[x265] [PATCH] [PATCH] threadpool: use all available threads on each NUMA node with --pools=NULL option too
Mahesh Pittala
mahesh at multicorewareinc.com
Fri Dec 18 05:26:06 CET 2015
Thanks, it does better. both strncasecmp & stricmp compares strings without
case sensitivity so I'm taking stricmp.
On Fri, Dec 18, 2015 at 1:02 AM, David Wolstencroft <lordrpi at gmail.com>
wrote:
> strncasecmp?
>
> On Thu, Dec 17, 2015 at 9:59 AM, <mahesh at multicorewareinc.com> wrote:
>
>> # HG changeset patch
>> # User Mahesh Pittala <mahesh at multicorewareinc.com>
>> # Date 1450375170 -19800
>> # Thu Dec 17 23:29:30 2015 +0530
>> # Node ID d3451b996e26740e8a2e2f4f4d9b6d91cfa52958
>> # Parent 9c797a5915814a74663f4f1f4edf8a9fd17d5c42
>> [PATCH] threadpool: use all available threads on each NUMA node with
>> --pools=NULL option too
>>
>>
>> from x265.readthedocs, --pools = NULL, "", "*" are doing same
>> functionality.
>> --pools="" & --pools="*" are using all threads but when we pass
>> --pools=NULL, disabled
>> all threads pools but supposed to enable.
>>
>> diff -r 9c797a591581 -r d3451b996e26 source/common/threadpool.cpp
>> --- a/source/common/threadpool.cpp Thu Dec 10 21:37:01 2015 +0530
>> +++ b/source/common/threadpool.cpp Thu Dec 17 23:29:30 2015 +0530
>> @@ -268,6 +268,10 @@
>> if (p->numaPools && *p->numaPools)
>> {
>> const char *nodeStr = p->numaPools;
>> + char nodeStrupper[100];
>> + for(int i = 0; nodeStr[i]; i++)
>> + nodeStrupper[i] = toupper(nodeStr[i]);
>> + nodeStrupper[strlen(nodeStr)] = '\0';
>> for (int i = 0; i < numNumaNodes; i++)
>> {
>> if (!*nodeStr)
>> @@ -277,7 +281,7 @@
>> }
>> else if (*nodeStr == '-')
>> threadsPerPool[i] = 0;
>> - else if (*nodeStr == '*')
>> + else if (*nodeStr == '*' || !strcmp(nodeStrupper, "NULL"))
>> {
>> for (int j = i; j < numNumaNodes; j++)
>> {
>> _______________________________________________
>> x265-devel mailing list
>> x265-devel at videolan.org
>> https://mailman.videolan.org/listinfo/x265-devel
>>
>
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20151218/1032d062/attachment-0001.html>
More information about the x265-devel
mailing list