Mercurial > hg > Gears > GearsAgda
annotate src/parallel_execution/helper_cuda.h @ 303:1dbaef86593b
CUDAtwice.cbc
| author | ikkun |
|---|---|
| date | Mon, 13 Feb 2017 18:23:29 +0900 |
| parents | src/test/helper_cuda.h@609bf62768b9 |
| children | 9755206813cb |
| rev | line source |
|---|---|
| 291 | 1 /** |
| 2 * Copyright 1993-2013 NVIDIA Corporation. All rights reserved. | |
| 3 * | |
| 4 * Please refer to the NVIDIA end user license agreement (EULA) associated | |
| 5 * with this source code for terms and conditions that govern your use of | |
| 6 * this software. Any use, reproduction, disclosure, or distribution of | |
| 7 * this software and related documentation outside the terms of the EULA | |
| 8 * is strictly prohibited. | |
| 9 * | |
| 10 */ | |
| 11 | |
| 12 //////////////////////////////////////////////////////////////////////////////// | |
| 13 // These are CUDA Helper functions for initialization and error checking | |
| 14 | |
| 15 #ifndef HELPER_CUDA_H | |
| 16 #define HELPER_CUDA_H | |
| 17 | |
| 18 #pragma once | |
| 19 | |
| 20 #include <stdlib.h> | |
| 21 #include <stdio.h> | |
| 22 #include <string.h> | |
| 23 | |
| 24 #include "helper_string.h" | |
| 25 | |
| 26 #ifndef EXIT_WAIVED | |
| 27 #define EXIT_WAIVED 2 | |
| 28 #endif | |
| 29 | |
| 30 // Note, it is required that your SDK sample to include the proper header files, please | |
| 31 // refer the CUDA examples for examples of the needed CUDA headers, which may change depending | |
| 32 // on which CUDA functions are used. | |
| 33 | |
| 34 // CUDA Runtime error messages | |
| 35 #ifdef __DRIVER_TYPES_H__ | |
| 36 static const char *_cudaGetErrorEnum(cudaError_t error) | |
| 37 { | |
| 38 switch (error) | |
| 39 { | |
| 40 case cudaSuccess: | |
| 41 return "cudaSuccess"; | |
| 42 | |
| 43 case cudaErrorMissingConfiguration: | |
| 44 return "cudaErrorMissingConfiguration"; | |
| 45 | |
| 46 case cudaErrorMemoryAllocation: | |
| 47 return "cudaErrorMemoryAllocation"; | |
| 48 | |
| 49 case cudaErrorInitializationError: | |
| 50 return "cudaErrorInitializationError"; | |
| 51 | |
| 52 case cudaErrorLaunchFailure: | |
| 53 return "cudaErrorLaunchFailure"; | |
| 54 | |
| 55 case cudaErrorPriorLaunchFailure: | |
| 56 return "cudaErrorPriorLaunchFailure"; | |
| 57 | |
| 58 case cudaErrorLaunchTimeout: | |
| 59 return "cudaErrorLaunchTimeout"; | |
| 60 | |
| 61 case cudaErrorLaunchOutOfResources: | |
| 62 return "cudaErrorLaunchOutOfResources"; | |
| 63 | |
| 64 case cudaErrorInvalidDeviceFunction: | |
| 65 return "cudaErrorInvalidDeviceFunction"; | |
| 66 | |
| 67 case cudaErrorInvalidConfiguration: | |
| 68 return "cudaErrorInvalidConfiguration"; | |
| 69 | |
| 70 case cudaErrorInvalidDevice: | |
| 71 return "cudaErrorInvalidDevice"; | |
| 72 | |
| 73 case cudaErrorInvalidValue: | |
| 74 return "cudaErrorInvalidValue"; | |
| 75 | |
| 76 case cudaErrorInvalidPitchValue: | |
| 77 return "cudaErrorInvalidPitchValue"; | |
| 78 | |
| 79 case cudaErrorInvalidSymbol: | |
| 80 return "cudaErrorInvalidSymbol"; | |
| 81 | |
| 82 case cudaErrorMapBufferObjectFailed: | |
| 83 return "cudaErrorMapBufferObjectFailed"; | |
| 84 | |
| 85 case cudaErrorUnmapBufferObjectFailed: | |
| 86 return "cudaErrorUnmapBufferObjectFailed"; | |
| 87 | |
| 88 case cudaErrorInvalidHostPointer: | |
| 89 return "cudaErrorInvalidHostPointer"; | |
| 90 | |
| 91 case cudaErrorInvalidDevicePointer: | |
| 92 return "cudaErrorInvalidDevicePointer"; | |
| 93 | |
| 94 case cudaErrorInvalidTexture: | |
| 95 return "cudaErrorInvalidTexture"; | |
| 96 | |
| 97 case cudaErrorInvalidTextureBinding: | |
| 98 return "cudaErrorInvalidTextureBinding"; | |
| 99 | |
| 100 case cudaErrorInvalidChannelDescriptor: | |
| 101 return "cudaErrorInvalidChannelDescriptor"; | |
| 102 | |
| 103 case cudaErrorInvalidMemcpyDirection: | |
| 104 return "cudaErrorInvalidMemcpyDirection"; | |
| 105 | |
| 106 case cudaErrorAddressOfConstant: | |
| 107 return "cudaErrorAddressOfConstant"; | |
| 108 | |
| 109 case cudaErrorTextureFetchFailed: | |
| 110 return "cudaErrorTextureFetchFailed"; | |
| 111 | |
| 112 case cudaErrorTextureNotBound: | |
| 113 return "cudaErrorTextureNotBound"; | |
| 114 | |
| 115 case cudaErrorSynchronizationError: | |
| 116 return "cudaErrorSynchronizationError"; | |
| 117 | |
| 118 case cudaErrorInvalidFilterSetting: | |
| 119 return "cudaErrorInvalidFilterSetting"; | |
| 120 | |
| 121 case cudaErrorInvalidNormSetting: | |
| 122 return "cudaErrorInvalidNormSetting"; | |
| 123 | |
| 124 case cudaErrorMixedDeviceExecution: | |
| 125 return "cudaErrorMixedDeviceExecution"; | |
| 126 | |
| 127 case cudaErrorCudartUnloading: | |
| 128 return "cudaErrorCudartUnloading"; | |
| 129 | |
| 130 case cudaErrorUnknown: | |
| 131 return "cudaErrorUnknown"; | |
| 132 | |
| 133 case cudaErrorNotYetImplemented: | |
| 134 return "cudaErrorNotYetImplemented"; | |
| 135 | |
| 136 case cudaErrorMemoryValueTooLarge: | |
| 137 return "cudaErrorMemoryValueTooLarge"; | |
| 138 | |
| 139 case cudaErrorInvalidResourceHandle: | |
| 140 return "cudaErrorInvalidResourceHandle"; | |
| 141 | |
| 142 case cudaErrorNotReady: | |
| 143 return "cudaErrorNotReady"; | |
| 144 | |
| 145 case cudaErrorInsufficientDriver: | |
| 146 return "cudaErrorInsufficientDriver"; | |
| 147 | |
| 148 case cudaErrorSetOnActiveProcess: | |
| 149 return "cudaErrorSetOnActiveProcess"; | |
| 150 | |
| 151 case cudaErrorInvalidSurface: | |
| 152 return "cudaErrorInvalidSurface"; | |
| 153 | |
| 154 case cudaErrorNoDevice: | |
| 155 return "cudaErrorNoDevice"; | |
| 156 | |
| 157 case cudaErrorECCUncorrectable: | |
| 158 return "cudaErrorECCUncorrectable"; | |
| 159 | |
| 160 case cudaErrorSharedObjectSymbolNotFound: | |
| 161 return "cudaErrorSharedObjectSymbolNotFound"; | |
| 162 | |
| 163 case cudaErrorSharedObjectInitFailed: | |
| 164 return "cudaErrorSharedObjectInitFailed"; | |
| 165 | |
| 166 case cudaErrorUnsupportedLimit: | |
| 167 return "cudaErrorUnsupportedLimit"; | |
| 168 | |
| 169 case cudaErrorDuplicateVariableName: | |
| 170 return "cudaErrorDuplicateVariableName"; | |
| 171 | |
| 172 case cudaErrorDuplicateTextureName: | |
| 173 return "cudaErrorDuplicateTextureName"; | |
| 174 | |
| 175 case cudaErrorDuplicateSurfaceName: | |
| 176 return "cudaErrorDuplicateSurfaceName"; | |
| 177 | |
| 178 case cudaErrorDevicesUnavailable: | |
| 179 return "cudaErrorDevicesUnavailable"; | |
| 180 | |
| 181 case cudaErrorInvalidKernelImage: | |
| 182 return "cudaErrorInvalidKernelImage"; | |
| 183 | |
| 184 case cudaErrorNoKernelImageForDevice: | |
| 185 return "cudaErrorNoKernelImageForDevice"; | |
| 186 | |
| 187 case cudaErrorIncompatibleDriverContext: | |
| 188 return "cudaErrorIncompatibleDriverContext"; | |
| 189 | |
| 190 case cudaErrorPeerAccessAlreadyEnabled: | |
| 191 return "cudaErrorPeerAccessAlreadyEnabled"; | |
| 192 | |
| 193 case cudaErrorPeerAccessNotEnabled: | |
| 194 return "cudaErrorPeerAccessNotEnabled"; | |
| 195 | |
| 196 case cudaErrorDeviceAlreadyInUse: | |
| 197 return "cudaErrorDeviceAlreadyInUse"; | |
| 198 | |
| 199 case cudaErrorProfilerDisabled: | |
| 200 return "cudaErrorProfilerDisabled"; | |
| 201 | |
| 202 case cudaErrorProfilerNotInitialized: | |
| 203 return "cudaErrorProfilerNotInitialized"; | |
| 204 | |
| 205 case cudaErrorProfilerAlreadyStarted: | |
| 206 return "cudaErrorProfilerAlreadyStarted"; | |
| 207 | |
| 208 case cudaErrorProfilerAlreadyStopped: | |
| 209 return "cudaErrorProfilerAlreadyStopped"; | |
| 210 | |
| 211 /* Since CUDA 4.0*/ | |
| 212 case cudaErrorAssert: | |
| 213 return "cudaErrorAssert"; | |
| 214 | |
| 215 case cudaErrorTooManyPeers: | |
| 216 return "cudaErrorTooManyPeers"; | |
| 217 | |
| 218 case cudaErrorHostMemoryAlreadyRegistered: | |
| 219 return "cudaErrorHostMemoryAlreadyRegistered"; | |
| 220 | |
| 221 case cudaErrorHostMemoryNotRegistered: | |
| 222 return "cudaErrorHostMemoryNotRegistered"; | |
| 223 | |
| 224 /* Since CUDA 5.0 */ | |
| 225 case cudaErrorOperatingSystem: | |
| 226 return "cudaErrorOperatingSystem"; | |
| 227 | |
| 228 case cudaErrorPeerAccessUnsupported: | |
| 229 return "cudaErrorPeerAccessUnsupported"; | |
| 230 | |
| 231 case cudaErrorLaunchMaxDepthExceeded: | |
| 232 return "cudaErrorLaunchMaxDepthExceeded"; | |
| 233 | |
| 234 case cudaErrorLaunchFileScopedTex: | |
| 235 return "cudaErrorLaunchFileScopedTex"; | |
| 236 | |
| 237 case cudaErrorLaunchFileScopedSurf: | |
| 238 return "cudaErrorLaunchFileScopedSurf"; | |
| 239 | |
| 240 case cudaErrorSyncDepthExceeded: | |
| 241 return "cudaErrorSyncDepthExceeded"; | |
| 242 | |
| 243 case cudaErrorLaunchPendingCountExceeded: | |
| 244 return "cudaErrorLaunchPendingCountExceeded"; | |
| 245 | |
| 246 case cudaErrorNotPermitted: | |
| 247 return "cudaErrorNotPermitted"; | |
| 248 | |
| 249 case cudaErrorNotSupported: | |
| 250 return "cudaErrorNotSupported"; | |
| 251 | |
| 252 /* Since CUDA 6.0 */ | |
| 253 case cudaErrorHardwareStackError: | |
| 254 return "cudaErrorHardwareStackError"; | |
| 255 | |
| 256 case cudaErrorIllegalInstruction: | |
| 257 return "cudaErrorIllegalInstruction"; | |
| 258 | |
| 259 case cudaErrorMisalignedAddress: | |
| 260 return "cudaErrorMisalignedAddress"; | |
| 261 | |
| 262 case cudaErrorInvalidAddressSpace: | |
| 263 return "cudaErrorInvalidAddressSpace"; | |
| 264 | |
| 265 case cudaErrorInvalidPc: | |
| 266 return "cudaErrorInvalidPc"; | |
| 267 | |
| 268 case cudaErrorIllegalAddress: | |
| 269 return "cudaErrorIllegalAddress"; | |
| 270 | |
| 271 /* Since CUDA 6.5*/ | |
| 272 case cudaErrorInvalidPtx: | |
| 273 return "cudaErrorInvalidPtx"; | |
| 274 | |
| 275 case cudaErrorInvalidGraphicsContext: | |
| 276 return "cudaErrorInvalidGraphicsContext"; | |
| 277 | |
| 278 case cudaErrorStartupFailure: | |
| 279 return "cudaErrorStartupFailure"; | |
| 280 | |
| 281 case cudaErrorApiFailureBase: | |
| 282 return "cudaErrorApiFailureBase"; | |
| 283 | |
| 284 /* Since CUDA 8.0*/ | |
| 285 case cudaErrorNvlinkUncorrectable : | |
| 286 return "cudaErrorNvlinkUncorrectable"; | |
| 287 } | |
| 288 | |
| 289 return "<unknown>"; | |
| 290 } | |
| 291 #endif | |
| 292 | |
| 293 #ifdef __cuda_cuda_h__ | |
| 294 // CUDA Driver API errors | |
|
301
609bf62768b9
add -DUSE_CUDA=1 flag to cmake
Shinji KONO <kono@ie.u-ryukyu.ac.jp>
parents:
291
diff
changeset
|
295 const char *_cudaGetErrorEnum(CUresult error) |
| 291 | 296 { |
| 297 switch (error) | |
| 298 { | |
| 299 case CUDA_SUCCESS: | |
| 300 return "CUDA_SUCCESS"; | |
| 301 | |
| 302 case CUDA_ERROR_INVALID_VALUE: | |
| 303 return "CUDA_ERROR_INVALID_VALUE"; | |
| 304 | |
| 305 case CUDA_ERROR_OUT_OF_MEMORY: | |
| 306 return "CUDA_ERROR_OUT_OF_MEMORY"; | |
| 307 | |
| 308 case CUDA_ERROR_NOT_INITIALIZED: | |
| 309 return "CUDA_ERROR_NOT_INITIALIZED"; | |
| 310 | |
| 311 case CUDA_ERROR_DEINITIALIZED: | |
| 312 return "CUDA_ERROR_DEINITIALIZED"; | |
| 313 | |
| 314 case CUDA_ERROR_PROFILER_DISABLED: | |
| 315 return "CUDA_ERROR_PROFILER_DISABLED"; | |
| 316 | |
| 317 case CUDA_ERROR_PROFILER_NOT_INITIALIZED: | |
| 318 return "CUDA_ERROR_PROFILER_NOT_INITIALIZED"; | |
| 319 | |
| 320 case CUDA_ERROR_PROFILER_ALREADY_STARTED: | |
| 321 return "CUDA_ERROR_PROFILER_ALREADY_STARTED"; | |
| 322 | |
| 323 case CUDA_ERROR_PROFILER_ALREADY_STOPPED: | |
| 324 return "CUDA_ERROR_PROFILER_ALREADY_STOPPED"; | |
| 325 | |
| 326 case CUDA_ERROR_NO_DEVICE: | |
| 327 return "CUDA_ERROR_NO_DEVICE"; | |
| 328 | |
| 329 case CUDA_ERROR_INVALID_DEVICE: | |
| 330 return "CUDA_ERROR_INVALID_DEVICE"; | |
| 331 | |
| 332 case CUDA_ERROR_INVALID_IMAGE: | |
| 333 return "CUDA_ERROR_INVALID_IMAGE"; | |
| 334 | |
| 335 case CUDA_ERROR_INVALID_CONTEXT: | |
| 336 return "CUDA_ERROR_INVALID_CONTEXT"; | |
| 337 | |
| 338 case CUDA_ERROR_CONTEXT_ALREADY_CURRENT: | |
| 339 return "CUDA_ERROR_CONTEXT_ALREADY_CURRENT"; | |
| 340 | |
| 341 case CUDA_ERROR_MAP_FAILED: | |
| 342 return "CUDA_ERROR_MAP_FAILED"; | |
| 343 | |
| 344 case CUDA_ERROR_UNMAP_FAILED: | |
| 345 return "CUDA_ERROR_UNMAP_FAILED"; | |
| 346 | |
| 347 case CUDA_ERROR_ARRAY_IS_MAPPED: | |
| 348 return "CUDA_ERROR_ARRAY_IS_MAPPED"; | |
| 349 | |
| 350 case CUDA_ERROR_ALREADY_MAPPED: | |
| 351 return "CUDA_ERROR_ALREADY_MAPPED"; | |
| 352 | |
| 353 case CUDA_ERROR_NO_BINARY_FOR_GPU: | |
| 354 return "CUDA_ERROR_NO_BINARY_FOR_GPU"; | |
| 355 | |
| 356 case CUDA_ERROR_ALREADY_ACQUIRED: | |
| 357 return "CUDA_ERROR_ALREADY_ACQUIRED"; | |
| 358 | |
| 359 case CUDA_ERROR_NOT_MAPPED: | |
| 360 return "CUDA_ERROR_NOT_MAPPED"; | |
| 361 | |
| 362 case CUDA_ERROR_NOT_MAPPED_AS_ARRAY: | |
| 363 return "CUDA_ERROR_NOT_MAPPED_AS_ARRAY"; | |
| 364 | |
| 365 case CUDA_ERROR_NOT_MAPPED_AS_POINTER: | |
| 366 return "CUDA_ERROR_NOT_MAPPED_AS_POINTER"; | |
| 367 | |
| 368 case CUDA_ERROR_ECC_UNCORRECTABLE: | |
| 369 return "CUDA_ERROR_ECC_UNCORRECTABLE"; | |
| 370 | |
| 371 case CUDA_ERROR_UNSUPPORTED_LIMIT: | |
| 372 return "CUDA_ERROR_UNSUPPORTED_LIMIT"; | |
| 373 | |
| 374 case CUDA_ERROR_CONTEXT_ALREADY_IN_USE: | |
| 375 return "CUDA_ERROR_CONTEXT_ALREADY_IN_USE"; | |
| 376 | |
| 377 case CUDA_ERROR_PEER_ACCESS_UNSUPPORTED: | |
| 378 return "CUDA_ERROR_PEER_ACCESS_UNSUPPORTED"; | |
| 379 | |
| 380 case CUDA_ERROR_INVALID_PTX: | |
| 381 return "CUDA_ERROR_INVALID_PTX"; | |
| 382 | |
| 383 case CUDA_ERROR_INVALID_GRAPHICS_CONTEXT: | |
| 384 return "CUDA_ERROR_INVALID_GRAPHICS_CONTEXT"; | |
| 385 | |
| 386 case CUDA_ERROR_NVLINK_UNCORRECTABLE: | |
| 387 return "CUDA_ERROR_NVLINK_UNCORRECTABLE"; | |
| 388 | |
| 389 case CUDA_ERROR_INVALID_SOURCE: | |
| 390 return "CUDA_ERROR_INVALID_SOURCE"; | |
| 391 | |
| 392 case CUDA_ERROR_FILE_NOT_FOUND: | |
| 393 return "CUDA_ERROR_FILE_NOT_FOUND"; | |
| 394 | |
| 395 case CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND: | |
| 396 return "CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND"; | |
| 397 | |
| 398 case CUDA_ERROR_SHARED_OBJECT_INIT_FAILED: | |
| 399 return "CUDA_ERROR_SHARED_OBJECT_INIT_FAILED"; | |
| 400 | |
| 401 case CUDA_ERROR_OPERATING_SYSTEM: | |
| 402 return "CUDA_ERROR_OPERATING_SYSTEM"; | |
| 403 | |
| 404 case CUDA_ERROR_INVALID_HANDLE: | |
| 405 return "CUDA_ERROR_INVALID_HANDLE"; | |
| 406 | |
| 407 case CUDA_ERROR_NOT_FOUND: | |
| 408 return "CUDA_ERROR_NOT_FOUND"; | |
| 409 | |
| 410 case CUDA_ERROR_NOT_READY: | |
| 411 return "CUDA_ERROR_NOT_READY"; | |
| 412 | |
| 413 case CUDA_ERROR_ILLEGAL_ADDRESS: | |
| 414 return "CUDA_ERROR_ILLEGAL_ADDRESS"; | |
| 415 | |
| 416 case CUDA_ERROR_LAUNCH_FAILED: | |
| 417 return "CUDA_ERROR_LAUNCH_FAILED"; | |
| 418 | |
| 419 case CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES: | |
| 420 return "CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES"; | |
| 421 | |
| 422 case CUDA_ERROR_LAUNCH_TIMEOUT: | |
| 423 return "CUDA_ERROR_LAUNCH_TIMEOUT"; | |
| 424 | |
| 425 case CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING: | |
| 426 return "CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING"; | |
| 427 | |
| 428 case CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED: | |
| 429 return "CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED"; | |
| 430 | |
| 431 case CUDA_ERROR_PEER_ACCESS_NOT_ENABLED: | |
| 432 return "CUDA_ERROR_PEER_ACCESS_NOT_ENABLED"; | |
| 433 | |
| 434 case CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE: | |
| 435 return "CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE"; | |
| 436 | |
| 437 case CUDA_ERROR_CONTEXT_IS_DESTROYED: | |
| 438 return "CUDA_ERROR_CONTEXT_IS_DESTROYED"; | |
| 439 | |
| 440 case CUDA_ERROR_ASSERT: | |
| 441 return "CUDA_ERROR_ASSERT"; | |
| 442 | |
| 443 case CUDA_ERROR_TOO_MANY_PEERS: | |
| 444 return "CUDA_ERROR_TOO_MANY_PEERS"; | |
| 445 | |
| 446 case CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED: | |
| 447 return "CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED"; | |
| 448 | |
| 449 case CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED: | |
| 450 return "CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED"; | |
| 451 | |
| 452 case CUDA_ERROR_HARDWARE_STACK_ERROR: | |
| 453 return "CUDA_ERROR_HARDWARE_STACK_ERROR"; | |
| 454 | |
| 455 case CUDA_ERROR_ILLEGAL_INSTRUCTION: | |
| 456 return "CUDA_ERROR_ILLEGAL_INSTRUCTION"; | |
| 457 | |
| 458 case CUDA_ERROR_MISALIGNED_ADDRESS: | |
| 459 return "CUDA_ERROR_MISALIGNED_ADDRESS"; | |
| 460 | |
| 461 case CUDA_ERROR_INVALID_ADDRESS_SPACE: | |
| 462 return "CUDA_ERROR_INVALID_ADDRESS_SPACE"; | |
| 463 | |
| 464 case CUDA_ERROR_INVALID_PC: | |
| 465 return "CUDA_ERROR_INVALID_PC"; | |
| 466 | |
| 467 case CUDA_ERROR_NOT_PERMITTED: | |
| 468 return "CUDA_ERROR_NOT_PERMITTED"; | |
| 469 | |
| 470 case CUDA_ERROR_NOT_SUPPORTED: | |
| 471 return "CUDA_ERROR_NOT_SUPPORTED"; | |
| 472 | |
| 473 case CUDA_ERROR_UNKNOWN: | |
| 474 return "CUDA_ERROR_UNKNOWN"; | |
| 475 } | |
| 476 | |
| 477 return "<unknown>"; | |
| 478 } | |
| 479 #endif | |
| 480 | |
| 481 #ifdef CUBLAS_API_H_ | |
| 482 // cuBLAS API errors | |
| 483 static const char *_cudaGetErrorEnum(cublasStatus_t error) | |
| 484 { | |
| 485 switch (error) | |
| 486 { | |
| 487 case CUBLAS_STATUS_SUCCESS: | |
| 488 return "CUBLAS_STATUS_SUCCESS"; | |
| 489 | |
| 490 case CUBLAS_STATUS_NOT_INITIALIZED: | |
| 491 return "CUBLAS_STATUS_NOT_INITIALIZED"; | |
| 492 | |
| 493 case CUBLAS_STATUS_ALLOC_FAILED: | |
| 494 return "CUBLAS_STATUS_ALLOC_FAILED"; | |
| 495 | |
| 496 case CUBLAS_STATUS_INVALID_VALUE: | |
| 497 return "CUBLAS_STATUS_INVALID_VALUE"; | |
| 498 | |
| 499 case CUBLAS_STATUS_ARCH_MISMATCH: | |
| 500 return "CUBLAS_STATUS_ARCH_MISMATCH"; | |
| 501 | |
| 502 case CUBLAS_STATUS_MAPPING_ERROR: | |
| 503 return "CUBLAS_STATUS_MAPPING_ERROR"; | |
| 504 | |
| 505 case CUBLAS_STATUS_EXECUTION_FAILED: | |
| 506 return "CUBLAS_STATUS_EXECUTION_FAILED"; | |
| 507 | |
| 508 case CUBLAS_STATUS_INTERNAL_ERROR: | |
| 509 return "CUBLAS_STATUS_INTERNAL_ERROR"; | |
| 510 | |
| 511 case CUBLAS_STATUS_NOT_SUPPORTED: | |
| 512 return "CUBLAS_STATUS_NOT_SUPPORTED"; | |
| 513 | |
| 514 case CUBLAS_STATUS_LICENSE_ERROR: | |
| 515 return "CUBLAS_STATUS_LICENSE_ERROR"; | |
| 516 } | |
| 517 | |
| 518 return "<unknown>"; | |
| 519 } | |
| 520 #endif | |
| 521 | |
| 522 #ifdef _CUFFT_H_ | |
| 523 // cuFFT API errors | |
| 524 static const char *_cudaGetErrorEnum(cufftResult error) | |
| 525 { | |
| 526 switch (error) | |
| 527 { | |
| 528 case CUFFT_SUCCESS: | |
| 529 return "CUFFT_SUCCESS"; | |
| 530 | |
| 531 case CUFFT_INVALID_PLAN: | |
| 532 return "CUFFT_INVALID_PLAN"; | |
| 533 | |
| 534 case CUFFT_ALLOC_FAILED: | |
| 535 return "CUFFT_ALLOC_FAILED"; | |
| 536 | |
| 537 case CUFFT_INVALID_TYPE: | |
| 538 return "CUFFT_INVALID_TYPE"; | |
| 539 | |
| 540 case CUFFT_INVALID_VALUE: | |
| 541 return "CUFFT_INVALID_VALUE"; | |
| 542 | |
| 543 case CUFFT_INTERNAL_ERROR: | |
| 544 return "CUFFT_INTERNAL_ERROR"; | |
| 545 | |
| 546 case CUFFT_EXEC_FAILED: | |
| 547 return "CUFFT_EXEC_FAILED"; | |
| 548 | |
| 549 case CUFFT_SETUP_FAILED: | |
| 550 return "CUFFT_SETUP_FAILED"; | |
| 551 | |
| 552 case CUFFT_INVALID_SIZE: | |
| 553 return "CUFFT_INVALID_SIZE"; | |
| 554 | |
| 555 case CUFFT_UNALIGNED_DATA: | |
| 556 return "CUFFT_UNALIGNED_DATA"; | |
| 557 | |
| 558 case CUFFT_INCOMPLETE_PARAMETER_LIST: | |
| 559 return "CUFFT_INCOMPLETE_PARAMETER_LIST"; | |
| 560 | |
| 561 case CUFFT_INVALID_DEVICE: | |
| 562 return "CUFFT_INVALID_DEVICE"; | |
| 563 | |
| 564 case CUFFT_PARSE_ERROR: | |
| 565 return "CUFFT_PARSE_ERROR"; | |
| 566 | |
| 567 case CUFFT_NO_WORKSPACE: | |
| 568 return "CUFFT_NO_WORKSPACE"; | |
| 569 | |
| 570 case CUFFT_NOT_IMPLEMENTED: | |
| 571 return "CUFFT_NOT_IMPLEMENTED"; | |
| 572 | |
| 573 case CUFFT_LICENSE_ERROR: | |
| 574 return "CUFFT_LICENSE_ERROR"; | |
| 575 | |
| 576 case CUFFT_NOT_SUPPORTED: | |
| 577 return "CUFFT_NOT_SUPPORTED"; | |
| 578 } | |
| 579 | |
| 580 return "<unknown>"; | |
| 581 } | |
| 582 #endif | |
| 583 | |
| 584 | |
| 585 #ifdef CUSPARSEAPI | |
| 586 // cuSPARSE API errors | |
| 587 static const char *_cudaGetErrorEnum(cusparseStatus_t error) | |
| 588 { | |
| 589 switch (error) | |
| 590 { | |
| 591 case CUSPARSE_STATUS_SUCCESS: | |
| 592 return "CUSPARSE_STATUS_SUCCESS"; | |
| 593 | |
| 594 case CUSPARSE_STATUS_NOT_INITIALIZED: | |
| 595 return "CUSPARSE_STATUS_NOT_INITIALIZED"; | |
| 596 | |
| 597 case CUSPARSE_STATUS_ALLOC_FAILED: | |
| 598 return "CUSPARSE_STATUS_ALLOC_FAILED"; | |
| 599 | |
| 600 case CUSPARSE_STATUS_INVALID_VALUE: | |
| 601 return "CUSPARSE_STATUS_INVALID_VALUE"; | |
| 602 | |
| 603 case CUSPARSE_STATUS_ARCH_MISMATCH: | |
| 604 return "CUSPARSE_STATUS_ARCH_MISMATCH"; | |
| 605 | |
| 606 case CUSPARSE_STATUS_MAPPING_ERROR: | |
| 607 return "CUSPARSE_STATUS_MAPPING_ERROR"; | |
| 608 | |
| 609 case CUSPARSE_STATUS_EXECUTION_FAILED: | |
| 610 return "CUSPARSE_STATUS_EXECUTION_FAILED"; | |
| 611 | |
| 612 case CUSPARSE_STATUS_INTERNAL_ERROR: | |
| 613 return "CUSPARSE_STATUS_INTERNAL_ERROR"; | |
| 614 | |
| 615 case CUSPARSE_STATUS_MATRIX_TYPE_NOT_SUPPORTED: | |
| 616 return "CUSPARSE_STATUS_MATRIX_TYPE_NOT_SUPPORTED"; | |
| 617 } | |
| 618 | |
| 619 return "<unknown>"; | |
| 620 } | |
| 621 #endif | |
| 622 | |
| 623 #ifdef CUSOLVER_COMMON_H_ | |
| 624 //cuSOLVER API errors | |
| 625 static const char *_cudaGetErrorEnum(cusolverStatus_t error) | |
| 626 { | |
| 627 switch(error) | |
| 628 { | |
| 629 case CUSOLVER_STATUS_SUCCESS: | |
| 630 return "CUSOLVER_STATUS_SUCCESS"; | |
| 631 case CUSOLVER_STATUS_NOT_INITIALIZED: | |
| 632 return "CUSOLVER_STATUS_NOT_INITIALIZED"; | |
| 633 case CUSOLVER_STATUS_ALLOC_FAILED: | |
| 634 return "CUSOLVER_STATUS_ALLOC_FAILED"; | |
| 635 case CUSOLVER_STATUS_INVALID_VALUE: | |
| 636 return "CUSOLVER_STATUS_INVALID_VALUE"; | |
| 637 case CUSOLVER_STATUS_ARCH_MISMATCH: | |
| 638 return "CUSOLVER_STATUS_ARCH_MISMATCH"; | |
| 639 case CUSOLVER_STATUS_MAPPING_ERROR: | |
| 640 return "CUSOLVER_STATUS_MAPPING_ERROR"; | |
| 641 case CUSOLVER_STATUS_EXECUTION_FAILED: | |
| 642 return "CUSOLVER_STATUS_EXECUTION_FAILED"; | |
| 643 case CUSOLVER_STATUS_INTERNAL_ERROR: | |
| 644 return "CUSOLVER_STATUS_INTERNAL_ERROR"; | |
| 645 case CUSOLVER_STATUS_MATRIX_TYPE_NOT_SUPPORTED: | |
| 646 return "CUSOLVER_STATUS_MATRIX_TYPE_NOT_SUPPORTED"; | |
| 647 case CUSOLVER_STATUS_NOT_SUPPORTED : | |
| 648 return "CUSOLVER_STATUS_NOT_SUPPORTED "; | |
| 649 case CUSOLVER_STATUS_ZERO_PIVOT: | |
| 650 return "CUSOLVER_STATUS_ZERO_PIVOT"; | |
| 651 case CUSOLVER_STATUS_INVALID_LICENSE: | |
| 652 return "CUSOLVER_STATUS_INVALID_LICENSE"; | |
| 653 } | |
| 654 | |
| 655 return "<unknown>"; | |
| 656 | |
| 657 } | |
| 658 #endif | |
| 659 | |
| 660 #ifdef CURAND_H_ | |
| 661 // cuRAND API errors | |
| 662 static const char *_cudaGetErrorEnum(curandStatus_t error) | |
| 663 { | |
| 664 switch (error) | |
| 665 { | |
| 666 case CURAND_STATUS_SUCCESS: | |
| 667 return "CURAND_STATUS_SUCCESS"; | |
| 668 | |
| 669 case CURAND_STATUS_VERSION_MISMATCH: | |
| 670 return "CURAND_STATUS_VERSION_MISMATCH"; | |
| 671 | |
| 672 case CURAND_STATUS_NOT_INITIALIZED: | |
| 673 return "CURAND_STATUS_NOT_INITIALIZED"; | |
| 674 | |
| 675 case CURAND_STATUS_ALLOCATION_FAILED: | |
| 676 return "CURAND_STATUS_ALLOCATION_FAILED"; | |
| 677 | |
| 678 case CURAND_STATUS_TYPE_ERROR: | |
| 679 return "CURAND_STATUS_TYPE_ERROR"; | |
| 680 | |
| 681 case CURAND_STATUS_OUT_OF_RANGE: | |
| 682 return "CURAND_STATUS_OUT_OF_RANGE"; | |
| 683 | |
| 684 case CURAND_STATUS_LENGTH_NOT_MULTIPLE: | |
| 685 return "CURAND_STATUS_LENGTH_NOT_MULTIPLE"; | |
| 686 | |
| 687 case CURAND_STATUS_DOUBLE_PRECISION_REQUIRED: | |
| 688 return "CURAND_STATUS_DOUBLE_PRECISION_REQUIRED"; | |
| 689 | |
| 690 case CURAND_STATUS_LAUNCH_FAILURE: | |
| 691 return "CURAND_STATUS_LAUNCH_FAILURE"; | |
| 692 | |
| 693 case CURAND_STATUS_PREEXISTING_FAILURE: | |
| 694 return "CURAND_STATUS_PREEXISTING_FAILURE"; | |
| 695 | |
| 696 case CURAND_STATUS_INITIALIZATION_FAILED: | |
| 697 return "CURAND_STATUS_INITIALIZATION_FAILED"; | |
| 698 | |
| 699 case CURAND_STATUS_ARCH_MISMATCH: | |
| 700 return "CURAND_STATUS_ARCH_MISMATCH"; | |
| 701 | |
| 702 case CURAND_STATUS_INTERNAL_ERROR: | |
| 703 return "CURAND_STATUS_INTERNAL_ERROR"; | |
| 704 } | |
| 705 | |
| 706 return "<unknown>"; | |
| 707 } | |
| 708 #endif | |
| 709 | |
| 710 #ifdef NV_NPPIDEFS_H | |
| 711 // NPP API errors | |
| 712 static const char *_cudaGetErrorEnum(NppStatus error) | |
| 713 { | |
| 714 switch (error) | |
| 715 { | |
| 716 case NPP_NOT_SUPPORTED_MODE_ERROR: | |
| 717 return "NPP_NOT_SUPPORTED_MODE_ERROR"; | |
| 718 | |
| 719 case NPP_ROUND_MODE_NOT_SUPPORTED_ERROR: | |
| 720 return "NPP_ROUND_MODE_NOT_SUPPORTED_ERROR"; | |
| 721 | |
| 722 case NPP_RESIZE_NO_OPERATION_ERROR: | |
| 723 return "NPP_RESIZE_NO_OPERATION_ERROR"; | |
| 724 | |
| 725 case NPP_NOT_SUFFICIENT_COMPUTE_CAPABILITY: | |
| 726 return "NPP_NOT_SUFFICIENT_COMPUTE_CAPABILITY"; | |
| 727 | |
| 728 #if ((NPP_VERSION_MAJOR << 12) + (NPP_VERSION_MINOR << 4)) <= 0x5000 | |
| 729 | |
| 730 case NPP_BAD_ARG_ERROR: | |
| 731 return "NPP_BAD_ARGUMENT_ERROR"; | |
| 732 | |
| 733 case NPP_COEFF_ERROR: | |
| 734 return "NPP_COEFFICIENT_ERROR"; | |
| 735 | |
| 736 case NPP_RECT_ERROR: | |
| 737 return "NPP_RECTANGLE_ERROR"; | |
| 738 | |
| 739 case NPP_QUAD_ERROR: | |
| 740 return "NPP_QUADRANGLE_ERROR"; | |
| 741 | |
| 742 case NPP_MEM_ALLOC_ERR: | |
| 743 return "NPP_MEMORY_ALLOCATION_ERROR"; | |
| 744 | |
| 745 case NPP_HISTO_NUMBER_OF_LEVELS_ERROR: | |
| 746 return "NPP_HISTOGRAM_NUMBER_OF_LEVELS_ERROR"; | |
| 747 | |
| 748 case NPP_INVALID_INPUT: | |
| 749 return "NPP_INVALID_INPUT"; | |
| 750 | |
| 751 case NPP_POINTER_ERROR: | |
| 752 return "NPP_POINTER_ERROR"; | |
| 753 | |
| 754 case NPP_WARNING: | |
| 755 return "NPP_WARNING"; | |
| 756 | |
| 757 case NPP_ODD_ROI_WARNING: | |
| 758 return "NPP_ODD_ROI_WARNING"; | |
| 759 #else | |
| 760 | |
| 761 // These are for CUDA 5.5 or higher | |
| 762 case NPP_BAD_ARGUMENT_ERROR: | |
| 763 return "NPP_BAD_ARGUMENT_ERROR"; | |
| 764 | |
| 765 case NPP_COEFFICIENT_ERROR: | |
| 766 return "NPP_COEFFICIENT_ERROR"; | |
| 767 | |
| 768 case NPP_RECTANGLE_ERROR: | |
| 769 return "NPP_RECTANGLE_ERROR"; | |
| 770 | |
| 771 case NPP_QUADRANGLE_ERROR: | |
| 772 return "NPP_QUADRANGLE_ERROR"; | |
| 773 | |
| 774 case NPP_MEMORY_ALLOCATION_ERR: | |
| 775 return "NPP_MEMORY_ALLOCATION_ERROR"; | |
| 776 | |
| 777 case NPP_HISTOGRAM_NUMBER_OF_LEVELS_ERROR: | |
| 778 return "NPP_HISTOGRAM_NUMBER_OF_LEVELS_ERROR"; | |
| 779 | |
| 780 case NPP_INVALID_HOST_POINTER_ERROR: | |
| 781 return "NPP_INVALID_HOST_POINTER_ERROR"; | |
| 782 | |
| 783 case NPP_INVALID_DEVICE_POINTER_ERROR: | |
| 784 return "NPP_INVALID_DEVICE_POINTER_ERROR"; | |
| 785 #endif | |
| 786 | |
| 787 case NPP_LUT_NUMBER_OF_LEVELS_ERROR: | |
| 788 return "NPP_LUT_NUMBER_OF_LEVELS_ERROR"; | |
| 789 | |
| 790 case NPP_TEXTURE_BIND_ERROR: | |
| 791 return "NPP_TEXTURE_BIND_ERROR"; | |
| 792 | |
| 793 case NPP_WRONG_INTERSECTION_ROI_ERROR: | |
| 794 return "NPP_WRONG_INTERSECTION_ROI_ERROR"; | |
| 795 | |
| 796 case NPP_NOT_EVEN_STEP_ERROR: | |
| 797 return "NPP_NOT_EVEN_STEP_ERROR"; | |
| 798 | |
| 799 case NPP_INTERPOLATION_ERROR: | |
| 800 return "NPP_INTERPOLATION_ERROR"; | |
| 801 | |
| 802 case NPP_RESIZE_FACTOR_ERROR: | |
| 803 return "NPP_RESIZE_FACTOR_ERROR"; | |
| 804 | |
| 805 case NPP_HAAR_CLASSIFIER_PIXEL_MATCH_ERROR: | |
| 806 return "NPP_HAAR_CLASSIFIER_PIXEL_MATCH_ERROR"; | |
| 807 | |
| 808 | |
| 809 #if ((NPP_VERSION_MAJOR << 12) + (NPP_VERSION_MINOR << 4)) <= 0x5000 | |
| 810 | |
| 811 case NPP_MEMFREE_ERR: | |
| 812 return "NPP_MEMFREE_ERR"; | |
| 813 | |
| 814 case NPP_MEMSET_ERR: | |
| 815 return "NPP_MEMSET_ERR"; | |
| 816 | |
| 817 case NPP_MEMCPY_ERR: | |
| 818 return "NPP_MEMCPY_ERROR"; | |
| 819 | |
| 820 case NPP_MIRROR_FLIP_ERR: | |
| 821 return "NPP_MIRROR_FLIP_ERR"; | |
| 822 #else | |
| 823 | |
| 824 case NPP_MEMFREE_ERROR: | |
| 825 return "NPP_MEMFREE_ERROR"; | |
| 826 | |
| 827 case NPP_MEMSET_ERROR: | |
| 828 return "NPP_MEMSET_ERROR"; | |
| 829 | |
| 830 case NPP_MEMCPY_ERROR: | |
| 831 return "NPP_MEMCPY_ERROR"; | |
| 832 | |
| 833 case NPP_MIRROR_FLIP_ERROR: | |
| 834 return "NPP_MIRROR_FLIP_ERROR"; | |
| 835 #endif | |
| 836 | |
| 837 case NPP_ALIGNMENT_ERROR: | |
| 838 return "NPP_ALIGNMENT_ERROR"; | |
| 839 | |
| 840 case NPP_STEP_ERROR: | |
| 841 return "NPP_STEP_ERROR"; | |
| 842 | |
| 843 case NPP_SIZE_ERROR: | |
| 844 return "NPP_SIZE_ERROR"; | |
| 845 | |
| 846 case NPP_NULL_POINTER_ERROR: | |
| 847 return "NPP_NULL_POINTER_ERROR"; | |
| 848 | |
| 849 case NPP_CUDA_KERNEL_EXECUTION_ERROR: | |
| 850 return "NPP_CUDA_KERNEL_EXECUTION_ERROR"; | |
| 851 | |
| 852 case NPP_NOT_IMPLEMENTED_ERROR: | |
| 853 return "NPP_NOT_IMPLEMENTED_ERROR"; | |
| 854 | |
| 855 case NPP_ERROR: | |
| 856 return "NPP_ERROR"; | |
| 857 | |
| 858 case NPP_SUCCESS: | |
| 859 return "NPP_SUCCESS"; | |
| 860 | |
| 861 case NPP_WRONG_INTERSECTION_QUAD_WARNING: | |
| 862 return "NPP_WRONG_INTERSECTION_QUAD_WARNING"; | |
| 863 | |
| 864 case NPP_MISALIGNED_DST_ROI_WARNING: | |
| 865 return "NPP_MISALIGNED_DST_ROI_WARNING"; | |
| 866 | |
| 867 case NPP_AFFINE_QUAD_INCORRECT_WARNING: | |
| 868 return "NPP_AFFINE_QUAD_INCORRECT_WARNING"; | |
| 869 | |
| 870 case NPP_DOUBLE_SIZE_WARNING: | |
| 871 return "NPP_DOUBLE_SIZE_WARNING"; | |
| 872 | |
| 873 case NPP_WRONG_INTERSECTION_ROI_WARNING: | |
| 874 return "NPP_WRONG_INTERSECTION_ROI_WARNING"; | |
| 875 | |
| 876 #if ((NPP_VERSION_MAJOR << 12) + (NPP_VERSION_MINOR << 4)) >= 0x6000 | |
| 877 /* These are 6.0 or higher */ | |
| 878 case NPP_LUT_PALETTE_BITSIZE_ERROR: | |
| 879 return "NPP_LUT_PALETTE_BITSIZE_ERROR"; | |
| 880 | |
| 881 case NPP_ZC_MODE_NOT_SUPPORTED_ERROR: | |
| 882 return "NPP_ZC_MODE_NOT_SUPPORTED_ERROR"; | |
| 883 | |
| 884 case NPP_QUALITY_INDEX_ERROR: | |
| 885 return "NPP_QUALITY_INDEX_ERROR"; | |
| 886 | |
| 887 case NPP_CHANNEL_ORDER_ERROR: | |
| 888 return "NPP_CHANNEL_ORDER_ERROR"; | |
| 889 | |
| 890 case NPP_ZERO_MASK_VALUE_ERROR: | |
| 891 return "NPP_ZERO_MASK_VALUE_ERROR"; | |
| 892 | |
| 893 case NPP_NUMBER_OF_CHANNELS_ERROR: | |
| 894 return "NPP_NUMBER_OF_CHANNELS_ERROR"; | |
| 895 | |
| 896 case NPP_COI_ERROR: | |
| 897 return "NPP_COI_ERROR"; | |
| 898 | |
| 899 case NPP_DIVISOR_ERROR: | |
| 900 return "NPP_DIVISOR_ERROR"; | |
| 901 | |
| 902 case NPP_CHANNEL_ERROR: | |
| 903 return "NPP_CHANNEL_ERROR"; | |
| 904 | |
| 905 case NPP_STRIDE_ERROR: | |
| 906 return "NPP_STRIDE_ERROR"; | |
| 907 | |
| 908 case NPP_ANCHOR_ERROR: | |
| 909 return "NPP_ANCHOR_ERROR"; | |
| 910 | |
| 911 case NPP_MASK_SIZE_ERROR: | |
| 912 return "NPP_MASK_SIZE_ERROR"; | |
| 913 | |
| 914 case NPP_MOMENT_00_ZERO_ERROR: | |
| 915 return "NPP_MOMENT_00_ZERO_ERROR"; | |
| 916 | |
| 917 case NPP_THRESHOLD_NEGATIVE_LEVEL_ERROR: | |
| 918 return "NPP_THRESHOLD_NEGATIVE_LEVEL_ERROR"; | |
| 919 | |
| 920 case NPP_THRESHOLD_ERROR: | |
| 921 return "NPP_THRESHOLD_ERROR"; | |
| 922 | |
| 923 case NPP_CONTEXT_MATCH_ERROR: | |
| 924 return "NPP_CONTEXT_MATCH_ERROR"; | |
| 925 | |
| 926 case NPP_FFT_FLAG_ERROR: | |
| 927 return "NPP_FFT_FLAG_ERROR"; | |
| 928 | |
| 929 case NPP_FFT_ORDER_ERROR: | |
| 930 return "NPP_FFT_ORDER_ERROR"; | |
| 931 | |
| 932 case NPP_SCALE_RANGE_ERROR: | |
| 933 return "NPP_SCALE_RANGE_ERROR"; | |
| 934 | |
| 935 case NPP_DATA_TYPE_ERROR: | |
| 936 return "NPP_DATA_TYPE_ERROR"; | |
| 937 | |
| 938 case NPP_OUT_OFF_RANGE_ERROR: | |
| 939 return "NPP_OUT_OFF_RANGE_ERROR"; | |
| 940 | |
| 941 case NPP_DIVIDE_BY_ZERO_ERROR: | |
| 942 return "NPP_DIVIDE_BY_ZERO_ERROR"; | |
| 943 | |
| 944 case NPP_RANGE_ERROR: | |
| 945 return "NPP_RANGE_ERROR"; | |
| 946 | |
| 947 case NPP_NO_MEMORY_ERROR: | |
| 948 return "NPP_NO_MEMORY_ERROR"; | |
| 949 | |
| 950 case NPP_ERROR_RESERVED: | |
| 951 return "NPP_ERROR_RESERVED"; | |
| 952 | |
| 953 case NPP_NO_OPERATION_WARNING: | |
| 954 return "NPP_NO_OPERATION_WARNING"; | |
| 955 | |
| 956 case NPP_DIVIDE_BY_ZERO_WARNING: | |
| 957 return "NPP_DIVIDE_BY_ZERO_WARNING"; | |
| 958 #endif | |
| 959 | |
| 960 #if ((NPP_VERSION_MAJOR << 12) + (NPP_VERSION_MINOR << 4)) >= 0x7000 | |
| 961 /* These are 7.0 or higher */ | |
| 962 case NPP_OVERFLOW_ERROR: | |
| 963 return "NPP_OVERFLOW_ERROR"; | |
| 964 | |
| 965 case NPP_CORRUPTED_DATA_ERROR: | |
| 966 return "NPP_CORRUPTED_DATA_ERROR"; | |
| 967 #endif | |
| 968 } | |
| 969 | |
| 970 return "<unknown>"; | |
| 971 } | |
| 972 #endif | |
| 973 | |
| 974 #ifdef __DRIVER_TYPES_H__ | |
| 975 #ifndef DEVICE_RESET | |
| 976 #define DEVICE_RESET cudaDeviceReset(); | |
| 977 #endif | |
| 978 #else | |
| 979 #ifndef DEVICE_RESET | |
| 980 #define DEVICE_RESET | |
| 981 #endif | |
| 982 #endif | |
| 983 | |
| 984 template< typename T > | |
| 985 void check(T result, char const *const func, const char *const file, int const line) | |
| 986 { | |
| 987 if (result) | |
| 988 { | |
| 989 fprintf(stderr, "CUDA error at %s:%d code=%d(%s) \"%s\" \n", | |
| 990 file, line, static_cast<unsigned int>(result), _cudaGetErrorEnum(result), func); | |
| 991 DEVICE_RESET | |
| 992 // Make sure we call CUDA Device Reset before exiting | |
| 993 exit(EXIT_FAILURE); | |
| 994 } | |
| 995 } | |
| 996 | |
| 997 #ifdef __DRIVER_TYPES_H__ | |
| 998 // This will output the proper CUDA error strings in the event that a CUDA host call returns an error | |
| 999 #define checkCudaErrors(val) check ( (val), #val, __FILE__, __LINE__ ) | |
| 1000 | |
| 1001 // This will output the proper error string when calling cudaGetLastError | |
| 1002 #define getLastCudaError(msg) __getLastCudaError (msg, __FILE__, __LINE__) | |
| 1003 | |
| 1004 inline void __getLastCudaError(const char *errorMessage, const char *file, const int line) | |
| 1005 { | |
| 1006 cudaError_t err = cudaGetLastError(); | |
| 1007 | |
| 1008 if (cudaSuccess != err) | |
| 1009 { | |
| 1010 fprintf(stderr, "%s(%i) : getLastCudaError() CUDA error : %s : (%d) %s.\n", | |
| 1011 file, line, errorMessage, (int)err, cudaGetErrorString(err)); | |
| 1012 DEVICE_RESET | |
| 1013 exit(EXIT_FAILURE); | |
| 1014 } | |
| 1015 } | |
| 1016 #endif | |
| 1017 | |
| 1018 #ifndef MAX | |
| 1019 #define MAX(a,b) (a > b ? a : b) | |
| 1020 #endif | |
| 1021 | |
| 1022 // Float To Int conversion | |
| 1023 inline int ftoi(float value) | |
| 1024 { | |
| 1025 return (value >= 0 ? (int)(value + 0.5) : (int)(value - 0.5)); | |
| 1026 } | |
| 1027 | |
| 1028 // Beginning of GPU Architecture definitions | |
| 1029 inline int _ConvertSMVer2Cores(int major, int minor) | |
| 1030 { | |
| 1031 // Defines for GPU Architecture types (using the SM version to determine the # of cores per SM | |
| 1032 typedef struct | |
| 1033 { | |
| 1034 int SM; // 0xMm (hexidecimal notation), M = SM Major version, and m = SM minor version | |
| 1035 int Cores; | |
| 1036 } sSMtoCores; | |
| 1037 | |
| 1038 sSMtoCores nGpuArchCoresPerSM[] = | |
| 1039 { | |
| 1040 { 0x20, 32 }, // Fermi Generation (SM 2.0) GF100 class | |
| 1041 { 0x21, 48 }, // Fermi Generation (SM 2.1) GF10x class | |
| 1042 { 0x30, 192}, // Kepler Generation (SM 3.0) GK10x class | |
| 1043 { 0x32, 192}, // Kepler Generation (SM 3.2) GK10x class | |
| 1044 { 0x35, 192}, // Kepler Generation (SM 3.5) GK11x class | |
| 1045 { 0x37, 192}, // Kepler Generation (SM 3.7) GK21x class | |
| 1046 { 0x50, 128}, // Maxwell Generation (SM 5.0) GM10x class | |
| 1047 { 0x52, 128}, // Maxwell Generation (SM 5.2) GM20x class | |
| 1048 { 0x53, 128}, // Maxwell Generation (SM 5.3) GM20x class | |
| 1049 { 0x60, 64 }, // Pascal Generation (SM 6.0) GP100 class | |
| 1050 { 0x61, 128}, // Pascal Generation (SM 6.1) GP10x class | |
| 1051 { 0x62, 128}, // Pascal Generation (SM 6.2) GP10x class | |
| 1052 { -1, -1 } | |
| 1053 }; | |
| 1054 | |
| 1055 int index = 0; | |
| 1056 | |
| 1057 while (nGpuArchCoresPerSM[index].SM != -1) | |
| 1058 { | |
| 1059 if (nGpuArchCoresPerSM[index].SM == ((major << 4) + minor)) | |
| 1060 { | |
| 1061 return nGpuArchCoresPerSM[index].Cores; | |
| 1062 } | |
| 1063 | |
| 1064 index++; | |
| 1065 } | |
| 1066 | |
| 1067 // If we don't find the values, we default use the previous one to run properly | |
| 1068 printf("MapSMtoCores for SM %d.%d is undefined. Default to use %d Cores/SM\n", major, minor, nGpuArchCoresPerSM[index-1].Cores); | |
| 1069 return nGpuArchCoresPerSM[index-1].Cores; | |
| 1070 } | |
| 1071 // end of GPU Architecture definitions | |
| 1072 | |
| 1073 #ifdef __CUDA_RUNTIME_H__ | |
| 1074 // General GPU Device CUDA Initialization | |
| 1075 inline int gpuDeviceInit(int devID) | |
| 1076 { | |
| 1077 int device_count; | |
| 1078 checkCudaErrors(cudaGetDeviceCount(&device_count)); | |
| 1079 | |
| 1080 if (device_count == 0) | |
| 1081 { | |
| 1082 fprintf(stderr, "gpuDeviceInit() CUDA error: no devices supporting CUDA.\n"); | |
| 1083 exit(EXIT_FAILURE); | |
| 1084 } | |
| 1085 | |
| 1086 if (devID < 0) | |
| 1087 { | |
| 1088 devID = 0; | |
| 1089 } | |
| 1090 | |
| 1091 if (devID > device_count-1) | |
| 1092 { | |
| 1093 fprintf(stderr, "\n"); | |
| 1094 fprintf(stderr, ">> %d CUDA capable GPU device(s) detected. <<\n", device_count); | |
| 1095 fprintf(stderr, ">> gpuDeviceInit (-device=%d) is not a valid GPU device. <<\n", devID); | |
| 1096 fprintf(stderr, "\n"); | |
| 1097 return -devID; | |
| 1098 } | |
| 1099 | |
| 1100 cudaDeviceProp deviceProp; | |
| 1101 checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID)); | |
| 1102 | |
| 1103 if (deviceProp.computeMode == cudaComputeModeProhibited) | |
| 1104 { | |
| 1105 fprintf(stderr, "Error: device is running in <Compute Mode Prohibited>, no threads can use ::cudaSetDevice().\n"); | |
| 1106 return -1; | |
| 1107 } | |
| 1108 | |
| 1109 if (deviceProp.major < 1) | |
| 1110 { | |
| 1111 fprintf(stderr, "gpuDeviceInit(): GPU device does not support CUDA.\n"); | |
| 1112 exit(EXIT_FAILURE); | |
| 1113 } | |
| 1114 | |
| 1115 checkCudaErrors(cudaSetDevice(devID)); | |
| 1116 printf("gpuDeviceInit() CUDA Device [%d]: \"%s\n", devID, deviceProp.name); | |
| 1117 | |
| 1118 return devID; | |
| 1119 } | |
| 1120 | |
| 1121 // This function returns the best GPU (with maximum GFLOPS) | |
| 1122 inline int gpuGetMaxGflopsDeviceId() | |
| 1123 { | |
| 1124 int current_device = 0, sm_per_multiproc = 0; | |
| 1125 int max_perf_device = 0; | |
| 1126 int device_count = 0, best_SM_arch = 0; | |
| 1127 int devices_prohibited = 0; | |
| 1128 | |
| 1129 unsigned long long max_compute_perf = 0; | |
| 1130 cudaDeviceProp deviceProp; | |
| 1131 cudaGetDeviceCount(&device_count); | |
| 1132 | |
| 1133 checkCudaErrors(cudaGetDeviceCount(&device_count)); | |
| 1134 | |
| 1135 if (device_count == 0) | |
| 1136 { | |
| 1137 fprintf(stderr, "gpuGetMaxGflopsDeviceId() CUDA error: no devices supporting CUDA.\n"); | |
| 1138 exit(EXIT_FAILURE); | |
| 1139 } | |
| 1140 | |
| 1141 // Find the best major SM Architecture GPU device | |
| 1142 while (current_device < device_count) | |
| 1143 { | |
| 1144 cudaGetDeviceProperties(&deviceProp, current_device); | |
| 1145 | |
| 1146 // If this GPU is not running on Compute Mode prohibited, then we can add it to the list | |
| 1147 if (deviceProp.computeMode != cudaComputeModeProhibited) | |
| 1148 { | |
| 1149 if (deviceProp.major > 0 && deviceProp.major < 9999) | |
| 1150 { | |
| 1151 best_SM_arch = MAX(best_SM_arch, deviceProp.major); | |
| 1152 } | |
| 1153 } | |
| 1154 else | |
| 1155 { | |
| 1156 devices_prohibited++; | |
| 1157 } | |
| 1158 | |
| 1159 current_device++; | |
| 1160 } | |
| 1161 | |
| 1162 if (devices_prohibited == device_count) | |
| 1163 { | |
| 1164 fprintf(stderr, "gpuGetMaxGflopsDeviceId() CUDA error: all devices have compute mode prohibited.\n"); | |
| 1165 exit(EXIT_FAILURE); | |
| 1166 } | |
| 1167 | |
| 1168 // Find the best CUDA capable GPU device | |
| 1169 current_device = 0; | |
| 1170 | |
| 1171 while (current_device < device_count) | |
| 1172 { | |
| 1173 cudaGetDeviceProperties(&deviceProp, current_device); | |
| 1174 | |
| 1175 // If this GPU is not running on Compute Mode prohibited, then we can add it to the list | |
| 1176 if (deviceProp.computeMode != cudaComputeModeProhibited) | |
| 1177 { | |
| 1178 if (deviceProp.major == 9999 && deviceProp.minor == 9999) | |
| 1179 { | |
| 1180 sm_per_multiproc = 1; | |
| 1181 } | |
| 1182 else | |
| 1183 { | |
| 1184 sm_per_multiproc = _ConvertSMVer2Cores(deviceProp.major, deviceProp.minor); | |
| 1185 } | |
| 1186 | |
| 1187 unsigned long long compute_perf = (unsigned long long) deviceProp.multiProcessorCount * sm_per_multiproc * deviceProp.clockRate; | |
| 1188 | |
| 1189 if (compute_perf > max_compute_perf) | |
| 1190 { | |
| 1191 // If we find GPU with SM major > 2, search only these | |
| 1192 if (best_SM_arch > 2) | |
| 1193 { | |
| 1194 // If our device==dest_SM_arch, choose this, or else pass | |
| 1195 if (deviceProp.major == best_SM_arch) | |
| 1196 { | |
| 1197 max_compute_perf = compute_perf; | |
| 1198 max_perf_device = current_device; | |
| 1199 } | |
| 1200 } | |
| 1201 else | |
| 1202 { | |
| 1203 max_compute_perf = compute_perf; | |
| 1204 max_perf_device = current_device; | |
| 1205 } | |
| 1206 } | |
| 1207 } | |
| 1208 | |
| 1209 ++current_device; | |
| 1210 } | |
| 1211 | |
| 1212 return max_perf_device; | |
| 1213 } | |
| 1214 | |
| 1215 | |
| 1216 // Initialization code to find the best CUDA Device | |
| 1217 inline int findCudaDevice(int argc, const char **argv) | |
| 1218 { | |
| 1219 cudaDeviceProp deviceProp; | |
| 1220 int devID = 0; | |
| 1221 | |
| 1222 // If the command-line has a device number specified, use it | |
| 1223 if (checkCmdLineFlag(argc, argv, "device")) | |
| 1224 { | |
| 1225 devID = getCmdLineArgumentInt(argc, argv, "device="); | |
| 1226 | |
| 1227 if (devID < 0) | |
| 1228 { | |
| 1229 printf("Invalid command line parameter\n "); | |
| 1230 exit(EXIT_FAILURE); | |
| 1231 } | |
| 1232 else | |
| 1233 { | |
| 1234 devID = gpuDeviceInit(devID); | |
| 1235 | |
| 1236 if (devID < 0) | |
| 1237 { | |
| 1238 printf("exiting...\n"); | |
| 1239 exit(EXIT_FAILURE); | |
| 1240 } | |
| 1241 } | |
| 1242 } | |
| 1243 else | |
| 1244 { | |
| 1245 // Otherwise pick the device with highest Gflops/s | |
| 1246 devID = gpuGetMaxGflopsDeviceId(); | |
| 1247 checkCudaErrors(cudaSetDevice(devID)); | |
| 1248 checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID)); | |
| 1249 printf("GPU Device %d: \"%s\" with compute capability %d.%d\n\n", devID, deviceProp.name, deviceProp.major, deviceProp.minor); | |
| 1250 } | |
| 1251 | |
| 1252 return devID; | |
| 1253 } | |
| 1254 | |
| 1255 // General check for CUDA GPU SM Capabilities | |
| 1256 inline bool checkCudaCapabilities(int major_version, int minor_version) | |
| 1257 { | |
| 1258 cudaDeviceProp deviceProp; | |
| 1259 deviceProp.major = 0; | |
| 1260 deviceProp.minor = 0; | |
| 1261 int dev; | |
| 1262 | |
| 1263 checkCudaErrors(cudaGetDevice(&dev)); | |
| 1264 checkCudaErrors(cudaGetDeviceProperties(&deviceProp, dev)); | |
| 1265 | |
| 1266 if ((deviceProp.major > major_version) || | |
| 1267 (deviceProp.major == major_version && deviceProp.minor >= minor_version)) | |
| 1268 { | |
| 1269 printf(" Device %d: <%16s >, Compute SM %d.%d detected\n", dev, deviceProp.name, deviceProp.major, deviceProp.minor); | |
| 1270 return true; | |
| 1271 } | |
| 1272 else | |
| 1273 { | |
| 1274 printf(" No GPU device was found that can support CUDA compute capability %d.%d.\n", major_version, minor_version); | |
| 1275 return false; | |
| 1276 } | |
| 1277 } | |
| 1278 #endif | |
| 1279 | |
| 1280 // end of CUDA Helper Functions | |
| 1281 | |
| 1282 | |
| 1283 #endif |