-
Notifications
You must be signed in to change notification settings - Fork 3
/
clang.patch
420 lines (395 loc) · 16.9 KB
/
clang.patch
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
diff -urN clang-3.2.src/include/clang/Basic/Attr.td clang/include/clang/Basic/Attr.td
--- clang-3.2.src/include/clang/Basic/Attr.td 2012-11-01 17:55:47.000000000 -0600
+++ clang/include/clang/Basic/Attr.td 2013-08-05 16:14:42.265182108 -0600
@@ -277,18 +277,46 @@
}
def CUDAConstant : InheritableAttr {
+ let Spellings = [GNU<"cudaconstant">];
+}
+
+def GPUConstant : InheritableAttr {
let Spellings = [GNU<"constant">];
}
def CUDADevice : InheritableAttr {
+ let Spellings = [GNU<"cudadevice">];
+}
+
+def GPUDevice : InheritableAttr {
let Spellings = [GNU<"device">];
}
+def CUDADeviceBuiltIn : InheritableAttr {
+ let Spellings = [GNU<"device_builtin">];
+}
+
+def CUDADeviceBuiltInTexture : InheritableAttr {
+ let Spellings = [GNU<"device_builtin_texture_type">];
+}
+
+def CUDADeviceBuiltInSurface : InheritableAttr {
+ let Spellings = [GNU<"device_builtin_surface_type">];
+}
+
def CUDAGlobal : InheritableAttr {
+ let Spellings = [GNU<"cudaglobal">];
+}
+
+def GPUGlobal : InheritableAttr {
let Spellings = [GNU<"global">];
}
def CUDAHost : InheritableAttr {
+ let Spellings = [GNU<"cudahost">];
+}
+
+def GPUHost : InheritableAttr {
let Spellings = [GNU<"host">];
}
@@ -298,6 +326,10 @@
}
def CUDAShared : InheritableAttr {
+ let Spellings = [GNU<"cudashared">];
+}
+
+def GPUShared : InheritableAttr {
let Spellings = [GNU<"shared">];
}
diff -urN clang-3.2.src/include/clang/Driver/CC1Options.td clang/include/clang/Driver/CC1Options.td
--- clang-3.2.src/include/clang/Driver/CC1Options.td 2012-11-21 01:33:40.000000000 -0700
+++ clang/include/clang/Driver/CC1Options.td 2013-08-05 23:44:41.857773022 -0600
@@ -532,4 +532,7 @@
def fcuda_is_device : Flag<["-"], "fcuda-is-device">,
HelpText<"Generate code for CUDA device">;
+def fgpu_device : Flag<["-"], "fgpu-device">,
+ HelpText<"Generate code for CUDA kernel device">;
+
} // let Flags = [CC1Option]
diff -urN clang-3.2.src/include/clang/Frontend/CodeGenOptions.def clang/include/clang/Frontend/CodeGenOptions.def
--- clang-3.2.src/include/clang/Frontend/CodeGenOptions.def 2012-11-09 11:27:01.000000000 -0700
+++ clang/include/clang/Frontend/CodeGenOptions.def 2013-08-05 16:03:02.233347627 -0600
@@ -31,6 +31,7 @@
CODEGENOPT(AsmVerbose , 1, 0) ///< -dA, -fverbose-asm.
CODEGENOPT(ObjCAutoRefCountExceptions , 1, 0) ///< Whether ARC should be EH-safe.
CODEGENOPT(CUDAIsDevice , 1, 0) ///< Set when compiling for CUDA device.
+CODEGENOPT(GPUDevice , 1, 0) ///< Set when compiling for CUDA device (Optional).
CODEGENOPT(CXAAtExit , 1, 1) ///< Use __cxa_atexit for calling destructors.
CODEGENOPT(CXXCtorDtorAliases, 1, 0) ///< Emit complete ctors/dtors as linker
///< aliases to base ctors when possible.
diff -urN clang-3.2.src/lib/CodeGen/CodeGenModule.cpp clang/lib/CodeGen/CodeGenModule.cpp
--- clang-3.2.src/lib/CodeGen/CodeGenModule.cpp 2012-11-21 23:10:56.000000000 -0700
+++ clang/lib/CodeGen/CodeGenModule.cpp 2013-08-05 16:49:18.866523041 -0600
@@ -45,6 +45,8 @@
#include "llvm/DataLayout.h"
#include "llvm/Support/CallSite.h"
#include "llvm/Support/ErrorHandling.h"
+#include <iostream>
+#include <fstream>
using namespace clang;
using namespace CodeGen;
@@ -953,6 +955,18 @@
} else {
const VarDecl *VD = cast<VarDecl>(Global);
assert(VD->isFileVarDecl() && "Cannot emit local var decl as global.");
+ if (VD->isThisDeclarationADefinition() != VarDecl::Definition) {
+ if (Global->hasAttr<GPUSharedAttr>()) {
+ // shared variables ...
+ std::ofstream extAttrFile;
+ extAttrFile.open("ExternAttribute.txt", std::ios::app);
+ std::string info = VD->getNameAsString();
+ info += " : shared";
+ extAttrFile << info.c_str() << std::endl;
+ extAttrFile.close();
+ }
+ return;
+ }
if (VD->isThisDeclarationADefinition() != VarDecl::Definition)
return;
@@ -1206,6 +1220,15 @@
Ty = getTypes().ConvertType(cast<ValueDecl>(GD.getDecl())->getType());
StringRef MangledName = getMangledName(GD);
+ if (LangOpts.CUDA
+ && CodeGenOpts.GPUDevice) {
+ if (GD.getDecl()->hasAttr<GPUGlobalAttr>()) {
+ std::ofstream kernelSetFile;
+ kernelSetFile.open("kernelSet.txt", std::ios::app);
+ kernelSetFile << MangledName.str() << std::endl;
+ kernelSetFile.close();
+ }
+ }
return GetOrCreateLLVMFunction(MangledName, Ty, GD, ForVTable);
}
diff -urN clang-3.2.src/lib/Frontend/CompilerInvocation.cpp clang/lib/Frontend/CompilerInvocation.cpp
--- clang-3.2.src/lib/Frontend/CompilerInvocation.cpp 2012-11-21 01:33:40.000000000 -0700
+++ clang/lib/Frontend/CompilerInvocation.cpp 2013-08-05 16:16:16.065203361 -0600
@@ -352,6 +352,7 @@
Opts.AsmVerbose = Args.hasArg(OPT_masm_verbose);
Opts.ObjCAutoRefCountExceptions = Args.hasArg(OPT_fobjc_arc_exceptions);
Opts.CUDAIsDevice = Args.hasArg(OPT_fcuda_is_device);
+ Opts.GPUDevice = Args.hasArg(OPT_fgpu_device);
Opts.CXAAtExit = !Args.hasArg(OPT_fno_use_cxa_atexit);
Opts.CXXCtorDtorAliases = Args.hasArg(OPT_mconstructor_aliases);
Opts.CodeModel = Args.getLastArgValue(OPT_mcode_model);
diff -urN clang-3.2.src/lib/Sema/SemaDeclAttr.cpp clang/lib/Sema/SemaDeclAttr.cpp
--- clang-3.2.src/lib/Sema/SemaDeclAttr.cpp 2012-11-07 14:31:36.000000000 -0700
+++ clang/lib/Sema/SemaDeclAttr.cpp 2013-08-05 16:41:54.473297482 -0600
@@ -3421,7 +3421,8 @@
S.Context));
}
-static void handleConstantAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+static void handleConstantAttr(Sema &S, Decl *D, const AttributeList &Attr,
+ bool isCUDADevice) {
if (S.LangOpts.CUDA) {
// check the attribute arguments.
if (Attr.hasParameterOrArguments()) {
@@ -3435,13 +3436,17 @@
return;
}
- D->addAttr(::new (S.Context) CUDAConstantAttr(Attr.getRange(), S.Context));
+ if (isCUDADevice)
+ D->addAttr(::new (S.Context) CUDAConstantAttr(Attr.getRange(), S.Context));
+ else
+ D->addAttr(::new (S.Context) GPUConstantAttr(Attr.getRange(), S.Context));
} else {
S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "constant";
}
}
-static void handleDeviceAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+static void handleDeviceAttr(Sema &S, Decl *D, const AttributeList &Attr,
+ bool isCUDADevice) {
if (S.LangOpts.CUDA) {
// check the attribute arguments.
if (Attr.getNumArgs() != 0) {
@@ -3455,13 +3460,77 @@
return;
}
- D->addAttr(::new (S.Context) CUDADeviceAttr(Attr.getRange(), S.Context));
+ if (isCUDADevice)
+ D->addAttr(::new (S.Context) CUDADeviceAttr(Attr.getRange(), S.Context));
+ else
+ D->addAttr(::new (S.Context) GPUDeviceAttr(Attr.getRange(), S.Context));
} else {
S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "device";
}
}
-static void handleGlobalAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+static void handleDeviceBuiltInAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+ if (S.LangOpts.CUDA) {
+ // check the attribute arguments.
+ if (Attr.getNumArgs() != 0) {
+ S.Diag(Attr.getLoc(), diag::err_attribute_wrong_number_arguments) << 0;
+ return;
+ }
+
+ if (!isa<FunctionDecl>(D) && !isa<VarDecl>(D) && !isa<TypeDecl>(D)) {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_wrong_decl_type)
+ << Attr.getName() << ExpectedVariableOrFunction;
+ return;
+ }
+
+ D->addAttr(::new (S.Context) CUDADeviceBuiltInAttr(Attr.getRange(), S.Context));
+ } else {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "device builtin";
+ }
+}
+
+static void handleDeviceBuiltInTextureAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+ if (S.LangOpts.CUDA) {
+ // check the attribute arguments.
+ if (Attr.getNumArgs() != 0) {
+ S.Diag(Attr.getLoc(), diag::err_attribute_wrong_number_arguments) << 0;
+ return;
+ }
+
+ if (!isa<TypeDecl>(D)) {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_wrong_decl_type)
+ << Attr.getName() << ExpectedVariableOrFunction;
+ return;
+ }
+
+ D->addAttr(::new (S.Context) CUDADeviceBuiltInTextureAttr(Attr.getRange(), S.Context));
+ } else {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "device builtin texture";
+ }
+}
+
+static void handleDeviceBuiltInSurfaceAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+ if (S.LangOpts.CUDA) {
+ // check the attribute arguments.
+ if (Attr.getNumArgs() != 0) {
+ S.Diag(Attr.getLoc(), diag::err_attribute_wrong_number_arguments) << 0;
+ return;
+ }
+
+ if (!isa<TypeDecl>(D)) {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_wrong_decl_type)
+ << Attr.getName() << ExpectedVariableOrFunction;
+ return;
+ }
+
+ D->addAttr(::new (S.Context) CUDADeviceBuiltInSurfaceAttr(Attr.getRange(), S.Context));
+ } else {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "device builtin surface";
+ }
+}
+
+static void handleGlobalAttr(Sema &S, Decl *D, const AttributeList &Attr,
+ bool isCUDADevice) {
if (S.LangOpts.CUDA) {
// check the attribute arguments.
if (!checkAttributeNumArgs(S, Attr, 0))
@@ -3488,13 +3557,17 @@
return;
}
- D->addAttr(::new (S.Context) CUDAGlobalAttr(Attr.getRange(), S.Context));
+ if (isCUDADevice)
+ D->addAttr(::new (S.Context) CUDAGlobalAttr(Attr.getRange(), S.Context));
+ else
+ D->addAttr(::new (S.Context) GPUGlobalAttr(Attr.getRange(), S.Context));
} else {
S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "global";
}
}
-static void handleHostAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+static void handleHostAttr(Sema &S, Decl *D, const AttributeList &Attr,
+ bool isCUDADevice) {
if (S.LangOpts.CUDA) {
// check the attribute arguments.
if (!checkAttributeNumArgs(S, Attr, 0))
@@ -3507,13 +3580,17 @@
return;
}
- D->addAttr(::new (S.Context) CUDAHostAttr(Attr.getRange(), S.Context));
+ if (isCUDADevice)
+ D->addAttr(::new (S.Context) CUDAHostAttr(Attr.getRange(), S.Context));
+ else
+ D->addAttr(::new (S.Context) GPUHostAttr(Attr.getRange(), S.Context));
} else {
S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "host";
}
}
-static void handleSharedAttr(Sema &S, Decl *D, const AttributeList &Attr) {
+static void handleSharedAttr(Sema &S, Decl *D, const AttributeList &Attr,
+ bool isCUDADevice) {
if (S.LangOpts.CUDA) {
// check the attribute arguments.
if (!checkAttributeNumArgs(S, Attr, 0))
@@ -3526,7 +3603,10 @@
return;
}
- D->addAttr(::new (S.Context) CUDASharedAttr(Attr.getRange(), S.Context));
+ if (isCUDADevice)
+ D->addAttr(::new (S.Context) CUDASharedAttr(Attr.getRange(), S.Context));
+ else
+ D->addAttr(::new (S.Context) GPUSharedAttr(Attr.getRange(), S.Context));
} else {
S.Diag(Attr.getLoc(), diag::warn_attribute_ignored) << "shared";
}
@@ -4250,8 +4330,17 @@
static void ProcessNonInheritableDeclAttr(Sema &S, Scope *scope, Decl *D,
const AttributeList &Attr) {
switch (Attr.getKind()) {
- case AttributeList::AT_CUDADevice: handleDeviceAttr (S, D, Attr); break;
- case AttributeList::AT_CUDAHost: handleHostAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDADevice:
+ handleDeviceAttr (S, D, Attr, true); break;
+ case AttributeList::AT_GPUDevice:
+ handleDeviceAttr (S, D, Attr, false); break;
+ case AttributeList::AT_CUDADeviceBuiltIn: handleDeviceBuiltInAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDADeviceBuiltInTexture: handleDeviceBuiltInTextureAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDADeviceBuiltInSurface: handleDeviceBuiltInSurfaceAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDAHost:
+ handleHostAttr (S, D, Attr, true); break;
+ case AttributeList::AT_GPUHost:
+ handleHostAttr (S, D, Attr, false); break;
case AttributeList::AT_Overloadable:handleOverloadableAttr(S, D, Attr); break;
default:
break;
@@ -4275,7 +4364,12 @@
// ProcessTypeAttributes.
break;
case AttributeList::AT_CUDADevice:
+ case AttributeList::AT_GPUDevice:
+ case AttributeList::AT_CUDADeviceBuiltIn:
+ case AttributeList::AT_CUDADeviceBuiltInTexture:
+ case AttributeList::AT_CUDADeviceBuiltInSurface:
case AttributeList::AT_CUDAHost:
+ case AttributeList::AT_GPUHost:
case AttributeList::AT_Overloadable:
// Ignore, this is a non-inheritable attribute, handled
// by ProcessNonInheritableDeclAttr.
@@ -4293,7 +4387,8 @@
case AttributeList::AT_CarriesDependency:
handleDependencyAttr (S, D, Attr); break;
case AttributeList::AT_Common: handleCommonAttr (S, D, Attr); break;
- case AttributeList::AT_CUDAConstant:handleConstantAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDAConstant:handleConstantAttr (S, D, Attr, true); break;
+ case AttributeList::AT_GPUConstant: handleConstantAttr (S, D, Attr, false); break;
case AttributeList::AT_Constructor: handleConstructorAttr (S, D, Attr); break;
case AttributeList::AT_Deprecated:
handleAttrWithMessage<DeprecatedAttr>(S, D, Attr, "deprecated");
@@ -4307,7 +4402,8 @@
break;
case AttributeList::AT_Format: handleFormatAttr (S, D, Attr); break;
case AttributeList::AT_FormatArg: handleFormatArgAttr (S, D, Attr); break;
- case AttributeList::AT_CUDAGlobal: handleGlobalAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDAGlobal: handleGlobalAttr (S, D, Attr, true); break;
+ case AttributeList::AT_GPUGlobal: handleGlobalAttr (S, D, Attr, false); break;
case AttributeList::AT_GNUInline: handleGNUInlineAttr (S, D, Attr); break;
case AttributeList::AT_CUDALaunchBounds:
handleLaunchBoundsAttr(S, D, Attr);
@@ -4326,7 +4422,8 @@
case AttributeList::AT_Naked: handleNakedAttr (S, D, Attr); break;
case AttributeList::AT_NoReturn: handleNoReturnAttr (S, D, Attr); break;
case AttributeList::AT_NoThrow: handleNothrowAttr (S, D, Attr); break;
- case AttributeList::AT_CUDAShared: handleSharedAttr (S, D, Attr); break;
+ case AttributeList::AT_CUDAShared: handleSharedAttr (S, D, Attr, true); break;
+ case AttributeList::AT_GPUShared: handleSharedAttr (S, D, Attr, false); break;
case AttributeList::AT_VecReturn: handleVecReturnAttr (S, D, Attr); break;
case AttributeList::AT_ObjCOwnership:
diff -urN clang-3.2.src/lib/Sema/SemaOverload.cpp clang/lib/Sema/SemaOverload.cpp
--- clang-3.2.src/lib/Sema/SemaOverload.cpp 2012-11-29 07:56:07.000000000 -0700
+++ clang/lib/Sema/SemaOverload.cpp 2013-08-05 16:51:48.284577297 -0600
@@ -5219,6 +5219,29 @@
return DefaultLvalueConversion(From);
}
+static bool functionBelongsToKleeFamily(std::string fName) {
+ return (fName.compare("klee_define_fixed_object") == 0
+ || fName.compare("klee_make_symbolic") == 0
+ || fName.compare("klee_range") == 0
+ || fName.compare("klee_int") == 0
+ || fName.compare("klee_silent_exit") == 0
+ || fName.compare("klee_abort") == 0
+ || fName.compare("klee_report_error") == 0
+ || fName.compare("klee_get_obj_size") == 0
+ || fName.compare("klee_print_expr") == 0
+ || fName.compare("klee_choose") == 0
+ || fName.compare("klee_is_symbolic") == 0
+ || fName.compare("klee_assume") == 0
+ || fName.find("klee_warning") != std::string::npos
+ || fName.compare("klee_prefer_cex") == 0
+ || fName.compare("klee_mark_global") == 0
+ || fName.find("klee_get_value") != std::string::npos
+ || fName.compare("klee_check_memory_access") == 0
+ || fName.compare("klee_set_forking") == 0
+ || fName.compare("klee_alias_function") == 0
+ || fName.compare("klee_stack_trace") == 0);
+}
+
/// AddOverloadCandidate - Adds the given function to the set of
/// candidate functions, using the given function call arguments. If
/// @p SuppressUserConversions, then don't allow user-defined
@@ -5314,10 +5337,13 @@
// (CUDA B.1): Check for invalid calls between targets.
if (getLangOpts().CUDA)
if (const FunctionDecl *Caller = dyn_cast<FunctionDecl>(CurContext))
- if (CheckCUDATarget(Caller, Function)) {
- Candidate.Viable = false;
- Candidate.FailureKind = ovl_fail_bad_target;
- return;
+ if (Function->getNameAsString().compare("printf") != 0
+ && !functionBelongsToKleeFamily(Function->getNameAsString())) {
+ if (CheckCUDATarget(Caller, Function)) {
+ Candidate.Viable = false;
+ Candidate.FailureKind = ovl_fail_bad_target;
+ return;
+ }
}
// Determine the implicit conversion sequences for each of the