1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2
3; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42
4; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
5; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
6
7target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
8
9@src64 = common global [4 x i64] zeroinitializer, align 32
10@dst64 = common global [4 x i64] zeroinitializer, align 32
11@src32 = common global [8 x i32] zeroinitializer, align 32
12@dst32 = common global [8 x i32] zeroinitializer, align 32
13@src16 = common global [16 x i16] zeroinitializer, align 32
14@dst16 = common global [16 x i16] zeroinitializer, align 32
15@src8  = common global [32 x i8] zeroinitializer, align 32
16@dst8  = common global [32 x i8] zeroinitializer, align 32
17
18declare i64 @llvm.ctlz.i64(i64, i1)
19declare i32 @llvm.ctlz.i32(i32, i1)
20declare i16 @llvm.ctlz.i16(i16, i1)
21declare  i8 @llvm.ctlz.i8(i8, i1)
22
23;
24; CTLZ
25;
26
27define void @ctlz_2i64() #0 {
28; CHECK-LABEL: @ctlz_2i64(
29; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
30; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
31; CHECK-NEXT:    [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 false)
32; CHECK-NEXT:    [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 false)
33; CHECK-NEXT:    store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
34; CHECK-NEXT:    store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
35; CHECK-NEXT:    ret void
36;
37  %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
38  %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
39  %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 0)
40  %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 0)
41  store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
42  store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
43  ret void
44}
45
46define void @ctlz_4i64() #0 {
47; CHECK-LABEL: @ctlz_4i64(
48; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
49; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
50; CHECK-NEXT:    [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
51; CHECK-NEXT:    [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
52; CHECK-NEXT:    [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 false)
53; CHECK-NEXT:    [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 false)
54; CHECK-NEXT:    [[CTLZ2:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD2]], i1 false)
55; CHECK-NEXT:    [[CTLZ3:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD3]], i1 false)
56; CHECK-NEXT:    store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
57; CHECK-NEXT:    store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
58; CHECK-NEXT:    store i64 [[CTLZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
59; CHECK-NEXT:    store i64 [[CTLZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
60; CHECK-NEXT:    ret void
61;
62  %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
63  %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
64  %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
65  %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
66  %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 0)
67  %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 0)
68  %ctlz2 = call i64 @llvm.ctlz.i64(i64 %ld2, i1 0)
69  %ctlz3 = call i64 @llvm.ctlz.i64(i64 %ld3, i1 0)
70  store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
71  store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
72  store i64 %ctlz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
73  store i64 %ctlz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
74  ret void
75}
76
77define void @ctlz_4i32() #0 {
78; CHECK-LABEL: @ctlz_4i32(
79; CHECK-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
80; CHECK-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
81; CHECK-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
82; CHECK-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
83; CHECK-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
84; CHECK-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
85; CHECK-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
86; CHECK-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
87; CHECK-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
88; CHECK-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
89; CHECK-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
90; CHECK-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
91; CHECK-NEXT:    ret void
92;
93  %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
94  %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
95  %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
96  %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
97  %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 0)
98  %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 0)
99  %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 0)
100  %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 0)
101  store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
102  store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
103  store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
104  store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
105  ret void
106}
107
108define void @ctlz_8i32() #0 {
109; SSE-LABEL: @ctlz_8i32(
110; SSE-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
111; SSE-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
112; SSE-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
113; SSE-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
114; SSE-NEXT:    [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
115; SSE-NEXT:    [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
116; SSE-NEXT:    [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
117; SSE-NEXT:    [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
118; SSE-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
119; SSE-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
120; SSE-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
121; SSE-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
122; SSE-NEXT:    [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false)
123; SSE-NEXT:    [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false)
124; SSE-NEXT:    [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false)
125; SSE-NEXT:    [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false)
126; SSE-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
127; SSE-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
128; SSE-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
129; SSE-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
130; SSE-NEXT:    store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
131; SSE-NEXT:    store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
132; SSE-NEXT:    store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
133; SSE-NEXT:    store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
134; SSE-NEXT:    ret void
135;
136; AVX1-LABEL: @ctlz_8i32(
137; AVX1-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
138; AVX1-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
139; AVX1-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
140; AVX1-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
141; AVX1-NEXT:    [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
142; AVX1-NEXT:    [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
143; AVX1-NEXT:    [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
144; AVX1-NEXT:    [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
145; AVX1-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
146; AVX1-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
147; AVX1-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
148; AVX1-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
149; AVX1-NEXT:    [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false)
150; AVX1-NEXT:    [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false)
151; AVX1-NEXT:    [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false)
152; AVX1-NEXT:    [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false)
153; AVX1-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
154; AVX1-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
155; AVX1-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
156; AVX1-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
157; AVX1-NEXT:    store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
158; AVX1-NEXT:    store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
159; AVX1-NEXT:    store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
160; AVX1-NEXT:    store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
161; AVX1-NEXT:    ret void
162;
163; AVX2-LABEL: @ctlz_8i32(
164; AVX2-NEXT:    [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
165; AVX2-NEXT:    [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 false)
166; AVX2-NEXT:    store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
167; AVX2-NEXT:    ret void
168;
169  %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
170  %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
171  %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
172  %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
173  %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
174  %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
175  %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
176  %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
177  %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 0)
178  %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 0)
179  %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 0)
180  %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 0)
181  %ctlz4 = call i32 @llvm.ctlz.i32(i32 %ld4, i1 0)
182  %ctlz5 = call i32 @llvm.ctlz.i32(i32 %ld5, i1 0)
183  %ctlz6 = call i32 @llvm.ctlz.i32(i32 %ld6, i1 0)
184  %ctlz7 = call i32 @llvm.ctlz.i32(i32 %ld7, i1 0)
185  store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
186  store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
187  store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
188  store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
189  store i32 %ctlz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
190  store i32 %ctlz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
191  store i32 %ctlz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
192  store i32 %ctlz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
193  ret void
194}
195
196define void @ctlz_8i16() #0 {
197; CHECK-LABEL: @ctlz_8i16(
198; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
199; CHECK-NEXT:    [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false)
200; CHECK-NEXT:    store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
201; CHECK-NEXT:    ret void
202;
203  %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
204  %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
205  %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
206  %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
207  %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
208  %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
209  %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
210  %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
211  %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 0)
212  %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 0)
213  %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 0)
214  %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 0)
215  %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 0)
216  %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 0)
217  %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 0)
218  %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 0)
219  store i16 %ctlz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
220  store i16 %ctlz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
221  store i16 %ctlz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
222  store i16 %ctlz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
223  store i16 %ctlz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
224  store i16 %ctlz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
225  store i16 %ctlz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
226  store i16 %ctlz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
227  ret void
228}
229
230define void @ctlz_16i16() #0 {
231; SSE-LABEL: @ctlz_16i16(
232; SSE-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
233; SSE-NEXT:    [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
234; SSE-NEXT:    [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false)
235; SSE-NEXT:    [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 false)
236; SSE-NEXT:    store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
237; SSE-NEXT:    store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
238; SSE-NEXT:    ret void
239;
240; AVX-LABEL: @ctlz_16i16(
241; AVX-NEXT:    [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
242; AVX-NEXT:    [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 false)
243; AVX-NEXT:    store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
244; AVX-NEXT:    ret void
245;
246  %ld0  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  0), align 2
247  %ld1  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  1), align 2
248  %ld2  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  2), align 2
249  %ld3  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  3), align 2
250  %ld4  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  4), align 2
251  %ld5  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  5), align 2
252  %ld6  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  6), align 2
253  %ld7  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  7), align 2
254  %ld8  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  8), align 2
255  %ld9  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  9), align 2
256  %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
257  %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
258  %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
259  %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
260  %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
261  %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
262  %ctlz0  = call i16 @llvm.ctlz.i16(i16 %ld0, i1 0)
263  %ctlz1  = call i16 @llvm.ctlz.i16(i16 %ld1, i1 0)
264  %ctlz2  = call i16 @llvm.ctlz.i16(i16 %ld2, i1 0)
265  %ctlz3  = call i16 @llvm.ctlz.i16(i16 %ld3, i1 0)
266  %ctlz4  = call i16 @llvm.ctlz.i16(i16 %ld4, i1 0)
267  %ctlz5  = call i16 @llvm.ctlz.i16(i16 %ld5, i1 0)
268  %ctlz6  = call i16 @llvm.ctlz.i16(i16 %ld6, i1 0)
269  %ctlz7  = call i16 @llvm.ctlz.i16(i16 %ld7, i1 0)
270  %ctlz8  = call i16 @llvm.ctlz.i16(i16 %ld8, i1 0)
271  %ctlz9  = call i16 @llvm.ctlz.i16(i16 %ld9, i1 0)
272  %ctlz10 = call i16 @llvm.ctlz.i16(i16 %ld10, i1 0)
273  %ctlz11 = call i16 @llvm.ctlz.i16(i16 %ld11, i1 0)
274  %ctlz12 = call i16 @llvm.ctlz.i16(i16 %ld12, i1 0)
275  %ctlz13 = call i16 @llvm.ctlz.i16(i16 %ld13, i1 0)
276  %ctlz14 = call i16 @llvm.ctlz.i16(i16 %ld14, i1 0)
277  %ctlz15 = call i16 @llvm.ctlz.i16(i16 %ld15, i1 0)
278  store i16 %ctlz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  0), align 2
279  store i16 %ctlz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  1), align 2
280  store i16 %ctlz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  2), align 2
281  store i16 %ctlz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  3), align 2
282  store i16 %ctlz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  4), align 2
283  store i16 %ctlz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  5), align 2
284  store i16 %ctlz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  6), align 2
285  store i16 %ctlz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  7), align 2
286  store i16 %ctlz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  8), align 2
287  store i16 %ctlz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  9), align 2
288  store i16 %ctlz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
289  store i16 %ctlz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
290  store i16 %ctlz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
291  store i16 %ctlz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
292  store i16 %ctlz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
293  store i16 %ctlz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
294  ret void
295}
296
297define void @ctlz_16i8() #0 {
298; CHECK-LABEL: @ctlz_16i8(
299; CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
300; CHECK-NEXT:    [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false)
301; CHECK-NEXT:    store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
302; CHECK-NEXT:    ret void
303;
304  %ld0  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  0), align 1
305  %ld1  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  1), align 1
306  %ld2  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  2), align 1
307  %ld3  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  3), align 1
308  %ld4  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  4), align 1
309  %ld5  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  5), align 1
310  %ld6  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  6), align 1
311  %ld7  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  7), align 1
312  %ld8  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  8), align 1
313  %ld9  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  9), align 1
314  %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
315  %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
316  %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
317  %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
318  %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
319  %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
320  %ctlz0  = call i8 @llvm.ctlz.i8(i8 %ld0, i1 0)
321  %ctlz1  = call i8 @llvm.ctlz.i8(i8 %ld1, i1 0)
322  %ctlz2  = call i8 @llvm.ctlz.i8(i8 %ld2, i1 0)
323  %ctlz3  = call i8 @llvm.ctlz.i8(i8 %ld3, i1 0)
324  %ctlz4  = call i8 @llvm.ctlz.i8(i8 %ld4, i1 0)
325  %ctlz5  = call i8 @llvm.ctlz.i8(i8 %ld5, i1 0)
326  %ctlz6  = call i8 @llvm.ctlz.i8(i8 %ld6, i1 0)
327  %ctlz7  = call i8 @llvm.ctlz.i8(i8 %ld7, i1 0)
328  %ctlz8  = call i8 @llvm.ctlz.i8(i8 %ld8, i1 0)
329  %ctlz9  = call i8 @llvm.ctlz.i8(i8 %ld9, i1 0)
330  %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 0)
331  %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 0)
332  %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 0)
333  %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 0)
334  %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 0)
335  %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 0)
336  store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  0), align 1
337  store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  1), align 1
338  store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  2), align 1
339  store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  3), align 1
340  store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  4), align 1
341  store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  5), align 1
342  store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  6), align 1
343  store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  7), align 1
344  store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  8), align 1
345  store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  9), align 1
346  store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
347  store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
348  store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
349  store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
350  store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
351  store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
352  ret void
353}
354
355define void @ctlz_32i8() #0 {
356; CHECK-LABEL: @ctlz_32i8(
357; CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
358; CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
359; CHECK-NEXT:    [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false)
360; CHECK-NEXT:    [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 false)
361; CHECK-NEXT:    store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
362; CHECK-NEXT:    store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
363; CHECK-NEXT:    ret void
364;
365  %ld0  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  0), align 1
366  %ld1  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  1), align 1
367  %ld2  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  2), align 1
368  %ld3  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  3), align 1
369  %ld4  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  4), align 1
370  %ld5  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  5), align 1
371  %ld6  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  6), align 1
372  %ld7  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  7), align 1
373  %ld8  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  8), align 1
374  %ld9  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  9), align 1
375  %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
376  %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
377  %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
378  %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
379  %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
380  %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
381  %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
382  %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
383  %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
384  %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
385  %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
386  %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
387  %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
388  %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
389  %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
390  %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
391  %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
392  %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
393  %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
394  %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
395  %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
396  %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
397  %ctlz0  = call i8 @llvm.ctlz.i8(i8 %ld0, i1 0)
398  %ctlz1  = call i8 @llvm.ctlz.i8(i8 %ld1, i1 0)
399  %ctlz2  = call i8 @llvm.ctlz.i8(i8 %ld2, i1 0)
400  %ctlz3  = call i8 @llvm.ctlz.i8(i8 %ld3, i1 0)
401  %ctlz4  = call i8 @llvm.ctlz.i8(i8 %ld4, i1 0)
402  %ctlz5  = call i8 @llvm.ctlz.i8(i8 %ld5, i1 0)
403  %ctlz6  = call i8 @llvm.ctlz.i8(i8 %ld6, i1 0)
404  %ctlz7  = call i8 @llvm.ctlz.i8(i8 %ld7, i1 0)
405  %ctlz8  = call i8 @llvm.ctlz.i8(i8 %ld8, i1 0)
406  %ctlz9  = call i8 @llvm.ctlz.i8(i8 %ld9, i1 0)
407  %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 0)
408  %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 0)
409  %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 0)
410  %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 0)
411  %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 0)
412  %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 0)
413  %ctlz16 = call i8 @llvm.ctlz.i8(i8 %ld16, i1 0)
414  %ctlz17 = call i8 @llvm.ctlz.i8(i8 %ld17, i1 0)
415  %ctlz18 = call i8 @llvm.ctlz.i8(i8 %ld18, i1 0)
416  %ctlz19 = call i8 @llvm.ctlz.i8(i8 %ld19, i1 0)
417  %ctlz20 = call i8 @llvm.ctlz.i8(i8 %ld20, i1 0)
418  %ctlz21 = call i8 @llvm.ctlz.i8(i8 %ld21, i1 0)
419  %ctlz22 = call i8 @llvm.ctlz.i8(i8 %ld22, i1 0)
420  %ctlz23 = call i8 @llvm.ctlz.i8(i8 %ld23, i1 0)
421  %ctlz24 = call i8 @llvm.ctlz.i8(i8 %ld24, i1 0)
422  %ctlz25 = call i8 @llvm.ctlz.i8(i8 %ld25, i1 0)
423  %ctlz26 = call i8 @llvm.ctlz.i8(i8 %ld26, i1 0)
424  %ctlz27 = call i8 @llvm.ctlz.i8(i8 %ld27, i1 0)
425  %ctlz28 = call i8 @llvm.ctlz.i8(i8 %ld28, i1 0)
426  %ctlz29 = call i8 @llvm.ctlz.i8(i8 %ld29, i1 0)
427  %ctlz30 = call i8 @llvm.ctlz.i8(i8 %ld30, i1 0)
428  %ctlz31 = call i8 @llvm.ctlz.i8(i8 %ld31, i1 0)
429  store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  0), align 1
430  store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  1), align 1
431  store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  2), align 1
432  store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  3), align 1
433  store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  4), align 1
434  store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  5), align 1
435  store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  6), align 1
436  store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  7), align 1
437  store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  8), align 1
438  store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  9), align 1
439  store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
440  store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
441  store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
442  store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
443  store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
444  store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
445  store i8 %ctlz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
446  store i8 %ctlz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
447  store i8 %ctlz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
448  store i8 %ctlz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
449  store i8 %ctlz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
450  store i8 %ctlz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
451  store i8 %ctlz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
452  store i8 %ctlz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
453  store i8 %ctlz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
454  store i8 %ctlz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
455  store i8 %ctlz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
456  store i8 %ctlz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
457  store i8 %ctlz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
458  store i8 %ctlz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
459  store i8 %ctlz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
460  store i8 %ctlz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
461  ret void
462}
463
464;
465; CTLZ_ZERO_UNDEF
466;
467
468define void @ctlz_undef_2i64() #0 {
469; CHECK-LABEL: @ctlz_undef_2i64(
470; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
471; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
472; CHECK-NEXT:    [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 true)
473; CHECK-NEXT:    [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 true)
474; CHECK-NEXT:    store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
475; CHECK-NEXT:    store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
476; CHECK-NEXT:    ret void
477;
478  %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
479  %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
480  %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 -1)
481  %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 -1)
482  store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
483  store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
484  ret void
485}
486
487define void @ctlz_undef_4i64() #0 {
488; CHECK-LABEL: @ctlz_undef_4i64(
489; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
490; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
491; CHECK-NEXT:    [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
492; CHECK-NEXT:    [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
493; CHECK-NEXT:    [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 true)
494; CHECK-NEXT:    [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 true)
495; CHECK-NEXT:    [[CTLZ2:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD2]], i1 true)
496; CHECK-NEXT:    [[CTLZ3:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD3]], i1 true)
497; CHECK-NEXT:    store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
498; CHECK-NEXT:    store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
499; CHECK-NEXT:    store i64 [[CTLZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
500; CHECK-NEXT:    store i64 [[CTLZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
501; CHECK-NEXT:    ret void
502;
503  %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
504  %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
505  %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
506  %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
507  %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 -1)
508  %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 -1)
509  %ctlz2 = call i64 @llvm.ctlz.i64(i64 %ld2, i1 -1)
510  %ctlz3 = call i64 @llvm.ctlz.i64(i64 %ld3, i1 -1)
511  store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
512  store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
513  store i64 %ctlz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
514  store i64 %ctlz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
515  ret void
516}
517
518define void @ctlz_undef_4i32() #0 {
519; CHECK-LABEL: @ctlz_undef_4i32(
520; CHECK-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
521; CHECK-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
522; CHECK-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
523; CHECK-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
524; CHECK-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
525; CHECK-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
526; CHECK-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
527; CHECK-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
528; CHECK-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
529; CHECK-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
530; CHECK-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
531; CHECK-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
532; CHECK-NEXT:    ret void
533;
534  %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
535  %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
536  %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
537  %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
538  %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 -1)
539  %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 -1)
540  %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 -1)
541  %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 -1)
542  store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
543  store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
544  store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
545  store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
546  ret void
547}
548
549define void @ctlz_undef_8i32() #0 {
550; SSE-LABEL: @ctlz_undef_8i32(
551; SSE-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
552; SSE-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
553; SSE-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
554; SSE-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
555; SSE-NEXT:    [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
556; SSE-NEXT:    [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
557; SSE-NEXT:    [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
558; SSE-NEXT:    [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
559; SSE-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
560; SSE-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
561; SSE-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
562; SSE-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
563; SSE-NEXT:    [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true)
564; SSE-NEXT:    [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true)
565; SSE-NEXT:    [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true)
566; SSE-NEXT:    [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true)
567; SSE-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
568; SSE-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
569; SSE-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
570; SSE-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
571; SSE-NEXT:    store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
572; SSE-NEXT:    store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
573; SSE-NEXT:    store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
574; SSE-NEXT:    store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
575; SSE-NEXT:    ret void
576;
577; AVX1-LABEL: @ctlz_undef_8i32(
578; AVX1-NEXT:    [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
579; AVX1-NEXT:    [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
580; AVX1-NEXT:    [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
581; AVX1-NEXT:    [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
582; AVX1-NEXT:    [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
583; AVX1-NEXT:    [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
584; AVX1-NEXT:    [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
585; AVX1-NEXT:    [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
586; AVX1-NEXT:    [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
587; AVX1-NEXT:    [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
588; AVX1-NEXT:    [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
589; AVX1-NEXT:    [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
590; AVX1-NEXT:    [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true)
591; AVX1-NEXT:    [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true)
592; AVX1-NEXT:    [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true)
593; AVX1-NEXT:    [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true)
594; AVX1-NEXT:    store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
595; AVX1-NEXT:    store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
596; AVX1-NEXT:    store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
597; AVX1-NEXT:    store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
598; AVX1-NEXT:    store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
599; AVX1-NEXT:    store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
600; AVX1-NEXT:    store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
601; AVX1-NEXT:    store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
602; AVX1-NEXT:    ret void
603;
604; AVX2-LABEL: @ctlz_undef_8i32(
605; AVX2-NEXT:    [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
606; AVX2-NEXT:    [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 true)
607; AVX2-NEXT:    store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
608; AVX2-NEXT:    ret void
609;
610  %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
611  %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
612  %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
613  %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
614  %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
615  %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
616  %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
617  %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
618  %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 -1)
619  %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 -1)
620  %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 -1)
621  %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 -1)
622  %ctlz4 = call i32 @llvm.ctlz.i32(i32 %ld4, i1 -1)
623  %ctlz5 = call i32 @llvm.ctlz.i32(i32 %ld5, i1 -1)
624  %ctlz6 = call i32 @llvm.ctlz.i32(i32 %ld6, i1 -1)
625  %ctlz7 = call i32 @llvm.ctlz.i32(i32 %ld7, i1 -1)
626  store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
627  store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
628  store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
629  store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
630  store i32 %ctlz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
631  store i32 %ctlz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
632  store i32 %ctlz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
633  store i32 %ctlz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
634  ret void
635}
636
637define void @ctlz_undef_8i16() #0 {
638; CHECK-LABEL: @ctlz_undef_8i16(
639; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
640; CHECK-NEXT:    [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true)
641; CHECK-NEXT:    store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
642; CHECK-NEXT:    ret void
643;
644  %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
645  %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
646  %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
647  %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
648  %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
649  %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
650  %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
651  %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
652  %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 -1)
653  %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 -1)
654  %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 -1)
655  %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 -1)
656  %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 -1)
657  %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 -1)
658  %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 -1)
659  %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 -1)
660  store i16 %ctlz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
661  store i16 %ctlz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
662  store i16 %ctlz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
663  store i16 %ctlz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
664  store i16 %ctlz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
665  store i16 %ctlz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
666  store i16 %ctlz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
667  store i16 %ctlz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
668  ret void
669}
670
671define void @ctlz_undef_16i16() #0 {
672; SSE-LABEL: @ctlz_undef_16i16(
673; SSE-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
674; SSE-NEXT:    [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
675; SSE-NEXT:    [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true)
676; SSE-NEXT:    [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 true)
677; SSE-NEXT:    store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
678; SSE-NEXT:    store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
679; SSE-NEXT:    ret void
680;
681; AVX-LABEL: @ctlz_undef_16i16(
682; AVX-NEXT:    [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
683; AVX-NEXT:    [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 true)
684; AVX-NEXT:    store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
685; AVX-NEXT:    ret void
686;
687  %ld0  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  0), align 2
688  %ld1  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  1), align 2
689  %ld2  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  2), align 2
690  %ld3  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  3), align 2
691  %ld4  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  4), align 2
692  %ld5  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  5), align 2
693  %ld6  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  6), align 2
694  %ld7  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  7), align 2
695  %ld8  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  8), align 2
696  %ld9  = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64  9), align 2
697  %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
698  %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
699  %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
700  %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
701  %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
702  %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
703  %ctlz0  = call i16 @llvm.ctlz.i16(i16 %ld0, i1 -1)
704  %ctlz1  = call i16 @llvm.ctlz.i16(i16 %ld1, i1 -1)
705  %ctlz2  = call i16 @llvm.ctlz.i16(i16 %ld2, i1 -1)
706  %ctlz3  = call i16 @llvm.ctlz.i16(i16 %ld3, i1 -1)
707  %ctlz4  = call i16 @llvm.ctlz.i16(i16 %ld4, i1 -1)
708  %ctlz5  = call i16 @llvm.ctlz.i16(i16 %ld5, i1 -1)
709  %ctlz6  = call i16 @llvm.ctlz.i16(i16 %ld6, i1 -1)
710  %ctlz7  = call i16 @llvm.ctlz.i16(i16 %ld7, i1 -1)
711  %ctlz8  = call i16 @llvm.ctlz.i16(i16 %ld8, i1 -1)
712  %ctlz9  = call i16 @llvm.ctlz.i16(i16 %ld9, i1 -1)
713  %ctlz10 = call i16 @llvm.ctlz.i16(i16 %ld10, i1 -1)
714  %ctlz11 = call i16 @llvm.ctlz.i16(i16 %ld11, i1 -1)
715  %ctlz12 = call i16 @llvm.ctlz.i16(i16 %ld12, i1 -1)
716  %ctlz13 = call i16 @llvm.ctlz.i16(i16 %ld13, i1 -1)
717  %ctlz14 = call i16 @llvm.ctlz.i16(i16 %ld14, i1 -1)
718  %ctlz15 = call i16 @llvm.ctlz.i16(i16 %ld15, i1 -1)
719  store i16 %ctlz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  0), align 2
720  store i16 %ctlz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  1), align 2
721  store i16 %ctlz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  2), align 2
722  store i16 %ctlz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  3), align 2
723  store i16 %ctlz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  4), align 2
724  store i16 %ctlz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  5), align 2
725  store i16 %ctlz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  6), align 2
726  store i16 %ctlz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  7), align 2
727  store i16 %ctlz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  8), align 2
728  store i16 %ctlz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64  9), align 2
729  store i16 %ctlz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
730  store i16 %ctlz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
731  store i16 %ctlz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
732  store i16 %ctlz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
733  store i16 %ctlz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
734  store i16 %ctlz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
735  ret void
736}
737
738define void @ctlz_undef_16i8() #0 {
739; CHECK-LABEL: @ctlz_undef_16i8(
740; CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
741; CHECK-NEXT:    [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true)
742; CHECK-NEXT:    store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
743; CHECK-NEXT:    ret void
744;
745  %ld0  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  0), align 1
746  %ld1  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  1), align 1
747  %ld2  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  2), align 1
748  %ld3  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  3), align 1
749  %ld4  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  4), align 1
750  %ld5  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  5), align 1
751  %ld6  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  6), align 1
752  %ld7  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  7), align 1
753  %ld8  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  8), align 1
754  %ld9  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  9), align 1
755  %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
756  %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
757  %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
758  %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
759  %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
760  %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
761  %ctlz0  = call i8 @llvm.ctlz.i8(i8 %ld0, i1 -1)
762  %ctlz1  = call i8 @llvm.ctlz.i8(i8 %ld1, i1 -1)
763  %ctlz2  = call i8 @llvm.ctlz.i8(i8 %ld2, i1 -1)
764  %ctlz3  = call i8 @llvm.ctlz.i8(i8 %ld3, i1 -1)
765  %ctlz4  = call i8 @llvm.ctlz.i8(i8 %ld4, i1 -1)
766  %ctlz5  = call i8 @llvm.ctlz.i8(i8 %ld5, i1 -1)
767  %ctlz6  = call i8 @llvm.ctlz.i8(i8 %ld6, i1 -1)
768  %ctlz7  = call i8 @llvm.ctlz.i8(i8 %ld7, i1 -1)
769  %ctlz8  = call i8 @llvm.ctlz.i8(i8 %ld8, i1 -1)
770  %ctlz9  = call i8 @llvm.ctlz.i8(i8 %ld9, i1 -1)
771  %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 -1)
772  %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 -1)
773  %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 -1)
774  %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 -1)
775  %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 -1)
776  %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 -1)
777  store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  0), align 1
778  store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  1), align 1
779  store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  2), align 1
780  store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  3), align 1
781  store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  4), align 1
782  store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  5), align 1
783  store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  6), align 1
784  store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  7), align 1
785  store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  8), align 1
786  store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  9), align 1
787  store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
788  store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
789  store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
790  store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
791  store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
792  store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
793  ret void
794}
795
796define void @ctlz_undef_32i8() #0 {
797; CHECK-LABEL: @ctlz_undef_32i8(
798; CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
799; CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
800; CHECK-NEXT:    [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true)
801; CHECK-NEXT:    [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 true)
802; CHECK-NEXT:    store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
803; CHECK-NEXT:    store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
804; CHECK-NEXT:    ret void
805;
806  %ld0  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  0), align 1
807  %ld1  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  1), align 1
808  %ld2  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  2), align 1
809  %ld3  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  3), align 1
810  %ld4  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  4), align 1
811  %ld5  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  5), align 1
812  %ld6  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  6), align 1
813  %ld7  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  7), align 1
814  %ld8  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  8), align 1
815  %ld9  = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64  9), align 1
816  %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
817  %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
818  %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
819  %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
820  %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
821  %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
822  %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
823  %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
824  %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
825  %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
826  %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
827  %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
828  %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
829  %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
830  %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
831  %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
832  %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
833  %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
834  %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
835  %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
836  %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
837  %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
838  %ctlz0  = call i8 @llvm.ctlz.i8(i8 %ld0, i1 -1)
839  %ctlz1  = call i8 @llvm.ctlz.i8(i8 %ld1, i1 -1)
840  %ctlz2  = call i8 @llvm.ctlz.i8(i8 %ld2, i1 -1)
841  %ctlz3  = call i8 @llvm.ctlz.i8(i8 %ld3, i1 -1)
842  %ctlz4  = call i8 @llvm.ctlz.i8(i8 %ld4, i1 -1)
843  %ctlz5  = call i8 @llvm.ctlz.i8(i8 %ld5, i1 -1)
844  %ctlz6  = call i8 @llvm.ctlz.i8(i8 %ld6, i1 -1)
845  %ctlz7  = call i8 @llvm.ctlz.i8(i8 %ld7, i1 -1)
846  %ctlz8  = call i8 @llvm.ctlz.i8(i8 %ld8, i1 -1)
847  %ctlz9  = call i8 @llvm.ctlz.i8(i8 %ld9, i1 -1)
848  %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 -1)
849  %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 -1)
850  %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 -1)
851  %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 -1)
852  %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 -1)
853  %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 -1)
854  %ctlz16 = call i8 @llvm.ctlz.i8(i8 %ld16, i1 -1)
855  %ctlz17 = call i8 @llvm.ctlz.i8(i8 %ld17, i1 -1)
856  %ctlz18 = call i8 @llvm.ctlz.i8(i8 %ld18, i1 -1)
857  %ctlz19 = call i8 @llvm.ctlz.i8(i8 %ld19, i1 -1)
858  %ctlz20 = call i8 @llvm.ctlz.i8(i8 %ld20, i1 -1)
859  %ctlz21 = call i8 @llvm.ctlz.i8(i8 %ld21, i1 -1)
860  %ctlz22 = call i8 @llvm.ctlz.i8(i8 %ld22, i1 -1)
861  %ctlz23 = call i8 @llvm.ctlz.i8(i8 %ld23, i1 -1)
862  %ctlz24 = call i8 @llvm.ctlz.i8(i8 %ld24, i1 -1)
863  %ctlz25 = call i8 @llvm.ctlz.i8(i8 %ld25, i1 -1)
864  %ctlz26 = call i8 @llvm.ctlz.i8(i8 %ld26, i1 -1)
865  %ctlz27 = call i8 @llvm.ctlz.i8(i8 %ld27, i1 -1)
866  %ctlz28 = call i8 @llvm.ctlz.i8(i8 %ld28, i1 -1)
867  %ctlz29 = call i8 @llvm.ctlz.i8(i8 %ld29, i1 -1)
868  %ctlz30 = call i8 @llvm.ctlz.i8(i8 %ld30, i1 -1)
869  %ctlz31 = call i8 @llvm.ctlz.i8(i8 %ld31, i1 -1)
870  store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  0), align 1
871  store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  1), align 1
872  store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  2), align 1
873  store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  3), align 1
874  store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  4), align 1
875  store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  5), align 1
876  store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  6), align 1
877  store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  7), align 1
878  store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  8), align 1
879  store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64  9), align 1
880  store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
881  store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
882  store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
883  store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
884  store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
885  store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
886  store i8 %ctlz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
887  store i8 %ctlz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
888  store i8 %ctlz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
889  store i8 %ctlz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
890  store i8 %ctlz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
891  store i8 %ctlz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
892  store i8 %ctlz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
893  store i8 %ctlz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
894  store i8 %ctlz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
895  store i8 %ctlz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
896  store i8 %ctlz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
897  store i8 %ctlz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
898  store i8 %ctlz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
899  store i8 %ctlz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
900  store i8 %ctlz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
901  store i8 %ctlz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
902  ret void
903}
904
905attributes #0 = { nounwind }
906