Mercurial > hg > truffle
annotate src/gpu/ptx/vm/gpu_ptx.hpp @ 11484:2aac62d79af4
Truffle-DSL: fixed bug if node class was used also as child.
| author | Christian Humer <christian.humer@gmail.com> |
|---|---|
| date | Fri, 30 Aug 2013 17:31:59 +0200 |
| parents | 4a6b03a16808 |
| children | 49bb1bc983c6 |
| rev | line source |
|---|---|
| 10879 | 1 /* |
| 2 * Copyright (c) 2013, Oracle and/or its affiliates. All rights reserved. | |
| 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. | |
| 4 * | |
| 5 * This code is free software; you can redistribute it and/or modify it | |
| 6 * under the terms of the GNU General Public License version 2 only, as | |
| 7 * published by the Free Software Foundation. | |
| 8 * | |
| 9 * This code is distributed in the hope that it will be useful, but WITHOUT | |
| 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
| 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
| 12 * version 2 for more details (a copy is included in the LICENSE file that | |
| 13 * accompanied this code). | |
| 14 * | |
| 15 * You should have received a copy of the GNU General Public License version | |
| 16 * 2 along with this work; if not, write to the Free Software Foundation, | |
| 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. | |
| 18 * | |
| 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA | |
| 20 * or visit www.oracle.com if you need additional information or have any | |
| 21 * questions. | |
| 22 * | |
| 23 */ | |
| 24 | |
| 25 #ifndef GPU_PTX_HPP | |
| 26 #define GPU_PTX_HPP | |
| 27 | |
| 28 /* | |
| 29 * Some useful macro definitions from publicly available cuda.h. | |
| 30 * These definitions are for convenience. | |
| 31 */ | |
| 32 #define GRAAL_CUDA_SUCCESS 0 | |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
33 /**< Device shares a unified address space with the host */ |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
34 #define GRAAL_CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING 41 |
| 10879 | 35 #define GRAAL_CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR 75 |
| 36 #define GRAAL_CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR 76 | |
| 37 #define GRAAL_CU_JIT_MAX_REGISTERS 0 | |
| 38 #define GRAAL_CU_JIT_THREADS_PER_BLOCK 1 | |
| 39 #define GRAAL_CU_JIT_INFO_LOG_BUFFER 3 | |
| 40 #define GRAAL_CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES 4 | |
| 41 #define GRAAL_CUDA_ERROR_NO_BINARY_FOR_GPU 209 | |
| 42 | |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
43 /** |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
44 * End of array terminator for the extra parameter to |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
45 * ::cuLaunchKernel |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
46 */ |
|
11291
4a6b03a16808
Fix PTX test to prevent crash; trivial code cleanup
bharadwaj
parents:
11283
diff
changeset
|
47 #define GRAAL_CU_LAUNCH_PARAM_END ((void *) 0x00) |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
48 |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
49 /** |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
50 * Indicator that the next value in the extra parameter to |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
51 * ::cuLaunchKernel will be a pointer to a buffer containing all kernel |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
52 * parameters used for launching kernel f. This buffer needs to |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
53 * honor all alignment/padding requirements of the individual parameters. |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
54 * If ::GRAAL_CU_LAUNCH_PARAM_BUFFER_SIZE is not also specified in the |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
55 * extra array, then ::GRAAL_CU_LAUNCH_PARAM_BUFFER_POINTER will have no |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
56 * effect. |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
57 */ |
|
11291
4a6b03a16808
Fix PTX test to prevent crash; trivial code cleanup
bharadwaj
parents:
11283
diff
changeset
|
58 #define GRAAL_CU_LAUNCH_PARAM_BUFFER_POINTER ((void *) 0x01) |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
59 |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
60 /** |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
61 * Indicator that the next value in the extra parameter to |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
62 * ::cuLaunchKernel will be a pointer to a size_t which contains the |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
63 * size of the buffer specified with ::GRAAL_CU_LAUNCH_PARAM_BUFFER_POINTER. |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
64 * It is required that ::GRAAL_CU_LAUNCH_PARAM_BUFFER_POINTER also be specified |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
65 * in the extra array if the value associated with |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
66 * ::GRAAL_CU_LAUNCH_PARAM_BUFFER_SIZE is not zero. |
|
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
67 */ |
|
11291
4a6b03a16808
Fix PTX test to prevent crash; trivial code cleanup
bharadwaj
parents:
11283
diff
changeset
|
68 #define GRAAL_CU_LAUNCH_PARAM_BUFFER_SIZE ((void *) 0x02) |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
69 |
| 10879 | 70 class Ptx { |
| 71 friend class gpu; | |
| 72 | |
| 73 protected: | |
| 74 static bool probe_linkage(); | |
| 75 static bool initialize_gpu(); | |
| 76 static void * generate_kernel(unsigned char *code, int code_len, const char *name); | |
|
11283
1cd1f8ff70a1
CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents:
10879
diff
changeset
|
77 static bool execute_kernel(address kernel, JavaCallArguments *); |
| 10879 | 78 |
| 79 private: | |
| 80 typedef int (*cuda_cu_init_func_t)(unsigned int); | |
| 81 typedef int (*cuda_cu_ctx_create_func_t)(void *, int, int); | |
| 82 typedef int (*cuda_cu_ctx_detach_func_t)(int *); | |
| 83 typedef int (*cuda_cu_ctx_synchronize_func_t)(int *); | |
| 84 typedef int (*cuda_cu_device_get_count_func_t)(int *); | |
| 85 typedef int (*cuda_cu_device_get_name_func_t)(char *, int, int); | |
| 86 typedef int (*cuda_cu_device_get_func_t)(int *, int); | |
| 87 typedef int (*cuda_cu_device_compute_capability_func_t)(int *, int *, int); | |
| 88 typedef int (*cuda_cu_device_get_attribute_func_t)(int *, int, int); | |
| 89 typedef int (*cuda_cu_launch_kernel_func_t)(void *, | |
| 90 unsigned int, unsigned int, unsigned int, | |
| 91 unsigned int, unsigned int, unsigned int, | |
| 92 unsigned int, void *, void **, void **); | |
| 93 typedef int (*cuda_cu_module_get_function_func_t)(void *, void *, const char *); | |
| 94 typedef int (*cuda_cu_module_load_data_ex_func_t)(void *, void *, unsigned int, void *, void **); | |
| 95 | |
| 96 static cuda_cu_init_func_t _cuda_cu_init; | |
| 97 static cuda_cu_ctx_create_func_t _cuda_cu_ctx_create; | |
| 98 static cuda_cu_ctx_detach_func_t _cuda_cu_ctx_detach; | |
| 99 static cuda_cu_ctx_synchronize_func_t _cuda_cu_ctx_synchronize; | |
| 100 static cuda_cu_device_get_count_func_t _cuda_cu_device_get_count; | |
| 101 static cuda_cu_device_get_name_func_t _cuda_cu_device_get_name; | |
| 102 static cuda_cu_device_get_func_t _cuda_cu_device_get; | |
| 103 static cuda_cu_device_compute_capability_func_t _cuda_cu_device_compute_capability; /* Deprecated as of CUDA 5.0 */ | |
| 104 static cuda_cu_device_get_attribute_func_t _cuda_cu_device_get_attribute; | |
| 105 static cuda_cu_launch_kernel_func_t _cuda_cu_launch_kernel; | |
| 106 static cuda_cu_module_get_function_func_t _cuda_cu_module_get_function; | |
| 107 static cuda_cu_module_load_data_ex_func_t _cuda_cu_module_load_data_ex; | |
| 108 | |
| 109 protected: | |
| 110 static void * _device_context; | |
| 111 }; | |
| 112 #endif // GPU_PTX_HPP |