1 | //===--- cuda/dynamic_cuda/cuda.h --------------------------------- C++ -*-===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | // |
9 | // The parts of the cuda api that are presently in use by the openmp cuda plugin |
10 | // |
11 | //===----------------------------------------------------------------------===// |
12 | |
13 | #ifndef DYNAMIC_CUDA_CUDA_H_INCLUDED |
14 | #define DYNAMIC_CUDA_CUDA_H_INCLUDED |
15 | |
16 | #include <cstddef> |
17 | #include <cstdint> |
18 | |
19 | typedef int CUdevice; |
20 | typedef uintptr_t CUdeviceptr; |
21 | typedef struct CUmod_st *CUmodule; |
22 | typedef struct CUctx_st *CUcontext; |
23 | typedef struct CUfunc_st *CUfunction; |
24 | typedef struct CUstream_st *CUstream; |
25 | typedef struct CUevent_st *CUevent; |
26 | |
27 | #define CU_DEVICE_INVALID ((CUdevice)-2) |
28 | |
29 | typedef unsigned long long CUmemGenericAllocationHandle_v1; |
30 | typedef CUmemGenericAllocationHandle_v1 CUmemGenericAllocationHandle; |
31 | |
32 | #define CU_DEVICE_INVALID ((CUdevice)-2) |
33 | |
34 | typedef enum CUmemAllocationGranularity_flags_enum { |
35 | CU_MEM_ALLOC_GRANULARITY_MINIMUM = 0x0, |
36 | CU_MEM_ALLOC_GRANULARITY_RECOMMENDED = 0x1 |
37 | } CUmemAllocationGranularity_flags; |
38 | |
39 | typedef enum CUmemAccess_flags_enum { |
40 | CU_MEM_ACCESS_FLAGS_PROT_NONE = 0x0, |
41 | CU_MEM_ACCESS_FLAGS_PROT_READ = 0x1, |
42 | CU_MEM_ACCESS_FLAGS_PROT_READWRITE = 0x3, |
43 | CU_MEM_ACCESS_FLAGS_PROT_MAX = 0x7FFFFFFF |
44 | } CUmemAccess_flags; |
45 | |
46 | typedef enum CUmemLocationType_enum { |
47 | CU_MEM_LOCATION_TYPE_INVALID = 0x0, |
48 | CU_MEM_LOCATION_TYPE_DEVICE = 0x1, |
49 | CU_MEM_LOCATION_TYPE_MAX = 0x7FFFFFFF |
50 | } CUmemLocationType; |
51 | |
52 | typedef struct CUmemLocation_st { |
53 | CUmemLocationType type; |
54 | int id; |
55 | } CUmemLocation_v1; |
56 | typedef CUmemLocation_v1 CUmemLocation; |
57 | |
58 | typedef struct CUmemAccessDesc_st { |
59 | CUmemLocation location; |
60 | CUmemAccess_flags flags; |
61 | } CUmemAccessDesc_v1; |
62 | |
63 | typedef CUmemAccessDesc_v1 CUmemAccessDesc; |
64 | |
65 | typedef enum CUmemAllocationType_enum { |
66 | CU_MEM_ALLOCATION_TYPE_INVALID = 0x0, |
67 | CU_MEM_ALLOCATION_TYPE_PINNED = 0x1, |
68 | CU_MEM_ALLOCATION_TYPE_MAX = 0x7FFFFFFF |
69 | } CUmemAllocationType; |
70 | |
71 | typedef enum CUmemAllocationHandleType_enum { |
72 | CU_MEM_HANDLE_TYPE_NONE = 0x0, |
73 | CU_MEM_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR = 0x1, |
74 | CU_MEM_HANDLE_TYPE_WIN32 = 0x2, |
75 | CU_MEM_HANDLE_TYPE_WIN32_KMT = 0x4, |
76 | CU_MEM_HANDLE_TYPE_MAX = 0x7FFFFFFF |
77 | } CUmemAllocationHandleType; |
78 | |
79 | typedef struct CUmemAllocationProp_st { |
80 | CUmemAllocationType type; |
81 | CUmemAllocationHandleType requestedHandleTypes; |
82 | CUmemLocation location; |
83 | |
84 | void *win32HandleMetaData; |
85 | struct { |
86 | unsigned char compressionType; |
87 | unsigned char gpuDirectRDMACapable; |
88 | unsigned short usage; |
89 | unsigned char reserved[4]; |
90 | } allocFlags; |
91 | } CUmemAllocationProp_v1; |
92 | typedef CUmemAllocationProp_v1 CUmemAllocationProp; |
93 | |
94 | typedef enum cudaError_enum { |
95 | CUDA_SUCCESS = 0, |
96 | CUDA_ERROR_INVALID_VALUE = 1, |
97 | CUDA_ERROR_NO_DEVICE = 100, |
98 | CUDA_ERROR_INVALID_HANDLE = 400, |
99 | CUDA_ERROR_NOT_READY = 600, |
100 | CUDA_ERROR_TOO_MANY_PEERS = 711, |
101 | } CUresult; |
102 | |
103 | typedef enum CUstream_flags_enum { |
104 | CU_STREAM_DEFAULT = 0x0, |
105 | CU_STREAM_NON_BLOCKING = 0x1, |
106 | } CUstream_flags; |
107 | |
108 | typedef enum CUlimit_enum { |
109 | CU_LIMIT_STACK_SIZE = 0x0, |
110 | CU_LIMIT_PRINTF_FIFO_SIZE = 0x1, |
111 | CU_LIMIT_MALLOC_HEAP_SIZE = 0x2, |
112 | CU_LIMIT_DEV_RUNTIME_SYNC_DEPTH = 0x3, |
113 | CU_LIMIT_DEV_RUNTIME_PENDING_LAUNCH_COUNT = 0x4, |
114 | CU_LIMIT_MAX_L2_FETCH_GRANULARITY = 0x5, |
115 | CU_LIMIT_PERSISTING_L2_CACHE_SIZE = 0x6, |
116 | CU_LIMIT_MAX |
117 | } CUlimit; |
118 | |
119 | typedef enum CUdevice_attribute_enum { |
120 | CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK = 1, |
121 | CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X = 2, |
122 | CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y = 3, |
123 | CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z = 4, |
124 | CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X = 5, |
125 | CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y = 6, |
126 | CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z = 7, |
127 | CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK = 8, |
128 | CU_DEVICE_ATTRIBUTE_SHARED_MEMORY_PER_BLOCK = 8, |
129 | CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY = 9, |
130 | CU_DEVICE_ATTRIBUTE_WARP_SIZE = 10, |
131 | CU_DEVICE_ATTRIBUTE_MAX_PITCH = 11, |
132 | CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_BLOCK = 12, |
133 | CU_DEVICE_ATTRIBUTE_REGISTERS_PER_BLOCK = 12, |
134 | CU_DEVICE_ATTRIBUTE_CLOCK_RATE = 13, |
135 | CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT = 14, |
136 | CU_DEVICE_ATTRIBUTE_GPU_OVERLAP = 15, |
137 | CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT = 16, |
138 | CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT = 17, |
139 | CU_DEVICE_ATTRIBUTE_INTEGRATED = 18, |
140 | CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY = 19, |
141 | CU_DEVICE_ATTRIBUTE_COMPUTE_MODE = 20, |
142 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_WIDTH = 21, |
143 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_WIDTH = 22, |
144 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_HEIGHT = 23, |
145 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH = 24, |
146 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT = 25, |
147 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH = 26, |
148 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_WIDTH = 27, |
149 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_HEIGHT = 28, |
150 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_LAYERS = 29, |
151 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_ARRAY_WIDTH = 27, |
152 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_ARRAY_HEIGHT = 28, |
153 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_ARRAY_NUMSLICES = 29, |
154 | CU_DEVICE_ATTRIBUTE_SURFACE_ALIGNMENT = 30, |
155 | CU_DEVICE_ATTRIBUTE_CONCURRENT_KERNELS = 31, |
156 | CU_DEVICE_ATTRIBUTE_ECC_ENABLED = 32, |
157 | CU_DEVICE_ATTRIBUTE_PCI_BUS_ID = 33, |
158 | CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID = 34, |
159 | CU_DEVICE_ATTRIBUTE_TCC_DRIVER = 35, |
160 | CU_DEVICE_ATTRIBUTE_MEMORY_CLOCK_RATE = 36, |
161 | CU_DEVICE_ATTRIBUTE_GLOBAL_MEMORY_BUS_WIDTH = 37, |
162 | CU_DEVICE_ATTRIBUTE_L2_CACHE_SIZE = 38, |
163 | CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_MULTIPROCESSOR = 39, |
164 | CU_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT = 40, |
165 | CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING = 41, |
166 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_WIDTH = 42, |
167 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_LAYERS = 43, |
168 | CU_DEVICE_ATTRIBUTE_CAN_TEX2D_GATHER = 44, |
169 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_GATHER_WIDTH = 45, |
170 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_GATHER_HEIGHT = 46, |
171 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH_ALTERNATE = 47, |
172 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT_ALTERNATE = 48, |
173 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH_ALTERNATE = 49, |
174 | CU_DEVICE_ATTRIBUTE_PCI_DOMAIN_ID = 50, |
175 | CU_DEVICE_ATTRIBUTE_TEXTURE_PITCH_ALIGNMENT = 51, |
176 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_WIDTH = 52, |
177 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_WIDTH = 53, |
178 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_LAYERS = 54, |
179 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE1D_WIDTH = 55, |
180 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE2D_WIDTH = 56, |
181 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE2D_HEIGHT = 57, |
182 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE3D_WIDTH = 58, |
183 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE3D_HEIGHT = 59, |
184 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE3D_DEPTH = 60, |
185 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE1D_LAYERED_WIDTH = 61, |
186 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE1D_LAYERED_LAYERS = 62, |
187 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE2D_LAYERED_WIDTH = 63, |
188 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE2D_LAYERED_HEIGHT = 64, |
189 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE2D_LAYERED_LAYERS = 65, |
190 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACECUBEMAP_WIDTH = 66, |
191 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACECUBEMAP_LAYERED_WIDTH = 67, |
192 | CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACECUBEMAP_LAYERED_LAYERS = 68, |
193 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LINEAR_WIDTH = 69, |
194 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_WIDTH = 70, |
195 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_HEIGHT = 71, |
196 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_PITCH = 72, |
197 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_WIDTH = 73, |
198 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_HEIGHT = 74, |
199 | CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR = 75, |
200 | CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR = 76, |
201 | CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH = 77, |
202 | CU_DEVICE_ATTRIBUTE_STREAM_PRIORITIES_SUPPORTED = 78, |
203 | CU_DEVICE_ATTRIBUTE_GLOBAL_L1_CACHE_SUPPORTED = 79, |
204 | CU_DEVICE_ATTRIBUTE_LOCAL_L1_CACHE_SUPPORTED = 80, |
205 | CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR = 81, |
206 | CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR = 82, |
207 | CU_DEVICE_ATTRIBUTE_MANAGED_MEMORY = 83, |
208 | CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD = 84, |
209 | CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD_GROUP_ID = 85, |
210 | CU_DEVICE_ATTRIBUTE_HOST_NATIVE_ATOMIC_SUPPORTED = 86, |
211 | CU_DEVICE_ATTRIBUTE_SINGLE_TO_DOUBLE_PRECISION_PERF_RATIO = 87, |
212 | CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS = 88, |
213 | CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS = 89, |
214 | CU_DEVICE_ATTRIBUTE_COMPUTE_PREEMPTION_SUPPORTED = 90, |
215 | CU_DEVICE_ATTRIBUTE_CAN_USE_HOST_POINTER_FOR_REGISTERED_MEM = 91, |
216 | CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_MEM_OPS = 92, |
217 | CU_DEVICE_ATTRIBUTE_CAN_USE_64_BIT_STREAM_MEM_OPS = 93, |
218 | CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_WAIT_VALUE_NOR = 94, |
219 | CU_DEVICE_ATTRIBUTE_COOPERATIVE_LAUNCH = 95, |
220 | CU_DEVICE_ATTRIBUTE_COOPERATIVE_MULTI_DEVICE_LAUNCH = 96, |
221 | CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK_OPTIN = 97, |
222 | CU_DEVICE_ATTRIBUTE_CAN_FLUSH_REMOTE_WRITES = 98, |
223 | CU_DEVICE_ATTRIBUTE_HOST_REGISTER_SUPPORTED = 99, |
224 | CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES = 100, |
225 | CU_DEVICE_ATTRIBUTE_DIRECT_MANAGED_MEM_ACCESS_FROM_HOST = 101, |
226 | CU_DEVICE_ATTRIBUTE_VIRTUAL_ADDRESS_MANAGEMENT_SUPPORTED = 102, |
227 | CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED = 102, |
228 | CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR_SUPPORTED = 103, |
229 | CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_WIN32_HANDLE_SUPPORTED = 104, |
230 | CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_WIN32_KMT_HANDLE_SUPPORTED = 105, |
231 | CU_DEVICE_ATTRIBUTE_MAX_BLOCKS_PER_MULTIPROCESSOR = 106, |
232 | CU_DEVICE_ATTRIBUTE_GENERIC_COMPRESSION_SUPPORTED = 107, |
233 | CU_DEVICE_ATTRIBUTE_MAX_PERSISTING_L2_CACHE_SIZE = 108, |
234 | CU_DEVICE_ATTRIBUTE_MAX_ACCESS_POLICY_WINDOW_SIZE = 109, |
235 | CU_DEVICE_ATTRIBUTE_GPU_DIRECT_RDMA_WITH_CUDA_VMM_SUPPORTED = 110, |
236 | CU_DEVICE_ATTRIBUTE_RESERVED_SHARED_MEMORY_PER_BLOCK = 111, |
237 | CU_DEVICE_ATTRIBUTE_SPARSE_CUDA_ARRAY_SUPPORTED = 112, |
238 | CU_DEVICE_ATTRIBUTE_READ_ONLY_HOST_REGISTER_SUPPORTED = 113, |
239 | CU_DEVICE_ATTRIBUTE_TIMELINE_SEMAPHORE_INTEROP_SUPPORTED = 114, |
240 | CU_DEVICE_ATTRIBUTE_MEMORY_POOLS_SUPPORTED = 115, |
241 | CU_DEVICE_ATTRIBUTE_GPU_DIRECT_RDMA_SUPPORTED = 116, |
242 | CU_DEVICE_ATTRIBUTE_GPU_DIRECT_RDMA_FLUSH_WRITES_OPTIONS = 117, |
243 | CU_DEVICE_ATTRIBUTE_GPU_DIRECT_RDMA_WRITES_ORDERING = 118, |
244 | CU_DEVICE_ATTRIBUTE_MEMPOOL_SUPPORTED_HANDLE_TYPES = 119, |
245 | CU_DEVICE_ATTRIBUTE_MAX, |
246 | } CUdevice_attribute; |
247 | |
248 | typedef enum CUfunction_attribute_enum { |
249 | CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK = 0, |
250 | } CUfunction_attribute; |
251 | |
252 | typedef enum CUctx_flags_enum { |
253 | CU_CTX_SCHED_BLOCKING_SYNC = 0x04, |
254 | CU_CTX_SCHED_MASK = 0x07, |
255 | } CUctx_flags; |
256 | |
257 | typedef enum CUmemAttach_flags_enum { |
258 | CU_MEM_ATTACH_GLOBAL = 0x1, |
259 | CU_MEM_ATTACH_HOST = 0x2, |
260 | CU_MEM_ATTACH_SINGLE = 0x4, |
261 | } CUmemAttach_flags; |
262 | |
263 | typedef enum CUcomputeMode_enum { |
264 | CU_COMPUTEMODE_DEFAULT = 0, |
265 | CU_COMPUTEMODE_PROHIBITED = 2, |
266 | CU_COMPUTEMODE_EXCLUSIVE_PROCESS = 3, |
267 | } CUcompute_mode; |
268 | |
269 | typedef enum CUevent_flags_enum { |
270 | CU_EVENT_DEFAULT = 0x0, |
271 | CU_EVENT_BLOCKING_SYNC = 0x1, |
272 | CU_EVENT_DISABLE_TIMING = 0x2, |
273 | CU_EVENT_INTERPROCESS = 0x4 |
274 | } CUevent_flags; |
275 | |
276 | CUresult cuCtxGetDevice(CUdevice *); |
277 | CUresult cuDeviceGet(CUdevice *, int); |
278 | CUresult cuDeviceGetAttribute(int *, CUdevice_attribute, CUdevice); |
279 | CUresult cuDeviceGetCount(int *); |
280 | CUresult cuFuncGetAttribute(int *, CUfunction_attribute, CUfunction); |
281 | |
282 | // Device info |
283 | CUresult cuDeviceGetName(char *, int, CUdevice); |
284 | CUresult cuDeviceTotalMem(size_t *, CUdevice); |
285 | CUresult cuDriverGetVersion(int *); |
286 | |
287 | CUresult cuGetErrorString(CUresult, const char **); |
288 | CUresult cuInit(unsigned); |
289 | CUresult cuLaunchKernel(CUfunction, unsigned, unsigned, unsigned, unsigned, |
290 | unsigned, unsigned, unsigned, CUstream, void **, |
291 | void **); |
292 | |
293 | CUresult cuMemAlloc(CUdeviceptr *, size_t); |
294 | CUresult cuMemAllocHost(void **, size_t); |
295 | CUresult cuMemAllocManaged(CUdeviceptr *, size_t, unsigned int); |
296 | CUresult cuMemAllocAsync(CUdeviceptr *, size_t, CUstream); |
297 | |
298 | CUresult cuMemcpyDtoDAsync(CUdeviceptr, CUdeviceptr, size_t, CUstream); |
299 | CUresult cuMemcpyDtoH(void *, CUdeviceptr, size_t); |
300 | CUresult cuMemcpyDtoHAsync(void *, CUdeviceptr, size_t, CUstream); |
301 | CUresult cuMemcpyHtoD(CUdeviceptr, const void *, size_t); |
302 | CUresult cuMemcpyHtoDAsync(CUdeviceptr, const void *, size_t, CUstream); |
303 | |
304 | CUresult cuMemFree(CUdeviceptr); |
305 | CUresult cuMemFreeHost(void *); |
306 | CUresult cuMemFreeAsync(CUdeviceptr, CUstream); |
307 | |
308 | CUresult cuModuleGetFunction(CUfunction *, CUmodule, const char *); |
309 | CUresult cuModuleGetGlobal(CUdeviceptr *, size_t *, CUmodule, const char *); |
310 | |
311 | CUresult cuModuleUnload(CUmodule); |
312 | CUresult cuStreamCreate(CUstream *, unsigned); |
313 | CUresult cuStreamDestroy(CUstream); |
314 | CUresult cuStreamSynchronize(CUstream); |
315 | CUresult cuStreamQuery(CUstream); |
316 | CUresult cuCtxSetCurrent(CUcontext); |
317 | CUresult cuDevicePrimaryCtxRelease(CUdevice); |
318 | CUresult cuDevicePrimaryCtxGetState(CUdevice, unsigned *, int *); |
319 | CUresult cuDevicePrimaryCtxSetFlags(CUdevice, unsigned); |
320 | CUresult cuDevicePrimaryCtxRetain(CUcontext *, CUdevice); |
321 | CUresult cuModuleLoadDataEx(CUmodule *, const void *, unsigned, void *, |
322 | void **); |
323 | |
324 | CUresult cuDeviceCanAccessPeer(int *, CUdevice, CUdevice); |
325 | CUresult cuCtxEnablePeerAccess(CUcontext, unsigned); |
326 | CUresult cuMemcpyPeerAsync(CUdeviceptr, CUcontext, CUdeviceptr, CUcontext, |
327 | size_t, CUstream); |
328 | |
329 | CUresult cuCtxGetLimit(size_t *, CUlimit); |
330 | CUresult cuCtxSetLimit(CUlimit, size_t); |
331 | |
332 | CUresult cuEventCreate(CUevent *, unsigned int); |
333 | CUresult cuEventRecord(CUevent, CUstream); |
334 | CUresult cuStreamWaitEvent(CUstream, CUevent, unsigned int); |
335 | CUresult cuEventSynchronize(CUevent); |
336 | CUresult cuEventDestroy(CUevent); |
337 | |
338 | CUresult cuMemUnmap(CUdeviceptr ptr, size_t size); |
339 | CUresult cuMemRelease(CUmemGenericAllocationHandle handle); |
340 | CUresult cuMemAddressFree(CUdeviceptr ptr, size_t size); |
341 | CUresult cuMemGetInfo(size_t *free, size_t *total); |
342 | CUresult cuMemAddressReserve(CUdeviceptr *ptr, size_t size, size_t alignment, |
343 | CUdeviceptr addr, unsigned long long flags); |
344 | CUresult cuMemMap(CUdeviceptr ptr, size_t size, size_t offset, |
345 | CUmemGenericAllocationHandle handle, |
346 | unsigned long long flags); |
347 | CUresult cuMemCreate(CUmemGenericAllocationHandle *handle, size_t size, |
348 | const CUmemAllocationProp *prop, unsigned long long flags); |
349 | CUresult cuMemSetAccess(CUdeviceptr ptr, size_t size, |
350 | const CUmemAccessDesc *desc, size_t count); |
351 | CUresult cuMemGetAllocationGranularity(size_t *granularity, |
352 | const CUmemAllocationProp *prop, |
353 | CUmemAllocationGranularity_flags option); |
354 | |
355 | #endif |
356 | |