Re: [Qemu-devel] [PATCH v3 05/11] qemu-thread: add simple test-and-set spinlock

[Richard Henderson <rth@twiddle.net>]

On 04/20/2016 11:11 AM, Emilio G. Cota wrote:
> On Wed, Apr 20, 2016 at 10:55:45 -0700, Richard Henderson wrote:
>> On 04/20/2016 10:17 AM, Emilio G. Cota wrote:
>>> I've tried to find a GCC intrinsic for test-and-set, and I've only found
>>> lock_test_and_set, which is what we use for atomic_xchg (except on ppc)
>>> because it really is an atomic exchange:
>>>   "This builtin, as described by Intel, is not a traditional test-and-set
>>>    operation, but rather an atomic exchange operation."
>>>   https://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/Atomic-Builtins.html
>>
>> Please read the entire documentation, not just the first sentence.
>
> I did read it and I'm aware of the limitations of using xchg.
>
> My comment was related to this:
>
>> [...] do note that there are compiler primitives for test-and-set that
>> (can be) simpler for a cpu to implement than xchg.
>
> What compiler (I assume gcc) primitives are these? I couldn't find them.

__sync_lock_test_and_set and __atomic_test_and_set.

Both expand to ldstub on sparcv7, tas on coldfire, tas.b on sh.
None of these are xchg operations.

I had forgotten that there wasn't a __sync_exchange builtin, so 
__sync_lock_test_and_set plays double-duty as both xchg and test-and-set.

But when __atomic builtins are available, __atomic_exchange does not fall back; 
you must use __atomic_test_and_set for less capable hosts.

But of course none of this is really relevant for normal hosts.


r~