| 1 |
/* Copyright (C) 2015-2025 Free Software Foundation, Inc. |
| 2 |
Contributed by Alexander Monakov <[email protected]> |
| 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 |
/* This file handles maintenance of threads on NVPTX. */ |
| 27 |
|
| 28 |
#if defined __nvptx_softstack__ && defined __nvptx_unisimt__ |
| 29 |
|
| 30 |
#include "libgomp.h" |
| 31 |
#include <stdlib.h> |
| 32 |
#include <string.h> |
| 33 |
|
| 34 |
struct gomp_thread *nvptx_thrs __attribute__((shared,nocommon)); |
| 35 |
int __gomp_team_num __attribute__((shared,nocommon)); |
| 36 |
|
| 37 |
static void gomp_thread_start (struct gomp_thread_pool *); |
| 38 |
extern void build_indirect_map (void); |
| 39 |
|
| 40 |
/* There should be some .shared space reserved for us. There's no way to |
| 41 |
express this magic extern sizeless array in C so use asm. */ |
| 42 |
asm (".extern .shared .u8 __nvptx_lowlat_pool[];\n"); |
| 43 |
|
| 44 |
/* Defined in basic-allocator.c via config/nvptx/allocator.c. */ |
| 45 |
void __nvptx_lowlat_init (void *heap, size_t size); |
| 46 |
|
| 47 |
/* This externally visible function handles target region entry. It |
| 48 |
sets up a per-team thread pool and transfers control by calling FN (FN_DATA) |
| 49 |
in the master thread or gomp_thread_start in other threads. |
| 50 |
|
| 51 |
The name of this function is part of the interface with the compiler: for |
| 52 |
each target region, GCC emits a PTX .kernel function that sets up soft-stack |
| 53 |
and uniform-simt state and calls this function, passing in FN the original |
| 54 |
function outlined for the target region. */ |
| 55 |
|
| 56 |
void |
| 57 |
gomp_nvptx_main (void (*fn) (void *), void *fn_data) |
| 58 |
{ |
| 59 |
int tid, ntids; |
| 60 |
asm ("mov.u32 %0, %%tid.y;" : "=r" (tid)); |
| 61 |
asm ("mov.u32 %0, %%ntid.y;" : "=r" (ntids)); |
| 62 |
|
| 63 |
if (tid == 0) |
| 64 |
{ |
| 65 |
gomp_global_icv.nthreads_var = ntids; |
| 66 |
gomp_global_icv.thread_limit_var = ntids; |
| 67 |
/* Starting additional threads is not supported. */ |
| 68 |
gomp_global_icv.dyn_var = true; |
| 69 |
|
| 70 |
__gomp_team_num = 0; |
| 71 |
nvptx_thrs = alloca (ntids * sizeof (*nvptx_thrs)); |
| 72 |
memset (nvptx_thrs, 0, ntids * sizeof (*nvptx_thrs)); |
| 73 |
|
| 74 |
/* Initialize indirect function support. */ |
| 75 |
unsigned int block_id; |
| 76 |
asm ("mov.u32 %0, %%ctaid.x;" : "=r" (block_id)); |
| 77 |
if (block_id == 0) |
| 78 |
build_indirect_map (); |
| 79 |
|
| 80 |
/* Find the low-latency heap details .... */ |
| 81 |
uint32_t *shared_pool; |
| 82 |
uint32_t shared_pool_size = 0; |
| 83 |
asm ("cvta.shared.u64\t%0, __nvptx_lowlat_pool;" : "=r"(shared_pool)); |
| 84 |
#if __PTX_ISA_VERSION_MAJOR__ > 4 \ |
| 85 |
|| (__PTX_ISA_VERSION_MAJOR__ == 4 && __PTX_ISA_VERSION_MINOR__ >= 1) |
| 86 |
asm ("mov.u32\t%0, %%dynamic_smem_size;\n" |
| 87 |
: "=r"(shared_pool_size)); |
| 88 |
#endif |
| 89 |
__nvptx_lowlat_init (shared_pool, shared_pool_size); |
| 90 |
|
| 91 |
/* Initialize the thread pool. */ |
| 92 |
struct gomp_thread_pool *pool = alloca (sizeof (*pool)); |
| 93 |
pool->threads = alloca (ntids * sizeof (*pool->threads)); |
| 94 |
for (tid = 0; tid < ntids; tid++) |
| 95 |
pool->threads[tid] = nvptx_thrs + tid; |
| 96 |
pool->threads_size = ntids; |
| 97 |
pool->threads_used = ntids; |
| 98 |
pool->threads_busy = 1; |
| 99 |
pool->last_team = NULL; |
| 100 |
gomp_simple_barrier_init (&pool->threads_dock, ntids); |
| 101 |
|
| 102 |
nvptx_thrs[0].thread_pool = pool; |
| 103 |
asm ("bar.sync 0;"); |
| 104 |
fn (fn_data); |
| 105 |
|
| 106 |
gomp_free_thread (nvptx_thrs); |
| 107 |
} |
| 108 |
else |
| 109 |
{ |
| 110 |
asm ("bar.sync 0;"); |
| 111 |
gomp_thread_start (nvptx_thrs[0].thread_pool); |
| 112 |
} |
| 113 |
} |
| 114 |
|
| 115 |
/* This function contains the idle loop in which a thread waits |
| 116 |
to be called up to become part of a team. */ |
| 117 |
|
| 118 |
static void |
| 119 |
gomp_thread_start (struct gomp_thread_pool *pool) |
| 120 |
{ |
| 121 |
struct gomp_thread *thr = gomp_thread (); |
| 122 |
|
| 123 |
gomp_sem_init (&thr->release, 0); |
| 124 |
thr->thread_pool = pool; |
| 125 |
|
| 126 |
do |
| 127 |
{ |
| 128 |
gomp_simple_barrier_wait (&pool->threads_dock); |
| 129 |
if (!thr->fn) |
| 130 |
continue; |
| 131 |
thr->fn (thr->data); |
| 132 |
thr->fn = NULL; |
| 133 |
|
| 134 |
struct gomp_task *task = thr->task; |
| 135 |
gomp_team_barrier_wait_final (&thr->ts.team->barrier); |
| 136 |
gomp_finish_task (task); |
| 137 |
} |
| 138 |
/* Work around an NVIDIA driver bug: when generating sm_50 machine code, |
| 139 |
it can trash stack pointer R1 in loops lacking exit edges. Add a cheap |
| 140 |
artificial exit that the driver would not be able to optimize out. */ |
| 141 |
while (nvptx_thrs); |
| 142 |
} |
| 143 |
|
| 144 |
/* Launch a team. */ |
| 145 |
|
| 146 |
void |
| 147 |
gomp_team_start (void (*fn) (void *), void *data, unsigned nthreads, |
| 148 |
unsigned flags, struct gomp_team *team, |
| 149 |
struct gomp_taskgroup *taskgroup) |
| 150 |
{ |
| 151 |
struct gomp_thread *thr, *nthr; |
| 152 |
struct gomp_task *task; |
| 153 |
struct gomp_task_icv *icv; |
| 154 |
struct gomp_thread_pool *pool; |
| 155 |
unsigned long nthreads_var; |
| 156 |
|
| 157 |
thr = gomp_thread (); |
| 158 |
pool = thr->thread_pool; |
| 159 |
task = thr->task; |
| 160 |
icv = task ? &task->icv : &gomp_global_icv; |
| 161 |
|
| 162 |
/* Always save the previous state, even if this isn't a nested team. |
| 163 |
In particular, we should save any work share state from an outer |
| 164 |
orphaned work share construct. */ |
| 165 |
team->prev_ts = thr->ts; |
| 166 |
|
| 167 |
thr->ts.team = team; |
| 168 |
thr->ts.team_id = 0; |
| 169 |
++thr->ts.level; |
| 170 |
if (nthreads > 1) |
| 171 |
++thr->ts.active_level; |
| 172 |
thr->ts.work_share = &team->work_shares[0]; |
| 173 |
thr->ts.last_work_share = NULL; |
| 174 |
thr->ts.single_count = 0; |
| 175 |
thr->ts.static_trip = 0; |
| 176 |
thr->task = &team->implicit_task[0]; |
| 177 |
nthreads_var = icv->nthreads_var; |
| 178 |
gomp_init_task (thr->task, task, icv); |
| 179 |
team->implicit_task[0].icv.nthreads_var = nthreads_var; |
| 180 |
team->implicit_task[0].taskgroup = taskgroup; |
| 181 |
|
| 182 |
if (nthreads == 1) |
| 183 |
return; |
| 184 |
|
| 185 |
/* Release existing idle threads. */ |
| 186 |
for (unsigned i = 1; i < nthreads; ++i) |
| 187 |
{ |
| 188 |
nthr = pool->threads[i]; |
| 189 |
nthr->ts.team = team; |
| 190 |
nthr->ts.work_share = &team->work_shares[0]; |
| 191 |
nthr->ts.last_work_share = NULL; |
| 192 |
nthr->ts.team_id = i; |
| 193 |
nthr->ts.level = team->prev_ts.level + 1; |
| 194 |
nthr->ts.active_level = thr->ts.active_level; |
| 195 |
nthr->ts.single_count = 0; |
| 196 |
nthr->ts.static_trip = 0; |
| 197 |
nthr->task = &team->implicit_task[i]; |
| 198 |
gomp_init_task (nthr->task, task, icv); |
| 199 |
team->implicit_task[i].icv.nthreads_var = nthreads_var; |
| 200 |
team->implicit_task[i].taskgroup = taskgroup; |
| 201 |
nthr->fn = fn; |
| 202 |
nthr->data = data; |
| 203 |
team->ordered_release[i] = &nthr->release; |
| 204 |
} |
| 205 |
|
| 206 |
gomp_simple_barrier_wait (&pool->threads_dock); |
| 207 |
} |
| 208 |
|
| 209 |
int |
| 210 |
gomp_pause_host (void) |
| 211 |
{ |
| 212 |
return -1; |
| 213 |
} |
| 214 |
|
| 215 |
#include "../../team.c" |
| 216 |
#endif |