1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs
2 // RUN: %clang_cc1 -fopenmp-enable-irbuilder -verify -fopenmp -fopenmp-version=51 -x c -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
3 // expected-no-diagnostics
4 
5 #ifndef HEADER
6 #define HEADER
7 
8 // CHECK-LABEL: define {{.*}}@unroll_partial_factor(
9 // CHECK-NEXT:  [[ENTRY:.*]]:
10 // CHECK-NEXT:    %[[A_ADDR:.+]] = alloca float*, align 8
11 // CHECK-NEXT:    %[[B_ADDR:.+]] = alloca float*, align 8
12 // CHECK-NEXT:    %[[C_ADDR:.+]] = alloca float*, align 8
13 // CHECK-NEXT:    %[[D_ADDR:.+]] = alloca float*, align 8
14 // CHECK-NEXT:    %[[I:.+]] = alloca i32, align 4
15 // CHECK-NEXT:    %[[AGG_CAPTURED:.+]] = alloca %struct.anon, align 8
16 // CHECK-NEXT:    %[[AGG_CAPTURED1:.+]] = alloca %struct.anon.0, align 4
17 // CHECK-NEXT:    %[[DOTCOUNT_ADDR:.+]] = alloca i32, align 4
18 // CHECK-NEXT:    store float* %[[A:.+]], float** %[[A_ADDR]], align 8
19 // CHECK-NEXT:    store float* %[[B:.+]], float** %[[B_ADDR]], align 8
20 // CHECK-NEXT:    store float* %[[C:.+]], float** %[[C_ADDR]], align 8
21 // CHECK-NEXT:    store float* %[[D:.+]], float** %[[D_ADDR]], align 8
22 // CHECK-NEXT:    store i32 0, i32* %[[I]], align 4
23 // CHECK-NEXT:    %[[TMP0:.+]] = getelementptr inbounds %struct.anon, %struct.anon* %[[AGG_CAPTURED]], i32 0, i32 0
24 // CHECK-NEXT:    store i32* %[[I]], i32** %[[TMP0]], align 8
25 // CHECK-NEXT:    %[[TMP1:.+]] = getelementptr inbounds %struct.anon.0, %struct.anon.0* %[[AGG_CAPTURED1]], i32 0, i32 0
26 // CHECK-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[I]], align 4
27 // CHECK-NEXT:    store i32 %[[TMP2]], i32* %[[TMP1]], align 4
28 // CHECK-NEXT:    call void @__captured_stmt(i32* %[[DOTCOUNT_ADDR]], %struct.anon* %[[AGG_CAPTURED]])
29 // CHECK-NEXT:    %[[DOTCOUNT:.+]] = load i32, i32* %[[DOTCOUNT_ADDR]], align 4
30 // CHECK-NEXT:    br label %[[OMP_LOOP_PREHEADER:.+]]
31 // CHECK-EMPTY:
32 // CHECK-NEXT:  [[OMP_LOOP_PREHEADER]]:
33 // CHECK-NEXT:    br label %[[OMP_LOOP_HEADER:.+]]
34 // CHECK-EMPTY:
35 // CHECK-NEXT:  [[OMP_LOOP_HEADER]]:
36 // CHECK-NEXT:    %[[OMP_LOOP_IV:.+]] = phi i32 [ 0, %[[OMP_LOOP_PREHEADER]] ], [ %[[OMP_LOOP_NEXT:.+]], %[[OMP_LOOP_INC:.+]] ]
37 // CHECK-NEXT:    br label %[[OMP_LOOP_COND:.+]]
38 // CHECK-EMPTY:
39 // CHECK-NEXT:  [[OMP_LOOP_COND]]:
40 // CHECK-NEXT:    %[[OMP_LOOP_CMP:.+]] = icmp ult i32 %[[OMP_LOOP_IV]], %[[DOTCOUNT]]
41 // CHECK-NEXT:    br i1 %[[OMP_LOOP_CMP]], label %[[OMP_LOOP_BODY:.+]], label %[[OMP_LOOP_EXIT:.+]]
42 // CHECK-EMPTY:
43 // CHECK-NEXT:  [[OMP_LOOP_BODY]]:
44 // CHECK-NEXT:    call void @__captured_stmt.1(i32* %[[I]], i32 %[[OMP_LOOP_IV]], %struct.anon.0* %[[AGG_CAPTURED1]])
45 // CHECK-NEXT:    %[[TMP3:.+]] = load float*, float** %[[B_ADDR]], align 8
46 // CHECK-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[I]], align 4
47 // CHECK-NEXT:    %[[IDXPROM:.+]] = sext i32 %[[TMP4]] to i64
48 // CHECK-NEXT:    %[[ARRAYIDX:.+]] = getelementptr inbounds float, float* %[[TMP3]], i64 %[[IDXPROM]]
49 // CHECK-NEXT:    %[[TMP5:.+]] = load float, float* %[[ARRAYIDX]], align 4
50 // CHECK-NEXT:    %[[TMP6:.+]] = load float*, float** %[[C_ADDR]], align 8
51 // CHECK-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[I]], align 4
52 // CHECK-NEXT:    %[[IDXPROM2:.+]] = sext i32 %[[TMP7]] to i64
53 // CHECK-NEXT:    %[[ARRAYIDX3:.+]] = getelementptr inbounds float, float* %[[TMP6]], i64 %[[IDXPROM2]]
54 // CHECK-NEXT:    %[[TMP8:.+]] = load float, float* %[[ARRAYIDX3]], align 4
55 // CHECK-NEXT:    %[[MUL:.+]] = fmul float %[[TMP5]], %[[TMP8]]
56 // CHECK-NEXT:    %[[TMP9:.+]] = load float*, float** %[[D_ADDR]], align 8
57 // CHECK-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[I]], align 4
58 // CHECK-NEXT:    %[[IDXPROM4:.+]] = sext i32 %[[TMP10]] to i64
59 // CHECK-NEXT:    %[[ARRAYIDX5:.+]] = getelementptr inbounds float, float* %[[TMP9]], i64 %[[IDXPROM4]]
60 // CHECK-NEXT:    %[[TMP11:.+]] = load float, float* %[[ARRAYIDX5]], align 4
61 // CHECK-NEXT:    %[[MUL6:.+]] = fmul float %[[MUL]], %[[TMP11]]
62 // CHECK-NEXT:    %[[TMP12:.+]] = load float*, float** %[[A_ADDR]], align 8
63 // CHECK-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[I]], align 4
64 // CHECK-NEXT:    %[[IDXPROM7:.+]] = sext i32 %[[TMP13]] to i64
65 // CHECK-NEXT:    %[[ARRAYIDX8:.+]] = getelementptr inbounds float, float* %[[TMP12]], i64 %[[IDXPROM7]]
66 // CHECK-NEXT:    store float %[[MUL6]], float* %[[ARRAYIDX8]], align 4
67 // CHECK-NEXT:    br label %[[OMP_LOOP_INC]]
68 // CHECK-EMPTY:
69 // CHECK-NEXT:  [[OMP_LOOP_INC]]:
70 // CHECK-NEXT:    %[[OMP_LOOP_NEXT]] = add nuw i32 %[[OMP_LOOP_IV]], 1
71 // CHECK-NEXT:    br label %[[OMP_LOOP_HEADER]], !llvm.loop ![[LOOP3:[0-9]+]]
72 // CHECK-EMPTY:
73 // CHECK-NEXT:  [[OMP_LOOP_EXIT]]:
74 // CHECK-NEXT:    br label %[[OMP_LOOP_AFTER:.+]]
75 // CHECK-EMPTY:
76 // CHECK-NEXT:  [[OMP_LOOP_AFTER]]:
77 // CHECK-NEXT:    ret void
78 // CHECK-NEXT:  }
unroll_partial_factor(float * a,float * b,float * c,float * d)79 void unroll_partial_factor(float *a, float *b, float *c, float *d) {
80 #pragma omp unroll partial(3)
81   for (int i = 0; i < 2; i++) {
82     a[i] = b[i] * c[i] * d[i];
83   }
84 }
85 
86 #endif // HEADER
87 
88 // CHECK-LABEL: define {{.*}}@__captured_stmt(
89 // CHECK-NEXT:  [[ENTRY:.*]]:
90 // CHECK-NEXT:    %[[DISTANCE_ADDR:.+]] = alloca i32*, align 8
91 // CHECK-NEXT:    %[[__CONTEXT_ADDR:.+]] = alloca %struct.anon*, align 8
92 // CHECK-NEXT:    %[[DOTSTART:.+]] = alloca i32, align 4
93 // CHECK-NEXT:    %[[DOTSTOP:.+]] = alloca i32, align 4
94 // CHECK-NEXT:    %[[DOTSTEP:.+]] = alloca i32, align 4
95 // CHECK-NEXT:    store i32* %[[DISTANCE:.+]], i32** %[[DISTANCE_ADDR]], align 8
96 // CHECK-NEXT:    store %struct.anon* %[[__CONTEXT:.+]], %struct.anon** %[[__CONTEXT_ADDR]], align 8
97 // CHECK-NEXT:    %[[TMP0:.+]] = load %struct.anon*, %struct.anon** %[[__CONTEXT_ADDR]], align 8
98 // CHECK-NEXT:    %[[TMP1:.+]] = getelementptr inbounds %struct.anon, %struct.anon* %[[TMP0]], i32 0, i32 0
99 // CHECK-NEXT:    %[[TMP2:.+]] = load i32*, i32** %[[TMP1]], align 8
100 // CHECK-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[TMP2]], align 4
101 // CHECK-NEXT:    store i32 %[[TMP3]], i32* %[[DOTSTART]], align 4
102 // CHECK-NEXT:    store i32 2, i32* %[[DOTSTOP]], align 4
103 // CHECK-NEXT:    store i32 1, i32* %[[DOTSTEP]], align 4
104 // CHECK-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTSTART]], align 4
105 // CHECK-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTSTOP]], align 4
106 // CHECK-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP4]], %[[TMP5]]
107 // CHECK-NEXT:    br i1 %[[CMP]], label %[[COND_TRUE:.+]], label %[[COND_FALSE:.+]]
108 // CHECK-EMPTY:
109 // CHECK-NEXT:  [[COND_TRUE]]:
110 // CHECK-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTSTOP]], align 4
111 // CHECK-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTSTART]], align 4
112 // CHECK-NEXT:    %[[SUB:.+]] = sub nsw i32 %[[TMP6]], %[[TMP7]]
113 // CHECK-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTSTEP]], align 4
114 // CHECK-NEXT:    %[[DIV:.+]] = udiv i32 %[[SUB]], %[[TMP8]]
115 // CHECK-NEXT:    br label %[[COND_END:.+]]
116 // CHECK-EMPTY:
117 // CHECK-NEXT:  [[COND_FALSE]]:
118 // CHECK-NEXT:    br label %[[COND_END]]
119 // CHECK-EMPTY:
120 // CHECK-NEXT:  [[COND_END]]:
121 // CHECK-NEXT:    %[[COND:.+]] = phi i32 [ %[[DIV]], %[[COND_TRUE]] ], [ 0, %[[COND_FALSE]] ]
122 // CHECK-NEXT:    %[[TMP9:.+]] = load i32*, i32** %[[DISTANCE_ADDR]], align 8
123 // CHECK-NEXT:    store i32 %[[COND]], i32* %[[TMP9]], align 4
124 // CHECK-NEXT:    ret void
125 // CHECK-NEXT:  }
126 
127 
128 // CHECK-LABEL: define {{.*}}@__captured_stmt.1(
129 // CHECK-NEXT:  [[ENTRY:.*]]:
130 // CHECK-NEXT:    %[[LOOPVAR_ADDR:.+]] = alloca i32*, align 8
131 // CHECK-NEXT:    %[[LOGICAL_ADDR:.+]] = alloca i32, align 4
132 // CHECK-NEXT:    %[[__CONTEXT_ADDR:.+]] = alloca %struct.anon.0*, align 8
133 // CHECK-NEXT:    store i32* %[[LOOPVAR:.+]], i32** %[[LOOPVAR_ADDR]], align 8
134 // CHECK-NEXT:    store i32 %[[LOGICAL:.+]], i32* %[[LOGICAL_ADDR]], align 4
135 // CHECK-NEXT:    store %struct.anon.0* %[[__CONTEXT:.+]], %struct.anon.0** %[[__CONTEXT_ADDR]], align 8
136 // CHECK-NEXT:    %[[TMP0:.+]] = load %struct.anon.0*, %struct.anon.0** %[[__CONTEXT_ADDR]], align 8
137 // CHECK-NEXT:    %[[TMP1:.+]] = getelementptr inbounds %struct.anon.0, %struct.anon.0* %[[TMP0]], i32 0, i32 0
138 // CHECK-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[TMP1]], align 4
139 // CHECK-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[LOGICAL_ADDR]], align 4
140 // CHECK-NEXT:    %[[MUL:.+]] = mul i32 1, %[[TMP3]]
141 // CHECK-NEXT:    %[[ADD:.+]] = add i32 %[[TMP2]], %[[MUL]]
142 // CHECK-NEXT:    %[[TMP4:.+]] = load i32*, i32** %[[LOOPVAR_ADDR]], align 8
143 // CHECK-NEXT:    store i32 %[[ADD]], i32* %[[TMP4]], align 4
144 // CHECK-NEXT:    ret void
145 // CHECK-NEXT:  }
146 
147 
148 // CHECK: ![[META0:[0-9]+]] = !{i32 1, !"wchar_size", i32 4}
149 // CHECK: ![[META1:[0-9]+]] = !{i32 7, !"openmp", i32 51}
150 // CHECK: ![[META2:[0-9]+]] =
151 // CHECK: ![[LOOP3]] = distinct !{![[LOOP3]], ![[LOOPPROP4:[0-9]+]], ![[LOOPPROP5:[0-9]+]]}
152 // CHECK: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.enable"}
153 // CHECK: ![[LOOPPROP5]] = !{!"llvm.loop.unroll.count", i32 3}
154