annotate src/gpu/ptx/vm/gpu_ptx.hpp @ 11842:8d8f63069f58

PTX warp limiter to available GPU processors
author Morris Meyer <morris.meyer@oracle.com>
date Mon, 30 Sep 2013 13:03:47 -0400
parents 365d8f385fb5
children c7abc8411011
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
1 /*
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
2 * Copyright (c) 2013, Oracle and/or its affiliates. All rights reserved.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
4 *
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
5 * This code is free software; you can redistribute it and/or modify it
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
6 * under the terms of the GNU General Public License version 2 only, as
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
7 * published by the Free Software Foundation.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
8 *
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
9 * This code is distributed in the hope that it will be useful, but WITHOUT
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
12 * version 2 for more details (a copy is included in the LICENSE file that
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
13 * accompanied this code).
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
14 *
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
15 * You should have received a copy of the GNU General Public License version
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
16 * 2 along with this work; if not, write to the Free Software Foundation,
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
18 *
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
20 * or visit www.oracle.com if you need additional information or have any
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
21 * questions.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
22 *
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
23 */
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
24
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
25 #ifndef GPU_PTX_HPP
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
26 #define GPU_PTX_HPP
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
27
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
28 /*
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
29 * Some useful macro definitions from publicly available cuda.h.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
30 * These definitions are for convenience.
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
31 */
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
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
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
35 #define GRAAL_CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR 75
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
36 #define GRAAL_CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR 76
11842
8d8f63069f58 PTX warp limiter to available GPU processors
Morris Meyer <morris.meyer@oracle.com>
parents: 11822
diff changeset
37 #define GRAAL_CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT 16
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
38 #define GRAAL_CU_JIT_MAX_REGISTERS 0
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
39 #define GRAAL_CU_JIT_THREADS_PER_BLOCK 1
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
40 #define GRAAL_CU_JIT_INFO_LOG_BUFFER 3
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
41 #define GRAAL_CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES 4
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
42 #define GRAAL_CUDA_ERROR_NO_BINARY_FOR_GPU 209
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
43
11283
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
44 /**
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
45 * 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
46 * ::cuLaunchKernel
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
47 */
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
48 #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
49
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
50 /**
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
51 * 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
52 * ::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
53 * 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
54 * 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
55 * 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
56 * 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
57 * effect.
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
58 */
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
59 #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
60
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
61 /**
1cd1f8ff70a1 CR-20: PTX kernel invocation with arguments - from Bharadwaj
Morris Meyer <morris.meyer@oracle.com>
parents: 10879
diff changeset
62 * 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
63 * ::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
64 * 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
65 * 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
66 * 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
67 * ::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
68 */
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
69 #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
70
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
71 class Ptx {
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
72 friend class gpu;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
73
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
74 protected:
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
75 static bool probe_linkage();
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
76 static bool initialize_gpu();
11842
8d8f63069f58 PTX warp limiter to available GPU processors
Morris Meyer <morris.meyer@oracle.com>
parents: 11822
diff changeset
77 static unsigned int total_cores();
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
78 static void * generate_kernel(unsigned char *code, int code_len, const char *name);
11822
365d8f385fb5 PTX: one-dimensional parallel warp invocation, ParallelOver annotation
Morris Meyer <morris.meyer@oracle.com>
parents: 11527
diff changeset
79 static bool execute_warp(int dimX, int dimY, int dimZ, address kernel, PTXKernelArguments & ka, JavaValue &ret);
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
80 static bool execute_kernel(address kernel, PTXKernelArguments & ka, JavaValue &ret);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
81 public:
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
82 #if defined(__x86_64) || defined(AMD64) || defined(_M_AMD64)
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
83 typedef unsigned long long CUdeviceptr;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
84 #else
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
85 typedef unsigned int CUdeviceptr;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
86 #endif
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
87
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
88 private:
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
89 typedef int (*cuda_cu_init_func_t)(unsigned int);
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
90 typedef int (*cuda_cu_ctx_create_func_t)(void*, int, int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
91 typedef int (*cuda_cu_ctx_destroy_func_t)(void*);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
92 typedef int (*cuda_cu_ctx_synchronize_func_t)(void);
11527
c99e65785936 Improvements to PTX codegen; allows more PTX tests that run on the device to pass.
bharadwaj
parents: 11485
diff changeset
93 typedef int (*cuda_cu_ctx_set_current_func_t)(void*);
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
94 typedef int (*cuda_cu_device_get_count_func_t)(int*);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
95 typedef int (*cuda_cu_device_get_name_func_t)(char*, int, int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
96 typedef int (*cuda_cu_device_get_func_t)(int*, int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
97 typedef int (*cuda_cu_device_compute_capability_func_t)(int*, int*, int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
98 typedef int (*cuda_cu_device_get_attribute_func_t)(int*, int, int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
99 typedef int (*cuda_cu_launch_kernel_func_t)(struct CUfunc_st*,
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
100 unsigned int, unsigned int, unsigned int,
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
101 unsigned int, unsigned int, unsigned int,
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
102 unsigned int, void*, void**, void**);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
103 typedef int (*cuda_cu_module_get_function_func_t)(void*, void*, const char*);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
104 typedef int (*cuda_cu_module_load_data_ex_func_t)(void*, void*, unsigned int, void*, void**);
11527
c99e65785936 Improvements to PTX codegen; allows more PTX tests that run on the device to pass.
bharadwaj
parents: 11485
diff changeset
105 typedef int (*cuda_cu_memalloc_func_t)(void*, size_t);
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
106 typedef int (*cuda_cu_memfree_func_t)(gpu::Ptx::CUdeviceptr);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
107 typedef int (*cuda_cu_memcpy_htod_func_t)(gpu::Ptx::CUdeviceptr, const void*, unsigned int);
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
108 typedef int (*cuda_cu_memcpy_dtoh_func_t)(const void*, gpu::Ptx::CUdeviceptr, unsigned int);
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
109
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
110 public:
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
111 static cuda_cu_init_func_t _cuda_cu_init;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
112 static cuda_cu_ctx_create_func_t _cuda_cu_ctx_create;
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
113 static cuda_cu_ctx_destroy_func_t _cuda_cu_ctx_destroy;
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
114 static cuda_cu_ctx_synchronize_func_t _cuda_cu_ctx_synchronize;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
115 static cuda_cu_device_get_count_func_t _cuda_cu_device_get_count;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
116 static cuda_cu_device_get_name_func_t _cuda_cu_device_get_name;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
117 static cuda_cu_device_get_func_t _cuda_cu_device_get;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
118 static cuda_cu_device_compute_capability_func_t _cuda_cu_device_compute_capability; /* Deprecated as of CUDA 5.0 */
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
119 static cuda_cu_device_get_attribute_func_t _cuda_cu_device_get_attribute;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
120 static cuda_cu_launch_kernel_func_t _cuda_cu_launch_kernel;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
121 static cuda_cu_module_get_function_func_t _cuda_cu_module_get_function;
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
122 static cuda_cu_module_load_data_ex_func_t _cuda_cu_module_load_data_ex;
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
123 static cuda_cu_memalloc_func_t _cuda_cu_memalloc;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
124 static cuda_cu_memfree_func_t _cuda_cu_memfree;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
125 static cuda_cu_memcpy_htod_func_t _cuda_cu_memcpy_htod;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
126 static cuda_cu_memcpy_dtoh_func_t _cuda_cu_memcpy_dtoh;
11527
c99e65785936 Improvements to PTX codegen; allows more PTX tests that run on the device to pass.
bharadwaj
parents: 11485
diff changeset
127 static cuda_cu_ctx_set_current_func_t _cuda_cu_ctx_set_current;
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
128
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
129 protected:
11485
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
130 static void* _device_context;
49bb1bc983c6 Implement several missing PTX codegen features; return value capture and method args passing of java method executed on GPU.
bharadwaj
parents: 11291
diff changeset
131 static int _cu_device;
10879
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
132 };
d55f24eac4b1 PTX support for Linux
Morris Meyer <morris.meyer@oracle.com>
parents:
diff changeset
133 #endif // GPU_PTX_HPP