+ else // new logic:
exactly 'count' threads on all NUMAs
+ {
+ threadsPerPool[numNumaNodes] = X265_MIN(count,
numNumaNodes * MAX_POOL_THREADS);
+ nodeMaskPerPool[numNumaNodes] = ((uint64_t)-1 >>
(64 - numNumaNodes));
+ }
Using numNumaNodes * MAX_POOL_THREADS as the upper-bound on the # threads
per numa node isn't the cleanest. The right thing to do here is to compute
the min across all cpusPerNode[i] and use that as the second param in the
call to X265_MIN() so that in the case the user gives count > # cpus in any
NUMA node, we are still ok.
Also, can you please include the patch as inline in future so that it
becomes easier to review?
Pradeep.
On Mon, May 16, 2016 at 12:37 AM, Mateusz <[email protected]> wrote:
>
>
> _______________________________________________
> x265-devel mailing list
> [email protected]
> https://mailman.videolan.org/listinfo/x265-devel
>
>
_______________________________________________
x265-devel mailing list
[email protected]
https://mailman.videolan.org/listinfo/x265-devel
