On 01/04/2017 12:22 PM, Laurent Vivier wrote:
Yes, it's a good idea, but is it really expensive to always enable the
parallel_cpus flag?
It is somewhat expensive when the host does support the atomics being used; it
is very expensive if the host does not. Of course the common case is x86_64
Le 04/01/2017 à 19:39, Alex Bennée a écrit :
>
> Laurent Vivier writes:
>
>> We always need real atomics, as we can have shared memory between
>> processes.
>>
>> A good test case is the example from futex(2), futex_demo.c:
>>
>> the use case is
>>
>> mmap(...);
>> fork();
>>
>> Pare
Le 04/01/2017 à 20:35, Richard Henderson a écrit :
> On 12/28/2016 09:37 AM, Laurent Vivier wrote:
>> the use case is
>>
>> mmap(...);
>> fork();
>
> While true, we can notice that mmap contains MAP_SHARED, and trigger it
> then. Similarly for shmat.
Yes, it's a good idea, but is it reall
On 12/28/2016 09:37 AM, Laurent Vivier wrote:
the use case is
mmap(...);
fork();
While true, we can notice that mmap contains MAP_SHARED, and trigger it then.
Similarly for shmat.
r~
Laurent Vivier writes:
> We always need real atomics, as we can have shared memory between
> processes.
>
> A good test case is the example from futex(2), futex_demo.c:
>
> the use case is
>
> mmap(...);
> fork();
>
> Parent and Child:
>
> while(...)
> __sync_bool_compare
We always need real atomics, as we can have shared memory between
processes.
A good test case is the example from futex(2), futex_demo.c:
the use case is
mmap(...);
fork();
Parent and Child:
while(...)
__sync_bool_compare_and_swap(...)
...
futex(...)
In
