Hash :
b8d85c69
Author :
Date :
2022-11-30T12:51:59
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309
/*
Simple DirectMedia Layer
Copyright (C) 1997-2022 Sam Lantinga <slouken@libsdl.org>
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
#include "../SDL_internal.h"
#include "SDL_atomic.h"
#if defined(_MSC_VER) && (_MSC_VER >= 1500)
#include <intrin.h>
#define HAVE_MSC_ATOMICS 1
#endif
#if defined(__MACOSX__) /* !!! FIXME: should we favor gcc atomics? */
#include <libkern/OSAtomic.h>
#endif
#if !defined(HAVE_GCC_ATOMICS) && defined(__SOLARIS__)
#include <atomic.h>
#endif
/* The __atomic_load_n() intrinsic showed up in different times for different compilers. */
#if defined(__clang__)
#if __has_builtin(__atomic_load_n) || defined(HAVE_GCC_ATOMICS)
/* !!! FIXME: this advertises as available in the NDK but uses an external symbol we don't have.
It might be in a later NDK or we might need an extra library? --ryan. */
#if !defined(__ANDROID__)
#define HAVE_ATOMIC_LOAD_N 1
#endif
#endif
#elif defined(__GNUC__)
#if (__GNUC__ >= 5)
#define HAVE_ATOMIC_LOAD_N 1
#endif
#endif
/* *INDENT-OFF* */ /* clang-format off */
#if defined(__WATCOMC__) && defined(__386__)
SDL_COMPILE_TIME_ASSERT(intsize, 4==sizeof(int));
#define HAVE_WATCOM_ATOMICS
extern __inline int _SDL_xchg_watcom(volatile int *a, int v);
#pragma aux _SDL_xchg_watcom = \
"lock xchg [ecx], eax" \
parm [ecx] [eax] \
value [eax] \
modify exact [eax];
extern __inline unsigned char _SDL_cmpxchg_watcom(volatile int *a, int newval, int oldval);
#pragma aux _SDL_cmpxchg_watcom = \
"lock cmpxchg [edx], ecx" \
"setz al" \
parm [edx] [ecx] [eax] \
value [al] \
modify exact [eax];
extern __inline int _SDL_xadd_watcom(volatile int *a, int v);
#pragma aux _SDL_xadd_watcom = \
"lock xadd [ecx], eax" \
parm [ecx] [eax] \
value [eax] \
modify exact [eax];
#endif /* __WATCOMC__ && __386__ */
/* *INDENT-ON* */ /* clang-format on */
/*
If any of the operations are not provided then we must emulate some
of them. That means we need a nice implementation of spin locks
that avoids the "one big lock" problem. We use a vector of spin
locks and pick which one to use based on the address of the operand
of the function.
To generate the index of the lock we first shift by 3 bits to get
rid on the zero bits that result from 32 and 64 bit allignment of
data. We then mask off all but 5 bits and use those 5 bits as an
index into the table.
Picking the lock this way insures that accesses to the same data at
the same time will go to the same lock. OTOH, accesses to different
data have only a 1/32 chance of hitting the same lock. That should
pretty much eliminate the chances of several atomic operations on
different data from waiting on the same "big lock". If it isn't
then the table of locks can be expanded to a new size so long as
the new size is a power of two.
Contributed by Bob Pendleton, bob@pendleton.com
*/
#if !defined(HAVE_MSC_ATOMICS) && !defined(HAVE_GCC_ATOMICS) && !defined(__MACOSX__) && !defined(__SOLARIS__) && !defined(HAVE_WATCOM_ATOMICS)
#define EMULATE_CAS 1
#endif
#if EMULATE_CAS
static SDL_SpinLock locks[32];
static SDL_INLINE void enterLock(void *a)
{
uintptr_t index = ((((uintptr_t)a) >> 3) & 0x1f);
SDL_AtomicLock(&locks[index]);
}
static SDL_INLINE void leaveLock(void *a)
{
uintptr_t index = ((((uintptr_t)a) >> 3) & 0x1f);
SDL_AtomicUnlock(&locks[index]);
}
#endif
SDL_bool
SDL_AtomicCAS(SDL_atomic_t *a, int oldval, int newval)
{
#ifdef HAVE_MSC_ATOMICS
SDL_COMPILE_TIME_ASSERT(atomic_cas, sizeof(long) == sizeof(a->value));
return _InterlockedCompareExchange((long *)&a->value, (long)newval, (long)oldval) == (long)oldval;
#elif defined(HAVE_WATCOM_ATOMICS)
return (SDL_bool)_SDL_cmpxchg_watcom(&a->value, newval, oldval);
#elif defined(HAVE_GCC_ATOMICS)
return (SDL_bool) __sync_bool_compare_and_swap(&a->value, oldval, newval);
#elif defined(__MACOSX__) /* this is deprecated in 10.12 sdk; favor gcc atomics. */
return (SDL_bool) OSAtomicCompareAndSwap32Barrier(oldval, newval, &a->value);
#elif defined(__SOLARIS__)
return (SDL_bool)((int)atomic_cas_uint((volatile uint_t *)&a->value, (uint_t)oldval, (uint_t)newval) == oldval);
#elif EMULATE_CAS
SDL_bool retval = SDL_FALSE;
enterLock(a);
if (a->value == oldval) {
a->value = newval;
retval = SDL_TRUE;
}
leaveLock(a);
return retval;
#else
#error Please define your platform.
#endif
}
SDL_bool
SDL_AtomicCASPtr(void **a, void *oldval, void *newval)
{
#if defined(HAVE_MSC_ATOMICS)
return _InterlockedCompareExchangePointer(a, newval, oldval) == oldval;
#elif defined(HAVE_WATCOM_ATOMICS)
return (SDL_bool)_SDL_cmpxchg_watcom((int *)a, (long)newval, (long)oldval);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_bool_compare_and_swap(a, oldval, newval);
#elif defined(__MACOSX__) && defined(__LP64__) /* this is deprecated in 10.12 sdk; favor gcc atomics. */
return (SDL_bool) OSAtomicCompareAndSwap64Barrier((int64_t)oldval, (int64_t)newval, (int64_t*) a);
#elif defined(__MACOSX__) && !defined(__LP64__) /* this is deprecated in 10.12 sdk; favor gcc atomics. */
return (SDL_bool) OSAtomicCompareAndSwap32Barrier((int32_t)oldval, (int32_t)newval, (int32_t*) a);
#elif defined(__SOLARIS__)
return (SDL_bool)(atomic_cas_ptr(a, oldval, newval) == oldval);
#elif EMULATE_CAS
SDL_bool retval = SDL_FALSE;
enterLock(a);
if (*a == oldval) {
*a = newval;
retval = SDL_TRUE;
}
leaveLock(a);
return retval;
#else
#error Please define your platform.
#endif
}
int SDL_AtomicSet(SDL_atomic_t *a, int v)
{
#ifdef HAVE_MSC_ATOMICS
SDL_COMPILE_TIME_ASSERT(atomic_set, sizeof(long) == sizeof(a->value));
return _InterlockedExchange((long *)&a->value, v);
#elif defined(HAVE_WATCOM_ATOMICS)
return _SDL_xchg_watcom(&a->value, v);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_lock_test_and_set(&a->value, v);
#elif defined(__SOLARIS__)
return (int)atomic_swap_uint((volatile uint_t *)&a->value, v);
#else
int value;
do {
value = a->value;
} while (!SDL_AtomicCAS(a, value, v));
return value;
#endif
}
void *
SDL_AtomicSetPtr(void **a, void *v)
{
#if defined(HAVE_MSC_ATOMICS)
return _InterlockedExchangePointer(a, v);
#elif defined(HAVE_WATCOM_ATOMICS)
return (void *)_SDL_xchg_watcom((int *)a, (long)v);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_lock_test_and_set(a, v);
#elif defined(__SOLARIS__)
return atomic_swap_ptr(a, v);
#else
void *value;
do {
value = *a;
} while (!SDL_AtomicCASPtr(a, value, v));
return value;
#endif
}
int SDL_AtomicAdd(SDL_atomic_t *a, int v)
{
#ifdef HAVE_MSC_ATOMICS
SDL_COMPILE_TIME_ASSERT(atomic_add, sizeof(long) == sizeof(a->value));
return _InterlockedExchangeAdd((long *)&a->value, v);
#elif defined(HAVE_WATCOM_ATOMICS)
return _SDL_xadd_watcom(&a->value, v);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_fetch_and_add(&a->value, v);
#elif defined(__SOLARIS__)
int pv = a->value;
membar_consumer();
atomic_add_int((volatile uint_t *)&a->value, v);
return pv;
#else
int value;
do {
value = a->value;
} while (!SDL_AtomicCAS(a, value, (value + v)));
return value;
#endif
}
int SDL_AtomicGet(SDL_atomic_t *a)
{
#ifdef HAVE_ATOMIC_LOAD_N
return __atomic_load_n(&a->value, __ATOMIC_SEQ_CST);
#elif defined(HAVE_MSC_ATOMICS)
SDL_COMPILE_TIME_ASSERT(atomic_get, sizeof(long) == sizeof(a->value));
return _InterlockedOr((long *)&a->value, 0);
#elif defined(HAVE_WATCOM_ATOMICS)
return _SDL_xadd_watcom(&a->value, 0);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_or_and_fetch(&a->value, 0);
#elif defined(__MACOSX__) /* this is deprecated in 10.12 sdk; favor gcc atomics. */
return sizeof(a->value) == sizeof(uint32_t) ? OSAtomicOr32Barrier(0, (volatile uint32_t *)&a->value) : OSAtomicAdd64Barrier(0, (volatile int64_t *)&a->value);
#elif defined(__SOLARIS__)
return atomic_or_uint((volatile uint_t *)&a->value, 0);
#else
int value;
do {
value = a->value;
} while (!SDL_AtomicCAS(a, value, value));
return value;
#endif
}
void *
SDL_AtomicGetPtr(void **a)
{
#ifdef HAVE_ATOMIC_LOAD_N
return __atomic_load_n(a, __ATOMIC_SEQ_CST);
#elif defined(HAVE_MSC_ATOMICS)
return _InterlockedCompareExchangePointer(a, NULL, NULL);
#elif defined(HAVE_GCC_ATOMICS)
return __sync_val_compare_and_swap(a, (void *)0, (void *)0);
#elif defined(__SOLARIS__)
return atomic_cas_ptr(a, (void *)0, (void *)0);
#else
void *value;
do {
value = *a;
} while (!SDL_AtomicCASPtr(a, value, value));
return value;
#endif
}
#ifdef SDL_MEMORY_BARRIER_USES_FUNCTION
#error This file should be built in arm mode so the mcr instruction is available for memory barriers
#endif
void SDL_MemoryBarrierReleaseFunction(void)
{
SDL_MemoryBarrierRelease();
}
void SDL_MemoryBarrierAcquireFunction(void)
{
SDL_MemoryBarrierAcquire();
}
/* vi: set ts=4 sw=4 expandtab: */