From 50f448ee499448bf017327d495884de6f52c9762 Mon Sep 17 00:00:00 2001 From: Chia-I Wu Date: Thu, 23 Sep 2021 17:22:56 -0700 Subject: [PATCH] mesa: update to the latest u_atomic.h This requires gcc 4.1. Signed-off-by: Chia-I Wu Reviewed-by: Yiwei Zhang Reviewed-by: Ryan Neph Acked-by: Gert Wollny --- meson.build | 4 + src/gallium/auxiliary/util/u_atomic.h | 352 -------------------------- src/gallium/meson.build | 1 - src/mesa/util/u_atomic.h | 272 ++++++++++++++++++++ 4 files changed, 276 insertions(+), 353 deletions(-) delete mode 100644 src/gallium/auxiliary/util/u_atomic.h create mode 100644 src/mesa/util/u_atomic.h diff --git a/meson.build b/meson.build index 082e0b5..f6ed468 100644 --- a/meson.build +++ b/meson.build @@ -44,6 +44,10 @@ revision = 3 cc = meson.get_compiler('c') +if cc.get_id() == 'gcc' and cc.version().version_compare('< 4.1') + error('When using GCC, version 4.1 or later is required.') +endif + warnings = [ '-Werror=implicit-function-declaration', '-Werror=missing-prototypes', diff --git a/src/gallium/auxiliary/util/u_atomic.h b/src/gallium/auxiliary/util/u_atomic.h deleted file mode 100644 index fbe936c..0000000 --- a/src/gallium/auxiliary/util/u_atomic.h +++ /dev/null @@ -1,352 +0,0 @@ -/** - * Many similar implementations exist. See for example libwsbm - * or the linux kernel include/atomic.h - * - * No copyright claimed on this file. - * - */ - -#ifndef U_ATOMIC_H -#define U_ATOMIC_H - -#include "pipe/p_compiler.h" -#include "pipe/p_defines.h" - -/* Favor OS-provided implementations. - * - * Where no OS-provided implementation is available, fall back to - * locally coded assembly, compiler intrinsic or ultimately a - * mutex-based implementation. - */ -#if defined(PIPE_OS_SOLARIS) -#define PIPE_ATOMIC_OS_SOLARIS -#elif defined(PIPE_CC_MSVC) -#define PIPE_ATOMIC_MSVC_INTRINSIC -#elif (defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)) -#define PIPE_ATOMIC_ASM_MSVC_X86 -#elif (defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)) -#define PIPE_ATOMIC_ASM_GCC_X86 -#elif (defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86_64)) -#define PIPE_ATOMIC_ASM_GCC_X86_64 -#elif defined(PIPE_CC_GCC) && (PIPE_CC_GCC_VERSION >= 401) -#define PIPE_ATOMIC_GCC_INTRINSIC -#else -#error "Unsupported platform" -#endif - - -#if defined(PIPE_ATOMIC_ASM_GCC_X86_64) -#define PIPE_ATOMIC "GCC x86_64 assembly" - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -__attribute__((no_sanitize("memory"))) -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - unsigned char c; - - __asm__ __volatile__("lock; decl %0; sete %1":"+m"(*v), "=qm"(c) - ::"memory"); - - return c != 0; -} - -static inline void -p_atomic_inc(int32_t *v) -{ - __asm__ __volatile__("lock; incl %0":"+m"(*v)); -} - -static inline void -p_atomic_dec(int32_t *v) -{ - __asm__ __volatile__("lock; decl %0":"+m"(*v)); -} - -static inline int32_t -p_atomic_cmpxchg(int32_t *v, int32_t old, int32_t _new) -{ - return __sync_val_compare_and_swap(v, old, _new); -} - -#ifdef __cplusplus -} -#endif - -#endif /* PIPE_ATOMIC_ASM_GCC_X86_64 */ - - -#if defined(PIPE_ATOMIC_ASM_GCC_X86) - -#define PIPE_ATOMIC "GCC x86 assembly" - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -__attribute__((no_sanitize("memory"))) -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - unsigned char c; - - __asm__ __volatile__("lock; decl %0; sete %1":"+m"(*v), "=qm"(c) - ::"memory"); - - return c != 0; -} - -static inline void -p_atomic_inc(int32_t *v) -{ - __asm__ __volatile__("lock; incl %0":"+m"(*v)); -} - -static inline void -p_atomic_dec(int32_t *v) -{ - __asm__ __volatile__("lock; decl %0":"+m"(*v)); -} - -static inline int32_t -p_atomic_cmpxchg(int32_t *v, int32_t old, int32_t _new) -{ - return __sync_val_compare_and_swap(v, old, _new); -} - -#ifdef __cplusplus -} -#endif - -#endif - - - -/* Implementation using GCC-provided synchronization intrinsics - */ -#if defined(PIPE_ATOMIC_GCC_INTRINSIC) - -#define PIPE_ATOMIC "GCC Sync Intrinsics" - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - return (__sync_sub_and_fetch(v, 1) == 0); -} - -static inline void -p_atomic_inc(int32_t *v) -{ - (void) __sync_add_and_fetch(v, 1); -} - -static inline void -p_atomic_dec(int32_t *v) -{ - (void) __sync_sub_and_fetch(v, 1); -} - -static inline int32_t -p_atomic_cmpxchg(int32_t *v, int32_t old, int32_t _new) -{ - return __sync_val_compare_and_swap(v, old, _new); -} - -#ifdef __cplusplus -} -#endif - -#endif - - - -/* Unlocked version for single threaded environments, such as some - * windows kernel modules. - */ -#if defined(PIPE_ATOMIC_OS_UNLOCKED) - -#define PIPE_ATOMIC "Unlocked" - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) -#define p_atomic_dec_zero(_v) ((boolean) --(*(_v))) -#define p_atomic_inc(_v) ((void) (*(_v))++) -#define p_atomic_dec(_v) ((void) (*(_v))--) -#define p_atomic_cmpxchg(_v, old, _new) (*(_v) == old ? *(_v) = (_new) : *(_v)) - -#endif - - -/* Locally coded assembly for MSVC on x86: - */ -#if defined(PIPE_ATOMIC_ASM_MSVC_X86) - -#define PIPE_ATOMIC "MSVC x86 assembly" - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -__attribute__((no_sanitize("memory"))) -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - unsigned char c; - - __asm { - mov eax, [v] - lock dec dword ptr [eax] - sete byte ptr [c] - } - - return c != 0; -} - -static inline void -p_atomic_inc(int32_t *v) -{ - __asm { - mov eax, [v] - lock inc dword ptr [eax] - } -} - -static inline void -p_atomic_dec(int32_t *v) -{ - __asm { - mov eax, [v] - lock dec dword ptr [eax] - } -} - -static inline int32_t -p_atomic_cmpxchg(int32_t *v, int32_t old, int32_t _new) -{ - int32_t orig; - - __asm { - mov ecx, [v] - mov eax, [old] - mov edx, [_new] - lock cmpxchg [ecx], edx - mov [orig], eax - } - - return orig; -} - -#ifdef __cplusplus -} -#endif - -#endif - - -#if defined(PIPE_ATOMIC_MSVC_INTRINSIC) - -#define PIPE_ATOMIC "MSVC Intrinsics" - -#include - -#pragma intrinsic(_InterlockedIncrement) -#pragma intrinsic(_InterlockedDecrement) -#pragma intrinsic(_InterlockedCompareExchange) - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - return _InterlockedDecrement((long *)v) == 0; -} - -static inline void -p_atomic_inc(int32_t *v) -{ - _InterlockedIncrement((long *)v); -} - -static inline void -p_atomic_dec(int32_t *v) -{ - _InterlockedDecrement((long *)v); -} - -static inline int32_t -p_atomic_cmpxchg(int32_t *v, int32_t old, int32_t _new) -{ - return _InterlockedCompareExchange((long *)v, _new, old); -} - -#ifdef __cplusplus -} -#endif - -#endif - -#if defined(PIPE_ATOMIC_OS_SOLARIS) - -#define PIPE_ATOMIC "Solaris OS atomic functions" - -#include - -#ifdef __cplusplus -extern "C" { -#endif - -#define p_atomic_set(_v, _i) (*(_v) = (_i)) -#define p_atomic_read(_v) (*(_v)) - -static inline boolean -p_atomic_dec_zero(int32_t *v) -{ - uint32_t n = atomic_dec_32_nv((uint32_t *) v); - - return n != 0; -} - -#define p_atomic_inc(_v) atomic_inc_32((uint32_t *) _v) -#define p_atomic_dec(_v) atomic_dec_32((uint32_t *) _v) - -#define p_atomic_cmpxchg(_v, _old, _new) \ - atomic_cas_32( (uint32_t *) _v, (uint32_t) _old, (uint32_t) _new) - -#ifdef __cplusplus -} -#endif - -#endif - - -#ifndef PIPE_ATOMIC -#error "No pipe_atomic implementation selected" -#endif - - - -#endif /* U_ATOMIC_H */ diff --git a/src/gallium/meson.build b/src/gallium/meson.build index a7276df..a5b7588 100644 --- a/src/gallium/meson.build +++ b/src/gallium/meson.build @@ -53,7 +53,6 @@ sources_libgallium = [ 'auxiliary/util/u_half.h', 'auxiliary/util/u_prim.h', 'auxiliary/util/u_debug_describe.c', - 'auxiliary/util/u_atomic.h', 'auxiliary/util/xxhash.h', 'auxiliary/cso_cache/cso_hash.h', 'auxiliary/cso_cache/cso_cache.h', diff --git a/src/mesa/util/u_atomic.h b/src/mesa/util/u_atomic.h new file mode 100644 index 0000000..0bd6a4a --- /dev/null +++ b/src/mesa/util/u_atomic.h @@ -0,0 +1,272 @@ +/** + * Many similar implementations exist. See for example libwsbm + * or the linux kernel include/atomic.h + * + * No copyright claimed on this file. + * + */ + +#include "no_extern_c.h" + +#ifndef U_ATOMIC_H +#define U_ATOMIC_H + +#include +#include + +/* Favor OS-provided implementations. + * + * Where no OS-provided implementation is available, fall back to + * locally coded assembly, compiler intrinsic or ultimately a + * mutex-based implementation. + */ +#if defined(__sun) +#define PIPE_ATOMIC_OS_SOLARIS +#elif defined(_MSC_VER) +#define PIPE_ATOMIC_MSVC_INTRINSIC +#elif defined(__GNUC__) +#define PIPE_ATOMIC_GCC_INTRINSIC +#else +#error "Unsupported platform" +#endif + + +/* Implementation using GCC-provided synchronization intrinsics + */ +#if defined(PIPE_ATOMIC_GCC_INTRINSIC) + +#define PIPE_ATOMIC "GCC Sync Intrinsics" + +#if defined(USE_GCC_ATOMIC_BUILTINS) + +/* The builtins with explicit memory model are available since GCC 4.7. */ +#define p_atomic_set(_v, _i) __atomic_store_n((_v), (_i), __ATOMIC_RELEASE) +#define p_atomic_read(_v) __atomic_load_n((_v), __ATOMIC_ACQUIRE) +#define p_atomic_read_relaxed(_v) __atomic_load_n((_v), __ATOMIC_RELAXED) +#define p_atomic_dec_zero(v) (__atomic_sub_fetch((v), 1, __ATOMIC_ACQ_REL) == 0) +#define p_atomic_inc(v) (void) __atomic_add_fetch((v), 1, __ATOMIC_ACQ_REL) +#define p_atomic_dec(v) (void) __atomic_sub_fetch((v), 1, __ATOMIC_ACQ_REL) +#define p_atomic_add(v, i) (void) __atomic_add_fetch((v), (i), __ATOMIC_ACQ_REL) +#define p_atomic_inc_return(v) __atomic_add_fetch((v), 1, __ATOMIC_ACQ_REL) +#define p_atomic_dec_return(v) __atomic_sub_fetch((v), 1, __ATOMIC_ACQ_REL) +#define p_atomic_add_return(v, i) __atomic_add_fetch((v), (i), __ATOMIC_ACQ_REL) +#define p_atomic_xchg(v, i) __atomic_exchange_n((v), (i), __ATOMIC_ACQ_REL) +#define PIPE_NATIVE_ATOMIC_XCHG + +#else + +#define p_atomic_set(_v, _i) (*(_v) = (_i)) +#define p_atomic_read(_v) (*(_v)) +#define p_atomic_read_relaxed(_v) (*(_v)) +#define p_atomic_dec_zero(v) (__sync_sub_and_fetch((v), 1) == 0) +#define p_atomic_inc(v) (void) __sync_add_and_fetch((v), 1) +#define p_atomic_dec(v) (void) __sync_sub_and_fetch((v), 1) +#define p_atomic_add(v, i) (void) __sync_add_and_fetch((v), (i)) +#define p_atomic_inc_return(v) __sync_add_and_fetch((v), 1) +#define p_atomic_dec_return(v) __sync_sub_and_fetch((v), 1) +#define p_atomic_add_return(v, i) __sync_add_and_fetch((v), (i)) + +#endif + +/* There is no __atomic_* compare and exchange that returns the current value. + * Also, GCC 5.4 seems unable to optimize a compound statement expression that + * uses an additional stack variable with __atomic_compare_exchange[_n]. + */ +#define p_atomic_cmpxchg(v, old, _new) \ + __sync_val_compare_and_swap((v), (old), (_new)) + +#endif + + + +/* Unlocked version for single threaded environments, such as some + * windows kernel modules. + */ +#if defined(PIPE_ATOMIC_OS_UNLOCKED) + +#define PIPE_ATOMIC "Unlocked" + +#define p_atomic_set(_v, _i) (*(_v) = (_i)) +#define p_atomic_read(_v) (*(_v)) +#define p_atomic_read_relaxed(_v) (*(_v)) +#define p_atomic_dec_zero(_v) (p_atomic_dec_return(_v) == 0) +#define p_atomic_inc(_v) ((void) p_atomic_inc_return(_v)) +#define p_atomic_dec(_v) ((void) p_atomic_dec_return(_v)) +#define p_atomic_add(_v, _i) ((void) p_atomic_add_return((_v), (_i))) +#define p_atomic_inc_return(_v) (++(*(_v))) +#define p_atomic_dec_return(_v) (--(*(_v))) +#define p_atomic_add_return(_v, _i) (*(_v) = *(_v) + (_i)) +#define p_atomic_cmpxchg(_v, _old, _new) (*(_v) == (_old) ? (*(_v) = (_new), (_old)) : *(_v)) + +#endif + + +#if defined(PIPE_ATOMIC_MSVC_INTRINSIC) + +#define PIPE_ATOMIC "MSVC Intrinsics" + +/* We use the Windows header's Interlocked*64 functions instead of the + * _Interlocked*64 intrinsics wherever we can, as support for the latter varies + * with target CPU, whereas Windows headers take care of all portability + * issues: using intrinsics where available, falling back to library + * implementations where not. + */ +#ifndef WIN32_LEAN_AND_MEAN +#define WIN32_LEAN_AND_MEAN 1 +#endif +#include +#include +#include + +/* MSVC supports decltype keyword, but it's only supported on C++ and doesn't + * quite work here; and if a C++-only solution is worthwhile, then it would be + * better to use templates / function overloading, instead of decltype magic. + * Therefore, we rely on implicit casting to LONGLONG for the functions that return + */ + +#define p_atomic_set(_v, _i) (*(_v) = (_i)) +#define p_atomic_read(_v) (*(_v)) +#define p_atomic_read_relaxed(_v) (*(_v)) + +#define p_atomic_dec_zero(_v) \ + (p_atomic_dec_return(_v) == 0) + +#define p_atomic_inc(_v) \ + ((void) p_atomic_inc_return(_v)) + +#define p_atomic_inc_return(_v) (\ + sizeof *(_v) == sizeof(short) ? _InterlockedIncrement16((short *) (_v)) : \ + sizeof *(_v) == sizeof(long) ? _InterlockedIncrement ((long *) (_v)) : \ + sizeof *(_v) == sizeof(__int64) ? InterlockedIncrement64 ((__int64 *)(_v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_dec(_v) \ + ((void) p_atomic_dec_return(_v)) + +#define p_atomic_dec_return(_v) (\ + sizeof *(_v) == sizeof(short) ? _InterlockedDecrement16((short *) (_v)) : \ + sizeof *(_v) == sizeof(long) ? _InterlockedDecrement ((long *) (_v)) : \ + sizeof *(_v) == sizeof(__int64) ? InterlockedDecrement64 ((__int64 *)(_v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_add(_v, _i) \ + ((void) p_atomic_add_return((_v), (_i))) + +#define p_atomic_add_return(_v, _i) (\ + sizeof *(_v) == sizeof(char) ? _InterlockedExchangeAdd8 ((char *) (_v), (_i)) : \ + sizeof *(_v) == sizeof(short) ? _InterlockedExchangeAdd16((short *) (_v), (_i)) : \ + sizeof *(_v) == sizeof(long) ? _InterlockedExchangeAdd ((long *) (_v), (_i)) : \ + sizeof *(_v) == sizeof(__int64) ? InterlockedExchangeAdd64((__int64 *)(_v), (_i)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_cmpxchg(_v, _old, _new) (\ + sizeof *(_v) == sizeof(char) ? _InterlockedCompareExchange8 ((char *) (_v), (char) (_new), (char) (_old)) : \ + sizeof *(_v) == sizeof(short) ? _InterlockedCompareExchange16((short *) (_v), (short) (_new), (short) (_old)) : \ + sizeof *(_v) == sizeof(long) ? _InterlockedCompareExchange ((long *) (_v), (long) (_new), (long) (_old)) : \ + sizeof *(_v) == sizeof(__int64) ? InterlockedCompareExchange64 ((__int64 *)(_v), (__int64)(_new), (__int64)(_old)) : \ + (assert(!"should not get here"), 0)) + +#endif + +#if defined(PIPE_ATOMIC_OS_SOLARIS) + +#define PIPE_ATOMIC "Solaris OS atomic functions" + +#include +#include + +#define p_atomic_set(_v, _i) (*(_v) = (_i)) +#define p_atomic_read(_v) (*(_v)) + +#define p_atomic_dec_zero(v) (\ + sizeof(*v) == sizeof(uint8_t) ? atomic_dec_8_nv ((uint8_t *)(v)) == 0 : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_dec_16_nv((uint16_t *)(v)) == 0 : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_dec_32_nv((uint32_t *)(v)) == 0 : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_dec_64_nv((uint64_t *)(v)) == 0 : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_inc(v) (void) (\ + sizeof(*v) == sizeof(uint8_t) ? atomic_inc_8 ((uint8_t *)(v)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_inc_16((uint16_t *)(v)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_inc_32((uint32_t *)(v)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_inc_64((uint64_t *)(v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_inc_return(v) (__typeof(*v))( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_inc_8_nv ((uint8_t *)(v)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_inc_16_nv((uint16_t *)(v)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_inc_32_nv((uint32_t *)(v)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_inc_64_nv((uint64_t *)(v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_dec(v) (void) ( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_dec_8 ((uint8_t *)(v)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_dec_16((uint16_t *)(v)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_dec_32((uint32_t *)(v)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_dec_64((uint64_t *)(v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_dec_return(v) (__typeof(*v))( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_dec_8_nv ((uint8_t *)(v)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_dec_16_nv((uint16_t *)(v)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_dec_32_nv((uint32_t *)(v)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_dec_64_nv((uint64_t *)(v)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_add(v, i) (void) ( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_add_8 ((uint8_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_add_16((uint16_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_add_32((uint32_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_add_64((uint64_t *)(v), (i)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_add_return(v, i) (__typeof(*v)) ( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_add_8_nv ((uint8_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_add_16_nv((uint16_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_add_32_nv((uint32_t *)(v), (i)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_add_64_nv((uint64_t *)(v), (i)) : \ + (assert(!"should not get here"), 0)) + +#define p_atomic_cmpxchg(v, old, _new) (__typeof(*v))( \ + sizeof(*v) == sizeof(uint8_t) ? atomic_cas_8 ((uint8_t *)(v), (uint8_t )(old), (uint8_t )(_new)) : \ + sizeof(*v) == sizeof(uint16_t) ? atomic_cas_16((uint16_t *)(v), (uint16_t)(old), (uint16_t)(_new)) : \ + sizeof(*v) == sizeof(uint32_t) ? atomic_cas_32((uint32_t *)(v), (uint32_t)(old), (uint32_t)(_new)) : \ + sizeof(*v) == sizeof(uint64_t) ? atomic_cas_64((uint64_t *)(v), (uint64_t)(old), (uint64_t)(_new)) : \ + (assert(!"should not get here"), 0)) + +#endif + +#ifndef PIPE_ATOMIC +#error "No pipe_atomic implementation selected" +#endif + +#ifndef PIPE_NATIVE_ATOMIC_XCHG +static inline uint32_t p_atomic_xchg_32(uint32_t *v, uint32_t i) +{ + uint32_t actual = p_atomic_read(v); + uint32_t expected; + do { + expected = actual; + actual = p_atomic_cmpxchg(v, expected, i); + } while (expected != actual); + return actual; +} + +static inline uint64_t p_atomic_xchg_64(uint64_t *v, uint64_t i) +{ + uint64_t actual = p_atomic_read(v); + uint64_t expected; + do { + expected = actual; + actual = p_atomic_cmpxchg(v, expected, i); + } while (expected != actual); + return actual; +} + +#define p_atomic_xchg(v, i) (__typeof(*(v)))( \ + sizeof(*(v)) == sizeof(uint32_t) ? p_atomic_xchg_32((uint32_t *)(v), (uint32_t)(i)) : \ + sizeof(*(v)) == sizeof(uint64_t) ? p_atomic_xchg_64((uint64_t *)(v), (uint64_t)(i)) : \ + (assert(!"should not get here"), 0)) +#endif + +#endif /* U_ATOMIC_H */