1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=aarch64--linux-gnu | FileCheck %s
3
4define <8 x i8> @load4_v4i8_add(float %tmp, <4 x i8> *%a, <4 x i8> *%b) {
5; CHECK-LABEL: load4_v4i8_add:
6; CHECK:       // %bb.0:
7; CHECK-NEXT:    ldp s0, s1, [x0]
8; CHECK-NEXT:    ld1 { v0.s }[1], [x1], #4
9; CHECK-NEXT:    ld1 { v1.s }[1], [x1]
10; CHECK-NEXT:    add v0.8b, v0.8b, v1.8b
11; CHECK-NEXT:    ret
12  %la = load <4 x i8>, <4 x i8> *%a
13  %lb = load <4 x i8>, <4 x i8> *%b
14  %c = getelementptr <4 x i8>, <4 x i8> *%a, i64 1
15  %d = getelementptr <4 x i8>, <4 x i8> *%b, i64 1
16  %lc = load <4 x i8>, <4 x i8> *%c
17  %ld = load <4 x i8>, <4 x i8> *%d
18  %s1 = shufflevector <4 x i8> %la, <4 x i8> %lb, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
19  %s2 = shufflevector <4 x i8> %lc, <4 x i8> %ld, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
20  %add = add <8 x i8> %s1, %s2
21  ret <8 x i8> %add
22}
23
24define <8 x i16> @load4_v4i8_zext_add(float %tmp, <4 x i8> *%a, <4 x i8> *%b) {
25; CHECK-LABEL: load4_v4i8_zext_add:
26; CHECK:       // %bb.0:
27; CHECK-NEXT:    ldp s0, s1, [x0]
28; CHECK-NEXT:    ld1 { v0.s }[1], [x1], #4
29; CHECK-NEXT:    ld1 { v1.s }[1], [x1]
30; CHECK-NEXT:    uaddl v0.8h, v0.8b, v1.8b
31; CHECK-NEXT:    ret
32  %la = load <4 x i8>, <4 x i8> *%a
33  %lb = load <4 x i8>, <4 x i8> *%b
34  %c = getelementptr <4 x i8>, <4 x i8> *%a, i64 1
35  %d = getelementptr <4 x i8>, <4 x i8> *%b, i64 1
36  %lc = load <4 x i8>, <4 x i8> *%c
37  %ld = load <4 x i8>, <4 x i8> *%d
38  %s1 = shufflevector <4 x i8> %la, <4 x i8> %lb, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
39  %s2 = shufflevector <4 x i8> %lc, <4 x i8> %ld, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
40  %z1 = zext <8 x i8> %s1 to <8 x i16>
41  %z2 = zext <8 x i8> %s2 to <8 x i16>
42  %add = add <8 x i16> %z1, %z2
43  ret <8 x i16> %add
44}
45
46define i32 @large(i8* nocapture noundef readonly %p1, i32 noundef %st1, i8* nocapture noundef readonly %p2, i32 noundef %st2) {
47; CHECK-LABEL: large:
48; CHECK:       // %bb.0: // %entry
49; CHECK-NEXT:    // kill: def $w1 killed $w1 def $x1
50; CHECK-NEXT:    sxtw x8, w1
51; CHECK-NEXT:    // kill: def $w3 killed $w3 def $x3
52; CHECK-NEXT:    sxtw x11, w3
53; CHECK-NEXT:    add x9, x0, x8
54; CHECK-NEXT:    add x12, x2, x11
55; CHECK-NEXT:    add x10, x9, x8
56; CHECK-NEXT:    add x13, x12, x11
57; CHECK-NEXT:    add x8, x10, x8
58; CHECK-NEXT:    add x11, x13, x11
59; CHECK-NEXT:    ldp s1, s5, [x9]
60; CHECK-NEXT:    ldp s0, s4, [x8]
61; CHECK-NEXT:    ld1 { v0.s }[1], [x10], #4
62; CHECK-NEXT:    ld1 { v1.s }[1], [x0], #4
63; CHECK-NEXT:    ldp s2, s6, [x11]
64; CHECK-NEXT:    ldp s3, s7, [x12]
65; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
66; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
67; CHECK-NEXT:    ld1 { v2.s }[1], [x13], #4
68; CHECK-NEXT:    ld1 { v3.s }[1], [x2], #4
69; CHECK-NEXT:    ld1 { v4.s }[1], [x10]
70; CHECK-NEXT:    ld1 { v5.s }[1], [x0]
71; CHECK-NEXT:    ld1 { v6.s }[1], [x13]
72; CHECK-NEXT:    ld1 { v7.s }[1], [x2]
73; CHECK-NEXT:    ushll v2.8h, v2.8b, #0
74; CHECK-NEXT:    ushll v3.8h, v3.8b, #0
75; CHECK-NEXT:    usubl v16.4s, v0.4h, v2.4h
76; CHECK-NEXT:    usubl2 v0.4s, v0.8h, v2.8h
77; CHECK-NEXT:    usubl v2.4s, v1.4h, v3.4h
78; CHECK-NEXT:    usubl2 v1.4s, v1.8h, v3.8h
79; CHECK-NEXT:    ushll v3.8h, v4.8b, #0
80; CHECK-NEXT:    ushll v4.8h, v5.8b, #0
81; CHECK-NEXT:    ushll v5.8h, v6.8b, #0
82; CHECK-NEXT:    ushll v6.8h, v7.8b, #0
83; CHECK-NEXT:    usubl2 v7.4s, v3.8h, v5.8h
84; CHECK-NEXT:    usubl v3.4s, v3.4h, v5.4h
85; CHECK-NEXT:    usubl2 v5.4s, v4.8h, v6.8h
86; CHECK-NEXT:    usubl v4.4s, v4.4h, v6.4h
87; CHECK-NEXT:    shl v6.4s, v7.4s, #16
88; CHECK-NEXT:    shl v5.4s, v5.4s, #16
89; CHECK-NEXT:    shl v3.4s, v3.4s, #16
90; CHECK-NEXT:    shl v4.4s, v4.4s, #16
91; CHECK-NEXT:    add v1.4s, v5.4s, v1.4s
92; CHECK-NEXT:    add v2.4s, v4.4s, v2.4s
93; CHECK-NEXT:    add v0.4s, v6.4s, v0.4s
94; CHECK-NEXT:    rev64 v6.4s, v2.4s
95; CHECK-NEXT:    rev64 v17.4s, v1.4s
96; CHECK-NEXT:    add v3.4s, v3.4s, v16.4s
97; CHECK-NEXT:    rev64 v5.4s, v0.4s
98; CHECK-NEXT:    rev64 v4.4s, v3.4s
99; CHECK-NEXT:    addp v16.4s, v2.4s, v1.4s
100; CHECK-NEXT:    sub v1.4s, v1.4s, v17.4s
101; CHECK-NEXT:    sub v2.4s, v2.4s, v6.4s
102; CHECK-NEXT:    addp v7.4s, v3.4s, v0.4s
103; CHECK-NEXT:    sub v0.4s, v0.4s, v5.4s
104; CHECK-NEXT:    zip1 v5.4s, v2.4s, v1.4s
105; CHECK-NEXT:    sub v3.4s, v3.4s, v4.4s
106; CHECK-NEXT:    ext v18.16b, v7.16b, v7.16b, #8
107; CHECK-NEXT:    zip2 v4.4s, v0.4s, v3.4s
108; CHECK-NEXT:    mov v0.s[1], v3.s[0]
109; CHECK-NEXT:    ext v3.16b, v2.16b, v5.16b, #8
110; CHECK-NEXT:    mov v2.s[3], v1.s[2]
111; CHECK-NEXT:    uzp2 v19.4s, v7.4s, v16.4s
112; CHECK-NEXT:    uzp1 v6.4s, v7.4s, v16.4s
113; CHECK-NEXT:    uzp1 v7.4s, v18.4s, v16.4s
114; CHECK-NEXT:    uzp2 v1.4s, v18.4s, v16.4s
115; CHECK-NEXT:    mov v0.d[1], v3.d[1]
116; CHECK-NEXT:    mov v4.d[1], v2.d[1]
117; CHECK-NEXT:    add v5.4s, v19.4s, v6.4s
118; CHECK-NEXT:    sub v1.4s, v7.4s, v1.4s
119; CHECK-NEXT:    rev64 v2.4s, v5.4s
120; CHECK-NEXT:    sub v6.4s, v0.4s, v4.4s
121; CHECK-NEXT:    add v0.4s, v4.4s, v0.4s
122; CHECK-NEXT:    rev64 v3.4s, v1.4s
123; CHECK-NEXT:    rev64 v4.4s, v6.4s
124; CHECK-NEXT:    rev64 v7.4s, v0.4s
125; CHECK-NEXT:    addp v16.4s, v1.4s, v6.4s
126; CHECK-NEXT:    addp v17.4s, v5.4s, v0.4s
127; CHECK-NEXT:    sub v4.4s, v6.4s, v4.4s
128; CHECK-NEXT:    sub v0.4s, v0.4s, v7.4s
129; CHECK-NEXT:    sub v1.4s, v1.4s, v3.4s
130; CHECK-NEXT:    sub v2.4s, v5.4s, v2.4s
131; CHECK-NEXT:    ext v3.16b, v1.16b, v16.16b, #8
132; CHECK-NEXT:    ext v5.16b, v17.16b, v0.16b, #4
133; CHECK-NEXT:    ext v6.16b, v16.16b, v4.16b, #4
134; CHECK-NEXT:    zip1 v7.4s, v17.4s, v17.4s
135; CHECK-NEXT:    ext v18.16b, v3.16b, v1.16b, #4
136; CHECK-NEXT:    zip2 v5.4s, v5.4s, v17.4s
137; CHECK-NEXT:    zip2 v6.4s, v6.4s, v16.4s
138; CHECK-NEXT:    trn2 v7.4s, v7.4s, v2.4s
139; CHECK-NEXT:    ext v2.16b, v2.16b, v17.16b, #4
140; CHECK-NEXT:    mov v1.s[2], v16.s[1]
141; CHECK-NEXT:    ext v5.16b, v0.16b, v5.16b, #12
142; CHECK-NEXT:    ext v6.16b, v4.16b, v6.16b, #12
143; CHECK-NEXT:    uzp2 v3.4s, v3.4s, v18.4s
144; CHECK-NEXT:    mov v4.s[2], v16.s[3]
145; CHECK-NEXT:    mov v0.s[2], v17.s[3]
146; CHECK-NEXT:    ext v2.16b, v2.16b, v2.16b, #4
147; CHECK-NEXT:    sub v18.4s, v1.4s, v3.4s
148; CHECK-NEXT:    sub v19.4s, v4.4s, v6.4s
149; CHECK-NEXT:    sub v20.4s, v0.4s, v5.4s
150; CHECK-NEXT:    sub v21.4s, v7.4s, v2.4s
151; CHECK-NEXT:    mov v4.s[1], v16.s[2]
152; CHECK-NEXT:    mov v0.s[1], v17.s[2]
153; CHECK-NEXT:    mov v2.s[0], v17.s[1]
154; CHECK-NEXT:    mov v1.s[1], v16.s[0]
155; CHECK-NEXT:    add v4.4s, v4.4s, v6.4s
156; CHECK-NEXT:    add v0.4s, v0.4s, v5.4s
157; CHECK-NEXT:    add v2.4s, v7.4s, v2.4s
158; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
159; CHECK-NEXT:    mov v2.d[1], v21.d[1]
160; CHECK-NEXT:    mov v1.d[1], v18.d[1]
161; CHECK-NEXT:    mov v4.d[1], v19.d[1]
162; CHECK-NEXT:    mov v0.d[1], v20.d[1]
163; CHECK-NEXT:    movi v3.8h, #1
164; CHECK-NEXT:    movi v17.2d, #0x00ffff0000ffff
165; CHECK-NEXT:    ushr v5.4s, v4.4s, #15
166; CHECK-NEXT:    ushr v6.4s, v2.4s, #15
167; CHECK-NEXT:    ushr v7.4s, v0.4s, #15
168; CHECK-NEXT:    ushr v16.4s, v1.4s, #15
169; CHECK-NEXT:    and v6.16b, v6.16b, v3.16b
170; CHECK-NEXT:    and v16.16b, v16.16b, v3.16b
171; CHECK-NEXT:    and v7.16b, v7.16b, v3.16b
172; CHECK-NEXT:    and v3.16b, v5.16b, v3.16b
173; CHECK-NEXT:    mul v5.4s, v6.4s, v17.4s
174; CHECK-NEXT:    mul v6.4s, v16.4s, v17.4s
175; CHECK-NEXT:    mul v3.4s, v3.4s, v17.4s
176; CHECK-NEXT:    mul v7.4s, v7.4s, v17.4s
177; CHECK-NEXT:    add v2.4s, v5.4s, v2.4s
178; CHECK-NEXT:    add v1.4s, v6.4s, v1.4s
179; CHECK-NEXT:    add v4.4s, v3.4s, v4.4s
180; CHECK-NEXT:    add v0.4s, v7.4s, v0.4s
181; CHECK-NEXT:    eor v3.16b, v4.16b, v3.16b
182; CHECK-NEXT:    eor v0.16b, v0.16b, v7.16b
183; CHECK-NEXT:    eor v1.16b, v1.16b, v6.16b
184; CHECK-NEXT:    eor v2.16b, v2.16b, v5.16b
185; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
186; CHECK-NEXT:    add v0.4s, v0.4s, v3.4s
187; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
188; CHECK-NEXT:    addv s0, v0.4s
189; CHECK-NEXT:    fmov w8, s0
190; CHECK-NEXT:    lsr w9, w8, #16
191; CHECK-NEXT:    add w8, w9, w8, uxth
192; CHECK-NEXT:    lsr w0, w8, #1
193; CHECK-NEXT:    ret
194entry:
195  %idx.ext = sext i32 %st1 to i64
196  %idx.ext63 = sext i32 %st2 to i64
197  %arrayidx3 = getelementptr inbounds i8, i8* %p1, i64 4
198  %arrayidx5 = getelementptr inbounds i8, i8* %p2, i64 4
199  %0 = bitcast i8* %p1 to <4 x i8>*
200  %1 = load <4 x i8>, <4 x i8>* %0, align 1
201  %2 = bitcast i8* %p2 to <4 x i8>*
202  %3 = load <4 x i8>, <4 x i8>* %2, align 1
203  %4 = bitcast i8* %arrayidx3 to <4 x i8>*
204  %5 = load <4 x i8>, <4 x i8>* %4, align 1
205  %6 = bitcast i8* %arrayidx5 to <4 x i8>*
206  %7 = load <4 x i8>, <4 x i8>* %6, align 1
207  %add.ptr = getelementptr inbounds i8, i8* %p1, i64 %idx.ext
208  %add.ptr64 = getelementptr inbounds i8, i8* %p2, i64 %idx.ext63
209  %arrayidx3.1 = getelementptr inbounds i8, i8* %add.ptr, i64 4
210  %arrayidx5.1 = getelementptr inbounds i8, i8* %add.ptr64, i64 4
211  %8 = bitcast i8* %add.ptr to <4 x i8>*
212  %9 = load <4 x i8>, <4 x i8>* %8, align 1
213  %10 = bitcast i8* %add.ptr64 to <4 x i8>*
214  %11 = load <4 x i8>, <4 x i8>* %10, align 1
215  %12 = bitcast i8* %arrayidx3.1 to <4 x i8>*
216  %13 = load <4 x i8>, <4 x i8>* %12, align 1
217  %14 = bitcast i8* %arrayidx5.1 to <4 x i8>*
218  %15 = load <4 x i8>, <4 x i8>* %14, align 1
219  %add.ptr.1 = getelementptr inbounds i8, i8* %add.ptr, i64 %idx.ext
220  %add.ptr64.1 = getelementptr inbounds i8, i8* %add.ptr64, i64 %idx.ext63
221  %arrayidx3.2 = getelementptr inbounds i8, i8* %add.ptr.1, i64 4
222  %arrayidx5.2 = getelementptr inbounds i8, i8* %add.ptr64.1, i64 4
223  %16 = bitcast i8* %add.ptr.1 to <4 x i8>*
224  %17 = load <4 x i8>, <4 x i8>* %16, align 1
225  %18 = bitcast i8* %add.ptr64.1 to <4 x i8>*
226  %19 = load <4 x i8>, <4 x i8>* %18, align 1
227  %20 = bitcast i8* %arrayidx3.2 to <4 x i8>*
228  %21 = load <4 x i8>, <4 x i8>* %20, align 1
229  %22 = bitcast i8* %arrayidx5.2 to <4 x i8>*
230  %23 = load <4 x i8>, <4 x i8>* %22, align 1
231  %add.ptr.2 = getelementptr inbounds i8, i8* %add.ptr.1, i64 %idx.ext
232  %add.ptr64.2 = getelementptr inbounds i8, i8* %add.ptr64.1, i64 %idx.ext63
233  %arrayidx3.3 = getelementptr inbounds i8, i8* %add.ptr.2, i64 4
234  %arrayidx5.3 = getelementptr inbounds i8, i8* %add.ptr64.2, i64 4
235  %24 = bitcast i8* %add.ptr.2 to <4 x i8>*
236  %25 = load <4 x i8>, <4 x i8>* %24, align 1
237  %26 = shufflevector <4 x i8> %25, <4 x i8> %17, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
238  %27 = shufflevector <4 x i8> %9, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
239  %28 = shufflevector <16 x i8> %26, <16 x i8> %27, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 undef, i32 undef, i32 undef, i32 undef>
240  %29 = shufflevector <4 x i8> %1, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
241  %30 = shufflevector <16 x i8> %28, <16 x i8> %29, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
242  %31 = zext <16 x i8> %30 to <16 x i32>
243  %32 = bitcast i8* %add.ptr64.2 to <4 x i8>*
244  %33 = load <4 x i8>, <4 x i8>* %32, align 1
245  %34 = shufflevector <4 x i8> %33, <4 x i8> %19, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
246  %35 = shufflevector <4 x i8> %11, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
247  %36 = shufflevector <16 x i8> %34, <16 x i8> %35, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 undef, i32 undef, i32 undef, i32 undef>
248  %37 = shufflevector <4 x i8> %3, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
249  %38 = shufflevector <16 x i8> %36, <16 x i8> %37, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
250  %39 = zext <16 x i8> %38 to <16 x i32>
251  %40 = sub nsw <16 x i32> %31, %39
252  %41 = bitcast i8* %arrayidx3.3 to <4 x i8>*
253  %42 = load <4 x i8>, <4 x i8>* %41, align 1
254  %43 = shufflevector <4 x i8> %42, <4 x i8> %21, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
255  %44 = shufflevector <4 x i8> %13, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
256  %45 = shufflevector <16 x i8> %43, <16 x i8> %44, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 undef, i32 undef, i32 undef, i32 undef>
257  %46 = shufflevector <4 x i8> %5, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
258  %47 = shufflevector <16 x i8> %45, <16 x i8> %46, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
259  %48 = zext <16 x i8> %47 to <16 x i32>
260  %49 = bitcast i8* %arrayidx5.3 to <4 x i8>*
261  %50 = load <4 x i8>, <4 x i8>* %49, align 1
262  %51 = shufflevector <4 x i8> %50, <4 x i8> %23, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
263  %52 = shufflevector <4 x i8> %15, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
264  %53 = shufflevector <16 x i8> %51, <16 x i8> %52, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 undef, i32 undef, i32 undef, i32 undef>
265  %54 = shufflevector <4 x i8> %7, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
266  %55 = shufflevector <16 x i8> %53, <16 x i8> %54, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
267  %56 = zext <16 x i8> %55 to <16 x i32>
268  %57 = sub nsw <16 x i32> %48, %56
269  %58 = shl nsw <16 x i32> %57, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
270  %59 = add nsw <16 x i32> %58, %40
271  %60 = shufflevector <16 x i32> %59, <16 x i32> undef, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
272  %61 = add nsw <16 x i32> %59, %60
273  %62 = sub nsw <16 x i32> %59, %60
274  %63 = shufflevector <16 x i32> %61, <16 x i32> %62, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 22, i32 18, i32 26, i32 30, i32 5, i32 1, i32 9, i32 13, i32 20, i32 16, i32 24, i32 28>
275  %64 = shufflevector <16 x i32> %61, <16 x i32> %62, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 20, i32 16, i32 24, i32 28, i32 7, i32 3, i32 11, i32 15, i32 22, i32 18, i32 26, i32 30>
276  %65 = add nsw <16 x i32> %63, %64
277  %66 = sub nsw <16 x i32> %63, %64
278  %67 = shufflevector <16 x i32> %65, <16 x i32> %66, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
279  %68 = shufflevector <16 x i32> %65, <16 x i32> %66, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 25, i32 24, i32 27, i32 26, i32 29, i32 28, i32 31, i32 30>
280  %69 = add nsw <16 x i32> %67, %68
281  %70 = sub nsw <16 x i32> %67, %68
282  %71 = shufflevector <16 x i32> %69, <16 x i32> %70, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 20, i32 5, i32 6, i32 23, i32 24, i32 9, i32 10, i32 27, i32 28, i32 13, i32 14, i32 31>
283  %72 = shufflevector <16 x i32> %69, <16 x i32> %70, <16 x i32> <i32 2, i32 19, i32 0, i32 17, i32 23, i32 6, i32 5, i32 20, i32 27, i32 10, i32 9, i32 24, i32 31, i32 14, i32 13, i32 28>
284  %73 = add nsw <16 x i32> %71, %72
285  %74 = sub nsw <16 x i32> %71, %72
286  %75 = shufflevector <16 x i32> %73, <16 x i32> %74, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 4, i32 5, i32 22, i32 23, i32 8, i32 9, i32 26, i32 27, i32 12, i32 13, i32 30, i32 31>
287  %76 = lshr <16 x i32> %75, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
288  %77 = and <16 x i32> %76, <i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537, i32 65537>
289  %78 = mul nuw <16 x i32> %77, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
290  %79 = add <16 x i32> %78, %75
291  %80 = xor <16 x i32> %79, %78
292  %81 = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %80)
293  %conv118 = and i32 %81, 65535
294  %shr = lshr i32 %81, 16
295  %add119 = add nuw nsw i32 %conv118, %shr
296  %shr120 = lshr i32 %add119, 1
297  ret i32 %shr120
298}
299
300declare i32 @llvm.vector.reduce.add.v16i32(<16 x i32>)
301