i am running a patch with videoin and contour, two modules with linear spreads, random spreads, gaussian spreads, one attractor, oscillator, some quads with and without texture, 4 renderer nodes, etc. i would say not to complicated. finally i send a lot of data (not bundled, because sending bundled interrupts the processing) via osc and udp to max/msp which is working fine. but after 10-15 minutes i run out of memory (Exception EOutOfMemory in Modul vvvv.exe bei 00004863. Zu wenig Arbeitsspeicher)!
i already put one undo node in one of the subpatches and set its last pin to 0 but there is still the same error. how exactly should i use the undo node? put in the main patch or in every subpatch, connect with something? can i fix this problem with the undo node? thanks a lot for help and information! best, armin
i don´t think your problem is related to “the undo issue”. Watch the vvvv process in windows taskmanager. Reduce your patch node by node, if the increase in memory usage stops after you´ve deleted (a) certain node(s) you have most certainly found the cause. Then try to make a patch with as few nodes as possible which shows the same behavior. Post this patch and your findings here.
Try that :
Acces to your virtual memory
Control Panel -> System -> Advanced -> Performances -> Settings
Tab n°2 Advanced and Virtual Memory Modify.
You will see all your hard drive.
You need to choose first another disk than system C: ( best perf)
Cancel the virtual memory on C:
Choose another disk and put a static value equal to twice your RAM for both values
Ex : if 1024 Ram put 2048
if 2048 RAm put 4096
If you’ve got already 4Go only put 4096
Restart after done
@bjoern: thanks for that. i checked and found that the OSCEncoder node increases the memory when i send data not as a bundle (Spread as Bundle pin to 0). but when i activate the pin and send data as bundle the performance of the renderer goes down. often the output of the renderer just stucks for a sec when a huge bundle got sent. you can even see it when you open the helpfile of the OSCEncoder node and set both bundle pins of the OSC nodes to 0. i think what also counts is that i am sending lots of integers (4 strings each 192 ints) and some floats…but i am not sure about this.
here is my osc patch:
osc-udp-sender001.v4p (19.8 kB)
thanks for this. the memoryleak was introduced with the rewrite of the oscnodes for beta19. that problem should not exist with beta18. anyway i have now fixed it for beta>19.1
- but when i activate the pin and send data as bundle the performance of the renderer goes down.
i cannot see that problem in your patch. can you provide a demo for that behaviour too?
note also that the status output may write out OK, while it is actually not. fixed that also for beta>19.1
thats nice to hear. when do you think will you put the beta>19.1 for official download? i tried the osc in beta18, it works well. but there is another problem in beta18 with the videoin node. just tried the help file and connected a unibrain fire-i cam (no drivers from unibrain installed, just the standard 1394 Desktop Video Camera). in beta18 no image is available, the Video Format pin shows nil and the program reacts crazy slow. in beta19.1 its o.k. then i tested the dragonfly2 cam (with and without driver from pointgreyresearch.com). it is not working in beta18 and beta19.1. i guess i should post this more detailed in another section of this forum…
everything works fine in beta18! it was my fault. i just installed the wrong drivers.
when i finished the patches i will post them here to show u what i mean with the stuck of rendering while bundles are sent via osc udp. anyway its fine in beta18 now.
thanks for the quick answers.