On 11/12/2014 10:18 PM, Iain Sandoe wrote:
> * config/darwin/host-config.h New.
> * config/darwin/lock.c New.
> * configure.tgt (DEFAULT_X86_CPU): New, (target): New entry for darwin.
Looks pretty good.
> +# ifndef USE_ATOMIC
> +# define USE_ATOMIC 1
> +# endif
Why would USE_ATOMIC be defined previously?
> +inline static void LockUnlock(uint32_t *l) {
> + __atomic_store_4((_Atomic(uint32_t)*)l, 0, __ATOMIC_RELEASE);
> +}
Gnu coding style, please. All through the file here.
> +# define LOCK_SIZE sizeof(uint32_t)
> +# define NLOCKS (PAGE_SIZE / LOCK_SIZE)
> +static uint32_t locks[NLOCKS];
Um, surely not LOCK_SIZE, but CACHELINE_SIZE. It's the granularity of the
target region that's at issue, not the size of the lock itself.
> +#if USE_ATOMIC
> + LockLock (&locks[addr_hash (ptr, 1)]);
> +#else
> + OSSpinLockLock(&locks[addr_hash (ptr, 1)]);
> +#endif
Better to #define LockLock OSSpinLockLock within the area above, so as to
avoid the ifdefs here.
r~
