1// RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -x c -emit-llvm %s -o - | FileCheck %s
2// RUN: %clang_cc1 -fopenmp -x c -triple x86_64-apple-darwin10 -emit-pch -o %t %s
3// RUN: %clang_cc1 -fopenmp -x c -triple x86_64-apple-darwin10 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
4// expected-no-diagnostics
5// REQUIRES: x86-registered-target
6#ifndef HEADER
7#define HEADER
8
9_Bool bv, bx;
10char cv, cx;
11unsigned char ucv, ucx;
12short sv, sx;
13unsigned short usv, usx;
14int iv, ix;
15unsigned int uiv, uix;
16long lv, lx;
17unsigned long ulv, ulx;
18long long llv, llx;
19unsigned long long ullv, ullx;
20float fv, fx;
21double dv, dx;
22long double ldv, ldx;
23_Complex int civ, cix;
24_Complex float cfv, cfx;
25_Complex double cdv, cdx;
26
27typedef int int4 __attribute__((__vector_size__(16)));
28int4 int4x;
29
30struct BitFields {
31  int : 32;
32  int a : 31;
33} bfx;
34
35struct BitFields_packed {
36  int : 32;
37  int a : 31;
38} __attribute__ ((__packed__)) bfx_packed;
39
40struct BitFields2 {
41  int : 31;
42  int a : 1;
43} bfx2;
44
45struct BitFields2_packed {
46  int : 31;
47  int a : 1;
48} __attribute__ ((__packed__)) bfx2_packed;
49
50struct BitFields3 {
51  int : 11;
52  int a : 14;
53} bfx3;
54
55struct BitFields3_packed {
56  int : 11;
57  int a : 14;
58} __attribute__ ((__packed__)) bfx3_packed;
59
60struct BitFields4 {
61  short : 16;
62  int a: 1;
63  long b : 7;
64} bfx4;
65
66struct BitFields4_packed {
67  short : 16;
68  int a: 1;
69  long b : 7;
70} __attribute__ ((__packed__)) bfx4_packed;
71
72typedef float float2 __attribute__((ext_vector_type(2)));
73float2 float2x;
74
75// Register "0" is currently an invalid register for global register variables.
76// Use "esp" instead of "0".
77// register int rix __asm__("0");
78register int rix __asm__("esp");
79
80int main() {
81// CHECK: load atomic i8, i8*
82// CHECK: store i8
83#pragma omp atomic read
84  bv = bx;
85// CHECK: load atomic i8, i8*
86// CHECK: store i8
87#pragma omp atomic read
88  cv = cx;
89// CHECK: load atomic i8, i8*
90// CHECK: store i8
91#pragma omp atomic read
92  ucv = ucx;
93// CHECK: load atomic i16, i16*
94// CHECK: store i16
95#pragma omp atomic read
96  sv = sx;
97// CHECK: load atomic i16, i16*
98// CHECK: store i16
99#pragma omp atomic read
100  usv = usx;
101// CHECK: load atomic i32, i32*
102// CHECK: store i32
103#pragma omp atomic read
104  iv = ix;
105// CHECK: load atomic i32, i32*
106// CHECK: store i32
107#pragma omp atomic read
108  uiv = uix;
109// CHECK: load atomic i64, i64*
110// CHECK: store i64
111#pragma omp atomic read
112  lv = lx;
113// CHECK: load atomic i64, i64*
114// CHECK: store i64
115#pragma omp atomic read
116  ulv = ulx;
117// CHECK: load atomic i64, i64*
118// CHECK: store i64
119#pragma omp atomic read
120  llv = llx;
121// CHECK: load atomic i64, i64*
122// CHECK: store i64
123#pragma omp atomic read
124  ullv = ullx;
125// CHECK: load atomic i32, i32* bitcast (float*
126// CHECK: bitcast i32 {{.*}} to float
127// CHECK: store float
128#pragma omp atomic read
129  fv = fx;
130// CHECK: load atomic i64, i64* bitcast (double*
131// CHECK: bitcast i64 {{.*}} to double
132// CHECK: store double
133#pragma omp atomic read
134  dv = dx;
135// CHECK: [[LD:%.+]] = load atomic i128, i128* bitcast (x86_fp80*
136// CHECK: [[BITCAST:%.+]] = bitcast x86_fp80* [[LDTEMP:%.*]] to i128*
137// CHECK: store i128 [[LD]], i128* [[BITCAST]]
138// CHECK: [[LD:%.+]] = load x86_fp80, x86_fp80* [[LDTEMP]]
139// CHECK: store x86_fp80 [[LD]]
140#pragma omp atomic read
141  ldv = ldx;
142// CHECK: call{{.*}} void @__atomic_load(i64 8,
143// CHECK: store i32
144// CHECK: store i32
145#pragma omp atomic read
146  civ = cix;
147// CHECK: call{{.*}} void @__atomic_load(i64 8,
148// CHECK: store float
149// CHECK: store float
150#pragma omp atomic read
151  cfv = cfx;
152// CHECK: call{{.*}} void @__atomic_load(i64 16,
153// CHECK: call{{.*}} @__kmpc_flush(
154// CHECK: store double
155// CHECK: store double
156#pragma omp atomic seq_cst read
157  cdv = cdx;
158// CHECK: load atomic i64, i64*
159// CHECK: store i8
160#pragma omp atomic read
161  bv = ulx;
162// CHECK: load atomic i8, i8*
163// CHECK: store i8
164#pragma omp atomic read
165  cv = bx;
166// CHECK: load atomic i8, i8*
167// CHECK: call{{.*}} @__kmpc_flush(
168// CHECK: store i8
169#pragma omp atomic read, seq_cst
170  ucv = cx;
171// CHECK: load atomic i64, i64*
172// CHECK: store i16
173#pragma omp atomic read
174  sv = ulx;
175// CHECK: load atomic i64, i64*
176// CHECK: store i16
177#pragma omp atomic read
178  usv = lx;
179// CHECK: load atomic i32, i32*
180// CHECK: call{{.*}} @__kmpc_flush(
181// CHECK: store i32
182#pragma omp atomic seq_cst, read
183  iv = uix;
184// CHECK: load atomic i32, i32*
185// CHECK: store i32
186#pragma omp atomic read
187  uiv = ix;
188// CHECK: call{{.*}} void @__atomic_load(i64 8,
189// CHECK: store i64
190#pragma omp atomic read
191  lv = cix;
192// CHECK: load atomic i32, i32*
193// CHECK: store i64
194#pragma omp atomic read
195  ulv = fx;
196// CHECK: load atomic i64, i64*
197// CHECK: store i64
198#pragma omp atomic read
199  llv = dx;
200// CHECK: load atomic i128, i128*
201// CHECK: store i64
202#pragma omp atomic read
203  ullv = ldx;
204// CHECK: call{{.*}} void @__atomic_load(i64 8,
205// CHECK: store float
206#pragma omp atomic read
207  fv = cix;
208// CHECK: load atomic i16, i16*
209// CHECK: store double
210#pragma omp atomic read
211  dv = sx;
212// CHECK: load atomic i8, i8*
213// CHECK: store x86_fp80
214#pragma omp atomic read
215  ldv = bx;
216// CHECK: load atomic i8, i8*
217// CHECK: store i32
218// CHECK: store i32
219#pragma omp atomic read
220  civ = bx;
221// CHECK: load atomic i16, i16*
222// CHECK: store float
223// CHECK: store float
224#pragma omp atomic read
225  cfv = usx;
226// CHECK: load atomic i64, i64*
227// CHECK: store double
228// CHECK: store double
229#pragma omp atomic read
230  cdv = llx;
231// CHECK: [[I128VAL:%.+]] = load atomic i128, i128* bitcast (<4 x i32>* @{{.+}} to i128*) monotonic
232// CHECK: [[I128PTR:%.+]] = bitcast <4 x i32>* [[LDTEMP:%.+]] to i128*
233// CHECK: store i128 [[I128VAL]], i128* [[I128PTR]]
234// CHECK: [[LD:%.+]] = load <4 x i32>, <4 x i32>* [[LDTEMP]]
235// CHECK: extractelement <4 x i32> [[LD]]
236// CHECK: store i8
237#pragma omp atomic read
238  bv = int4x[0];
239// CHECK: [[LD:%.+]] = load atomic i32, i32* bitcast (i8* getelementptr (i8, i8* bitcast (%{{.+}}* @{{.+}} to i8*), i64 4) to i32*) monotonic
240// CHECK: store i32 [[LD]], i32* [[LDTEMP:%.+]]
241// CHECK: [[LD:%.+]] = load i32, i32* [[LDTEMP]]
242// CHECK: [[SHL:%.+]] = shl i32 [[LD]], 1
243// CHECK: ashr i32 [[SHL]], 1
244// CHECK: store x86_fp80
245#pragma omp atomic read
246  ldv = bfx.a;
247// CHECK: [[LDTEMP_VOID_PTR:%.+]] = bitcast i32* [[LDTEMP:%.+]] to i8*
248// CHECK: call void @__atomic_load(i64 4, i8* getelementptr (i8, i8* bitcast (%struct.BitFields_packed* @bfx_packed to i8*), i64 4), i8* [[LDTEMP_VOID_PTR]], i32 0)
249// CHECK: [[LD:%.+]] = load i32, i32* [[LDTEMP]]
250// CHECK: [[SHL:%.+]] = shl i32 [[LD]], 1
251// CHECK: ashr i32 [[SHL]], 1
252// CHECK: store x86_fp80
253#pragma omp atomic read
254  ldv = bfx_packed.a;
255// CHECK: [[LD:%.+]] = load atomic i32, i32* getelementptr inbounds (%struct.BitFields2, %struct.BitFields2* @bfx2, i32 0, i32 0) monotonic
256// CHECK: store i32 [[LD]], i32* [[LDTEMP:%.+]]
257// CHECK: [[LD:%.+]] = load i32, i32* [[LDTEMP]]
258// CHECK: ashr i32 [[LD]], 31
259// CHECK: store x86_fp80
260#pragma omp atomic read
261  ldv = bfx2.a;
262// CHECK: [[LD:%.+]] = load atomic i8, i8* getelementptr (i8, i8* bitcast (%struct.BitFields2_packed* @bfx2_packed to i8*), i64 3) monotonic
263// CHECK: store i8 [[LD]], i8* [[LDTEMP:%.+]]
264// CHECK: [[LD:%.+]] = load i8, i8* [[LDTEMP]]
265// CHECK: ashr i8 [[LD]], 7
266// CHECK: store x86_fp80
267#pragma omp atomic read
268  ldv = bfx2_packed.a;
269// CHECK: [[LD:%.+]] = load atomic i32, i32* getelementptr inbounds (%struct.BitFields3, %struct.BitFields3* @bfx3, i32 0, i32 0) monotonic
270// CHECK: store i32 [[LD]], i32* [[LDTEMP:%.+]]
271// CHECK: [[LD:%.+]] = load i32, i32* [[LDTEMP]]
272// CHECK: [[SHL:%.+]] = shl i32 [[LD]], 7
273// CHECK: ashr i32 [[SHL]], 18
274// CHECK: store x86_fp80
275#pragma omp atomic read
276  ldv = bfx3.a;
277// CHECK: [[LDTEMP_VOID_PTR:%.+]] = bitcast i24* [[LDTEMP:%.+]] to i8*
278// CHECK: call void @__atomic_load(i64 3, i8* getelementptr (i8, i8* bitcast (%struct.BitFields3_packed* @bfx3_packed to i8*), i64 1), i8* [[LDTEMP_VOID_PTR]], i32 0)
279// CHECK: [[LD:%.+]] = load i24, i24* [[LDTEMP]]
280// CHECK: [[SHL:%.+]] = shl i24 [[LD]], 7
281// CHECK: [[ASHR:%.+]] = ashr i24 [[SHL]], 10
282// CHECK: sext i24 [[ASHR]] to i32
283// CHECK: store x86_fp80
284#pragma omp atomic read
285  ldv = bfx3_packed.a;
286// CHECK: [[LD:%.+]] = load atomic i64, i64* bitcast (%struct.BitFields4* @bfx4 to i64*) monotonic
287// CHECK: store i64 [[LD]], i64* [[LDTEMP:%.+]]
288// CHECK: [[LD:%.+]] = load i64, i64* [[LDTEMP]]
289// CHECK: [[SHL:%.+]] = shl i64 [[LD]], 47
290// CHECK: [[ASHR:%.+]] = ashr i64 [[SHL]], 63
291// CHECK: trunc i64 [[ASHR]] to i32
292// CHECK: store x86_fp80
293#pragma omp atomic read
294  ldv = bfx4.a;
295// CHECK: [[LD:%.+]] = load atomic i8, i8* getelementptr inbounds (%struct.BitFields4_packed, %struct.BitFields4_packed* @bfx4_packed, i32 0, i32 0, i64 2) monotonic
296// CHECK: store i8 [[LD]], i8* [[LDTEMP:%.+]]
297// CHECK: [[LD:%.+]] = load i8, i8* [[LDTEMP]]
298// CHECK: [[SHL:%.+]] = shl i8 [[LD]], 7
299// CHECK: [[ASHR:%.+]] = ashr i8 [[SHL]], 7
300// CHECK: sext i8 [[ASHR]] to i32
301// CHECK: store x86_fp80
302#pragma omp atomic read
303  ldv = bfx4_packed.a;
304// CHECK: [[LD:%.+]] = load atomic i64, i64* bitcast (%struct.BitFields4* @bfx4 to i64*) monotonic
305// CHECK: store i64 [[LD]], i64* [[LDTEMP:%.+]]
306// CHECK: [[LD:%.+]] = load i64, i64* [[LDTEMP]]
307// CHECK: [[SHL:%.+]] = shl i64 [[LD]], 40
308// CHECK: [[ASHR:%.+]] = ashr i64 [[SHL]], 57
309// CHECK: store x86_fp80
310#pragma omp atomic read
311  ldv = bfx4.b;
312// CHECK: [[LD:%.+]] = load atomic i8, i8* getelementptr inbounds (%struct.BitFields4_packed, %struct.BitFields4_packed* @bfx4_packed, i32 0, i32 0, i64 2) monotonic
313// CHECK: store i8 [[LD]], i8* [[LDTEMP:%.+]]
314// CHECK: [[LD:%.+]] = load i8, i8* [[LDTEMP]]
315// CHECK: [[ASHR:%.+]] = ashr i8 [[LD]], 1
316// CHECK: sext i8 [[ASHR]] to i64
317// CHECK: store x86_fp80
318#pragma omp atomic read
319  ldv = bfx4_packed.b;
320// CHECK: [[LD:%.+]] = load atomic i64, i64* bitcast (<2 x float>* @{{.+}} to i64*) monotonic
321// CHECK: [[BITCAST:%.+]] = bitcast <2 x float>* [[LDTEMP:%.+]] to i64*
322// CHECK: store i64 [[LD]], i64* [[BITCAST]]
323// CHECK: [[LD:%.+]] = load <2 x float>, <2 x float>* [[LDTEMP]]
324// CHECK: extractelement <2 x float> [[LD]]
325// CHECK: store i64
326#pragma omp atomic read
327  ulv = float2x.x;
328// CHECK: call{{.*}} i{{[0-9]+}} @llvm.read_register
329// CHECK: call{{.*}} @__kmpc_flush(
330// CHECK: store double
331#pragma omp atomic read seq_cst
332  dv = rix;
333  return 0;
334}
335
336#endif
337