Are you using lightmapping?

afecelis · Post by **afecelis** » Sat Jan 28, 2006 6:16 pm

I love the idea of "complete freedom" you get with this powerful tool and blender.

Definitely a project that must not be left into oblivion.

sRc · Post by **sRc** » Tue Feb 07, 2006 4:35 pm

afecelis wrote:I love the idea of "complete freedom" you get with this powerful tool and blender.

Definitely a project that must not be left into oblivion.

i agree. ive been looking into stuff like OpenBSP the last couple days (gonna try making a loarder for irrlicht, that is if I can get Quark to actually compile the stupid maps >_>), and everywhere i go I keep ending back up at Murphy's projects, MIM and others.

dont let it die!

Triple-J · Post by **Triple-J** » Tue Feb 14, 2006 7:21 am

The Anaconda wrote: dont let it die!

exactly!! We need u, I need u...

Any news???

Triple-J · Post by **Triple-J** » Tue Feb 14, 2006 7:30 am

and again, Murphy! Murphy! o// o// and everybody same time with waves, MURPHY! MURPHY! \o/

Murphy · Post by **Murphy** » Sat Feb 18, 2006 8:50 am

Sorry, still not much progress to speak of. I just haven't had the time to figure out how to do all the floating point stuff in the GPU.

If anyone has a link to a tutorial showing (simply) how to render and retrieve pixels in floats that aren't clamped 0.0 - 1.0 on ATI hardware, feel free to pass it along.

Pretty much every demo/tutorial I can find that uses floating point is huge and has a bunch o' other stuff I'd have to sort through to find what I need, and I just haven't had the time and inclination.

I'm still planning to complete the multi-pass (bouncing) algorithm someday, but it may not be until this summer, or I may go ahead and implement it the way I was originally planning which will doubtlessly be considerably slower...

vermeer · Post by **vermeer** » Sun Feb 19, 2006 3:44 pm

I have said it many times. It's my best hope in lightmapping in general, not that I expect it to kill Vray or FinalRender, but as the fsrad uv/lightmap packing like features (and the smooth normals surfaces) is simply nowhere at that functionality. Not to mention it'd rock totally in the free tools land.

You have my permanent vote

I am sorry I am not a coder, I don't know where you can get that info.

But how slow is slow? How much would it take? longer than fsrad ? (yup, seems that one didn't compute normals, anyway...)

As if it just is slower, fsrad times, I'd bet better go the slow renderer route if it can make a first alpha come sooner...

Who knows, maybe later we find some gpu docs and can update the core renderer or something...

I only can think at ATI site, developers section. I am an artist only...

omaremad_ · Post by **omaremad_** » Sun Feb 19, 2006 6:47 pm

hey are u trying somesort of hdr light maps (points at floating pts)

well i managed some faked hdrs by sampling the stuff down into the 1-0 range but that leads to inaccuracings

for real hdr use floating pt rts which are not very common on todays hw and require ps 3 for most of the decoding operartions

u can use the blender lightbaker script and archimap uv to create HDR and GI lightmaps u know

Guest · Post by **Guest** » Sun Feb 19, 2006 10:39 pm

omaremad_ wrote:hey are u trying somesort of hdr light maps (points at floating pts)

Sort of. The results will end up put into RGB bytes, but I need unclamped floating point values while calculating the light.

My original plan was to fake this by having each texel be a different color and use that as a look up table to floating point values stored in an array in main memory (I thought this was a somewhat novel idea, but since found out it's been done before). But since then I've decided I'd rather do it all floating point in hardware and just have a fairly recent GPU as a requirement (I made sure to get a decent GPU in my new laptop for just this reason).

Unfortunately, I really don't have time to become a GPU programming expert at the moment. I don't think I really need that much info to get it working, but I just can't seem to find it.

Ah well. I'll keep chipping away or just go back to my original concept.

Thanks for the continued vote of support, Vermeer (and others!). And we're talking pretty slow, I think. Far slower than FSRad. FSRad busts out pretty respectable results in like a minute. Mine would be more along the lines of an hour, probably. But yes, starting with the older method and then switching to the hardware render later shouldn't be that hard -- it's just more work and more time and I hate to throw away time on the project since I don't have that much to begin with.