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
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
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
6 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512
8 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
10 @src64 = common global [8 x double] zeroinitializer, align 64
11 @src32 = common global [16 x float] zeroinitializer, align 64
12 @dst64 = common global [8 x double] zeroinitializer, align 64
13 @dst32 = common global [16 x float] zeroinitializer, align 64
15 declare float @llvm.sqrt.f32(float)
16 declare double @llvm.sqrt.f64(double)
22 define void @sqrt_2f64() #0 {
23 ; CHECK-LABEL: @sqrt_2f64(
24 ; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8
25 ; CHECK-NEXT: [[TMP2:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP1]])
26 ; CHECK-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
27 ; CHECK-NEXT: ret void
29 %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
30 %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
31 %sqrt0 = call double @llvm.sqrt.f64(double %a0)
32 %sqrt1 = call double @llvm.sqrt.f64(double %a1)
33 store double %sqrt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
34 store double %sqrt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
38 define void @sqrt_4f64() #0 {
39 ; SSE-LABEL: @sqrt_4f64(
40 ; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8
41 ; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8
42 ; SSE-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP1]])
43 ; SSE-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP2]])
44 ; SSE-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
45 ; 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 8
48 ; AVX-LABEL: @sqrt_4f64(
49 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
50 ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.sqrt.v4f64(<4 x double> [[TMP1]])
51 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
54 %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
55 %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
56 %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
57 %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
58 %sqrt0 = call double @llvm.sqrt.f64(double %a0)
59 %sqrt1 = call double @llvm.sqrt.f64(double %a1)
60 %sqrt2 = call double @llvm.sqrt.f64(double %a2)
61 %sqrt3 = call double @llvm.sqrt.f64(double %a3)
62 store double %sqrt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
63 store double %sqrt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
64 store double %sqrt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
65 store double %sqrt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
69 define void @sqrt_8f64() #0 {
70 ; SSE-LABEL: @sqrt_8f64(
71 ; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 4
72 ; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 4
73 ; SSE-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 4
74 ; SSE-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 4
75 ; SSE-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP1]])
76 ; SSE-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP2]])
77 ; SSE-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP3]])
78 ; SSE-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP4]])
79 ; SSE-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 4
80 ; 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 4
81 ; 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 4
82 ; 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 4
85 ; AVX256-LABEL: @sqrt_8f64(
86 ; AVX256-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 4
87 ; AVX256-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 4
88 ; AVX256-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.sqrt.v4f64(<4 x double> [[TMP1]])
89 ; AVX256-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.sqrt.v4f64(<4 x double> [[TMP2]])
90 ; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 4
91 ; 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 4
92 ; AVX256-NEXT: ret void
94 ; AVX512-LABEL: @sqrt_8f64(
95 ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 4
96 ; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.sqrt.v8f64(<8 x double> [[TMP1]])
97 ; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 4
98 ; AVX512-NEXT: ret void
100 %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 4
101 %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 4
102 %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 4
103 %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 4
104 %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 4
105 %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 4
106 %a6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 4
107 %a7 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 7), align 4
108 %sqrt0 = call double @llvm.sqrt.f64(double %a0)
109 %sqrt1 = call double @llvm.sqrt.f64(double %a1)
110 %sqrt2 = call double @llvm.sqrt.f64(double %a2)
111 %sqrt3 = call double @llvm.sqrt.f64(double %a3)
112 %sqrt4 = call double @llvm.sqrt.f64(double %a4)
113 %sqrt5 = call double @llvm.sqrt.f64(double %a5)
114 %sqrt6 = call double @llvm.sqrt.f64(double %a6)
115 %sqrt7 = call double @llvm.sqrt.f64(double %a7)
116 store double %sqrt0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 4
117 store double %sqrt1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 4
118 store double %sqrt2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 4
119 store double %sqrt3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 4
120 store double %sqrt4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 4
121 store double %sqrt5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 4
122 store double %sqrt6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 4
123 store double %sqrt7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 4
127 define void @sqrt_4f32() #0 {
128 ; CHECK-LABEL: @sqrt_4f32(
129 ; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
130 ; CHECK-NEXT: [[TMP2:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP1]])
131 ; CHECK-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
132 ; CHECK-NEXT: ret void
134 %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
135 %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
136 %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
137 %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
138 %sqrt0 = call float @llvm.sqrt.f32(float %a0)
139 %sqrt1 = call float @llvm.sqrt.f32(float %a1)
140 %sqrt2 = call float @llvm.sqrt.f32(float %a2)
141 %sqrt3 = call float @llvm.sqrt.f32(float %a3)
142 store float %sqrt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
143 store float %sqrt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
144 store float %sqrt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
145 store float %sqrt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
149 define void @sqrt_8f32() #0 {
150 ; SSE-LABEL: @sqrt_8f32(
151 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
152 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4
153 ; SSE-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP1]])
154 ; SSE-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP2]])
155 ; SSE-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
156 ; 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 4
159 ; AVX-LABEL: @sqrt_8f32(
160 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
161 ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.sqrt.v8f32(<8 x float> [[TMP1]])
162 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
165 %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
166 %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
167 %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
168 %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
169 %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
170 %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
171 %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
172 %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
173 %sqrt0 = call float @llvm.sqrt.f32(float %a0)
174 %sqrt1 = call float @llvm.sqrt.f32(float %a1)
175 %sqrt2 = call float @llvm.sqrt.f32(float %a2)
176 %sqrt3 = call float @llvm.sqrt.f32(float %a3)
177 %sqrt4 = call float @llvm.sqrt.f32(float %a4)
178 %sqrt5 = call float @llvm.sqrt.f32(float %a5)
179 %sqrt6 = call float @llvm.sqrt.f32(float %a6)
180 %sqrt7 = call float @llvm.sqrt.f32(float %a7)
181 store float %sqrt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
182 store float %sqrt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
183 store float %sqrt2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
184 store float %sqrt3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
185 store float %sqrt4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
186 store float %sqrt5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
187 store float %sqrt6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
188 store float %sqrt7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
192 define void @sqrt_16f32() #0 {
193 ; SSE-LABEL: @sqrt_16f32(
194 ; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
195 ; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4
196 ; SSE-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4
197 ; SSE-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4
198 ; SSE-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP1]])
199 ; SSE-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP2]])
200 ; SSE-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP3]])
201 ; SSE-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP4]])
202 ; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
203 ; 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 4
204 ; 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 4
205 ; 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 4
208 ; AVX256-LABEL: @sqrt_16f32(
209 ; AVX256-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
210 ; AVX256-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
211 ; AVX256-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.sqrt.v8f32(<8 x float> [[TMP1]])
212 ; AVX256-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.sqrt.v8f32(<8 x float> [[TMP2]])
213 ; AVX256-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
214 ; 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 4
215 ; AVX256-NEXT: ret void
217 ; AVX512-LABEL: @sqrt_16f32(
218 ; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
219 ; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.sqrt.v16f32(<16 x float> [[TMP1]])
220 ; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
221 ; AVX512-NEXT: ret void
223 %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
224 %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
225 %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
226 %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
227 %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
228 %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
229 %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
230 %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
231 %a8 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8), align 4
232 %a9 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 9), align 4
233 %a10 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 10), align 4
234 %a11 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 11), align 4
235 %a12 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12), align 4
236 %a13 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 13), align 4
237 %a14 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 14), align 4
238 %a15 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 15), align 4
239 %sqrt0 = call float @llvm.sqrt.f32(float %a0 )
240 %sqrt1 = call float @llvm.sqrt.f32(float %a1 )
241 %sqrt2 = call float @llvm.sqrt.f32(float %a2 )
242 %sqrt3 = call float @llvm.sqrt.f32(float %a3 )
243 %sqrt4 = call float @llvm.sqrt.f32(float %a4 )
244 %sqrt5 = call float @llvm.sqrt.f32(float %a5 )
245 %sqrt6 = call float @llvm.sqrt.f32(float %a6 )
246 %sqrt7 = call float @llvm.sqrt.f32(float %a7 )
247 %sqrt8 = call float @llvm.sqrt.f32(float %a8 )
248 %sqrt9 = call float @llvm.sqrt.f32(float %a9 )
249 %sqrt10 = call float @llvm.sqrt.f32(float %a10)
250 %sqrt11 = call float @llvm.sqrt.f32(float %a11)
251 %sqrt12 = call float @llvm.sqrt.f32(float %a12)
252 %sqrt13 = call float @llvm.sqrt.f32(float %a13)
253 %sqrt14 = call float @llvm.sqrt.f32(float %a14)
254 %sqrt15 = call float @llvm.sqrt.f32(float %a15)
255 store float %sqrt0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
256 store float %sqrt1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
257 store float %sqrt2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
258 store float %sqrt3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
259 store float %sqrt4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
260 store float %sqrt5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
261 store float %sqrt6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
262 store float %sqrt7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
263 store float %sqrt8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8), align 4
264 store float %sqrt9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9), align 4
265 store float %sqrt10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 4
266 store float %sqrt11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
267 store float %sqrt12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
268 store float %sqrt13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
269 store float %sqrt14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
270 store float %sqrt15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
274 attributes #0 = { nounwind }