why start_thread consumes so much memory ?
Locked
 
|
Hi All,
I am trying to locate which part of our application that causes a memory spike. Considering that our application is composed of about two hundred elements which are scattered over 70 threads, I use Massif, one of Valgrind tools to check heap memory profile during the execution of our application. I list the detailed information of snapshot 48 of Massif output. As you can see that starting "this" thread consumes most of the memory. Also I found that this thread is not such thread that is created by "queue" element, because I can see that "queue" thread are created after the memory spike. It seems like it is a main thread that handles preliminary things. Can anyone give me some explanation about it. Any help is appreciated. Thanks! GB 1.079^ # | :@:::@# | :@:::::@:::@:::@# | ::::::::::::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | :::::::::::: :: ::@:::::@:::@:::@# | ::@:::::::::::::::::::::@@@@::::::::::::::::::::: :: ::@:::::@:::@:::@# 0 +----------------------------------------------------------------------->Gi 0 619.3 Number of snapshots: 52 Detailed snapshots: [3, 9, 32, 38, 43, 47, 48 (peak), 51] -------------------------------------------------------------------------------- n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B) -------------------------------------------------------------------------------- 48 661,339,638,719 1,158,568,168 1,153,638,753 4,929,415 0 99.57% (1,153,638,753B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc. ->71.34% (826,503,872B) 0x3054430468: gst_buffer_try_new_and_alloc (in /usr/lib64/libgstreamer-0.10.so.0.29.0) | ->71.34% (826,503,872B) 0x30544593E4: ??? (in /usr/lib64/libgstreamer-0.10.so.0.29.0) | ->67.01% (776,396,992B) 0x3054C2E34A: ??? (in /usr/lib64/libgstbase-0.10.so.0.29.0) | | ->67.01% (776,396,992B) 0x3054C2EA6A: ??? (in /usr/lib64/libgstbase-0.10.so.0.29.0) | | ->67.01% (776,396,992B) 0x305445DEC8: gst_pad_push (in /usr/lib64/libgstreamer-0.10.so.0.29.0) | | | ->67.01% (776,396,800B) 0x2851A728: ??? (in /usr/lib64/gstreamer-0.10/libgstcoreelements.so) | | | | ->67.01% (776,396,800B) 0x3054484362: ??? (in /usr/lib64/libgstreamer-0.10.so.0.29.0) | | | | ->67.01% (776,396,800B) 0x303526C746: ??? (in /lib64/libglib-2.0.so.0.3000.2) | | | | ->67.01% (776,396,800B) 0x303526A224: ??? (in /lib64/libglib-2.0.so.0.3000.2) | | | | ->67.01% (776,396,800B) 0x3033607D8E: start_thread (in /lib64/libpthread-2.14.90.so) | | | | | | | ->00.00% (192B) in 1+ places, all below ms_print's threshold (01.00%) | | | | | ->00.00% (0B) in 1+ places, all below ms_print's threshold (01.00%) | | -- Qi Chu, Ph.D student Australian International Gravitational Research Centre School of physics The University of Western Australia M013, 35 Stirling Hwy Crawley, WA 6009 Australia Email: [hidden email] _______________________________________________ gstreamer-devel mailing list [hidden email] http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
Re: why start_thread consumes so much memory ?
|
|
To answer the question in the subject, start_thread() is the base function for all your 70 threads. So the result from start_thread() being the main source of allocation is quite expected. It only means that most allocation are being done in threads.
Now, to get more useful result from your profiling, I would recommend you to run it on a Gstreamer build that has debugging symbols. This will replace the "??? (in /usr/lib64/libgstreamer-0.10.so.0.29.0)" in your backtraces with specific functions. best regards, Nicolas _______________________________________________ gstreamer-devel mailing list [hidden email] http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Nicolas,
Not sure why I missed your email until today. Thank you very much ! I will try that. Cheers! Qi
On Wed, Sep 26, 2012 at 9:50 PM, Nicolas Dufresne <[hidden email]> wrote:
Qi Chu, Ph.D student Australian International Gravitational Research Centre School of physics The University of Western Australia M013, 35 Stirling Hwy Crawley, WA 6009 Australia Email: [hidden email] _______________________________________________ gstreamer-devel mailing list [hidden email] http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
«
Return to GStreamer-devel
|
1 view|%1 views
| Free forum by Nabble | Edit this page |