1 ; RUN: opt %loadPolly %defaultOpts -polly-codegen -enable-polly-vector -dce -S %s | FileCheck %s
2 ; Obtained from C source as:
3 ; clang -S -emit-llvm -O0 vect-strided-u32-mult.c
4 ; opt -correlated-propagation -mem2reg -instcombine -loop-simplify -indvars -instnamer vect-strided-u32-mult.s > vect-strided-u32-mult.ll
6 ; ModuleID = 'vect-strided-u32-mult.s'
7 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
8 target triple = "x86_64-unknown-linux-gnu"
10 %struct.ii = type { i32, i32 }
12 define i32 @main1(i16* %arr, %struct.ii* %iarr) nounwind uwtable noinline {
14 %res = alloca [128 x i16], align 16
15 %ires = alloca [128 x %struct.ii], align 16
18 for.cond: ; preds = %for.inc, %entry
19 %indvar8 = phi i64 [ %indvar.next9, %for.inc ], [ 0, %entry ]
20 %ptr.0 = getelementptr i16* %arr, i64 %indvar8
21 %scevgep11 = getelementptr %struct.ii* %iarr, i64 %indvar8
22 %a = bitcast %struct.ii* %scevgep11 to i32*
23 %b = getelementptr %struct.ii* %iarr, i64 %indvar8, i32 1
24 %arrayidx27 = getelementptr [128 x i16]* %res, i64 0, i64 %indvar8
25 %b22 = getelementptr [128 x %struct.ii]* %ires, i64 0, i64 %indvar8, i32 1
26 %scevgep16 = getelementptr [128 x %struct.ii]* %ires, i64 0, i64 %indvar8
27 %a12 = bitcast %struct.ii* %scevgep16 to i32*
28 %exitcond = icmp ne i64 %indvar8, 128
29 br i1 %exitcond, label %for.body, label %for.end
31 for.body: ; preds = %for.cond
32 %tmp8 = load i32* %b, align 4
33 %tmp10 = load i32* %a, align 4
34 %sub = sub i32 %tmp8, %tmp10
35 store i32 %sub, i32* %a12, align 8
36 %tmp15 = load i32* %b, align 4
37 %tmp18 = load i32* %a, align 4
38 %add = add i32 %tmp15, %tmp18
39 store i32 %add, i32* %b22, align 4
40 %tmp24 = load i16* %ptr.0, align 2
41 store i16 %tmp24, i16* %arrayidx27, align 2
44 for.inc: ; preds = %for.body
45 %indvar.next9 = add i64 %indvar8, 1
48 for.end: ; preds = %for.cond
51 for.cond32: ; preds = %for.inc89, %for.end
52 %indvar = phi i64 [ %indvar.next, %for.inc89 ], [ 0, %for.end ]
53 %scevgep = getelementptr %struct.ii* %iarr, i64 %indvar
54 %a63 = bitcast %struct.ii* %scevgep to i32*
55 %b57 = getelementptr %struct.ii* %iarr, i64 %indvar, i32 1
56 %b72 = getelementptr [128 x %struct.ii]* %ires, i64 0, i64 %indvar, i32 1
57 %scevgep4 = getelementptr [128 x %struct.ii]* %ires, i64 0, i64 %indvar
58 %a51 = bitcast %struct.ii* %scevgep4 to i32*
59 %arrayidx43 = getelementptr i16* %arr, i64 %indvar
60 %arrayidx38 = getelementptr [128 x i16]* %res, i64 0, i64 %indvar
61 %i.1 = trunc i64 %indvar to i32
62 %cmp34 = icmp slt i32 %i.1, 128
63 br i1 %cmp34, label %for.body35, label %for.end92
65 for.body35: ; preds = %for.cond32
66 %tmp39 = load i16* %arrayidx38, align 2
67 %tmp44 = load i16* %arrayidx43, align 2
68 %cmp46 = icmp eq i16 %tmp39, %tmp44
69 br i1 %cmp46, label %lor.lhs.false, label %if.then
71 lor.lhs.false: ; preds = %for.body35
72 %tmp52 = load i32* %a51, align 8
73 %tmp58 = load i32* %b57, align 4
74 %tmp64 = load i32* %a63, align 4
75 %sub65 = sub i32 %tmp58, %tmp64
76 %cmp66 = icmp eq i32 %tmp52, %sub65
77 br i1 %cmp66, label %lor.lhs.false68, label %if.then
79 lor.lhs.false68: ; preds = %lor.lhs.false
80 %tmp73 = load i32* %b72, align 4
81 %tmp79 = load i32* %b57, align 4
82 %tmp85 = load i32* %a63, align 4
83 %add86 = add i32 %tmp79, %tmp85
84 %cmp87 = icmp eq i32 %tmp73, %add86
85 br i1 %cmp87, label %if.end, label %if.then
87 if.then: ; preds = %lor.lhs.false68, %lor.lhs.false, %for.body35
88 call void @abort() noreturn nounwind
91 if.end: ; preds = %lor.lhs.false68
94 for.inc89: ; preds = %if.end
95 %indvar.next = add i64 %indvar, 1
98 for.end92: ; preds = %for.cond32
102 declare void @abort() noreturn
104 define i32 @main() nounwind uwtable {
106 %arr = alloca [128 x i16], align 16
107 %iarr = alloca [128 x %struct.ii], align 16
108 call void @check_vect()
111 for.cond: ; preds = %for.inc, %entry
112 %indvar = phi i64 [ %indvar.next, %for.inc ], [ 0, %entry ]
113 %b = getelementptr [128 x %struct.ii]* %iarr, i64 0, i64 %indvar, i32 1
114 %tmp = mul i64 %indvar, 3
115 %mul = trunc i64 %tmp to i32
116 %scevgep = getelementptr [128 x %struct.ii]* %iarr, i64 0, i64 %indvar
117 %a = bitcast %struct.ii* %scevgep to i32*
118 %i.0 = trunc i64 %indvar to i32
119 %arrayidx = getelementptr [128 x i16]* %arr, i64 0, i64 %indvar
120 %conv = trunc i64 %indvar to i16
121 %exitcond = icmp ne i64 %indvar, 128
122 br i1 %exitcond, label %for.body, label %for.end
124 for.body: ; preds = %for.cond
125 store i16 %conv, i16* %arrayidx, align 2
126 store i32 %i.0, i32* %a, align 8
127 store i32 %mul, i32* %b, align 4
128 call void asm sideeffect "", "~{dirflag},~{fpsr},~{flags}"() nounwind, !srcloc !0
131 for.inc: ; preds = %for.body
132 %indvar.next = add i64 %indvar, 1
135 for.end: ; preds = %for.cond
136 %arraydecay = getelementptr inbounds [128 x i16]* %arr, i64 0, i64 0
137 %arraydecay12 = getelementptr inbounds [128 x %struct.ii]* %iarr, i64 0, i64 0
138 %call = call i32 @main1(i16* %arraydecay, %struct.ii* %arraydecay12)
142 define internal void @check_vect() nounwind uwtable noinline {
144 %a = alloca i32, align 4
145 %b = alloca i32, align 4
146 %c = alloca i32, align 4
147 %d = alloca i32, align 4
148 %call = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* @sig_ill_handler) nounwind
149 %call1 = call i32 @__get_cpuid(i32 1, i32* %a, i32* %b, i32* %c, i32* %d)
150 %tobool = icmp eq i32 %call1, 0
151 br i1 %tobool, label %if.then, label %lor.lhs.false
153 lor.lhs.false: ; preds = %entry
154 %tmp4 = load i32* %d, align 4
155 %and6 = and i32 %tmp4, 67108864
156 %cmp = icmp eq i32 %and6, 0
157 br i1 %cmp, label %if.then, label %if.end
159 if.then: ; preds = %entry, %lor.lhs.false
160 call void @exit(i32 0) noreturn nounwind
163 if.end: ; preds = %lor.lhs.false
164 %call7 = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* null) nounwind
168 declare void (i32)* @signal(i32, void (i32)*) nounwind
170 define internal void @sig_ill_handler(i32 %sig) nounwind uwtable {
172 call void @exit(i32 0) noreturn nounwind
175 return: ; No predecessors!
179 define internal i32 @__get_cpuid(i32 %__level, i32* %__eax, i32* %__ebx, i32* %__ecx, i32* %__edx) nounwind uwtable inlinehint {
181 %and = and i32 %__level, -2147483648
182 %call = call i32 @__get_cpuid_max(i32 %and, i32* null)
183 %cmp = icmp ult i32 %call, %__level
184 br i1 %cmp, label %if.then, label %if.end
186 if.then: ; preds = %entry
189 if.end: ; preds = %entry
190 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__level) nounwind, !srcloc !1
191 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
192 %asmresult8 = extractvalue { i32, i32, i32, i32 } %tmp, 1
193 %asmresult9 = extractvalue { i32, i32, i32, i32 } %tmp, 2
194 %asmresult10 = extractvalue { i32, i32, i32, i32 } %tmp, 3
195 store i32 %asmresult, i32* %__eax, align 4
196 store i32 %asmresult8, i32* %__ebx, align 4
197 store i32 %asmresult9, i32* %__ecx, align 4
198 store i32 %asmresult10, i32* %__edx, align 4
201 return: ; preds = %if.end, %if.then
202 %retval.0 = phi i32 [ 0, %if.then ], [ 1, %if.end ]
206 declare void @exit(i32) noreturn
208 define internal i32 @__get_cpuid_max(i32 %__ext, i32* %__sig) nounwind uwtable inlinehint {
210 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__ext) nounwind, !srcloc !2
211 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
212 %tobool = icmp eq i32* %__sig, null
213 br i1 %tobool, label %if.end, label %if.then
215 if.then: ; preds = %entry
216 %asmresult1 = extractvalue { i32, i32, i32, i32 } %tmp, 1
217 store i32 %asmresult1, i32* %__sig, align 4
220 if.end: ; preds = %entry, %if.then
224 !0 = metadata !{i32 926}
225 !1 = metadata !{i32 -2147343196, i32 -2147343188}
226 !2 = metadata !{i32 -2147343319, i32 -2147343311}