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-complex-2.c
4 ; opt -correlated-propagation -mem2reg -instcombine -loop-simplify -indvars -instnamer vect-complex-2.s > vect-complex-2.ll
6 ; ModuleID = 'vect-complex-2.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 @a = global [16 x { double, double }] [{ double, double } { double 1.000000e+01, double 2.000000e+01 }, { double, double } { double 1.100000e+01, double 2.100000e+01 }, { double, double } { double 1.200000e+01, double 2.200000e+01 }, { double, double } { double 1.300000e+01, double 2.300000e+01 }, { double, double } { double 1.400000e+01, double 2.400000e+01 }, { double, double } { double 1.500000e+01, double 2.500000e+01 }, { double, double } { double 1.600000e+01, double 2.600000e+01 }, { double, double } { double 1.700000e+01, double 2.700000e+01 }, { double, double } { double 1.800000e+01, double 2.800000e+01 }, { double, double } { double 1.900000e+01, double 2.900000e+01 }, { double, double } { double 2.000000e+01, double 3.000000e+01 }, { double, double } { double 2.100000e+01, double 3.100000e+01 }, { double, double } { double 2.200000e+01, double 3.200000e+01 }, { double, double } { double 2.300000e+01, double 3.300000e+01 }, { double, double } { double 2.400000e+01, double 3.400000e+01 }, { double, double } { double 2.500000e+01, double 3.500000e+01 }], align 16
11 @b = global [16 x { double, double }] [{ double, double } { double 3.000000e+01, double 4.000000e+01 }, { double, double } { double 3.100000e+01, double 4.100000e+01 }, { double, double } { double 3.200000e+01, double 4.200000e+01 }, { double, double } { double 3.300000e+01, double 4.300000e+01 }, { double, double } { double 3.400000e+01, double 4.400000e+01 }, { double, double } { double 3.500000e+01, double 4.500000e+01 }, { double, double } { double 3.600000e+01, double 4.600000e+01 }, { double, double } { double 3.700000e+01, double 4.700000e+01 }, { double, double } { double 3.800000e+01, double 4.800000e+01 }, { double, double } { double 3.900000e+01, double 4.900000e+01 }, { double, double } { double 4.000000e+01, double 5.000000e+01 }, { double, double } { double 4.100000e+01, double 5.100000e+01 }, { double, double } { double 4.200000e+01, double 5.200000e+01 }, { double, double } { double 4.300000e+01, double 5.300000e+01 }, { double, double } { double 4.400000e+01, double 5.400000e+01 }, { double, double } { double 4.500000e+01, double 5.500000e+01 }], align 16
12 @res = global [16 x { double, double }] [{ double, double } { double 4.000000e+01, double 6.000000e+01 }, { double, double } { double 4.200000e+01, double 6.200000e+01 }, { double, double } { double 4.400000e+01, double 6.400000e+01 }, { double, double } { double 4.600000e+01, double 6.600000e+01 }, { double, double } { double 4.800000e+01, double 6.800000e+01 }, { double, double } { double 5.000000e+01, double 7.000000e+01 }, { double, double } { double 5.200000e+01, double 7.200000e+01 }, { double, double } { double 5.400000e+01, double 7.400000e+01 }, { double, double } { double 5.600000e+01, double 7.600000e+01 }, { double, double } { double 5.800000e+01, double 7.800000e+01 }, { double, double } { double 6.000000e+01, double 8.000000e+01 }, { double, double } { double 6.200000e+01, double 8.200000e+01 }, { double, double } { double 6.400000e+01, double 8.400000e+01 }, { double, double } { double 6.600000e+01, double 8.600000e+01 }, { double, double } { double 6.800000e+01, double 8.800000e+01 }, { double, double } { double 7.000000e+01, double 9.000000e+01 }], align 16
13 @c = common global [16 x { double, double }] zeroinitializer, align 16
15 define void @foo() nounwind uwtable noinline {
19 for.cond: ; preds = %for.inc, %entry
20 %indvar = phi i64 [ %indvar.next, %for.inc ], [ 0, %entry ]
21 %scevgep = getelementptr [16 x { double, double }]* @c, i64 0, i64 %indvar
22 %real = bitcast { double, double }* %scevgep to double*
23 %imag = getelementptr [16 x { double, double }]* @c, i64 0, i64 %indvar, i32 1
24 %scevgep3 = getelementptr [16 x { double, double }]* @b, i64 0, i64 %indvar
25 %arrayidx4.realp = bitcast { double, double }* %scevgep3 to double*
26 %arrayidx4.imagp = getelementptr [16 x { double, double }]* @b, i64 0, i64 %indvar, i32 1
27 %scevgep6 = getelementptr [16 x { double, double }]* @a, i64 0, i64 %indvar
28 %arrayidx.realp = bitcast { double, double }* %scevgep6 to double*
29 %arrayidx.imagp = getelementptr [16 x { double, double }]* @a, i64 0, i64 %indvar, i32 1
30 %exitcond = icmp ne i64 %indvar, 16
31 br i1 %exitcond, label %for.body, label %for.end
33 for.body: ; preds = %for.cond
34 %arrayidx.real = load double* %arrayidx.realp, align 16
35 %arrayidx.imag = load double* %arrayidx.imagp, align 8
36 %arrayidx4.real = load double* %arrayidx4.realp, align 16
37 %arrayidx4.imag = load double* %arrayidx4.imagp, align 8
38 %add.r = fadd double %arrayidx.real, %arrayidx4.real
39 %add.i = fadd double %arrayidx.imag, %arrayidx4.imag
40 store double %add.r, double* %real, align 16
41 store double %add.i, double* %imag, align 8
44 for.inc: ; preds = %for.body
45 %indvar.next = add i64 %indvar, 1
48 for.end: ; preds = %for.cond
52 define i32 @main() nounwind uwtable {
54 call void @check_vect()
58 for.cond: ; preds = %for.inc, %entry
59 %indvar = phi i64 [ %indvar.next, %for.inc ], [ 0, %entry ]
60 %scevgep = getelementptr [16 x { double, double }]* @res, i64 0, i64 %indvar
61 %arrayidx4.realp = bitcast { double, double }* %scevgep to double*
62 %arrayidx4.imagp = getelementptr [16 x { double, double }]* @res, i64 0, i64 %indvar, i32 1
63 %scevgep3 = getelementptr [16 x { double, double }]* @c, i64 0, i64 %indvar
64 %arrayidx.realp = bitcast { double, double }* %scevgep3 to double*
65 %arrayidx.imagp = getelementptr [16 x { double, double }]* @c, i64 0, i64 %indvar, i32 1
66 %i.0 = trunc i64 %indvar to i32
67 %cmp = icmp slt i32 %i.0, 16
68 br i1 %cmp, label %for.body, label %for.end
70 for.body: ; preds = %for.cond
71 %arrayidx.real = load double* %arrayidx.realp, align 16
72 %arrayidx.imag = load double* %arrayidx.imagp, align 8
73 %arrayidx4.real = load double* %arrayidx4.realp, align 16
74 %arrayidx4.imag = load double* %arrayidx4.imagp, align 8
75 %cmp.r = fcmp une double %arrayidx.real, %arrayidx4.real
76 %cmp.i = fcmp une double %arrayidx.imag, %arrayidx4.imag
77 %or.ri = or i1 %cmp.r, %cmp.i
78 br i1 %or.ri, label %if.then, label %if.end
80 if.then: ; preds = %for.body
81 call void @abort() noreturn nounwind
84 if.end: ; preds = %for.body
87 for.inc: ; preds = %if.end
88 %indvar.next = add i64 %indvar, 1
91 for.end: ; preds = %for.cond
95 define internal void @check_vect() nounwind uwtable noinline {
97 %a = alloca i32, align 4
98 %b = alloca i32, align 4
99 %c = alloca i32, align 4
100 %d = alloca i32, align 4
101 %call = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* @sig_ill_handler) nounwind
102 %call1 = call i32 @__get_cpuid(i32 1, i32* %a, i32* %b, i32* %c, i32* %d)
103 %tobool = icmp eq i32 %call1, 0
104 br i1 %tobool, label %if.then, label %lor.lhs.false
106 lor.lhs.false: ; preds = %entry
107 %tmp4 = load i32* %d, align 4
108 %and6 = and i32 %tmp4, 67108864
109 %cmp = icmp eq i32 %and6, 0
110 br i1 %cmp, label %if.then, label %if.end
112 if.then: ; preds = %entry, %lor.lhs.false
113 call void @exit(i32 0) noreturn nounwind
116 if.end: ; preds = %lor.lhs.false
117 %call7 = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* null) nounwind
121 declare void @abort() noreturn
123 declare void (i32)* @signal(i32, void (i32)*) nounwind
125 define internal void @sig_ill_handler(i32 %sig) nounwind uwtable {
127 call void @exit(i32 0) noreturn nounwind
130 return: ; No predecessors!
134 define internal i32 @__get_cpuid(i32 %__level, i32* %__eax, i32* %__ebx, i32* %__ecx, i32* %__edx) nounwind uwtable inlinehint {
136 %and = and i32 %__level, -2147483648
137 %call = call i32 @__get_cpuid_max(i32 %and, i32* null)
138 %cmp = icmp ult i32 %call, %__level
139 br i1 %cmp, label %if.then, label %if.end
141 if.then: ; preds = %entry
144 if.end: ; preds = %entry
145 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__level) nounwind, !srcloc !0
146 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
147 %asmresult8 = extractvalue { i32, i32, i32, i32 } %tmp, 1
148 %asmresult9 = extractvalue { i32, i32, i32, i32 } %tmp, 2
149 %asmresult10 = extractvalue { i32, i32, i32, i32 } %tmp, 3
150 store i32 %asmresult, i32* %__eax, align 4
151 store i32 %asmresult8, i32* %__ebx, align 4
152 store i32 %asmresult9, i32* %__ecx, align 4
153 store i32 %asmresult10, i32* %__edx, align 4
156 return: ; preds = %if.end, %if.then
157 %retval.0 = phi i32 [ 0, %if.then ], [ 1, %if.end ]
161 declare void @exit(i32) noreturn
163 define internal i32 @__get_cpuid_max(i32 %__ext, i32* %__sig) nounwind uwtable inlinehint {
165 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__ext) nounwind, !srcloc !1
166 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
167 %tobool = icmp eq i32* %__sig, null
168 br i1 %tobool, label %if.end, label %if.then
170 if.then: ; preds = %entry
171 %asmresult1 = extractvalue { i32, i32, i32, i32 } %tmp, 1
172 store i32 %asmresult1, i32* %__sig, align 4
175 if.end: ; preds = %entry, %if.then
179 !0 = metadata !{i32 -2147342792, i32 -2147342784}
180 !1 = metadata !{i32 -2147342915, i32 -2147342907}