %op.extra, %thresh
115  br i1 %cmp25, label %if.then, label %if.end
116
117if.then:                                          ; preds = %for.body
118  unreachable
119
120if.end:                                           ; preds = %for.body
121  ret void
122}
123
124define void @test8(i8* nocapture readonly %sums, i32 signext %delta, i32 signext %thresh) {
125; CHECK-LABEL: test8:
126; CHECK-NOT: lbzux
127; CHECK: lxsibzx [[REG:[0-9]+]]
128; CHECK: vmrghb {{[0-9]+}}, {{[0-9]+}}, [[REG]]
129; P9BE-LABEL: test8:
130; P9BE-NOT: lbzux
131; P9BE: lxsibzx [[REG:[0-9]+]]
132; P9BE: vperm {{[0-9]+}}, {{[0-9]+}}, [[REG]]
133entry:
134  %idxprom = sext i32 %delta to i64
135  %add14 = add nsw i32 %delta, 8
136  %idxprom15 = sext i32 %add14 to i64
137  br label %for.body
138
139for.body:                                         ; preds = %entry
140  %arrayidx8 = getelementptr inbounds i8, i8* %sums, i64 %idxprom
141  %0 = load i8, i8* %arrayidx8, align 2
142  %arrayidx16 = getelementptr inbounds i8, i8* %sums, i64 %idxprom15
143  %1 = load i8, i8* %arrayidx16, align 2
144  %2 = insertelement <4 x i8> undef, i8 %0, i32 2
145  %3 = insertelement <4 x i8> %2, i8 %1, i32 3
146  %4 = zext <4 x i8> %3 to <4 x i32>
147  %5 = sub nsw <4 x i32> zeroinitializer, %4
148  %6 = sub nsw <4 x i32> zeroinitializer, %5
149  %7 = select <4 x i1> undef, <4 x i32> %6, <4 x i32> %5
150  %bin.rdx = add <4 x i32> %7, zeroinitializer
151  %rdx.shuf54 = shufflevector <4 x i32> %bin.rdx, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
152  %bin.rdx55 = add <4 x i32> %bin.rdx, %rdx.shuf54
153  %8 = extractelement <4 x i32> %bin.rdx55, i32 0
154  %op.extra = add nuw i32 %8, 0
155  %cmp25 = icmp slt i32 %op.extra, %thresh
156  br i1 %cmp25, label %if.then, label %if.end
157
158if.then:                                          ; preds = %for.body
159  unreachable
160
161if.end:                                           ; preds = %for.body
162  ret void
163}
164