| blob | 92ff1fbaae86d72042b241931e5b12b193a7c768 |
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
3 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=AVX256NODQ
4 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver1 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=AVX256NODQ
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=AVX256 --check-prefix=AVX256NODQ
6 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512
7 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=AVX256DQ
9 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
11 @src64 = common global [8 x i64] zeroinitializer, align 64
12 @src32 = common global [16 x i32] zeroinitializer, align 64
13 @src16 = common global [32 x i16] zeroinitializer, align 64
14 @src8 = common global [64 x i8] zeroinitializer, align 64
16 @dst64 = common global [8 x double] zeroinitializer, align 64
17 @dst32 = common global [16 x float] zeroinitializer, align 64
19 ;
20 ; UITOFP to vXf64
21 ;
23 define void @uitofp_2i64_2f64() #0 {
24 ; CHECK-LABEL: @uitofp_2i64_2f64(
25 ; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
26 ; CHECK-NEXT: [[TMP2:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x double>
27 ; CHECK-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
28 ; CHECK-NEXT: ret void
29 ;
30 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
31 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
32 %cvt0 = uitofp i64 %ld0 to double
33 %cvt1 = uitofp i64 %ld1 to double
34 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
35 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
36 ret void
37 }
39 define void @uitofp_4i64_4f64() #0 {
40 ; SSE-LABEL: @uitofp_4i64_4f64(
41 ; SSE-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
42 ; SSE-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2) to <2 x i64>*), align 16
43 ; SSE-NEXT: [[TMP3:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x double>
44 ; SSE-NEXT: [[TMP4:%.*]] = uitofp <2 x i64> [[TMP2]] to <2 x double>
45 ; SSE-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
46 ; SSE-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 16
47 ; SSE-NEXT: ret void
48 ;
49 ; AVX-LABEL: @uitofp_4i64_4f64(
50 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64
51 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double>
52 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
53 ; AVX-NEXT: ret void
54 ;
55 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
56 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
57 %ld2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
58 %ld3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
59 %cvt0 = uitofp i64 %ld0 to double
60 %cvt1 = uitofp i64 %ld1 to double
61 %cvt2 = uitofp i64 %ld2 to double
62 %cvt3 = uitofp i64 %ld3 to double
63 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
64 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
65 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
66 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
67 ret void
68 }
70 define void @uitofp_8i64_8f64() #0 {
71 ; SSE-LABEL: @uitofp_8i64_8f64(
72 ; SSE-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
73 ; SSE-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2) to <2 x i64>*), align 16
74 ; SSE-NEXT: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4) to <2 x i64>*), align 32
75 ; SSE-NEXT: [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6) to <2 x i64>*), align 16
76 ; SSE-NEXT: [[TMP5:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x double>
77 ; SSE-NEXT: [[TMP6:%.*]] = uitofp <2 x i64> [[TMP2]] to <2 x double>
78 ; SSE-NEXT: [[TMP7:%.*]] = uitofp <2 x i64> [[TMP3]] to <2 x double>
79 ; SSE-NEXT: [[TMP8:%.*]] = uitofp <2 x i64> [[TMP4]] to <2 x double>
80 ; SSE-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
81 ; SSE-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 16
82 ; SSE-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 32
83 ; SSE-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 16
84 ; SSE-NEXT: ret void
85 ;
86 ; AVX256-LABEL: @uitofp_8i64_8f64(
87 ; AVX256-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64
88 ; AVX256-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4) to <4 x i64>*), align 32
89 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double>
90 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <4 x i64> [[TMP2]] to <4 x double>
91 ; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
92 ; AVX256-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32
93 ; AVX256-NEXT: ret void
94 ;
95 ; AVX512-LABEL: @uitofp_8i64_8f64(
96 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @src64 to <8 x i64>*), align 64
97 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <8 x i64> [[TMP1]] to <8 x double>
98 ; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 64
99 ; AVX512-NEXT: ret void
100 ;
101 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
102 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
103 %ld2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
104 %ld3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
105 %ld4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32
106 %ld5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8
107 %ld6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16
108 %ld7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8
109 %cvt0 = uitofp i64 %ld0 to double
110 %cvt1 = uitofp i64 %ld1 to double
111 %cvt2 = uitofp i64 %ld2 to double
112 %cvt3 = uitofp i64 %ld3 to double
113 %cvt4 = uitofp i64 %ld4 to double
114 %cvt5 = uitofp i64 %ld5 to double
115 %cvt6 = uitofp i64 %ld6 to double
116 %cvt7 = uitofp i64 %ld7 to double
117 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
118 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
119 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
120 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
121 store double %cvt4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
122 store double %cvt5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
123 store double %cvt6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
124 store double %cvt7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
125 ret void
126 }
128 define void @uitofp_2i32_2f64() #0 {
129 ; SSE-LABEL: @uitofp_2i32_2f64(
130 ; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
131 ; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
132 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i32 [[LD0]] to double
133 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i32 [[LD1]] to double
134 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
135 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
136 ; SSE-NEXT: ret void
137 ;
138 ; AVX256NODQ-LABEL: @uitofp_2i32_2f64(
139 ; AVX256NODQ-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
140 ; AVX256NODQ-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
141 ; AVX256NODQ-NEXT: [[CVT0:%.*]] = uitofp i32 [[LD0]] to double
142 ; AVX256NODQ-NEXT: [[CVT1:%.*]] = uitofp i32 [[LD1]] to double
143 ; AVX256NODQ-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
144 ; AVX256NODQ-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
145 ; AVX256NODQ-NEXT: ret void
146 ;
147 ; AVX512-LABEL: @uitofp_2i32_2f64(
148 ; AVX512-NEXT: [[TMP1:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([16 x i32]* @src32 to <2 x i32>*), align 64
149 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <2 x i32> [[TMP1]] to <2 x double>
150 ; AVX512-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
151 ; AVX512-NEXT: ret void
152 ;
153 ; AVX256DQ-LABEL: @uitofp_2i32_2f64(
154 ; AVX256DQ-NEXT: [[TMP1:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([16 x i32]* @src32 to <2 x i32>*), align 64
155 ; AVX256DQ-NEXT: [[TMP2:%.*]] = uitofp <2 x i32> [[TMP1]] to <2 x double>
156 ; AVX256DQ-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
157 ; AVX256DQ-NEXT: ret void
158 ;
159 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
160 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
161 %cvt0 = uitofp i32 %ld0 to double
162 %cvt1 = uitofp i32 %ld1 to double
163 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
164 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
165 ret void
166 }
168 define void @uitofp_4i32_4f64() #0 {
169 ; SSE-LABEL: @uitofp_4i32_4f64(
170 ; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
171 ; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
172 ; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
173 ; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
174 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i32 [[LD0]] to double
175 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i32 [[LD1]] to double
176 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i32 [[LD2]] to double
177 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i32 [[LD3]] to double
178 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
179 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
180 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
181 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
182 ; SSE-NEXT: ret void
183 ;
184 ; AVX-LABEL: @uitofp_4i32_4f64(
185 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @src32 to <4 x i32>*), align 64
186 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x double>
187 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
188 ; AVX-NEXT: ret void
189 ;
190 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
191 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
192 %ld2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
193 %ld3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
194 %cvt0 = uitofp i32 %ld0 to double
195 %cvt1 = uitofp i32 %ld1 to double
196 %cvt2 = uitofp i32 %ld2 to double
197 %cvt3 = uitofp i32 %ld3 to double
198 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
199 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
200 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
201 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
202 ret void
203 }
205 define void @uitofp_8i32_8f64() #0 {
206 ; SSE-LABEL: @uitofp_8i32_8f64(
207 ; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
208 ; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
209 ; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
210 ; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
211 ; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4), align 16
212 ; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 5), align 4
213 ; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 6), align 8
214 ; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 7), align 4
215 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i32 [[LD0]] to double
216 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i32 [[LD1]] to double
217 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i32 [[LD2]] to double
218 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i32 [[LD3]] to double
219 ; SSE-NEXT: [[CVT4:%.*]] = uitofp i32 [[LD4]] to double
220 ; SSE-NEXT: [[CVT5:%.*]] = uitofp i32 [[LD5]] to double
221 ; SSE-NEXT: [[CVT6:%.*]] = uitofp i32 [[LD6]] to double
222 ; SSE-NEXT: [[CVT7:%.*]] = uitofp i32 [[LD7]] to double
223 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
224 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
225 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
226 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
227 ; SSE-NEXT: store double [[CVT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
228 ; SSE-NEXT: store double [[CVT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
229 ; SSE-NEXT: store double [[CVT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
230 ; SSE-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
231 ; SSE-NEXT: ret void
232 ;
233 ; AVX256-LABEL: @uitofp_8i32_8f64(
234 ; AVX256-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @src32 to <4 x i32>*), align 64
235 ; AVX256-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 16
236 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x double>
237 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <4 x i32> [[TMP2]] to <4 x double>
238 ; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
239 ; AVX256-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32
240 ; AVX256-NEXT: ret void
241 ;
242 ; AVX512-LABEL: @uitofp_8i32_8f64(
243 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @src32 to <8 x i32>*), align 64
244 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <8 x i32> [[TMP1]] to <8 x double>
245 ; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 64
246 ; AVX512-NEXT: ret void
247 ;
248 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
249 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
250 %ld2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
251 %ld3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
252 %ld4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4), align 16
253 %ld5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 5), align 4
254 %ld6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 6), align 8
255 %ld7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 7), align 4
256 %cvt0 = uitofp i32 %ld0 to double
257 %cvt1 = uitofp i32 %ld1 to double
258 %cvt2 = uitofp i32 %ld2 to double
259 %cvt3 = uitofp i32 %ld3 to double
260 %cvt4 = uitofp i32 %ld4 to double
261 %cvt5 = uitofp i32 %ld5 to double
262 %cvt6 = uitofp i32 %ld6 to double
263 %cvt7 = uitofp i32 %ld7 to double
264 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
265 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
266 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
267 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
268 store double %cvt4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
269 store double %cvt5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
270 store double %cvt6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
271 store double %cvt7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
272 ret void
273 }
275 define void @uitofp_2i16_2f64() #0 {
276 ; CHECK-LABEL: @uitofp_2i16_2f64(
277 ; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
278 ; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
279 ; CHECK-NEXT: [[CVT0:%.*]] = uitofp i16 [[LD0]] to double
280 ; CHECK-NEXT: [[CVT1:%.*]] = uitofp i16 [[LD1]] to double
281 ; CHECK-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
282 ; CHECK-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
283 ; CHECK-NEXT: ret void
284 ;
285 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
286 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
287 %cvt0 = uitofp i16 %ld0 to double
288 %cvt1 = uitofp i16 %ld1 to double
289 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
290 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
291 ret void
292 }
294 define void @uitofp_4i16_4f64() #0 {
295 ; SSE-LABEL: @uitofp_4i16_4f64(
296 ; SSE-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
297 ; SSE-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
298 ; SSE-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
299 ; SSE-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
300 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i16 [[LD0]] to double
301 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i16 [[LD1]] to double
302 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i16 [[LD2]] to double
303 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i16 [[LD3]] to double
304 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
305 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
306 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
307 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
308 ; SSE-NEXT: ret void
309 ;
310 ; AVX-LABEL: @uitofp_4i16_4f64(
311 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* bitcast ([32 x i16]* @src16 to <4 x i16>*), align 64
312 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <4 x i16> [[TMP1]] to <4 x double>
313 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
314 ; AVX-NEXT: ret void
315 ;
316 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
317 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
318 %ld2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
319 %ld3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
320 %cvt0 = uitofp i16 %ld0 to double
321 %cvt1 = uitofp i16 %ld1 to double
322 %cvt2 = uitofp i16 %ld2 to double
323 %cvt3 = uitofp i16 %ld3 to double
324 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
325 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
326 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
327 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
328 ret void
329 }
331 define void @uitofp_8i16_8f64() #0 {
332 ; SSE-LABEL: @uitofp_8i16_8f64(
333 ; SSE-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
334 ; SSE-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
335 ; SSE-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
336 ; SSE-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
337 ; SSE-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4), align 8
338 ; SSE-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 5), align 2
339 ; SSE-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 6), align 4
340 ; SSE-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 7), align 2
341 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i16 [[LD0]] to double
342 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i16 [[LD1]] to double
343 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i16 [[LD2]] to double
344 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i16 [[LD3]] to double
345 ; SSE-NEXT: [[CVT4:%.*]] = uitofp i16 [[LD4]] to double
346 ; SSE-NEXT: [[CVT5:%.*]] = uitofp i16 [[LD5]] to double
347 ; SSE-NEXT: [[CVT6:%.*]] = uitofp i16 [[LD6]] to double
348 ; SSE-NEXT: [[CVT7:%.*]] = uitofp i16 [[LD7]] to double
349 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
350 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
351 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
352 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
353 ; SSE-NEXT: store double [[CVT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
354 ; SSE-NEXT: store double [[CVT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
355 ; SSE-NEXT: store double [[CVT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
356 ; SSE-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
357 ; SSE-NEXT: ret void
358 ;
359 ; AVX256-LABEL: @uitofp_8i16_8f64(
360 ; AVX256-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* bitcast ([32 x i16]* @src16 to <4 x i16>*), align 64
361 ; AVX256-NEXT: [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4) to <4 x i16>*), align 8
362 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <4 x i16> [[TMP1]] to <4 x double>
363 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <4 x i16> [[TMP2]] to <4 x double>
364 ; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
365 ; AVX256-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32
366 ; AVX256-NEXT: ret void
367 ;
368 ; AVX512-LABEL: @uitofp_8i16_8f64(
369 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @src16 to <8 x i16>*), align 64
370 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <8 x i16> [[TMP1]] to <8 x double>
371 ; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 64
372 ; AVX512-NEXT: ret void
373 ;
374 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
375 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
376 %ld2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
377 %ld3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
378 %ld4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4), align 8
379 %ld5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 5), align 2
380 %ld6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 6), align 4
381 %ld7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 7), align 2
382 %cvt0 = uitofp i16 %ld0 to double
383 %cvt1 = uitofp i16 %ld1 to double
384 %cvt2 = uitofp i16 %ld2 to double
385 %cvt3 = uitofp i16 %ld3 to double
386 %cvt4 = uitofp i16 %ld4 to double
387 %cvt5 = uitofp i16 %ld5 to double
388 %cvt6 = uitofp i16 %ld6 to double
389 %cvt7 = uitofp i16 %ld7 to double
390 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
391 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
392 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
393 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
394 store double %cvt4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
395 store double %cvt5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
396 store double %cvt6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
397 store double %cvt7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
398 ret void
399 }
401 define void @uitofp_2i8_2f64() #0 {
402 ; SSE-LABEL: @uitofp_2i8_2f64(
403 ; SSE-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
404 ; SSE-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
405 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i8 [[LD0]] to double
406 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i8 [[LD1]] to double
407 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
408 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
409 ; SSE-NEXT: ret void
410 ;
411 ; AVX256NODQ-LABEL: @uitofp_2i8_2f64(
412 ; AVX256NODQ-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
413 ; AVX256NODQ-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
414 ; AVX256NODQ-NEXT: [[CVT0:%.*]] = uitofp i8 [[LD0]] to double
415 ; AVX256NODQ-NEXT: [[CVT1:%.*]] = uitofp i8 [[LD1]] to double
416 ; AVX256NODQ-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
417 ; AVX256NODQ-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
418 ; AVX256NODQ-NEXT: ret void
419 ;
420 ; AVX512-LABEL: @uitofp_2i8_2f64(
421 ; AVX512-NEXT: [[TMP1:%.*]] = load <2 x i8>, <2 x i8>* bitcast ([64 x i8]* @src8 to <2 x i8>*), align 64
422 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <2 x i8> [[TMP1]] to <2 x double>
423 ; AVX512-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
424 ; AVX512-NEXT: ret void
425 ;
426 ; AVX256DQ-LABEL: @uitofp_2i8_2f64(
427 ; AVX256DQ-NEXT: [[TMP1:%.*]] = load <2 x i8>, <2 x i8>* bitcast ([64 x i8]* @src8 to <2 x i8>*), align 64
428 ; AVX256DQ-NEXT: [[TMP2:%.*]] = uitofp <2 x i8> [[TMP1]] to <2 x double>
429 ; AVX256DQ-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 64
430 ; AVX256DQ-NEXT: ret void
431 ;
432 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
433 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
434 %cvt0 = uitofp i8 %ld0 to double
435 %cvt1 = uitofp i8 %ld1 to double
436 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
437 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
438 ret void
439 }
441 define void @uitofp_4i8_4f64() #0 {
442 ; SSE-LABEL: @uitofp_4i8_4f64(
443 ; SSE-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
444 ; SSE-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
445 ; SSE-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
446 ; SSE-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
447 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i8 [[LD0]] to double
448 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i8 [[LD1]] to double
449 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i8 [[LD2]] to double
450 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i8 [[LD3]] to double
451 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
452 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
453 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
454 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
455 ; SSE-NEXT: ret void
456 ;
457 ; AVX-LABEL: @uitofp_4i8_4f64(
458 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* bitcast ([64 x i8]* @src8 to <4 x i8>*), align 64
459 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <4 x i8> [[TMP1]] to <4 x double>
460 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
461 ; AVX-NEXT: ret void
462 ;
463 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
464 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
465 %ld2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
466 %ld3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
467 %cvt0 = uitofp i8 %ld0 to double
468 %cvt1 = uitofp i8 %ld1 to double
469 %cvt2 = uitofp i8 %ld2 to double
470 %cvt3 = uitofp i8 %ld3 to double
471 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
472 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
473 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
474 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
475 ret void
476 }
478 define void @uitofp_8i8_8f64() #0 {
479 ; SSE-LABEL: @uitofp_8i8_8f64(
480 ; SSE-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
481 ; SSE-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
482 ; SSE-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
483 ; SSE-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
484 ; SSE-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4), align 4
485 ; SSE-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 5), align 1
486 ; SSE-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 6), align 2
487 ; SSE-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 7), align 1
488 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i8 [[LD0]] to double
489 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i8 [[LD1]] to double
490 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i8 [[LD2]] to double
491 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i8 [[LD3]] to double
492 ; SSE-NEXT: [[CVT4:%.*]] = uitofp i8 [[LD4]] to double
493 ; SSE-NEXT: [[CVT5:%.*]] = uitofp i8 [[LD5]] to double
494 ; SSE-NEXT: [[CVT6:%.*]] = uitofp i8 [[LD6]] to double
495 ; SSE-NEXT: [[CVT7:%.*]] = uitofp i8 [[LD7]] to double
496 ; SSE-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
497 ; SSE-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
498 ; SSE-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
499 ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
500 ; SSE-NEXT: store double [[CVT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
501 ; SSE-NEXT: store double [[CVT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
502 ; SSE-NEXT: store double [[CVT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
503 ; SSE-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
504 ; SSE-NEXT: ret void
505 ;
506 ; AVX256-LABEL: @uitofp_8i8_8f64(
507 ; AVX256-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* bitcast ([64 x i8]* @src8 to <4 x i8>*), align 64
508 ; AVX256-NEXT: [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4) to <4 x i8>*), align 4
509 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <4 x i8> [[TMP1]] to <4 x double>
510 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <4 x i8> [[TMP2]] to <4 x double>
511 ; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64
512 ; AVX256-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32
513 ; AVX256-NEXT: ret void
514 ;
515 ; AVX512-LABEL: @uitofp_8i8_8f64(
516 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i8>, <8 x i8>* bitcast ([64 x i8]* @src8 to <8 x i8>*), align 64
517 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <8 x i8> [[TMP1]] to <8 x double>
518 ; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 64
519 ; AVX512-NEXT: ret void
520 ;
521 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
522 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
523 %ld2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
524 %ld3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
525 %ld4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4), align 4
526 %ld5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 5), align 1
527 %ld6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 6), align 2
528 %ld7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 7), align 1
529 %cvt0 = uitofp i8 %ld0 to double
530 %cvt1 = uitofp i8 %ld1 to double
531 %cvt2 = uitofp i8 %ld2 to double
532 %cvt3 = uitofp i8 %ld3 to double
533 %cvt4 = uitofp i8 %ld4 to double
534 %cvt5 = uitofp i8 %ld5 to double
535 %cvt6 = uitofp i8 %ld6 to double
536 %cvt7 = uitofp i8 %ld7 to double
537 store double %cvt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64
538 store double %cvt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
539 store double %cvt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16
540 store double %cvt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
541 store double %cvt4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32
542 store double %cvt5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
543 store double %cvt6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16
544 store double %cvt7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
545 ret void
546 }
548 ;
549 ; UITOFP to vXf32
550 ;
552 define void @uitofp_2i64_2f32() #0 {
553 ; SSE-LABEL: @uitofp_2i64_2f32(
554 ; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
555 ; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
556 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
557 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
558 ; SSE-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
559 ; SSE-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
560 ; SSE-NEXT: ret void
561 ;
562 ; AVX256NODQ-LABEL: @uitofp_2i64_2f32(
563 ; AVX256NODQ-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
564 ; AVX256NODQ-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
565 ; AVX256NODQ-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
566 ; AVX256NODQ-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
567 ; AVX256NODQ-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
568 ; AVX256NODQ-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
569 ; AVX256NODQ-NEXT: ret void
570 ;
571 ; AVX512-LABEL: @uitofp_2i64_2f32(
572 ; AVX512-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
573 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x float>
574 ; AVX512-NEXT: store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
575 ; AVX512-NEXT: ret void
576 ;
577 ; AVX256DQ-LABEL: @uitofp_2i64_2f32(
578 ; AVX256DQ-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
579 ; AVX256DQ-NEXT: [[TMP2:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x float>
580 ; AVX256DQ-NEXT: store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
581 ; AVX256DQ-NEXT: ret void
582 ;
583 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
584 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
585 %cvt0 = uitofp i64 %ld0 to float
586 %cvt1 = uitofp i64 %ld1 to float
587 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
588 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
589 ret void
590 }
592 define void @uitofp_4i64_4f32() #0 {
593 ; SSE-LABEL: @uitofp_4i64_4f32(
594 ; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
595 ; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
596 ; SSE-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
597 ; SSE-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
598 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
599 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
600 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to float
601 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to float
602 ; SSE-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
603 ; SSE-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
604 ; SSE-NEXT: store float [[CVT2]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
605 ; SSE-NEXT: store float [[CVT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
606 ; SSE-NEXT: ret void
607 ;
608 ; AVX256NODQ-LABEL: @uitofp_4i64_4f32(
609 ; AVX256NODQ-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
610 ; AVX256NODQ-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
611 ; AVX256NODQ-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
612 ; AVX256NODQ-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
613 ; AVX256NODQ-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
614 ; AVX256NODQ-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
615 ; AVX256NODQ-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to float
616 ; AVX256NODQ-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to float
617 ; AVX256NODQ-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
618 ; AVX256NODQ-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
619 ; AVX256NODQ-NEXT: store float [[CVT2]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
620 ; AVX256NODQ-NEXT: store float [[CVT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
621 ; AVX256NODQ-NEXT: ret void
622 ;
623 ; AVX512-LABEL: @uitofp_4i64_4f32(
624 ; AVX512-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64
625 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x float>
626 ; AVX512-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
627 ; AVX512-NEXT: ret void
628 ;
629 ; AVX256DQ-LABEL: @uitofp_4i64_4f32(
630 ; AVX256DQ-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64
631 ; AVX256DQ-NEXT: [[TMP2:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x float>
632 ; AVX256DQ-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
633 ; AVX256DQ-NEXT: ret void
634 ;
635 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
636 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
637 %ld2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
638 %ld3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
639 %cvt0 = uitofp i64 %ld0 to float
640 %cvt1 = uitofp i64 %ld1 to float
641 %cvt2 = uitofp i64 %ld2 to float
642 %cvt3 = uitofp i64 %ld3 to float
643 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
644 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
645 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
646 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
647 ret void
648 }
650 define void @uitofp_8i64_8f32() #0 {
651 ; SSE-LABEL: @uitofp_8i64_8f32(
652 ; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
653 ; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
654 ; SSE-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
655 ; SSE-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
656 ; SSE-NEXT: [[LD4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32
657 ; SSE-NEXT: [[LD5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8
658 ; SSE-NEXT: [[LD6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16
659 ; SSE-NEXT: [[LD7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8
660 ; SSE-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
661 ; SSE-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
662 ; SSE-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to float
663 ; SSE-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to float
664 ; SSE-NEXT: [[CVT4:%.*]] = uitofp i64 [[LD4]] to float
665 ; SSE-NEXT: [[CVT5:%.*]] = uitofp i64 [[LD5]] to float
666 ; SSE-NEXT: [[CVT6:%.*]] = uitofp i64 [[LD6]] to float
667 ; SSE-NEXT: [[CVT7:%.*]] = uitofp i64 [[LD7]] to float
668 ; SSE-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
669 ; SSE-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
670 ; SSE-NEXT: store float [[CVT2]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
671 ; SSE-NEXT: store float [[CVT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
672 ; SSE-NEXT: store float [[CVT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
673 ; SSE-NEXT: store float [[CVT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
674 ; SSE-NEXT: store float [[CVT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
675 ; SSE-NEXT: store float [[CVT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
676 ; SSE-NEXT: ret void
677 ;
678 ; AVX256NODQ-LABEL: @uitofp_8i64_8f32(
679 ; AVX256NODQ-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
680 ; AVX256NODQ-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
681 ; AVX256NODQ-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
682 ; AVX256NODQ-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
683 ; AVX256NODQ-NEXT: [[LD4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32
684 ; AVX256NODQ-NEXT: [[LD5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8
685 ; AVX256NODQ-NEXT: [[LD6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16
686 ; AVX256NODQ-NEXT: [[LD7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8
687 ; AVX256NODQ-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
688 ; AVX256NODQ-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
689 ; AVX256NODQ-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to float
690 ; AVX256NODQ-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to float
691 ; AVX256NODQ-NEXT: [[CVT4:%.*]] = uitofp i64 [[LD4]] to float
692 ; AVX256NODQ-NEXT: [[CVT5:%.*]] = uitofp i64 [[LD5]] to float
693 ; AVX256NODQ-NEXT: [[CVT6:%.*]] = uitofp i64 [[LD6]] to float
694 ; AVX256NODQ-NEXT: [[CVT7:%.*]] = uitofp i64 [[LD7]] to float
695 ; AVX256NODQ-NEXT: store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
696 ; AVX256NODQ-NEXT: store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
697 ; AVX256NODQ-NEXT: store float [[CVT2]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
698 ; AVX256NODQ-NEXT: store float [[CVT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
699 ; AVX256NODQ-NEXT: store float [[CVT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
700 ; AVX256NODQ-NEXT: store float [[CVT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
701 ; AVX256NODQ-NEXT: store float [[CVT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
702 ; AVX256NODQ-NEXT: store float [[CVT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
703 ; AVX256NODQ-NEXT: ret void
704 ;
705 ; AVX512-LABEL: @uitofp_8i64_8f32(
706 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @src64 to <8 x i64>*), align 64
707 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <8 x i64> [[TMP1]] to <8 x float>
708 ; AVX512-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
709 ; AVX512-NEXT: ret void
710 ;
711 ; AVX256DQ-LABEL: @uitofp_8i64_8f32(
712 ; AVX256DQ-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @src64 to <8 x i64>*), align 64
713 ; AVX256DQ-NEXT: [[TMP2:%.*]] = uitofp <8 x i64> [[TMP1]] to <8 x float>
714 ; AVX256DQ-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
715 ; AVX256DQ-NEXT: ret void
716 ;
717 %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
718 %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
719 %ld2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16
720 %ld3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8
721 %ld4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32
722 %ld5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8
723 %ld6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16
724 %ld7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8
725 %cvt0 = uitofp i64 %ld0 to float
726 %cvt1 = uitofp i64 %ld1 to float
727 %cvt2 = uitofp i64 %ld2 to float
728 %cvt3 = uitofp i64 %ld3 to float
729 %cvt4 = uitofp i64 %ld4 to float
730 %cvt5 = uitofp i64 %ld5 to float
731 %cvt6 = uitofp i64 %ld6 to float
732 %cvt7 = uitofp i64 %ld7 to float
733 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
734 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
735 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
736 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
737 store float %cvt4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
738 store float %cvt5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
739 store float %cvt6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
740 store float %cvt7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
741 ret void
742 }
744 define void @uitofp_4i32_4f32() #0 {
745 ; CHECK-LABEL: @uitofp_4i32_4f32(
746 ; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @src32 to <4 x i32>*), align 64
747 ; CHECK-NEXT: [[TMP2:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x float>
748 ; CHECK-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
749 ; CHECK-NEXT: ret void
750 ;
751 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
752 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
753 %ld2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
754 %ld3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
755 %cvt0 = uitofp i32 %ld0 to float
756 %cvt1 = uitofp i32 %ld1 to float
757 %cvt2 = uitofp i32 %ld2 to float
758 %cvt3 = uitofp i32 %ld3 to float
759 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
760 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
761 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
762 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
763 ret void
764 }
766 define void @uitofp_8i32_8f32() #0 {
767 ; SSE-LABEL: @uitofp_8i32_8f32(
768 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @src32 to <4 x i32>*), align 64
769 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 16
770 ; SSE-NEXT: [[TMP3:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x float>
771 ; SSE-NEXT: [[TMP4:%.*]] = uitofp <4 x i32> [[TMP2]] to <4 x float>
772 ; SSE-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
773 ; SSE-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
774 ; SSE-NEXT: ret void
775 ;
776 ; AVX-LABEL: @uitofp_8i32_8f32(
777 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @src32 to <8 x i32>*), align 64
778 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <8 x i32> [[TMP1]] to <8 x float>
779 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
780 ; AVX-NEXT: ret void
781 ;
782 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0), align 64
783 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1), align 4
784 %ld2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2), align 8
785 %ld3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3), align 4
786 %ld4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4), align 16
787 %ld5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 5), align 4
788 %ld6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 6), align 8
789 %ld7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 7), align 4
790 %cvt0 = uitofp i32 %ld0 to float
791 %cvt1 = uitofp i32 %ld1 to float
792 %cvt2 = uitofp i32 %ld2 to float
793 %cvt3 = uitofp i32 %ld3 to float
794 %cvt4 = uitofp i32 %ld4 to float
795 %cvt5 = uitofp i32 %ld5 to float
796 %cvt6 = uitofp i32 %ld6 to float
797 %cvt7 = uitofp i32 %ld7 to float
798 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
799 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
800 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
801 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
802 store float %cvt4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
803 store float %cvt5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
804 store float %cvt6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
805 store float %cvt7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
806 ret void
807 }
809 define void @uitofp_16i32_16f32() #0 {
810 ; SSE-LABEL: @uitofp_16i32_16f32(
811 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @src32 to <4 x i32>*), align 64
812 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 16
813 ; SSE-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 8) to <4 x i32>*), align 32
814 ; SSE-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 12) to <4 x i32>*), align 16
815 ; SSE-NEXT: [[TMP5:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x float>
816 ; SSE-NEXT: [[TMP6:%.*]] = uitofp <4 x i32> [[TMP2]] to <4 x float>
817 ; SSE-NEXT: [[TMP7:%.*]] = uitofp <4 x i32> [[TMP3]] to <4 x float>
818 ; SSE-NEXT: [[TMP8:%.*]] = uitofp <4 x i32> [[TMP4]] to <4 x float>
819 ; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
820 ; SSE-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
821 ; SSE-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 32
822 ; SSE-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 16
823 ; SSE-NEXT: ret void
824 ;
825 ; AVX256-LABEL: @uitofp_16i32_16f32(
826 ; AVX256-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @src32 to <8 x i32>*), align 64
827 ; AVX256-NEXT: [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 8) to <8 x i32>*), align 32
828 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <8 x i32> [[TMP1]] to <8 x float>
829 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <8 x i32> [[TMP2]] to <8 x float>
830 ; AVX256-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
831 ; AVX256-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 32
832 ; AVX256-NEXT: ret void
833 ;
834 ; AVX512-LABEL: @uitofp_16i32_16f32(
835 ; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @src32 to <16 x i32>*), align 64
836 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <16 x i32> [[TMP1]] to <16 x float>
837 ; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 64
838 ; AVX512-NEXT: ret void
839 ;
840 %ld0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 0 ), align 64
841 %ld1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 1 ), align 4
842 %ld2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 2 ), align 8
843 %ld3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 3 ), align 4
844 %ld4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 4 ), align 16
845 %ld5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 5 ), align 4
846 %ld6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 6 ), align 8
847 %ld7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 7 ), align 4
848 %ld8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 8 ), align 32
849 %ld9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 9 ), align 4
850 %ld10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 10), align 8
851 %ld11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 11), align 4
852 %ld12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 12), align 16
853 %ld13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 13), align 4
854 %ld14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 14), align 8
855 %ld15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @src32, i32 0, i64 15), align 4
856 %cvt0 = uitofp i32 %ld0 to float
857 %cvt1 = uitofp i32 %ld1 to float
858 %cvt2 = uitofp i32 %ld2 to float
859 %cvt3 = uitofp i32 %ld3 to float
860 %cvt4 = uitofp i32 %ld4 to float
861 %cvt5 = uitofp i32 %ld5 to float
862 %cvt6 = uitofp i32 %ld6 to float
863 %cvt7 = uitofp i32 %ld7 to float
864 %cvt8 = uitofp i32 %ld8 to float
865 %cvt9 = uitofp i32 %ld9 to float
866 %cvt10 = uitofp i32 %ld10 to float
867 %cvt11 = uitofp i32 %ld11 to float
868 %cvt12 = uitofp i32 %ld12 to float
869 %cvt13 = uitofp i32 %ld13 to float
870 %cvt14 = uitofp i32 %ld14 to float
871 %cvt15 = uitofp i32 %ld15 to float
872 store float %cvt0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0 ), align 64
873 store float %cvt1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1 ), align 4
874 store float %cvt2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2 ), align 8
875 store float %cvt3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3 ), align 4
876 store float %cvt4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4 ), align 16
877 store float %cvt5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5 ), align 4
878 store float %cvt6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6 ), align 8
879 store float %cvt7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7 ), align 4
880 store float %cvt8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8 ), align 32
881 store float %cvt9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9 ), align 4
882 store float %cvt10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 8
883 store float %cvt11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
884 store float %cvt12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 16
885 store float %cvt13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
886 store float %cvt14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 8
887 store float %cvt15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
888 ret void
889 }
891 define void @uitofp_4i16_4f32() #0 {
892 ; CHECK-LABEL: @uitofp_4i16_4f32(
893 ; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* bitcast ([32 x i16]* @src16 to <4 x i16>*), align 64
894 ; CHECK-NEXT: [[TMP2:%.*]] = uitofp <4 x i16> [[TMP1]] to <4 x float>
895 ; CHECK-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
896 ; CHECK-NEXT: ret void
897 ;
898 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
899 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
900 %ld2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
901 %ld3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
902 %cvt0 = uitofp i16 %ld0 to float
903 %cvt1 = uitofp i16 %ld1 to float
904 %cvt2 = uitofp i16 %ld2 to float
905 %cvt3 = uitofp i16 %ld3 to float
906 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
907 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
908 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
909 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
910 ret void
911 }
913 define void @uitofp_8i16_8f32() #0 {
914 ; SSE-LABEL: @uitofp_8i16_8f32(
915 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* bitcast ([32 x i16]* @src16 to <4 x i16>*), align 64
916 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4) to <4 x i16>*), align 8
917 ; SSE-NEXT: [[TMP3:%.*]] = uitofp <4 x i16> [[TMP1]] to <4 x float>
918 ; SSE-NEXT: [[TMP4:%.*]] = uitofp <4 x i16> [[TMP2]] to <4 x float>
919 ; SSE-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
920 ; SSE-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
921 ; SSE-NEXT: ret void
922 ;
923 ; AVX-LABEL: @uitofp_8i16_8f32(
924 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @src16 to <8 x i16>*), align 64
925 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <8 x i16> [[TMP1]] to <8 x float>
926 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
927 ; AVX-NEXT: ret void
928 ;
929 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0), align 64
930 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1), align 2
931 %ld2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2), align 4
932 %ld3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3), align 2
933 %ld4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4), align 8
934 %ld5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 5), align 2
935 %ld6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 6), align 4
936 %ld7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 7), align 2
937 %cvt0 = uitofp i16 %ld0 to float
938 %cvt1 = uitofp i16 %ld1 to float
939 %cvt2 = uitofp i16 %ld2 to float
940 %cvt3 = uitofp i16 %ld3 to float
941 %cvt4 = uitofp i16 %ld4 to float
942 %cvt5 = uitofp i16 %ld5 to float
943 %cvt6 = uitofp i16 %ld6 to float
944 %cvt7 = uitofp i16 %ld7 to float
945 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
946 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
947 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
948 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
949 store float %cvt4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
950 store float %cvt5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
951 store float %cvt6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
952 store float %cvt7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
953 ret void
954 }
956 define void @uitofp_16i16_16f32() #0 {
957 ; SSE-LABEL: @uitofp_16i16_16f32(
958 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* bitcast ([32 x i16]* @src16 to <4 x i16>*), align 64
959 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4) to <4 x i16>*), align 8
960 ; SSE-NEXT: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 8) to <4 x i16>*), align 16
961 ; SSE-NEXT: [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 12) to <4 x i16>*), align 8
962 ; SSE-NEXT: [[TMP5:%.*]] = uitofp <4 x i16> [[TMP1]] to <4 x float>
963 ; SSE-NEXT: [[TMP6:%.*]] = uitofp <4 x i16> [[TMP2]] to <4 x float>
964 ; SSE-NEXT: [[TMP7:%.*]] = uitofp <4 x i16> [[TMP3]] to <4 x float>
965 ; SSE-NEXT: [[TMP8:%.*]] = uitofp <4 x i16> [[TMP4]] to <4 x float>
966 ; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
967 ; SSE-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
968 ; SSE-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 32
969 ; SSE-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 16
970 ; SSE-NEXT: ret void
971 ;
972 ; AVX256-LABEL: @uitofp_16i16_16f32(
973 ; AVX256-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @src16 to <8 x i16>*), align 64
974 ; AVX256-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 8) to <8 x i16>*), align 16
975 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <8 x i16> [[TMP1]] to <8 x float>
976 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <8 x i16> [[TMP2]] to <8 x float>
977 ; AVX256-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
978 ; AVX256-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 32
979 ; AVX256-NEXT: ret void
980 ;
981 ; AVX512-LABEL: @uitofp_16i16_16f32(
982 ; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @src16 to <16 x i16>*), align 64
983 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <16 x i16> [[TMP1]] to <16 x float>
984 ; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 64
985 ; AVX512-NEXT: ret void
986 ;
987 %ld0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 0 ), align 64
988 %ld1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 1 ), align 2
989 %ld2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 2 ), align 4
990 %ld3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 3 ), align 2
991 %ld4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 4 ), align 8
992 %ld5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 5 ), align 2
993 %ld6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 6 ), align 4
994 %ld7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 7 ), align 2
995 %ld8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 8 ), align 16
996 %ld9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 9 ), align 2
997 %ld10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 10), align 4
998 %ld11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 11), align 2
999 %ld12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 12), align 8
1000 %ld13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 13), align 2
1001 %ld14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 14), align 4
1002 %ld15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @src16, i32 0, i64 15), align 2
1003 %cvt0 = uitofp i16 %ld0 to float
1004 %cvt1 = uitofp i16 %ld1 to float
1005 %cvt2 = uitofp i16 %ld2 to float
1006 %cvt3 = uitofp i16 %ld3 to float
1007 %cvt4 = uitofp i16 %ld4 to float
1008 %cvt5 = uitofp i16 %ld5 to float
1009 %cvt6 = uitofp i16 %ld6 to float
1010 %cvt7 = uitofp i16 %ld7 to float
1011 %cvt8 = uitofp i16 %ld8 to float
1012 %cvt9 = uitofp i16 %ld9 to float
1013 %cvt10 = uitofp i16 %ld10 to float
1014 %cvt11 = uitofp i16 %ld11 to float
1015 %cvt12 = uitofp i16 %ld12 to float
1016 %cvt13 = uitofp i16 %ld13 to float
1017 %cvt14 = uitofp i16 %ld14 to float
1018 %cvt15 = uitofp i16 %ld15 to float
1019 store float %cvt0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0 ), align 64
1020 store float %cvt1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1 ), align 4
1021 store float %cvt2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2 ), align 8
1022 store float %cvt3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3 ), align 4
1023 store float %cvt4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4 ), align 16
1024 store float %cvt5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5 ), align 4
1025 store float %cvt6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6 ), align 8
1026 store float %cvt7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7 ), align 4
1027 store float %cvt8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8 ), align 32
1028 store float %cvt9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9 ), align 4
1029 store float %cvt10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 8
1030 store float %cvt11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
1031 store float %cvt12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 16
1032 store float %cvt13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
1033 store float %cvt14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 8
1034 store float %cvt15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
1035 ret void
1036 }
1038 define void @uitofp_4i8_4f32() #0 {
1039 ; CHECK-LABEL: @uitofp_4i8_4f32(
1040 ; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* bitcast ([64 x i8]* @src8 to <4 x i8>*), align 64
1041 ; CHECK-NEXT: [[TMP2:%.*]] = uitofp <4 x i8> [[TMP1]] to <4 x float>
1042 ; CHECK-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
1043 ; CHECK-NEXT: ret void
1044 ;
1045 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
1046 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
1047 %ld2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
1048 %ld3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
1049 %cvt0 = uitofp i8 %ld0 to float
1050 %cvt1 = uitofp i8 %ld1 to float
1051 %cvt2 = uitofp i8 %ld2 to float
1052 %cvt3 = uitofp i8 %ld3 to float
1053 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
1054 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
1055 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
1056 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
1057 ret void
1058 }
1060 define void @uitofp_8i8_8f32() #0 {
1061 ; SSE-LABEL: @uitofp_8i8_8f32(
1062 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* bitcast ([64 x i8]* @src8 to <4 x i8>*), align 64
1063 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4) to <4 x i8>*), align 4
1064 ; SSE-NEXT: [[TMP3:%.*]] = uitofp <4 x i8> [[TMP1]] to <4 x float>
1065 ; SSE-NEXT: [[TMP4:%.*]] = uitofp <4 x i8> [[TMP2]] to <4 x float>
1066 ; SSE-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
1067 ; SSE-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
1068 ; SSE-NEXT: ret void
1069 ;
1070 ; AVX-LABEL: @uitofp_8i8_8f32(
1071 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x i8>, <8 x i8>* bitcast ([64 x i8]* @src8 to <8 x i8>*), align 64
1072 ; AVX-NEXT: [[TMP2:%.*]] = uitofp <8 x i8> [[TMP1]] to <8 x float>
1073 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
1074 ; AVX-NEXT: ret void
1075 ;
1076 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0), align 64
1077 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1), align 1
1078 %ld2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2), align 2
1079 %ld3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3), align 1
1080 %ld4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4), align 4
1081 %ld5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 5), align 1
1082 %ld6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 6), align 2
1083 %ld7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 7), align 1
1084 %cvt0 = uitofp i8 %ld0 to float
1085 %cvt1 = uitofp i8 %ld1 to float
1086 %cvt2 = uitofp i8 %ld2 to float
1087 %cvt3 = uitofp i8 %ld3 to float
1088 %cvt4 = uitofp i8 %ld4 to float
1089 %cvt5 = uitofp i8 %ld5 to float
1090 %cvt6 = uitofp i8 %ld6 to float
1091 %cvt7 = uitofp i8 %ld7 to float
1092 store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
1093 store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
1094 store float %cvt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 8
1095 store float %cvt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
1096 store float %cvt4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 16
1097 store float %cvt5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
1098 store float %cvt6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 8
1099 store float %cvt7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
1100 ret void
1101 }
1103 define void @uitofp_16i8_16f32() #0 {
1104 ; SSE-LABEL: @uitofp_16i8_16f32(
1105 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* bitcast ([64 x i8]* @src8 to <4 x i8>*), align 64
1106 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4) to <4 x i8>*), align 4
1107 ; SSE-NEXT: [[TMP3:%.*]] = load <4 x i8>, <4 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 8) to <4 x i8>*), align 8
1108 ; SSE-NEXT: [[TMP4:%.*]] = load <4 x i8>, <4 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 12) to <4 x i8>*), align 4
1109 ; SSE-NEXT: [[TMP5:%.*]] = uitofp <4 x i8> [[TMP1]] to <4 x float>
1110 ; SSE-NEXT: [[TMP6:%.*]] = uitofp <4 x i8> [[TMP2]] to <4 x float>
1111 ; SSE-NEXT: [[TMP7:%.*]] = uitofp <4 x i8> [[TMP3]] to <4 x float>
1112 ; SSE-NEXT: [[TMP8:%.*]] = uitofp <4 x i8> [[TMP4]] to <4 x float>
1113 ; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 64
1114 ; SSE-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 16
1115 ; SSE-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 32
1116 ; SSE-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 16
1117 ; SSE-NEXT: ret void
1118 ;
1119 ; AVX256-LABEL: @uitofp_16i8_16f32(
1120 ; AVX256-NEXT: [[TMP1:%.*]] = load <8 x i8>, <8 x i8>* bitcast ([64 x i8]* @src8 to <8 x i8>*), align 64
1121 ; AVX256-NEXT: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 8) to <8 x i8>*), align 8
1122 ; AVX256-NEXT: [[TMP3:%.*]] = uitofp <8 x i8> [[TMP1]] to <8 x float>
1123 ; AVX256-NEXT: [[TMP4:%.*]] = uitofp <8 x i8> [[TMP2]] to <8 x float>
1124 ; AVX256-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 64
1125 ; AVX256-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 32
1126 ; AVX256-NEXT: ret void
1127 ;
1128 ; AVX512-LABEL: @uitofp_16i8_16f32(
1129 ; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @src8 to <16 x i8>*), align 64
1130 ; AVX512-NEXT: [[TMP2:%.*]] = uitofp <16 x i8> [[TMP1]] to <16 x float>
1131 ; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 64
1132 ; AVX512-NEXT: ret void
1133 ;
1134 %ld0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 0 ), align 64
1135 %ld1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 1 ), align 1
1136 %ld2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 2 ), align 2
1137 %ld3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 3 ), align 1
1138 %ld4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 4 ), align 4
1139 %ld5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 5 ), align 1
1140 %ld6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 6 ), align 2
1141 %ld7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 7 ), align 1
1142 %ld8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 8 ), align 8
1143 %ld9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 9 ), align 1
1144 %ld10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 10), align 2
1145 %ld11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 11), align 1
1146 %ld12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 12), align 4
1147 %ld13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 13), align 1
1148 %ld14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 14), align 2
1149 %ld15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @src8, i32 0, i64 15), align 1
1150 %cvt0 = uitofp i8 %ld0 to float
1151 %cvt1 = uitofp i8 %ld1 to float
1152 %cvt2 = uitofp i8 %ld2 to float
1153 %cvt3 = uitofp i8 %ld3 to float
1154 %cvt4 = uitofp i8 %ld4 to float
1155 %cvt5 = uitofp i8 %ld5 to float
1156 %cvt6 = uitofp i8 %ld6 to float
1157 %cvt7 = uitofp i8 %ld7 to float
1158 %cvt8 = uitofp i8 %ld8 to float
1159 %cvt9 = uitofp i8 %ld9 to float
1160 %cvt10 = uitofp i8 %ld10 to float
1161 %cvt11 = uitofp i8 %ld11 to float
1162 %cvt12 = uitofp i8 %ld12 to float
1163 %cvt13 = uitofp i8 %ld13 to float
1164 %cvt14 = uitofp i8 %ld14 to float
1165 %cvt15 = uitofp i8 %ld15 to float
1166 store float %cvt0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0 ), align 64
1167 store float %cvt1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1 ), align 4
1168 store float %cvt2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2 ), align 8
1169 store float %cvt3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3 ), align 4
1170 store float %cvt4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4 ), align 16
1171 store float %cvt5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5 ), align 4
1172 store float %cvt6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6 ), align 8
1173 store float %cvt7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7 ), align 4
1174 store float %cvt8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8 ), align 32
1175 store float %cvt9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9 ), align 4
1176 store float %cvt10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 8
1177 store float %cvt11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
1178 store float %cvt12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 16
1179 store float %cvt13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
1180 store float %cvt14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 8
1181 store float %cvt15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
1182 ret void
1183 }
1185 attributes #0 = { nounwind }