View previous topic :: View next topic |
Author |
Message |
cgbeige
Posts: 52
Joined: 20 May 2009
|
Posted: Tue Feb 16, 2010 4:21 pm Post subject: multithreading or OpenCL in the cards for UV Layout? |
|
|
I'm dealing with a really large mesh and have to work with this resolution so UV Layout is really really slow when doing the unfold. I noticed it's using one thread and I have a 16 thread machine (8-core Nehalem Xeon). Is there a chance that you'll be able to multithread this operation or is it the type of thing that can be done efficiently on the GPU?
Either way, I think UVL needs to be able to deal reasonably with large meshes since Maya uses all my cores for its relax/unfold operations. |
|
Back to top |
|
 |
cgbeige
Posts: 52
Joined: 20 May 2009
|
Posted: Tue Feb 16, 2010 4:34 pm Post subject: |
|
|
oops - my mistake. Maya's UV relax is multithreaded, but not the unfold. |
|
Back to top |
|
 |
headus Site Admin

Posts: 2902
Joined: 24 Mar 2005
Location: Perth, Australia
|
Posted: Tue Feb 16, 2010 6:15 pm Post subject: |
|
|
"a chance that you'll be able to multithread this operation"
There's definitely a chance on a couple of fronts.
I did start on something a while back to speed up the flattening of dense shells, but couldn't get it to work property at the time. Its a good idea though, so I was hoping to revisit that at some point to fix the problems.This would speed it up for everyone.
And I have been playing around with multithreading another of our products, so have the knowledge now to apply that to UVLayout flattening. Your chances will be really good if you happen to be using the Linux version of UVLayout, because that's where I've been doing all my playing around; I don't know yet if what I've been doing applies to Windows or OS X.
Phil |
|
Back to top |
|
 |
headus Site Admin

Posts: 2902
Joined: 24 Mar 2005
Location: Perth, Australia
|
Posted: Tue Feb 16, 2010 6:31 pm Post subject: |
|
|
Oh, I forgot one thing; if you haven't already, open up the Preferences and untick the "Slow Smooth Optimize" option. This is from the User Guide ...
Slow Smooth Optimize
By default this is on, and it means that when the Optimize is run, shells are redrawn at every flattening optimization step. This makes it look nice and smooth when its running, but the continuous redrawing does use a significant chunk of available CPU. By turning this off, the optimize will only redraw shells twice a second; even though it looks slower, the CPU is now spending more time flattening, and so it'll get to the optimal result quicker.
Phil |
|
Back to top |
|
 |
cgbeige
Posts: 52
Joined: 20 May 2009
|
Posted: Tue Feb 16, 2010 7:15 pm Post subject: |
|
|
Thanks for the tip.
What would it take for you to change your lead platform to OS X  |
|
Back to top |
|
 |
headus Site Admin

Posts: 2902
Joined: 24 Mar 2005
Location: Perth, Australia
|
Posted: Tue Feb 16, 2010 7:37 pm Post subject: |
|
|
Sorry, no chance, I like KDE too much :)
But actually, OS X uses gcc too, so everything I'm doing under Linux as far as the multi-threading goes may work fine under OS X. I'll have a look once I've done my homework (i.e. fixing that display bug).
Phil |
|
Back to top |
|
 |
|