Re: Multiprocessing and y4mdenoise

>--- On Mon, 6/15/09, Scott Smith <[email protected]> wrote:
>On 5/26/09, Steven Boswell II <[email protected]> wrote:
>Mutexes only hop into the kernel if they have to sleep.

In my experience, that varies heavily depending on the operating system.
I can think of at least three platforms I've programmed on recently whose
mutexes were slow to the point of having to rewrite most of the code to
synchronize with atomic operations.

Besides, I solved the performance problem that was motivating the
"pipelined" approach to y4mdenoise's multiprocessing.  My only other idea
involves separate threads processing "slices" of the frame, which may or
may not lead to visible artifacts along the boundaries between slices.
That method will use atomic operations for the necessary coordination
between threads (for moved regions that cross slice boundaries).

Steven Boswell

------------------------------------------------------------------------------
Are you an open source citizen? Join us for the Open Source Bridge conference!
Portland, OR, June 17-19. Two days of sessions, one day of unconference: $250.
Need another reason to go? 24-hour hacker lounge. Register today!
http://ad.doubleclick.net/clk;215844324;13503038;v?http://opensourcebridge.org

_______________________________________________
Mjpeg-developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mjpeg-developer