Page 1 of 1

[REQ] wait_for_slave_frames_on_stop tweak

Posted: Fri Dec 17, 2010 1:30 am
by Zom-B
A small request about this feature:

At the moment after reaching halt condition master stops rendering and grabs for last slave frame rendered.

I would like to see the master keeping on rendering while waiting for the frames.
This sounds maybe a little unimportant, but not if you take my out-of-production example.

4 Computers rendering a animation with 300spp as target.
The transfered igi file is 170MB big and takes from 25-30 seconds to transfer via Ethernet to 2min for a MacBook that has WLAN only (slow WLAN network around here).
After all the master needs to wait about 3min to get all the Slave frames together.... why not keep him rendering during this time???

Another approach to optimize this feature would be to allow sending data first for the slowest slaves, descending to the fastest slaves! In my example the 2min transfer would allow the other 2 slaves to render for 2min longer until sending their stuff. other way around they would idle 2min until the slow mac slave finished uploading.

Re: [REQ] wait_for_slave_frames_on_stop tweak

Posted: Sat Dec 18, 2010 11:32 am
by Soup
I would think the master would want to stop rendering when 'stop' is pressed, so they can gain control over their computer imeaditely. But as for the slaves, I suppose it wouldn't hurt to let them take turns uploading their final frames, and render in the mean time.
3 minutes seems insignificant, but I can imagine on large render farms it could really add up.

Re: [REQ] wait_for_slave_frames_on_stop tweak

Posted: Sun Dec 19, 2010 2:42 am
by Zom-B
regarding render farms the biggest issue is the merging of IGI files, that takes quite long!
This approach would help a lot!
Merging and transferring so much MB seems to be quite suboptimal.

Re: [REQ] wait_for_slave_frames_on_stop tweak

Posted: Tue Oct 02, 2012 9:15 am
by pixie
http://ge.tt/10urxhO/v/0?c

As ssen on the table I have strange numbers arising from setting it to true.
The latest collumn on the right shows the deviation between the halt value and the effective stop time. I've set a batch where I would stamp time on each time the render starts, the only difference from the setting itself.