Считывание c IP-камер на 16 ядерном процессоре (Bug #3324)
Description
Здравствуйте!
Постараюсь описать ситуацию:
Используем последнюю версию OpenCV для записи с IP камер. Процессор Intel Xeon (R) CPU E5-2690 (16 ядерный) Windows x64.
Windows установлена на виртуальную машину, поэтому есть возможность варьировать используемое число ядер.
Записываем с IP-камер в формате Full HD (1920х1080).
Итак, столкнулся с такими трудностями:
Выделяем, например 14 или 16 ядер, начинаем считывать и записывать одновременно в разных потоках с 10 камер, в итоге процедура считывания и соответственно записи на жесткий диск внезапно обрывается. Ошибки никакие отловить не удается.
Выделяем, например 8 или 10 ядер, начинаем записывать одновременно в разных потоках с 10 камер процедуры считывания и соответственно записи на жесткий диск идет нормально никаких проблем нет.
Отмечу, что используем библиотеки OpenCV из папки x86.
Подскажите, с чем могут быть связаны подобные баги и как их возможно исправить?
Associated revisions
Merge pull request #3324 from akarsakov:fix_lpsolver_tests
History
Updated by Maria Dimashova over 11 years ago
Alexey, next time please write in English. Even if almost all OpenCV team speak Russian, OpenCV users also follow issue tracker.
Alexander, could you please look at this?
- Assignee set to Alexander Smorkalov
Updated by Maria Dimashova over 11 years ago
- Category set to highgui-camera
Updated by Алексей Алексеевич Хрюстов over 11 years ago
Maria Dimashova wrote:
Alexey, next time please write in English. Even if almost all OpenCV team speak Russian, OpenCV users also follow issue tracker.
Okay. Thank you
Translate:
Hello !
I will try to describe the situation :
We use the latest version of OpenCV for recording from IP cameras. Processor Intel Xeon (R) CPU E5- 2690 ( 16 core ) Windows x64.
Windows is installed on a virtual machine , so We can vary the number of cores for using .
We write with IP-cameras in the format Full HD ( 1920x1080) .
So, We have such difficulties :
1) For example we use 14 or 16 cores , we begin to read and write at the same time in different threads with 10 cameras , as a result the procedure of reading and writing suddenly breaks off .
But if we begin to read and write at the same time in different threads with 6 cameras, then it work excellent.
Mistakes can not catch any .
2) For example we use 8 or 10 cores , we begin to read and write at the same time in different threads with 10 cameras , as a result the procedure of reading and writing works excellent .
But We need to use 16 cores.
We use OpenCV library in the folder x86.
Please help me, How Can I fix trouble?
Updated by Dmitry Retinskiy over 11 years ago
Alexey, could you add an error log and a code snippet which has been used to reproduce the described case?
Thanks.
Updated by Dinar Ahmatnurov over 11 years ago
Hi, Alexey!
Any updates?
- Status changed from New to Open
Updated by Vadim Pisarevsky almost 10 years ago
well, it's unlikely we will work on this and considering the description (16 threads - problem; 8 threads - ok) and considering "age" of the bug let's close it
- Status changed from Open to Cancelled
- Affected version changed from 2.4.6 (latest release) to branch '2.4' (2.4-dev)