PDA

View Full Version : Clearing up the Prepass - i Just dont get it


Ric535
09-14-2005, 03:04 PM
When one does a radiosity render first there is the prepass and then it will render the image fully

What i have found out is that after the prepass is done the render will sometimes take a while just clearing up the prepass

I have found that if you stop the render after the prepass a .GI file will be written to the illum folder of your projetc and then you can rerender (provided yuo have checked recompute first time and save rad solution in the GI settings) and it will render the image much much quicker as it doesnt have to clear up the samples

The test i did was pretty crude but rendering all in one (prepass continuing to render) took 6 mins 20 and rendering just the prepass stopping and then rendering the saved solution took 4mins 50 in total - now on heavy renderings this would save a lot of time

I am wondering though maybe my test was too low quality as it was full of artifacts (too scared to try it on a full blown gi render incase it backfires), maybe this clearing up of the prepass is necessay to elliminate some artifacts i dont know, if not then i have found a pretty good if pretty annoying (to have to watch youtr render to see when one stage finishes to start the next) way of speeding up a GI render

What do you think

Thanks

Ric535
09-14-2005, 08:03 PM
read me somebody :)

STRAT
09-15-2005, 07:30 AM
read me somebody :)

hmm. interesting. i'll have a test too with a complicated archi model and report back

STRAT
09-15-2005, 08:11 AM
wtf? holy quivering sh**e batman!!! ok, did a test.


all in one pre-pass continuing to render (whilst saving the gi solution) -

pre-pass - 13mins
render - 10.5 mins
Total render time - 23.5 mins



pre-cached gi solution render -

pre-pass - 13 mins (rendered previously)
render - 1 min flat!!!



this is most strange. can someone explain this please???


also, and this should go for using pre-cached gi solutions in general, when you render the scene you MUST use EXACTLY the same scene settings as the full render, ie, res,samples etc etc - do not change anything. else the render speeds differ massively and the gi solution will look wrong.

it therory, after the gi prepass, the machine is just rendering the normal raytrace pass, but why should it be infinately more faster when using a pre-saved gi solution and not rendering all in one?

JamesMK
09-15-2005, 09:07 AM
Hmm, interesting find!

It seems like the benefits vary depending on what type of scene it is though - I just tried it on a very messy, high-frequency mesh and GI from a sky sphere... all in all totally different from an interior for instance, and I got:

Straight render prepass + render in one go -> 11 min 6 sec

Interrupted, render from illum cache -> 10 min 56 sec

So, no dramatic difference there... but the fact that there is a difference at all is quite interesting :curious:

STRAT
09-15-2005, 09:15 AM
Hmm, interesting find!

It seems like the benefits vary depending on what type of scene it is though - I just tried it on a very messy, high-frequency mesh and GI from a sky sphere... all in all totally different from an interior for instance, and I got:

Straight render prepass + render in one go -> 11 min 6 sec

Interrupted, render from illum cache -> 10 min 56 sec

So, no dramatic difference there... but the fact that there is a difference at all is quite interesting :curious:

absolutely. i did a couple more tests. in one instance i got another 600% time saving, whereas in another i got only a 20% time saving.

it would be nice if we could pre determin what type of scene would yeild what type of results. this way we would know if it's worth our while inturrupting the render.

if i was only going to save 2 mins over a 15 min render then i wouldnt bother, but as in my first test, if i could render out in 1 min instead of 10 mins then this is a massive advantage.

Ric535
09-15-2005, 11:53 AM
i have been wondering about it for a while now but have only just go round to testing it recently, i did some more tests and for some scenes as you say it is a real speed up for others just a little yet either eway its seems to be a good thing

(remember it will work best if your prepass was initaially 1/1 as that would create samples filling the whole render that would need to be cleaned up)

If C4D had an option in the render settings JUST to calculate the prepass (like some othere renderers do) then a batch render could be set up with the the same scene for jobs 1 and 2 which would use this method

Anyway this needs some proper explanation i think,

Thanks for testing :)

EDIT - also i think once the .gi file is saved but only the prepass is calculated, then the first time a render is continued when it is finished i think somehow the gi file is rewritten to as if you rerender once again (and again) then the following renders will be even quicker

STRAT
09-15-2005, 12:32 PM
thats a point. is there an option somewhere to just render out a pre-pass file and not the whole render?

Ernest Burden
09-15-2005, 01:12 PM
EDIT - also i think once the .gi file is saved but only the prepass is calculated, then the first time a render is continued when it is finished i think somehow the gi file is rewritten to as if you rerender once again (and again) then the following renders will be even quicker

That implies that the .gi file is ammended, not re-written. I have been pondering your post and do not know what to think about the differences you are observing.

Have you tried a bigger prepass instead of a same-size one? Can anyone explain what exactly is held in the .gi file to begin with?

Incarnadine
09-15-2005, 04:15 PM
I've seen this sometimes myself. (Attaching in case answers turn up!)

Ric535
09-15-2005, 04:54 PM
That implies that the .gi file is ammended, not re-written. I have been pondering your post and do not know what to think about the differences you are observing.

Have you tried a bigger prepass instead of a same-size one? Can anyone explain what exactly is held in the .gi file to begin with?

yes sorry amended, but i will have do do more tests with this

as for rendering a bigger size prepass, well i dont think thats possible

Ric535
09-20-2005, 09:55 PM
That implies that the .gi file is ammended, not re-written. I have been pondering your post and do not know what to think about the differences you are observing.

it seems as though i was right, after the prepass is calculated, if the render is stopped a .gi file is written. then if the render is started again using the saved prepass it will render a lot faster than if you had let it continue on from the original prepass

AND also the gi file IS ammended after the first full render is finished, example after the prepass is done the .gi file may be 27mb but after one full render it might go up to about 30 mb (in one test i did this was the case), it may be further ammeded for following renders but not by such a large margin, any further renders will run a lot faster than the first render after the prepass

I dont know if any of this makes sense..

Anyone else have any comments....

lllab
09-20-2005, 11:05 PM
very strange all this, but i dont get it.

srek could you explain what happens here?

thanks
stefan

CGTalk Moderation
09-20-2005, 11:05 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.