1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
2 ; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
3 ; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
5 %struct.zot = type { i32, i32, i32 }
7 define i1 @reorder_results(ptr %arg, i1 %arg1, ptr %arg2, i64 %arg3, ptr %arg4) {
8 ; CHECK-LABEL: define i1 @reorder_results(
9 ; CHECK-SAME: ptr [[ARG:%.*]], i1 [[ARG1:%.*]], ptr [[ARG2:%.*]], i64 [[ARG3:%.*]], ptr [[ARG4:%.*]]) {
11 ; CHECK-NEXT: [[LOAD:%.*]] = load ptr, ptr [[ARG4]], align 8
12 ; CHECK-NEXT: [[LOAD4:%.*]] = load i32, ptr [[LOAD]], align 4
13 ; CHECK-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr i8, ptr [[LOAD]], i64 4
14 ; CHECK-NEXT: [[LOAD5:%.*]] = load i32, ptr [[GETELEMENTPTR]], align 4
15 ; CHECK-NEXT: [[GETELEMENTPTR6:%.*]] = getelementptr i8, ptr [[LOAD]], i64 8
16 ; CHECK-NEXT: [[LOAD7:%.*]] = load i32, ptr [[GETELEMENTPTR6]], align 4
17 ; CHECK-NEXT: br i1 [[ARG1]], label [[BB12:%.*]], label [[BB9:%.*]]
19 ; CHECK-NEXT: ret i1 false
21 ; CHECK-NEXT: [[FREEZE:%.*]] = freeze ptr [[ARG]]
22 ; CHECK-NEXT: store i32 [[LOAD4]], ptr [[FREEZE]], align 4
23 ; CHECK-NEXT: [[GETELEMENTPTR10:%.*]] = getelementptr i8, ptr [[FREEZE]], i64 4
24 ; CHECK-NEXT: store i32 [[LOAD7]], ptr [[GETELEMENTPTR10]], align 4
25 ; CHECK-NEXT: [[GETELEMENTPTR11:%.*]] = getelementptr i8, ptr [[FREEZE]], i64 8
26 ; CHECK-NEXT: store i32 [[LOAD5]], ptr [[GETELEMENTPTR11]], align 4
27 ; CHECK-NEXT: br label [[BB8:%.*]]
29 ; CHECK-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr [[STRUCT_ZOT:%.*]], ptr [[ARG2]], i64 [[ARG3]]
30 ; CHECK-NEXT: store i32 [[LOAD4]], ptr [[GETELEMENTPTR13]], align 4
31 ; CHECK-NEXT: [[GETELEMENTPTR14:%.*]] = getelementptr i8, ptr [[GETELEMENTPTR13]], i64 4
32 ; CHECK-NEXT: store i32 [[LOAD7]], ptr [[GETELEMENTPTR14]], align 4
33 ; CHECK-NEXT: [[GETELEMENTPTR15:%.*]] = getelementptr i8, ptr [[GETELEMENTPTR13]], i64 8
34 ; CHECK-NEXT: store i32 [[LOAD5]], ptr [[GETELEMENTPTR15]], align 4
35 ; CHECK-NEXT: br label [[BB8]]
38 %load = load ptr, ptr %arg4, align 8
39 %load4 = load i32, ptr %load, align 4
40 %getelementptr = getelementptr i8, ptr %load, i64 4
41 %load5 = load i32, ptr %getelementptr, align 4
42 %getelementptr6 = getelementptr i8, ptr %load, i64 8
43 %load7 = load i32, ptr %getelementptr6, align 4
44 br i1 %arg1, label %bb12, label %bb9
46 bb8: ; preds = %bb12, %bb9
50 %freeze = freeze ptr %arg
51 store i32 %load4, ptr %freeze, align 4
52 %getelementptr10 = getelementptr i8, ptr %freeze, i64 4
53 store i32 %load7, ptr %getelementptr10, align 4
54 %getelementptr11 = getelementptr i8, ptr %freeze, i64 8
55 store i32 %load5, ptr %getelementptr11, align 4
59 %getelementptr13 = getelementptr %struct.zot, ptr %arg2, i64 %arg3
60 store i32 %load4, ptr %getelementptr13, align 4
61 %getelementptr14 = getelementptr i8, ptr %getelementptr13, i64 4
62 store i32 %load7, ptr %getelementptr14, align 4
63 %getelementptr15 = getelementptr i8, ptr %getelementptr13, i64 8
64 store i32 %load5, ptr %getelementptr15, align 4
68 define void @extract_mask(ptr %object, double %conv503, double %conv520) {
69 ; CHECK-LABEL: define void @extract_mask(
70 ; CHECK-SAME: ptr [[OBJECT:%.*]], double [[CONV503:%.*]], double [[CONV520:%.*]]) {
72 ; CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[OBJECT]], align 8
73 ; CHECK-NEXT: [[BBOX483:%.*]] = getelementptr float, ptr [[TMP0]]
74 ; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr [[BBOX483]], align 8
75 ; CHECK-NEXT: [[TMP2:%.*]] = fpext <2 x float> [[TMP1]] to <2 x double>
76 ; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> poison, <2 x i32> <i32 1, i32 0>
77 ; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP3]], double [[CONV503]], i32 0
78 ; CHECK-NEXT: [[TMP5:%.*]] = fcmp ogt <2 x double> [[TMP4]], <double 0.000000e+00, double -2.000000e+10>
79 ; CHECK-NEXT: [[TMP6:%.*]] = select <2 x i1> [[TMP5]], <2 x double> [[TMP3]], <2 x double> <double 0.000000e+00, double -2.000000e+10>
80 ; CHECK-NEXT: [[TMP7:%.*]] = fsub <2 x double> zeroinitializer, [[TMP6]]
81 ; CHECK-NEXT: [[TMP8:%.*]] = fptrunc <2 x double> [[TMP7]] to <2 x float>
82 ; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x float> [[TMP8]], i32 0
83 ; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x float> [[TMP8]], i32 1
84 ; CHECK-NEXT: [[MUL646:%.*]] = fmul float [[TMP9]], [[TMP10]]
85 ; CHECK-NEXT: [[CMP663:%.*]] = fcmp olt float [[MUL646]], 0.000000e+00
86 ; CHECK-NEXT: br i1 [[CMP663]], label [[IF_THEN665:%.*]], label [[IF_END668:%.*]]
88 ; CHECK-NEXT: [[ARRAYIDX656:%.*]] = getelementptr float, ptr [[OBJECT]], i64 10
89 ; CHECK-NEXT: [[BBOX651:%.*]] = getelementptr float, ptr [[OBJECT]]
90 ; CHECK-NEXT: [[CONV613:%.*]] = fptrunc double [[CONV503]] to float
91 ; CHECK-NEXT: store float [[CONV613]], ptr [[BBOX651]], align 8
92 ; CHECK-NEXT: [[BBOX_SROA_6_0_BBOX666_SROA_IDX:%.*]] = getelementptr float, ptr [[OBJECT]], i64 1
93 ; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x double> [[TMP6]], double [[CONV520]], i32 1
94 ; CHECK-NEXT: [[TMP12:%.*]] = fptrunc <2 x double> [[TMP11]] to <2 x float>
95 ; CHECK-NEXT: store <2 x float> [[TMP12]], ptr [[BBOX_SROA_6_0_BBOX666_SROA_IDX]], align 4
96 ; CHECK-NEXT: store <2 x float> [[TMP8]], ptr [[ARRAYIDX656]], align 8
97 ; CHECK-NEXT: br label [[IF_END668]]
99 ; CHECK-NEXT: ret void
102 %0 = load ptr, ptr %object, align 8
103 %bbox483 = getelementptr float, ptr %0
104 %1 = load float, ptr %bbox483, align 8
105 %conv486 = fpext float %1 to double
106 %cmp487 = fcmp ogt double %conv486, -2.000000e+10
107 %conv486.2 = select i1 %cmp487, double %conv486, double -2.000000e+10
108 %arrayidx502 = getelementptr float, ptr %0, i64 1
109 %2 = load float, ptr %arrayidx502, align 4
110 %conv5033 = fpext float %2 to double
111 %cmp504 = fcmp ogt double %conv503, 0.000000e+00
112 %cond514 = select i1 %cmp504, double %conv5033, double 0.000000e+00
113 %sub626 = fsub double 0.000000e+00, %conv486.2
114 %conv627 = fptrunc double %sub626 to float
115 %sub632 = fsub double 0.000000e+00, %cond514
116 %conv633 = fptrunc double %sub632 to float
117 %mul646 = fmul float %conv633, %conv627
118 %cmp663 = fcmp olt float %mul646, 0.000000e+00
119 br i1 %cmp663, label %if.then665, label %if.end668
121 if.then665: ; preds = %entry
122 %arrayidx656 = getelementptr float, ptr %object, i64 10
123 %lengths652 = getelementptr float, ptr %object, i64 11
124 %bbox651 = getelementptr float, ptr %object
125 %conv621 = fptrunc double %conv520 to float
126 %conv617 = fptrunc double %cond514 to float
127 %conv613 = fptrunc double %conv503 to float
128 store float %conv613, ptr %bbox651, align 8
129 %bbox.sroa.6.0.bbox666.sroa_idx = getelementptr float, ptr %object, i64 1
130 store float %conv617, ptr %bbox.sroa.6.0.bbox666.sroa_idx, align 4
131 %bbox.sroa.8.0.bbox666.sroa_idx = getelementptr float, ptr %object, i64 2
132 store float %conv621, ptr %bbox.sroa.8.0.bbox666.sroa_idx, align 8
133 store float %conv627, ptr %lengths652, align 4
134 store float %conv633, ptr %arrayidx656, align 8
137 if.end668: ; preds = %if.then665, %entry
141 define void @gather_2(ptr %mat1, float %0, float %1) {
142 ; NON-POW2-LABEL: define void @gather_2(
143 ; NON-POW2-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
144 ; NON-POW2-NEXT: entry:
145 ; NON-POW2-NEXT: [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
146 ; NON-POW2-NEXT: [[TMP4:%.*]] = shufflevector <3 x float> [[TMP2]], <3 x float> poison, <3 x i32> zeroinitializer
147 ; NON-POW2-NEXT: [[TMP5:%.*]] = insertelement <3 x float> <float 0.000000e+00, float poison, float poison>, float [[TMP1]], i32 1
148 ; NON-POW2-NEXT: [[TMP6:%.*]] = shufflevector <3 x float> [[TMP5]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
149 ; NON-POW2-NEXT: [[TMP7:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP4]], <3 x float> [[TMP6]], <3 x float> zeroinitializer)
150 ; NON-POW2-NEXT: [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
151 ; NON-POW2-NEXT: [[TMP8:%.*]] = fmul <3 x float> [[TMP7]], zeroinitializer
152 ; NON-POW2-NEXT: store <3 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
153 ; NON-POW2-NEXT: ret void
155 ; POW2-ONLY-LABEL: define void @gather_2(
156 ; POW2-ONLY-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
157 ; POW2-ONLY-NEXT: entry:
158 ; POW2-ONLY-NEXT: [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
159 ; POW2-ONLY-NEXT: [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
160 ; POW2-ONLY-NEXT: [[TMP4:%.*]] = insertelement <2 x float> <float 0.000000e+00, float poison>, float [[TMP1]], i32 1
161 ; POW2-ONLY-NEXT: [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
162 ; POW2-ONLY-NEXT: [[TMP6:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP1]], float 0.000000e+00)
163 ; POW2-ONLY-NEXT: [[TMP7:%.*]] = fmul float [[TMP6]], 0.000000e+00
164 ; POW2-ONLY-NEXT: [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
165 ; POW2-ONLY-NEXT: [[ARRAYIDX5_I_I_I280:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1, i64 2
166 ; POW2-ONLY-NEXT: [[TMP8:%.*]] = fmul <2 x float> [[TMP5]], zeroinitializer
167 ; POW2-ONLY-NEXT: store <2 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
168 ; POW2-ONLY-NEXT: store float [[TMP7]], ptr [[ARRAYIDX5_I_I_I280]], align 4
169 ; POW2-ONLY-NEXT: ret void
172 %2 = call float @llvm.fmuladd.f32(float %0, float 0.000000e+00, float 0.000000e+00)
173 %3 = call float @llvm.fmuladd.f32(float %1, float %0, float 0.000000e+00)
174 %4 = call float @llvm.fmuladd.f32(float %0, float %1, float 0.000000e+00)
175 %5 = fmul float %2, 0.000000e+00
176 %6 = fmul float %3, 0.000000e+00
177 %7 = fmul float %4, 0.000000e+00
178 %arrayidx163 = getelementptr [4 x [4 x float]], ptr %mat1, i64 0, i64 1
179 %arrayidx2.i.i.i278 = getelementptr [4 x [4 x float]], ptr %mat1, i64 0, i64 1, i64 1
180 %arrayidx5.i.i.i280 = getelementptr [4 x [4 x float]], ptr %mat1, i64 0, i64 1, i64 2
181 store float %5, ptr %arrayidx163, align 4
182 store float %6, ptr %arrayidx2.i.i.i278, align 4
183 store float %7, ptr %arrayidx5.i.i.i280, align 4
187 define i32 @reorder_indices_1(float %0) {
188 ; NON-POW2-LABEL: define i32 @reorder_indices_1(
189 ; NON-POW2-SAME: float [[TMP0:%.*]]) {
190 ; NON-POW2-NEXT: entry:
191 ; NON-POW2-NEXT: [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
192 ; NON-POW2-NEXT: [[TMP1:%.*]] = load <3 x float>, ptr [[NOR1]], align 4
193 ; NON-POW2-NEXT: [[TMP3:%.*]] = fneg <3 x float> [[TMP1]]
194 ; NON-POW2-NEXT: [[TMP4:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
195 ; NON-POW2-NEXT: [[TMP5:%.*]] = shufflevector <3 x float> [[TMP4]], <3 x float> poison, <3 x i32> zeroinitializer
196 ; NON-POW2-NEXT: [[TMP6:%.*]] = fmul <3 x float> [[TMP3]], [[TMP5]]
197 ; NON-POW2-NEXT: [[TMP10:%.*]] = shufflevector <3 x float> [[TMP6]], <3 x float> poison, <3 x i32> <i32 1, i32 2, i32 0>
198 ; NON-POW2-NEXT: [[TMP7:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP1]], <3 x float> zeroinitializer, <3 x float> [[TMP10]])
199 ; NON-POW2-NEXT: [[TMP8:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP5]], <3 x float> [[TMP7]], <3 x float> zeroinitializer)
200 ; NON-POW2-NEXT: [[TMP9:%.*]] = fmul <3 x float> [[TMP8]], zeroinitializer
201 ; NON-POW2-NEXT: store <3 x float> [[TMP9]], ptr [[NOR1]], align 4
202 ; NON-POW2-NEXT: ret i32 0
204 ; POW2-ONLY-LABEL: define i32 @reorder_indices_1(
205 ; POW2-ONLY-SAME: float [[TMP0:%.*]]) {
206 ; POW2-ONLY-NEXT: entry:
207 ; POW2-ONLY-NEXT: [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
208 ; POW2-ONLY-NEXT: [[ARRAYIDX2_I265:%.*]] = getelementptr float, ptr [[NOR1]], i64 2
209 ; POW2-ONLY-NEXT: [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2_I265]], align 4
210 ; POW2-ONLY-NEXT: [[TMP2:%.*]] = load <2 x float>, ptr [[NOR1]], align 4
211 ; POW2-ONLY-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
212 ; POW2-ONLY-NEXT: [[TMP4:%.*]] = fneg float [[TMP3]]
213 ; POW2-ONLY-NEXT: [[NEG11_I:%.*]] = fmul float [[TMP4]], [[TMP0]]
214 ; POW2-ONLY-NEXT: [[TMP5:%.*]] = call float @llvm.fmuladd.f32(float [[TMP1]], float 0.000000e+00, float [[NEG11_I]])
215 ; POW2-ONLY-NEXT: [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 poison>
216 ; POW2-ONLY-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
217 ; POW2-ONLY-NEXT: [[TMP8:%.*]] = fneg <2 x float> [[TMP7]]
218 ; POW2-ONLY-NEXT: [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
219 ; POW2-ONLY-NEXT: [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer
220 ; POW2-ONLY-NEXT: [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]]
221 ; POW2-ONLY-NEXT: [[TMP12:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> zeroinitializer, <2 x float> [[TMP11]])
222 ; POW2-ONLY-NEXT: [[TMP13:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP10]], <2 x float> [[TMP12]], <2 x float> zeroinitializer)
223 ; POW2-ONLY-NEXT: [[TMP14:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP5]], float 0.000000e+00)
224 ; POW2-ONLY-NEXT: [[TMP15:%.*]] = fmul <2 x float> [[TMP13]], zeroinitializer
225 ; POW2-ONLY-NEXT: [[MUL6_I_I_I:%.*]] = fmul float [[TMP14]], 0.000000e+00
226 ; POW2-ONLY-NEXT: store <2 x float> [[TMP15]], ptr [[NOR1]], align 4
227 ; POW2-ONLY-NEXT: store float [[MUL6_I_I_I]], ptr [[ARRAYIDX2_I265]], align 4
228 ; POW2-ONLY-NEXT: ret i32 0
231 %nor1 = alloca [0 x [3 x float]], i32 0, align 4
232 %arrayidx.i = getelementptr float, ptr %nor1, i64 1
233 %1 = load float, ptr %arrayidx.i, align 4
234 %arrayidx2.i265 = getelementptr float, ptr %nor1, i64 2
235 %2 = load float, ptr %arrayidx2.i265, align 4
237 %neg.i267 = fmul float %3, %0
238 %4 = call float @llvm.fmuladd.f32(float %1, float 0.000000e+00, float %neg.i267)
239 %5 = load float, ptr %nor1, align 4
241 %neg11.i = fmul float %6, %0
242 %7 = call float @llvm.fmuladd.f32(float %2, float 0.000000e+00, float %neg11.i)
244 %neg18.i = fmul float %8, %0
245 %9 = call float @llvm.fmuladd.f32(float %5, float 0.000000e+00, float %neg18.i)
246 %10 = call float @llvm.fmuladd.f32(float %0, float %9, float 0.000000e+00)
247 %11 = call float @llvm.fmuladd.f32(float %0, float %4, float 0.000000e+00)
248 %12 = call float @llvm.fmuladd.f32(float %0, float %7, float 0.000000e+00)
249 %mul.i.i.i = fmul float %10, 0.000000e+00
250 %mul3.i.i.i = fmul float %11, 0.000000e+00
251 %mul6.i.i.i = fmul float %12, 0.000000e+00
252 store float %mul.i.i.i, ptr %nor1, align 4
253 store float %mul3.i.i.i, ptr %arrayidx.i, align 4
254 store float %mul6.i.i.i, ptr %arrayidx2.i265, align 4
258 define void @reorder_indices_2(ptr %spoint) {
259 ; NON-POW2-LABEL: define void @reorder_indices_2(
260 ; NON-POW2-SAME: ptr [[SPOINT:%.*]]) {
261 ; NON-POW2-NEXT: entry:
262 ; NON-POW2-NEXT: [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
263 ; NON-POW2-NEXT: [[TMP0:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> zeroinitializer, <3 x float> zeroinitializer, <3 x float> zeroinitializer)
264 ; NON-POW2-NEXT: [[TMP1:%.*]] = fmul <3 x float> [[TMP0]], zeroinitializer
265 ; NON-POW2-NEXT: [[TMP2:%.*]] = shufflevector <3 x float> [[TMP1]], <3 x float> poison, <3 x i32> <i32 1, i32 2, i32 0>
266 ; NON-POW2-NEXT: store <3 x float> [[TMP2]], ptr [[DSCO]], align 4
267 ; NON-POW2-NEXT: ret void
269 ; POW2-ONLY-LABEL: define void @reorder_indices_2(
270 ; POW2-ONLY-SAME: ptr [[SPOINT:%.*]]) {
271 ; POW2-ONLY-NEXT: entry:
272 ; POW2-ONLY-NEXT: [[TMP0:%.*]] = extractelement <3 x float> zeroinitializer, i64 0
273 ; POW2-ONLY-NEXT: [[TMP1:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 0.000000e+00, float 0.000000e+00)
274 ; POW2-ONLY-NEXT: [[MUL4_I461:%.*]] = fmul float [[TMP1]], 0.000000e+00
275 ; POW2-ONLY-NEXT: [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
276 ; POW2-ONLY-NEXT: [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer, <2 x float> zeroinitializer)
277 ; POW2-ONLY-NEXT: [[TMP3:%.*]] = fmul <2 x float> [[TMP2]], zeroinitializer
278 ; POW2-ONLY-NEXT: store <2 x float> [[TMP3]], ptr [[DSCO]], align 4
279 ; POW2-ONLY-NEXT: [[ARRAYIDX5_I476:%.*]] = getelementptr float, ptr [[SPOINT]], i64 2
280 ; POW2-ONLY-NEXT: store float [[MUL4_I461]], ptr [[ARRAYIDX5_I476]], align 4
281 ; POW2-ONLY-NEXT: ret void
284 %0 = extractelement <3 x float> zeroinitializer, i64 1
285 %1 = extractelement <3 x float> zeroinitializer, i64 2
286 %2 = extractelement <3 x float> zeroinitializer, i64 0
287 %3 = tail call float @llvm.fmuladd.f32(float %0, float 0.000000e+00, float 0.000000e+00)
288 %4 = tail call float @llvm.fmuladd.f32(float %1, float 0.000000e+00, float 0.000000e+00)
289 %5 = tail call float @llvm.fmuladd.f32(float %2, float 0.000000e+00, float 0.000000e+00)
290 %mul.i457 = fmul float %3, 0.000000e+00
291 %mul2.i459 = fmul float %4, 0.000000e+00
292 %mul4.i461 = fmul float %5, 0.000000e+00
293 %dsco = getelementptr float, ptr %spoint, i64 0
294 store float %mul.i457, ptr %dsco, align 4
295 %arrayidx3.i474 = getelementptr float, ptr %spoint, i64 1
296 store float %mul2.i459, ptr %arrayidx3.i474, align 4
297 %arrayidx5.i476 = getelementptr float, ptr %spoint, i64 2
298 store float %mul4.i461, ptr %arrayidx5.i476, align 4
302 define void @reorder_indices_2x_load(ptr %png_ptr, ptr %info_ptr) {
303 ; CHECK-LABEL: define void @reorder_indices_2x_load(
304 ; CHECK-SAME: ptr [[PNG_PTR:%.*]], ptr [[INFO_PTR:%.*]]) {
306 ; CHECK-NEXT: [[BIT_DEPTH:%.*]] = getelementptr i8, ptr [[INFO_PTR]], i64 0
307 ; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[BIT_DEPTH]], align 4
308 ; CHECK-NEXT: [[COLOR_TYPE:%.*]] = getelementptr i8, ptr [[INFO_PTR]], i64 1
309 ; CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr [[COLOR_TYPE]], align 1
310 ; CHECK-NEXT: [[BIT_DEPTH37_I:%.*]] = getelementptr i8, ptr [[PNG_PTR]], i64 11
311 ; CHECK-NEXT: store i8 [[TMP0]], ptr [[BIT_DEPTH37_I]], align 1
312 ; CHECK-NEXT: [[COLOR_TYPE39_I:%.*]] = getelementptr i8, ptr [[PNG_PTR]], i64 10
313 ; CHECK-NEXT: store i8 [[TMP1]], ptr [[COLOR_TYPE39_I]], align 2
314 ; CHECK-NEXT: [[USR_BIT_DEPTH_I:%.*]] = getelementptr i8, ptr [[PNG_PTR]], i64 12
315 ; CHECK-NEXT: store i8 [[TMP0]], ptr [[USR_BIT_DEPTH_I]], align 8
316 ; CHECK-NEXT: ret void
319 %bit_depth = getelementptr i8, ptr %info_ptr, i64 0
320 %0 = load i8, ptr %bit_depth, align 4
321 %color_type = getelementptr i8, ptr %info_ptr, i64 1
322 %1 = load i8, ptr %color_type, align 1
323 %bit_depth37.i = getelementptr i8, ptr %png_ptr, i64 11
324 store i8 %0, ptr %bit_depth37.i, align 1
325 %color_type39.i = getelementptr i8, ptr %png_ptr, i64 10
326 store i8 %1, ptr %color_type39.i, align 2
327 %usr_bit_depth.i = getelementptr i8, ptr %png_ptr, i64 12
328 store i8 %0, ptr %usr_bit_depth.i, align 8
332 define void @reuse_shuffle_indidces_1(ptr %col, float %0, float %1) {
333 ; NON-POW2-LABEL: define void @reuse_shuffle_indidces_1(
334 ; NON-POW2-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
335 ; NON-POW2-NEXT: entry:
336 ; NON-POW2-NEXT: [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP1]], i32 0
337 ; NON-POW2-NEXT: [[TMP3:%.*]] = insertelement <3 x float> [[TMP2]], float [[TMP0]], i32 1
338 ; NON-POW2-NEXT: [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
339 ; NON-POW2-NEXT: [[TMP5:%.*]] = fmul <3 x float> [[TMP4]], zeroinitializer
340 ; NON-POW2-NEXT: [[TMP6:%.*]] = fadd <3 x float> [[TMP5]], zeroinitializer
341 ; NON-POW2-NEXT: store <3 x float> [[TMP6]], ptr [[COL]], align 4
342 ; NON-POW2-NEXT: ret void
344 ; POW2-ONLY-LABEL: define void @reuse_shuffle_indidces_1(
345 ; POW2-ONLY-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
346 ; POW2-ONLY-NEXT: entry:
347 ; POW2-ONLY-NEXT: [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
348 ; POW2-ONLY-NEXT: [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP0]], i32 1
349 ; POW2-ONLY-NEXT: [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
350 ; POW2-ONLY-NEXT: [[TMP5:%.*]] = fadd <2 x float> [[TMP4]], zeroinitializer
351 ; POW2-ONLY-NEXT: store <2 x float> [[TMP5]], ptr [[COL]], align 4
352 ; POW2-ONLY-NEXT: [[ARRAYIDX33:%.*]] = getelementptr float, ptr [[COL]], i64 2
353 ; POW2-ONLY-NEXT: [[MUL38:%.*]] = fmul float [[TMP0]], 0.000000e+00
354 ; POW2-ONLY-NEXT: [[TMP6:%.*]] = fadd float [[MUL38]], 0.000000e+00
355 ; POW2-ONLY-NEXT: store float [[TMP6]], ptr [[ARRAYIDX33]], align 4
356 ; POW2-ONLY-NEXT: ret void
359 %mul24 = fmul float %1, 0.000000e+00
360 %2 = fadd float %mul24, 0.000000e+00
361 store float %2, ptr %col, align 4
362 %arrayidx26 = getelementptr float, ptr %col, i64 1
363 %mul31 = fmul float %0, 0.000000e+00
364 %3 = fadd float %mul31, 0.000000e+00
365 store float %3, ptr %arrayidx26, align 4
366 %arrayidx33 = getelementptr float, ptr %col, i64 2
367 %mul38 = fmul float %0, 0.000000e+00
368 %4 = fadd float %mul38, 0.000000e+00
369 store float %4, ptr %arrayidx33, align 4
373 define void @reuse_shuffle_indices_2(ptr %inertia, double %0) {
374 ; CHECK-LABEL: define void @reuse_shuffle_indices_2(
375 ; CHECK-SAME: ptr [[INERTIA:%.*]], double [[TMP0:%.*]]) {
377 ; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> poison, double [[TMP0]], i32 0
378 ; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> zeroinitializer
379 ; CHECK-NEXT: [[TMP3:%.*]] = fptrunc <2 x double> [[TMP2]] to <2 x float>
380 ; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
381 ; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 poison>
382 ; CHECK-NEXT: [[TMP6:%.*]] = fadd <4 x float> [[TMP5]], <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
383 ; CHECK-NEXT: [[TMP7:%.*]] = fmul <4 x float> [[TMP6]], <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
384 ; CHECK-NEXT: [[TMP8:%.*]] = fadd <4 x float> [[TMP7]], <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
385 ; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x float> [[TMP8]], <4 x float> poison, <3 x i32> <i32 0, i32 1, i32 2>
386 ; CHECK-NEXT: store <3 x float> [[TMP9]], ptr [[INERTIA]], align 4
387 ; CHECK-NEXT: ret void
390 %1 = insertelement <2 x double> poison, double %0, i32 0
391 %2 = shufflevector <2 x double> %1, <2 x double> poison, <2 x i32> zeroinitializer
392 %3 = fptrunc <2 x double> %2 to <2 x float>
393 %4 = fmul <2 x float> %3, zeroinitializer
394 %5 = shufflevector <2 x float> %4, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 poison>
395 %6 = fadd <4 x float> %5, <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
396 %7 = fmul <4 x float> %6, <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
397 %8 = fadd <4 x float> %7, <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>
398 %9 = shufflevector <4 x float> %8, <4 x float> poison, <3 x i32> <i32 0, i32 1, i32 2>
399 store <3 x float> %9, ptr %inertia, align 4
403 define void @reuse_shuffle_indices_cost_crash_2(ptr %bezt, float %0) {
404 ; NON-POW2-LABEL: define void @reuse_shuffle_indices_cost_crash_2(
405 ; NON-POW2-SAME: ptr [[BEZT:%.*]], float [[TMP0:%.*]]) {
406 ; NON-POW2-NEXT: entry:
407 ; NON-POW2-NEXT: [[FNEG:%.*]] = fmul float [[TMP0]], 0.000000e+00
408 ; NON-POW2-NEXT: [[TMP1:%.*]] = insertelement <3 x float> poison, float [[FNEG]], i32 0
409 ; NON-POW2-NEXT: [[TMP2:%.*]] = shufflevector <3 x float> [[TMP1]], <3 x float> poison, <3 x i32> zeroinitializer
410 ; NON-POW2-NEXT: [[TMP3:%.*]] = insertelement <3 x float> <float poison, float poison, float 0.000000e+00>, float [[TMP0]], i32 0
411 ; NON-POW2-NEXT: [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> <i32 0, i32 0, i32 2>
412 ; NON-POW2-NEXT: [[TMP5:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP2]], <3 x float> [[TMP4]], <3 x float> zeroinitializer)
413 ; NON-POW2-NEXT: store <3 x float> [[TMP5]], ptr [[BEZT]], align 4
414 ; NON-POW2-NEXT: ret void
416 ; POW2-ONLY-LABEL: define void @reuse_shuffle_indices_cost_crash_2(
417 ; POW2-ONLY-SAME: ptr [[BEZT:%.*]], float [[TMP0:%.*]]) {
418 ; POW2-ONLY-NEXT: entry:
419 ; POW2-ONLY-NEXT: [[FNEG:%.*]] = fmul float [[TMP0]], 0.000000e+00
420 ; POW2-ONLY-NEXT: [[TMP1:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
421 ; POW2-ONLY-NEXT: [[TMP2:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> poison, <2 x i32> zeroinitializer
422 ; POW2-ONLY-NEXT: [[TMP3:%.*]] = insertelement <2 x float> poison, float [[FNEG]], i32 0
423 ; POW2-ONLY-NEXT: [[TMP4:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <2 x i32> zeroinitializer
424 ; POW2-ONLY-NEXT: [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
425 ; POW2-ONLY-NEXT: store <2 x float> [[TMP5]], ptr [[BEZT]], align 4
426 ; POW2-ONLY-NEXT: [[TMP6:%.*]] = tail call float @llvm.fmuladd.f32(float [[FNEG]], float 0.000000e+00, float 0.000000e+00)
427 ; POW2-ONLY-NEXT: [[ARRAYIDX8_I831:%.*]] = getelementptr float, ptr [[BEZT]], i64 2
428 ; POW2-ONLY-NEXT: store float [[TMP6]], ptr [[ARRAYIDX8_I831]], align 4
429 ; POW2-ONLY-NEXT: ret void
432 %fneg = fmul float %0, 0.000000e+00
433 %1 = tail call float @llvm.fmuladd.f32(float %0, float %fneg, float 0.000000e+00)
434 store float %1, ptr %bezt, align 4
435 %2 = tail call float @llvm.fmuladd.f32(float %0, float %fneg, float 0.000000e+00)
436 %arrayidx5.i = getelementptr float, ptr %bezt, i64 1
437 store float %2, ptr %arrayidx5.i, align 4
438 %3 = tail call float @llvm.fmuladd.f32(float %fneg, float 0.000000e+00, float 0.000000e+00)
439 %arrayidx8.i831 = getelementptr float, ptr %bezt, i64 2
440 store float %3, ptr %arrayidx8.i831, align 4
444 define void @reuse_shuffle_indices_cost_crash_3(ptr %m, double %conv, double %conv2) {
445 ; CHECK-LABEL: define void @reuse_shuffle_indices_cost_crash_3(
446 ; CHECK-SAME: ptr [[M:%.*]], double [[CONV:%.*]], double [[CONV2:%.*]]) {
448 ; CHECK-NEXT: [[SUB19:%.*]] = fsub double 0.000000e+00, [[CONV2]]
449 ; CHECK-NEXT: [[CONV20:%.*]] = fptrunc double [[SUB19]] to float
450 ; CHECK-NEXT: store float [[CONV20]], ptr [[M]], align 4
451 ; CHECK-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 0.000000e+00
452 ; CHECK-NEXT: [[CONV239:%.*]] = fptrunc double [[ADD]] to float
453 ; CHECK-NEXT: [[ARRAYIDX25:%.*]] = getelementptr [4 x float], ptr [[M]], i64 0, i64 1
454 ; CHECK-NEXT: store float [[CONV239]], ptr [[ARRAYIDX25]], align 4
455 ; CHECK-NEXT: [[ADD26:%.*]] = fsub double [[CONV]], [[CONV]]
456 ; CHECK-NEXT: [[CONV27:%.*]] = fptrunc double [[ADD26]] to float
457 ; CHECK-NEXT: [[ARRAYIDX29:%.*]] = getelementptr [4 x float], ptr [[M]], i64 0, i64 2
458 ; CHECK-NEXT: store float [[CONV27]], ptr [[ARRAYIDX29]], align 4
459 ; CHECK-NEXT: ret void
462 %sub19 = fsub double 0.000000e+00, %conv2
463 %conv20 = fptrunc double %sub19 to float
464 store float %conv20, ptr %m, align 4
465 %add = fadd double %conv, 0.000000e+00
466 %conv239 = fptrunc double %add to float
467 %arrayidx25 = getelementptr [4 x float], ptr %m, i64 0, i64 1
468 store float %conv239, ptr %arrayidx25, align 4
469 %add26 = fsub double %conv, %conv
470 %conv27 = fptrunc double %add26 to float
471 %arrayidx29 = getelementptr [4 x float], ptr %m, i64 0, i64 2
472 store float %conv27, ptr %arrayidx29, align 4
476 define void @reuse_shuffle_indices_cost_crash_4(double %conv7.i) {
477 ; CHECK-LABEL: define void @reuse_shuffle_indices_cost_crash_4(
478 ; CHECK-SAME: double [[CONV7_I:%.*]]) {
480 ; CHECK-NEXT: [[DATA_I111:%.*]] = alloca [0 x [0 x [0 x [3 x float]]]], i32 0, align 4
481 ; CHECK-NEXT: [[ARRAYIDX_2_I:%.*]] = getelementptr [3 x float], ptr [[DATA_I111]], i64 0, i64 2
482 ; CHECK-NEXT: [[MUL17_I_US:%.*]] = fmul double [[CONV7_I]], 0.000000e+00
483 ; CHECK-NEXT: [[MUL_2_I_I_US:%.*]] = fmul double [[MUL17_I_US]], 0.000000e+00
484 ; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> poison, double [[CONV7_I]], i32 0
485 ; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[TMP0]], <2 x double> poison, <2 x i32> zeroinitializer
486 ; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[TMP1]], zeroinitializer
487 ; CHECK-NEXT: [[ADD_2_I_I_US:%.*]] = fadd double [[MUL_2_I_I_US]], 0.000000e+00
488 ; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> [[TMP2]], [[TMP1]]
489 ; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP3]], zeroinitializer
490 ; CHECK-NEXT: [[TMP5:%.*]] = fptrunc <2 x double> [[TMP4]] to <2 x float>
491 ; CHECK-NEXT: store <2 x float> [[TMP5]], ptr [[DATA_I111]], align 4
492 ; CHECK-NEXT: [[CONV_2_I46_US:%.*]] = fptrunc double [[ADD_2_I_I_US]] to float
493 ; CHECK-NEXT: store float [[CONV_2_I46_US]], ptr [[ARRAYIDX_2_I]], align 4
494 ; CHECK-NEXT: [[CALL2_I_US:%.*]] = load volatile ptr, ptr [[DATA_I111]], align 8
495 ; CHECK-NEXT: ret void
498 %data.i111 = alloca [0 x [0 x [0 x [3 x float]]]], i32 0, align 4
499 %arrayidx.1.i = getelementptr [3 x float], ptr %data.i111, i64 0, i64 1
500 %arrayidx.2.i = getelementptr [3 x float], ptr %data.i111, i64 0, i64 2
501 %mul17.i.us = fmul double %conv7.i, 0.000000e+00
502 %mul.2.i.i.us = fmul double %mul17.i.us, 0.000000e+00
503 %add.i.i82.i.us = fadd double %conv7.i, 0.000000e+00
504 %add.1.i.i84.i.us = fadd double %conv7.i, 0.000000e+00
505 %mul.i.i91.i.us = fmul double %add.i.i82.i.us, %conv7.i
506 %mul.1.i.i92.i.us = fmul double %add.1.i.i84.i.us, %conv7.i
507 %add.i96.i.us = fadd double %mul.i.i91.i.us, 0.000000e+00
508 %add.1.i.i.us = fadd double %mul.1.i.i92.i.us, 0.000000e+00
509 %add.2.i.i.us = fadd double %mul.2.i.i.us, 0.000000e+00
510 %conv.i42.us = fptrunc double %add.i96.i.us to float
511 store float %conv.i42.us, ptr %data.i111, align 4
512 %conv.1.i44.us = fptrunc double %add.1.i.i.us to float
513 store float %conv.1.i44.us, ptr %arrayidx.1.i, align 4
514 %conv.2.i46.us = fptrunc double %add.2.i.i.us to float
515 store float %conv.2.i46.us, ptr %arrayidx.2.i, align 4
516 %call2.i.us = load volatile ptr, ptr %data.i111, align 8
520 define void @common_mask(ptr %m, double %conv, double %conv2) {
521 ; CHECK-LABEL: define void @common_mask(
522 ; CHECK-SAME: ptr [[M:%.*]], double [[CONV:%.*]], double [[CONV2:%.*]]) {
524 ; CHECK-NEXT: [[SUB19:%.*]] = fsub double [[CONV]], [[CONV]]
525 ; CHECK-NEXT: [[CONV20:%.*]] = fptrunc double [[SUB19]] to float
526 ; CHECK-NEXT: store float [[CONV20]], ptr [[M]], align 4
527 ; CHECK-NEXT: [[ADD:%.*]] = fadd double [[CONV2]], 0.000000e+00
528 ; CHECK-NEXT: [[CONV239:%.*]] = fptrunc double [[ADD]] to float
529 ; CHECK-NEXT: [[ARRAYIDX25:%.*]] = getelementptr [4 x float], ptr [[M]], i64 0, i64 1
530 ; CHECK-NEXT: store float [[CONV239]], ptr [[ARRAYIDX25]], align 4
531 ; CHECK-NEXT: [[ADD26:%.*]] = fsub double 0.000000e+00, [[CONV]]
532 ; CHECK-NEXT: [[CONV27:%.*]] = fptrunc double [[ADD26]] to float
533 ; CHECK-NEXT: [[ARRAYIDX29:%.*]] = getelementptr [4 x float], ptr [[M]], i64 0, i64 2
534 ; CHECK-NEXT: store float [[CONV27]], ptr [[ARRAYIDX29]], align 4
535 ; CHECK-NEXT: ret void
538 %sub19 = fsub double %conv, %conv
539 %conv20 = fptrunc double %sub19 to float
540 store float %conv20, ptr %m, align 4
541 %add = fadd double %conv2, 0.000000e+00
542 %conv239 = fptrunc double %add to float
543 %arrayidx25 = getelementptr [4 x float], ptr %m, i64 0, i64 1
544 store float %conv239, ptr %arrayidx25, align 4
545 %add26 = fsub double 0.000000e+00, %conv
546 %conv27 = fptrunc double %add26 to float
547 %arrayidx29 = getelementptr [4 x float], ptr %m, i64 0, i64 2
548 store float %conv27, ptr %arrayidx29, align 4
552 define void @vec3_extract(<3 x i16> %pixel.sroa.0.4.vec.insert606, ptr %call3.i536) {
553 ; NON-POW2-LABEL: define void @vec3_extract(
554 ; NON-POW2-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
555 ; NON-POW2-NEXT: entry:
556 ; NON-POW2-NEXT: store <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], ptr [[CALL3_I536]], align 2
557 ; NON-POW2-NEXT: ret void
559 ; POW2-ONLY-LABEL: define void @vec3_extract(
560 ; POW2-ONLY-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
561 ; POW2-ONLY-NEXT: entry:
562 ; POW2-ONLY-NEXT: [[PIXEL_SROA_0_4_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 2
563 ; POW2-ONLY-NEXT: [[RED668:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 2
564 ; POW2-ONLY-NEXT: store i16 [[PIXEL_SROA_0_4_VEC_EXTRACT]], ptr [[RED668]], align 2
565 ; POW2-ONLY-NEXT: [[TMP0:%.*]] = shufflevector <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
566 ; POW2-ONLY-NEXT: store <2 x i16> [[TMP0]], ptr [[CALL3_I536]], align 2
567 ; POW2-ONLY-NEXT: ret void
570 %pixel.sroa.0.4.vec.extract = extractelement <3 x i16> %pixel.sroa.0.4.vec.insert606, i64 2
571 %red668 = getelementptr i16, ptr %call3.i536, i64 2
572 store i16 %pixel.sroa.0.4.vec.extract, ptr %red668, align 2
573 %pixel.sroa.0.2.vec.extract = extractelement <3 x i16> %pixel.sroa.0.4.vec.insert606, i64 1
574 %green670 = getelementptr i16, ptr %call3.i536, i64 1
575 store i16 %pixel.sroa.0.2.vec.extract, ptr %green670, align 2
576 %pixel.sroa.0.0.vec.extract = extractelement <3 x i16> %pixel.sroa.0.4.vec.insert606, i64 0
577 store i16 %pixel.sroa.0.0.vec.extract, ptr %call3.i536, align 2
581 declare float @llvm.fmuladd.f32(float, float, float)