From: A. C. W. <acr...@gm...> - 2008-02-18 16:40:09
Attachments:
motion-buffer.patch
|
I have found that in my application of motion, the default buffer sizes for file output are limiting the framerate I am able to use to about 96 frames per minute for a 320 x 240 ppm in an embedded device using uclibc. I have a patch which uses a 1MB buffer which allows for a significantly higher frame-rate (in this case, 268 frames per minute). It might be worthwhile to add code to set the buffer size and the maximum number of buffers to use from the configuration file, but currently they are hard coded. The code currently doesn't free the buffers, as it is usually more efficient to re-use them instead. |
From: Jeffrey G. <jg...@kr...> - 2008-02-18 21:45:45
|
To whom it may concern: I applied this patch and have to say that my unscientific observations shows that the frame rates of my movies are improved. Much smoother with fewer skipped frames. One note about the patch though is that there was a section of the patched motion.c at line 2503 through 2507 where my compile barfed because of a ill-formed #if statement or ill-commented-out section of code -- I couldn't figure out which, so I commented-out that section of code using /* ... */ instead of the #. My environment is: Fedora Core 6 Latest FFMPEG 3.2.10 motion Thanks, Jeff G. On 2/18/08, A. Craig West <acr...@gm...> wrote: > I have found that in my application of motion, the default buffer > sizes for file output are limiting the framerate I am able to use to > about 96 frames per minute for a 320 x 240 ppm in an embedded device > using uclibc. I have a patch which uses a 1MB buffer which allows for > a significantly higher frame-rate (in this case, 268 frames per > minute). It might be worthwhile to add code to set the buffer size and > the maximum number of buffers to use from the configuration file, but > currently they are hard coded. The code currently doesn't free the > buffers, as it is usually more efficient to re-use them instead. > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Motion-user mailing list > Mot...@li... > https://lists.sourceforge.net/lists/listinfo/motion-user > http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome > > > -- Jeffrey A. Groves jg...@kr... Web Site: http://www.krenim.org/ |
From: A. C. W. <acr...@gm...> - 2008-02-19 00:08:50
Attachments:
motion-buffer.patch
|
Sorry about that, I made a last minute edit to the patch, and of course introduced an error... I believe this should fix it -Craig On Feb 18, 2008 4:45 PM, Jeffrey Groves <jg...@kr...> wrote: > To whom it may concern: > > I applied this patch and have to say that my unscientific observations > shows that the frame rates of my movies are improved. Much smoother > with fewer skipped frames. > > One note about the patch though is that there was a section of the > patched motion.c at line 2503 through 2507 where my compile barfed > because of a ill-formed #if statement or ill-commented-out section of > code -- I couldn't figure out which, so I commented-out that section > of code using /* ... */ instead of the #. > > My environment is: > Fedora Core 6 > Latest FFMPEG > 3.2.10 motion > > Thanks, > > Jeff G. > > > On 2/18/08, A. Craig West <acr...@gm...> wrote: > > I have found that in my application of motion, the default buffer > > sizes for file output are limiting the framerate I am able to use to > > about 96 frames per minute for a 320 x 240 ppm in an embedded device > > using uclibc. I have a patch which uses a 1MB buffer which allows for > > a significantly higher frame-rate (in this case, 268 frames per > > minute). It might be worthwhile to add code to set the buffer size and > > the maximum number of buffers to use from the configuration file, but > > currently they are hard coded. The code currently doesn't free the > > buffers, as it is usually more efficient to re-use them instead. > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Microsoft > > Defy all challenges. Microsoft(R) Visual Studio 2008. > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > > _______________________________________________ > > Motion-user mailing list > > Mot...@li... > > https://lists.sourceforge.net/lists/listinfo/motion-user > > http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome > > > > > > > > > -- > Jeffrey A. Groves jg...@kr... > Web Site: http://www.krenim.org/ > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Motion-user mailing list > Mot...@li... > https://lists.sourceforge.net/lists/listinfo/motion-user > http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome > |
From: A. C. W. <acr...@gm...> - 2008-02-28 13:17:05
|
I've had one user try my patch for increasing frame rates, and apparently he noticed significant improvements as well. The patch doesn't change the writing in the ffmpeg code, however. I am looking at making similar changes there, but was wondering if anybody knows why that code currently uses non-buiffered writes? -Craig |
From: Angel C. <ac...@te...> - 2008-03-06 13:25:15
|
Hi Craig, El jue, 28-02-2008 a las 08:16 -0500, A. Craig West escribió: > I've had one user try my patch for increasing frame rates, and > apparently he noticed significant improvements as well. The patch > doesn't change the writing in the ffmpeg code, however. I am looking > at making similar changes there, but was wondering if anybody knows > why that code currently uses non-buiffered writes? I've looked your patch and it improves performance , but still need more work to be included. Could you submit it to : http://www.lavrsen.dk/twiki/bin/view/Motion/MotionPatches ffmpeg code needs to be reviewed , i did some work on it but still many others things to improve. I'm working to add buffered writes using URLProtocol. > -Craig > Cheers, -- Angel Carpintero ack ( at ) telefonica ( dot ) net Key fingerprint = 3FD3 9C90 149E 7824 CECD 6BCF AC2C CA61 6EF1 B90D "No basta saber, hay que aplicar lo que se sabe; no basta querer hacerlas cosas, hay que hacerlas". "Knowing is not enough; we must apply. Willing is not enough; we must do" Johann Wolfgang von Goethe |
From: Roman G. <hac...@gm...> - 2008-03-30 13:38:13
|
On my unscientific test, it does too improve performance, please add this to motion :) On Thu, Mar 6, 2008 at 2:25 PM, Angel Carpintero <ac...@te...> wrote: > > Hi Craig, > > El jue, 28-02-2008 a las 08:16 -0500, A. Craig West escribió: > > > I've had one user try my patch for increasing frame rates, and > > apparently he noticed significant improvements as well. The patch > > doesn't change the writing in the ffmpeg code, however. I am looking > > at making similar changes there, but was wondering if anybody knows > > why that code currently uses non-buiffered writes? > > I've looked your patch and it improves performance , but still need more > work to be included. Could you submit it to : > > http://www.lavrsen.dk/twiki/bin/view/Motion/MotionPatches > > ffmpeg code needs to be reviewed , i did some work on it but still many > others things to improve. I'm working to add buffered writes using > URLProtocol. > > > -Craig > > > > Cheers, > -- > Angel Carpintero > ack ( at ) telefonica ( dot ) net > > Key fingerprint = 3FD3 9C90 149E 7824 CECD 6BCF AC2C CA61 6EF1 B90D > > "No basta saber, hay que aplicar lo que se sabe; > no basta querer hacerlas cosas, hay que hacerlas". > > "Knowing is not enough; we must apply. > Willing is not enough; we must do" > > Johann Wolfgang von Goethe > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Motion-user mailing list > Mot...@li... > https://lists.sourceforge.net/lists/listinfo/motion-user > http://www.lavrsen.dk/twiki/bin/view/Motion/WebHome > > |
From: A. C. W. <acr...@gm...> - 2009-05-18 18:02:28
|
2008/3/6 Angel Carpintero <ac...@te...>: > > Hi Craig, > I've looked your patch and it improves performance , but still need more > work to be included. Could you submit it to : > > http://www.lavrsen.dk/twiki/bin/view/Motion/MotionPatches It took me rather longer than anticipated, but I have now submitted my write buffers patch. This version should be an improvement over the previous one, in that I now include the ffmpeg code, which was arguably in the most need of it. I am getting significantly better frame rates using this code on an embedded device, so I think it is worthwhile... I am sure it still needs some work, so comments are of course appreciated... http://www.lavrsen.dk/foswiki/bin/view/Motion/MotionWriteBuffersPatch -Craig |