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-shift-3.c
4 ; opt -correlated-propagation -mem2reg -instcombine -loop-simplify -indvars -instnamer vect-shift-3.s > vect-shift-3.ll
6 ; ModuleID = 'vect-shift-3.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 @src = common global [32 x i16] zeroinitializer, align 32
11 @dst = common global [32 x i16] zeroinitializer, align 32
13 define void @array_shift() nounwind uwtable noinline {
17 for.cond: ; preds = %for.inc, %entry
18 %indvar = phi i64 [ %indvar.next, %for.inc ], [ 0, %entry ]
19 %arrayidx6 = getelementptr [32 x i16]* @dst, i64 0, i64 %indvar
20 %arrayidx = getelementptr [32 x i16]* @src, i64 0, i64 %indvar
21 %exitcond = icmp ne i64 %indvar, 32
22 br i1 %exitcond, label %for.body, label %for.end
24 for.body: ; preds = %for.cond
25 %tmp2 = load i16* %arrayidx, align 2
26 %shr1 = lshr i16 %tmp2, 3
27 store i16 %shr1, i16* %arrayidx6, align 2
30 for.inc: ; preds = %for.body
31 %indvar.next = add i64 %indvar, 1
34 for.end: ; preds = %for.cond
38 define i32 @main() nounwind uwtable {
40 %i = alloca i32, align 4
41 call void @check_vect()
42 volatile store i32 0, i32* %i, align 4
45 for.cond: ; preds = %for.inc, %entry
46 %tmp = volatile load i32* %i, align 4
47 %cmp = icmp slt i32 %tmp, 32
48 br i1 %cmp, label %for.body, label %for.end
50 for.body: ; preds = %for.cond
51 %tmp1 = volatile load i32* %i, align 4
52 %shl = shl i32 %tmp1, 3
53 %conv = trunc i32 %shl to i16
54 %tmp2 = volatile load i32* %i, align 4
55 %idxprom = sext i32 %tmp2 to i64
56 %arrayidx = getelementptr inbounds [32 x i16]* @src, i64 0, i64 %idxprom
57 store i16 %conv, i16* %arrayidx, align 2
60 for.inc: ; preds = %for.body
61 %tmp3 = volatile load i32* %i, align 4
62 %inc = add nsw i32 %tmp3, 1
63 volatile store i32 %inc, i32* %i, align 4
66 for.end: ; preds = %for.cond
67 call void @array_shift()
68 volatile store i32 0, i32* %i, align 4
71 for.cond4: ; preds = %for.inc17, %for.end
72 %tmp5 = volatile load i32* %i, align 4
73 %cmp6 = icmp slt i32 %tmp5, 32
74 br i1 %cmp6, label %for.body8, label %for.end20
76 for.body8: ; preds = %for.cond4
77 %tmp9 = volatile load i32* %i, align 4
78 %idxprom10 = sext i32 %tmp9 to i64
79 %arrayidx11 = getelementptr inbounds [32 x i16]* @dst, i64 0, i64 %idxprom10
80 %tmp12 = load i16* %arrayidx11, align 2
81 %conv13 = zext i16 %tmp12 to i32
82 %tmp14 = volatile load i32* %i, align 4
83 %cmp15 = icmp eq i32 %conv13, %tmp14
84 br i1 %cmp15, label %if.end, label %if.then
86 if.then: ; preds = %for.body8
87 call void @abort() noreturn nounwind
90 if.end: ; preds = %for.body8
93 for.inc17: ; preds = %if.end
94 %tmp18 = volatile load i32* %i, align 4
95 %inc19 = add nsw i32 %tmp18, 1
96 volatile store i32 %inc19, i32* %i, align 4
99 for.end20: ; preds = %for.cond4
103 define internal void @check_vect() nounwind uwtable noinline {
105 %a = alloca i32, align 4
106 %b = alloca i32, align 4
107 %c = alloca i32, align 4
108 %d = alloca i32, align 4
109 %call = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* @sig_ill_handler) nounwind
110 %call1 = call i32 @__get_cpuid(i32 1, i32* %a, i32* %b, i32* %c, i32* %d)
111 %tobool = icmp eq i32 %call1, 0
112 br i1 %tobool, label %if.then, label %lor.lhs.false
114 lor.lhs.false: ; preds = %entry
115 %tmp4 = load i32* %d, align 4
116 %and6 = and i32 %tmp4, 67108864
117 %cmp = icmp eq i32 %and6, 0
118 br i1 %cmp, label %if.then, label %if.end
120 if.then: ; preds = %entry, %lor.lhs.false
121 call void @exit(i32 0) noreturn nounwind
124 if.end: ; preds = %lor.lhs.false
125 %call7 = call void (i32)* (i32, void (i32)*)* @signal(i32 4, void (i32)* null) nounwind
129 declare void @abort() noreturn
131 declare void (i32)* @signal(i32, void (i32)*) nounwind
133 define internal void @sig_ill_handler(i32 %sig) nounwind uwtable {
135 call void @exit(i32 0) noreturn nounwind
138 return: ; No predecessors!
142 define internal i32 @__get_cpuid(i32 %__level, i32* %__eax, i32* %__ebx, i32* %__ecx, i32* %__edx) nounwind uwtable inlinehint {
144 %and = and i32 %__level, -2147483648
145 %call = call i32 @__get_cpuid_max(i32 %and, i32* null)
146 %cmp = icmp ult i32 %call, %__level
147 br i1 %cmp, label %if.then, label %if.end
149 if.then: ; preds = %entry
152 if.end: ; preds = %entry
153 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__level) nounwind, !srcloc !0
154 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
155 %asmresult8 = extractvalue { i32, i32, i32, i32 } %tmp, 1
156 %asmresult9 = extractvalue { i32, i32, i32, i32 } %tmp, 2
157 %asmresult10 = extractvalue { i32, i32, i32, i32 } %tmp, 3
158 store i32 %asmresult, i32* %__eax, align 4
159 store i32 %asmresult8, i32* %__ebx, align 4
160 store i32 %asmresult9, i32* %__ecx, align 4
161 store i32 %asmresult10, i32* %__edx, align 4
164 return: ; preds = %if.end, %if.then
165 %retval.0 = phi i32 [ 0, %if.then ], [ 1, %if.end ]
169 declare void @exit(i32) noreturn
171 define internal i32 @__get_cpuid_max(i32 %__ext, i32* %__sig) nounwind uwtable inlinehint {
173 %tmp = call { i32, i32, i32, i32 } asm "cpuid\0A\09", "={ax},={bx},={cx},={dx},0,~{dirflag},~{fpsr},~{flags}"(i32 %__ext) nounwind, !srcloc !1
174 %asmresult = extractvalue { i32, i32, i32, i32 } %tmp, 0
175 %tobool = icmp eq i32* %__sig, null
176 br i1 %tobool, label %if.end, label %if.then
178 if.then: ; preds = %entry
179 %asmresult1 = extractvalue { i32, i32, i32, i32 } %tmp, 1
180 store i32 %asmresult1, i32* %__sig, align 4
183 if.end: ; preds = %entry, %if.then
187 !0 = metadata !{i32 -2147345724, i32 -2147345716}
188 !1 = metadata !{i32 -2147345847, i32 -2147345839}