]>
Commit | Line | Data |
---|---|---|
a945c346 | 1 | /* Copyright (C) 2013-2024 Free Software Foundation, Inc. |
6103184e AM |
2 | Contributed by Jakub Jelinek <jakub@redhat.com>. |
3 | ||
4 | This file is part of the GNU Offloading and Multi Processing Library | |
5 | (libgomp). | |
6 | ||
7 | Libgomp is free software; you can redistribute it and/or modify it | |
8 | under the terms of the GNU General Public License as published by | |
9 | the Free Software Foundation; either version 3, or (at your option) | |
10 | any later version. | |
11 | ||
12 | Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY | |
13 | WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS | |
14 | FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
15 | more details. | |
16 | ||
17 | Under Section 7 of GPL version 3, you are granted additional | |
18 | permissions described in the GCC Runtime Library Exception, version | |
19 | 3.1, as published by the Free Software Foundation. | |
20 | ||
21 | You should have received a copy of the GNU General Public License and | |
22 | a copy of the GCC Runtime Library Exception along with this program; | |
23 | see the files COPYING3 and COPYING.RUNTIME respectively. If not, see | |
24 | <http://www.gnu.org/licenses/>. */ | |
25 | ||
26 | #include "libgomp.h" | |
131d18e9 | 27 | #include "libgomp-nvptx.h" /* For struct rev_offload + GOMP_REV_OFFLOAD_VAR. */ |
6103184e AM |
28 | #include <limits.h> |
29 | ||
9fa72756 | 30 | extern int __gomp_team_num __attribute__((shared)); |
131d18e9 TB |
31 | extern volatile struct gomp_offload_icvs GOMP_ADDITIONAL_ICVS; |
32 | volatile struct rev_offload *GOMP_REV_OFFLOAD_VAR; | |
9fa72756 | 33 | |
7d6da11f JJ |
34 | bool |
35 | GOMP_teams4 (unsigned int num_teams_lower, unsigned int num_teams_upper, | |
36 | unsigned int thread_limit, bool first) | |
6103184e | 37 | { |
9fa72756 JJ |
38 | unsigned int num_blocks, block_id; |
39 | asm ("mov.u32 %0, %%nctaid.x;" : "=r" (num_blocks)); | |
7d6da11f | 40 | if (!first) |
9fa72756 JJ |
41 | { |
42 | unsigned int team_num; | |
43 | if (num_blocks > gomp_num_teams_var) | |
44 | return false; | |
45 | team_num = __gomp_team_num; | |
46 | if (team_num > gomp_num_teams_var - num_blocks) | |
47 | return false; | |
48 | __gomp_team_num = team_num + num_blocks; | |
49 | return true; | |
50 | } | |
6103184e AM |
51 | if (thread_limit) |
52 | { | |
53 | struct gomp_task_icv *icv = gomp_icv (true); | |
54 | icv->thread_limit_var | |
55 | = thread_limit > INT_MAX ? UINT_MAX : thread_limit; | |
56 | } | |
9fa72756 | 57 | if (!num_teams_upper) |
ad0f80d9 TB |
58 | num_teams_upper = ((GOMP_ADDITIONAL_ICVS.nteams > 0 |
59 | && num_blocks > GOMP_ADDITIONAL_ICVS.nteams) | |
60 | ? GOMP_ADDITIONAL_ICVS.nteams : num_blocks); | |
9fa72756 JJ |
61 | else if (num_blocks < num_teams_lower) |
62 | num_teams_upper = num_teams_lower; | |
63 | else if (num_blocks < num_teams_upper) | |
64 | num_teams_upper = num_blocks; | |
65 | asm ("mov.u32 %0, %%ctaid.x;" : "=r" (block_id)); | |
66 | if (block_id >= num_teams_upper) | |
7d6da11f | 67 | return false; |
9fa72756 | 68 | __gomp_team_num = block_id; |
7d6da11f JJ |
69 | gomp_num_teams_var = num_teams_upper - 1; |
70 | return true; | |
6103184e | 71 | } |
28567c40 JJ |
72 | |
73 | int | |
74 | omp_pause_resource (omp_pause_resource_t kind, int device_num) | |
75 | { | |
76 | (void) kind; | |
77 | (void) device_num; | |
78 | return -1; | |
79 | } | |
80 | ||
81 | int | |
82 | omp_pause_resource_all (omp_pause_resource_t kind) | |
83 | { | |
84 | (void) kind; | |
85 | return -1; | |
86 | } | |
87 | ||
88 | ialias (omp_pause_resource) | |
89 | ialias (omp_pause_resource_all) | |
95d67762 JJ |
90 | |
91 | void | |
92 | GOMP_target_ext (int device, void (*fn) (void *), size_t mapnum, | |
93 | void **hostaddrs, size_t *sizes, unsigned short *kinds, | |
94 | unsigned int flags, void **depend, void **args) | |
95 | { | |
131d18e9 | 96 | static int lock = 0; /* == gomp_mutex_t lock; gomp_mutex_init (&lock); */ |
95d67762 JJ |
97 | (void) flags; |
98 | (void) depend; | |
99 | (void) args; | |
131d18e9 TB |
100 | |
101 | if (device != GOMP_DEVICE_HOST_FALLBACK | |
102 | || fn == NULL | |
103 | || GOMP_REV_OFFLOAD_VAR == NULL) | |
104 | return; | |
105 | ||
106 | gomp_mutex_lock (&lock); | |
107 | ||
108 | GOMP_REV_OFFLOAD_VAR->mapnum = mapnum; | |
109 | GOMP_REV_OFFLOAD_VAR->addrs = (uint64_t) hostaddrs; | |
110 | GOMP_REV_OFFLOAD_VAR->sizes = (uint64_t) sizes; | |
111 | GOMP_REV_OFFLOAD_VAR->kinds = (uint64_t) kinds; | |
112 | GOMP_REV_OFFLOAD_VAR->dev_num = GOMP_ADDITIONAL_ICVS.device_num; | |
113 | ||
114 | /* Set 'fn' to trigger processing on the host; wait for completion, | |
115 | which is flagged by setting 'fn' back to 0 on the host. */ | |
116 | uint64_t addr_struct_fn = (uint64_t) &GOMP_REV_OFFLOAD_VAR->fn; | |
117 | #if __PTX_SM__ >= 700 | |
118 | asm volatile ("st.global.release.sys.u64 [%0], %1;" | |
119 | : : "r"(addr_struct_fn), "r" (fn) : "memory"); | |
120 | #else | |
121 | __sync_synchronize (); /* membar.sys */ | |
122 | asm volatile ("st.volatile.global.u64 [%0], %1;" | |
123 | : : "r"(addr_struct_fn), "r" (fn) : "memory"); | |
124 | #endif | |
125 | ||
126 | #if __PTX_SM__ >= 700 | |
127 | uint64_t fn2; | |
128 | do | |
129 | { | |
130 | asm volatile ("ld.acquire.sys.global.u64 %0, [%1];" | |
131 | : "=r" (fn2) : "r" (addr_struct_fn) : "memory"); | |
132 | } | |
133 | while (fn2 != 0); | |
134 | #else | |
135 | /* ld.global.u64 %r64,[__gomp_rev_offload_var]; | |
136 | ld.u64 %r36,[%r64]; | |
137 | membar.sys; */ | |
138 | while (__atomic_load_n (&GOMP_REV_OFFLOAD_VAR->fn, __ATOMIC_ACQUIRE) != 0) | |
139 | ; /* spin */ | |
140 | #endif | |
141 | ||
142 | gomp_mutex_unlock (&lock); | |
95d67762 JJ |
143 | } |
144 | ||
145 | void | |
146 | GOMP_target_data_ext (int device, size_t mapnum, void **hostaddrs, | |
147 | size_t *sizes, unsigned short *kinds) | |
148 | { | |
149 | (void) device; | |
150 | (void) mapnum; | |
151 | (void) hostaddrs; | |
152 | (void) sizes; | |
153 | (void) kinds; | |
154 | __builtin_unreachable (); | |
155 | } | |
156 | ||
157 | void | |
158 | GOMP_target_end_data (void) | |
159 | { | |
160 | __builtin_unreachable (); | |
161 | } | |
162 | ||
163 | void | |
164 | GOMP_target_update_ext (int device, size_t mapnum, void **hostaddrs, | |
165 | size_t *sizes, unsigned short *kinds, | |
166 | unsigned int flags, void **depend) | |
167 | { | |
168 | (void) device; | |
169 | (void) mapnum; | |
170 | (void) hostaddrs; | |
171 | (void) sizes; | |
172 | (void) kinds; | |
173 | (void) flags; | |
174 | (void) depend; | |
175 | __builtin_unreachable (); | |
176 | } | |
177 | ||
178 | void | |
179 | GOMP_target_enter_exit_data (int device, size_t mapnum, void **hostaddrs, | |
180 | size_t *sizes, unsigned short *kinds, | |
181 | unsigned int flags, void **depend) | |
182 | { | |
183 | (void) device; | |
184 | (void) mapnum; | |
185 | (void) hostaddrs; | |
186 | (void) sizes; | |
187 | (void) kinds; | |
188 | (void) flags; | |
189 | (void) depend; | |
190 | __builtin_unreachable (); | |
191 | } |