Re: Re-4: [MBS] Downsize large picture with yielding w/low memory usage?
Am 25.11.2011 um 00:50 schrieb Kev: > Hi Christian. > > How did you check the results? Was it visually or did you compare the binary > output from the two versions? I compared the PNGs from both formats and didn't see a difference. > Did you notice any performance improvements when you switched to integers? Please test and report results soon. Greetings Christian -- See you in Orlando, Florida for Real World 2012 More details and registration here: http://www.realsoftware.com/community/realworld.php ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re: Re-4: [MBS] Downsize large picture with yielding w/low memory usage?
Hi Christian. How did you check the results? Was it visually or did you compare the binary output from the two versions? Did you notice any performance improvements when you switched to integers? Regards, Kev. - Original Message - From: Christian Schmitz To: MBS Real Studio Plugin List Cc: Sent: Thursday, 24 November 2011, 12:49 Subject: Re: Re-4: [MBS] Downsize large picture with yielding w/low memory usage? Am 24.11.2011 um 12:55 schrieb MGE Info: >> I did some comparisons and it looks like it doesn't have a visible effect. >> But please test when the next pr is available. > Wouldnt it be moreexisting code friendly, to add a Parameter using old and > new behavior? > In my case there would be to much to test. That would double my code base here. Really, the temp picture needs to store the pixel values. 0 to 255 per channel. We used to put them in doubles. So a value of 0.5 would be preserved in the temporary picture. But I made test here with scaling up and down with old vs. new plugin and I didn't see a difference in the results. Please test. Greetings Christian -- See you in Orlando, Florida for Real World 2012 More details and registration here: http://www.realsoftware.com/community/realworld.php ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re: [MBS] Downsize large picture with yielding w/low memory usage?
Am 24.11.2011 um 18:31 schrieb Michael Diehr: > Would it be possible to make the yield ticks value into a yieldMilliseconds > value instead? Or, to be more backwards compatible, how about the yieldTicks > value could be a double? Better use the MT variant in a thread. > Also - the MT - multithreaded version -- does it always use the current CPU > thread? The MT functions perform the actual work on a second worker thread. The Real Studio thread simply waits and gives away CPU time on other threads until the work thread is done. Greetings Christian -- See you in Orlando, Florida for Real World 2012 More details and registration here: http://www.realsoftware.com/community/realworld.php ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re: [MBS] Downsize large picture with yielding w/low memory usage?
On Nov 23, 2011, at 7:59 AM, Christian Schmitz wrote: > > Am 22.11.2011 um 17:30 schrieb Michael Diehr: > >> * does not require tons of memory > > I'll reduce the Scaling method memory need. > > For some reason we used doubles there and now I simply use UInt8 which > reduces memory requirements by more than 80%. That would help, thank you. Would it be possible to make the yield ticks value into a yieldMilliseconds value instead? Or, to be more backwards compatible, how about the yieldTicks value could be a double? Internally, I wonder if there is any overhad difference between calling Ticks() vs. Microseconds() a lot? Also - the MT - multithreaded version -- does it always use the current CPU thread? It would be interesting to have a MT version which only used the other CPUs, leaving the current thread's CPU idle (to give time to the rest of the foreground threads). ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re: Re-4: [MBS] Downsize large picture with yielding w/low memory usage?
Am 24.11.2011 um 12:55 schrieb MGE Info: >> I did some comparisons and it looks like it doesn't have a visible effect. >> But please test when the next pr is available. > Wouldnt it be moreexisting code friendly, to add a Parameter using old and > new behavior? > In my case there would be to much to test. That would double my code base here. Really, the temp picture needs to store the pixel values. 0 to 255 per channel. We used to put them in doubles. So a value of 0.5 would be preserved in the temporary picture. But I made test here with scaling up and down with old vs. new plugin and I didn't see a difference in the results. Please test. Greetings Christian -- See you in Orlando, Florida for Real World 2012 More details and registration here: http://www.realsoftware.com/community/realworld.php ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re-4: [MBS] Downsize large picture with yielding w/low memory usage?
> I did some comparisons and it looks like it doesn't have a visible effect. > But please test when the next pr is available. Wouldnt it be moreexisting code friendly, to add a Parameter using old and new behavior? In my case there would be to much to test. Thomas ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re: Re-2: [MBS] Downsize large picture with yielding w/low memory usage?
Am 24.11.2011 um 12:25 schrieb MGE Info: > Christian, > >> For some reason we used doubles there and now I simply use UInt8 which >> reduces memory requirements by more than 80%. > For all scaling filters? > Doesnt this break any excisting RB-Code or scaling quality? I did some comparisons and it looks like it doesn't have a visible effect. But please test when the next pr is available. Greetings Christian -- See you in Orlando, Florida for Real World 2012 More details and registration here: http://www.realsoftware.com/community/realworld.php ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
Re-2: [MBS] Downsize large picture with yielding w/low memory usage?
Christian, > For some reason we used doubles there and now I simply use UInt8 which > reduces memory requirements by more than 80%. For all scaling filters? Doesnt this break any excisting RB-Code or scaling quality? Thomas Original Message Subject: Re: [MBS] Downsize large picture with yielding w/low memory usage? (23-Nov-2011 16:59) From:Christian Schmitz To: mbsplugins@monkeybreadsoftware.info > > Am 22.11.2011 um 17:30 schrieb Michael Diehr: > > > * does not require tons of memory > > I'll reduce the Scaling method memory need. > > For some reason we used doubles there and now I simply use UInt8 which > reduces memory requirements by more than 80%. > > Greetings > Christian > > -- > See you in Orlando, Florida for Real World 2012 > > More details and registration here: > http://www.realsoftware.com/community/realworld.php > > > > ___ > Mbsplugins_monkeybreadsoftware.info mailing list > mbsplugins@monkeybreadsoftware.info > https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware. > info > ___ Mbsplugins_monkeybreadsoftware.info mailing list mbsplugins@monkeybreadsoftware.info https://ml01.ispgateway.de/mailman/listinfo/mbsplugins_monkeybreadsoftware.info
