1 /* Copyright (C) 2013-2023 Free Software Foundation, Inc.
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"
27 #include "libgomp-nvptx.h" /* For struct rev_offload + GOMP_REV_OFFLOAD_VAR. */
28 #include <limits.h>
29
30 extern int __gomp_team_num __attribute__((shared));
31 extern volatile struct gomp_offload_icvs GOMP_ADDITIONAL_ICVS;
32 volatile struct rev_offload *GOMP_REV_OFFLOAD_VAR;
33
34 bool
35 GOMP_teams4 (unsigned int num_teams_lower, unsigned int num_teams_upper,
36 unsigned int thread_limit, bool first)
37 {
38 unsigned int num_blocks, block_id;
39 asm ("mov.u32 %0, %%nctaid.x;" : "=r" (num_blocks));
40 if (!first)
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 }
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 }
57 if (!num_teams_upper)
58 num_teams_upper = num_blocks;
59 else if (num_blocks < num_teams_lower)
60 num_teams_upper = num_teams_lower;
61 else if (num_blocks < num_teams_upper)
62 num_teams_upper = num_blocks;
63 asm ("mov.u32 %0, %%ctaid.x;" : "=r" (block_id));
64 if (block_id >= num_teams_upper)
65 return false;
66 __gomp_team_num = block_id;
67 gomp_num_teams_var = num_teams_upper - 1;
68 return true;
69 }
70
71 int
72 omp_pause_resource (omp_pause_resource_t kind, int device_num)
73 {
74 (void) kind;
75 (void) device_num;
76 return -1;
77 }
78
79 int
80 omp_pause_resource_all (omp_pause_resource_t kind)
81 {
82 (void) kind;
83 return -1;
84 }
85
86 ialias (omp_pause_resource)
87 ialias (omp_pause_resource_all)
88
89 void
90 GOMP_target_ext (int device, void (*fn) (void *), size_t mapnum,
91 void **hostaddrs, size_t *sizes, unsigned short *kinds,
92 unsigned int flags, void **depend, void **args)
93 {
94 static int lock = 0; /* == gomp_mutex_t lock; gomp_mutex_init (&lock); */
95 (void) flags;
96 (void) depend;
97 (void) args;
98
99 if (device != GOMP_DEVICE_HOST_FALLBACK
100 || fn == NULL
101 || GOMP_REV_OFFLOAD_VAR == NULL)
102 return;
103
104 gomp_mutex_lock (&lock);
105
106 GOMP_REV_OFFLOAD_VAR->mapnum = mapnum;
107 GOMP_REV_OFFLOAD_VAR->addrs = (uint64_t) hostaddrs;
108 GOMP_REV_OFFLOAD_VAR->sizes = (uint64_t) sizes;
109 GOMP_REV_OFFLOAD_VAR->kinds = (uint64_t) kinds;
110 GOMP_REV_OFFLOAD_VAR->dev_num = GOMP_ADDITIONAL_ICVS.device_num;
111
112 /* Set 'fn' to trigger processing on the host; wait for completion,
113 which is flagged by setting 'fn' back to 0 on the host. */
114 uint64_t addr_struct_fn = (uint64_t) &GOMP_REV_OFFLOAD_VAR->fn;
115 #if __PTX_SM__ >= 700
116 asm volatile ("st.global.release.sys.u64 [%0], %1;"
117 : : "r"(addr_struct_fn), "r" (fn) : "memory");
118 #else
119 __sync_synchronize (); /* membar.sys */
120 asm volatile ("st.volatile.global.u64 [%0], %1;"
121 : : "r"(addr_struct_fn), "r" (fn) : "memory");
122 #endif
123
124 #if __PTX_SM__ >= 700
125 uint64_t fn2;
126 do
127 {
128 asm volatile ("ld.acquire.sys.global.u64 %0, [%1];"
129 : "=r" (fn2) : "r" (addr_struct_fn) : "memory");
130 }
131 while (fn2 != 0);
132 #else
133 /* ld.global.u64 %r64,[__gomp_rev_offload_var];
134 ld.u64 %r36,[%r64];
135 membar.sys; */
136 while (__atomic_load_n (&GOMP_REV_OFFLOAD_VAR->fn, __ATOMIC_ACQUIRE) != 0)
137 ; /* spin */
138 #endif
139
140 gomp_mutex_unlock (&lock);
141 }
142
143 void
144 GOMP_target_data_ext (int device, size_t mapnum, void **hostaddrs,
145 size_t *sizes, unsigned short *kinds)
146 {
147 (void) device;
148 (void) mapnum;
149 (void) hostaddrs;
150 (void) sizes;
151 (void) kinds;
152 __builtin_unreachable ();
153 }
154
155 void
156 GOMP_target_end_data (void)
157 {
158 __builtin_unreachable ();
159 }
160
161 void
162 GOMP_target_update_ext (int device, size_t mapnum, void **hostaddrs,
163 size_t *sizes, unsigned short *kinds,
164 unsigned int flags, void **depend)
165 {
166 (void) device;
167 (void) mapnum;
168 (void) hostaddrs;
169 (void) sizes;
170 (void) kinds;
171 (void) flags;
172 (void) depend;
173 __builtin_unreachable ();
174 }
175
176 void
177 GOMP_target_enter_exit_data (int device, size_t mapnum, void **hostaddrs,
178 size_t *sizes, unsigned short *kinds,
179 unsigned int flags, void **depend)
180 {
181 (void) device;
182 (void) mapnum;
183 (void) hostaddrs;
184 (void) sizes;
185 (void) kinds;
186 (void) flags;
187 (void) depend;
188 __builtin_unreachable ();
189 }