[Bf-committers] BIH acceleration structure?

[Yves Poissant]

Thanks Brecht for the patch. I will start with that and optimize it. It is 
better to already have something to kick-start with, especially since I'm 
new to the Blender source. When you say "that should work with the latest 
svn". I guess that means the latest v2.50 branch. Right?

Yes. you are right that optimizing the ray-triangle intersection test is an 
important consideration for speed. Wachter's thesis have a whole section 
about exactly that and he concludes that the Barycentric test is the most 
numerically stable, is one of the fastest and the barycentric representation 
can be encoded on a very small memory footprint. From my experience with BVH 
and kd-trees, though, optimizing memory layout and usage is the most 
important factor to get optimal speed.

Regards,
Yves

----- Original Message ----- 
From: "Brecht Van Lommel" <brechtvanlommel at pandora.be>
To: "bf-blender developers" <bf-committers at blender.org>
Sent: Tuesday, February 26, 2008 3:25 PM
Subject: Re: [Bf-committers] BIH acceleration structure?


>
> Hi,
>
> I integrated BIH code from Sunflow into Blender as a test, I didn't spend 
> time optimizing it to really get a good performance advantage out of it, 
> it was just a quick port from the java code. My guess is to really make 
> raytracing in Blender a lot faster you probably need to check on the ray 
> triangle intersection test too, and decently profile the code. But given 
> the results of those realtime raytracers, it should be possible to get a 
> big performance improvement. I've attached a patch of my code that should 
> work with the latest svn, if you're interested, though don't feel at all 
> obligated to use it, if you want to implement your own that's fine with 
> me.
>
> Cheers,
> Brecht.
>
>>----- Oorspronkelijk bericht -----
>>Van: Yves Poissant [mailto:ypoissant2 at videotron.ca]
>>Verzonden: dinsdag, februari 26, 2008 04:37 PM
>>Aan: 'bf-blender developers'
>>Onderwerp: Re: [Bf-committers] BIH acceleration structure?
>>
>>Thanks people with the welcome.
>>
>>I'll wait a bit here. I know that Brecht is really busy with Peach 
>>already.
>>In the meantime, I'm getting acqainted with Blender and I'm reading 
>>Carsten
>>Wachter's (the inventor of the BIH) PhD Thesis which was published just a
>>few weeks ago and where there is an in-depth discussion of the BIH and the
>>different strategies for its implementation. The Thesis is available at
>>http://vts.uni-ulm.de/doc.asp?id=6265 if anyone is interested.
>>
>>Yves
>>
>>----- Original Message ----- 
>>From: "Tom M" <letterrip at gmail.com>
>>To: "Yves Poissant" <ypoissant2 at videotron.ca>; "bf-blender developers"
>><bf-committers at blender.org>
>>Sent: Monday, February 25, 2008 5:13 PM
>>Subject: Re: [Bf-committers] BIH acceleration structure?
>>
>>
>>> On Sun, Feb 24, 2008 at 8:19 AM, Yves Poissant <ypoissant2 at videotron.ca>
>>> wrote:
>>>> Hi there,
>>>
>>> Hello,
>>>
>>> [snip]
>>>
>>> quite an impressive resume, we are happy to have you join us.
>>>
>>> Regarding BIH,
>>>
>>> as mentioned by joe, apparently Brecht worked on BIH some, also Kier
>>> did (as mentioned in the wiki - the University of Toronto students
>>> ended up doing composite nodes instead), but neither BIH effort came
>>> to fruition.  I contacted Kier, but apparently the coding he did is
>>> locked away on a laptop that he doesn't have access to and has no idea
>>> when he might have access to it again.
>>>
>>> Noone is working on BIH at this time as far as I know, but you might
>>> want to check with Brecht (he follows this mailing list but might be
>>> behind on his reading due to being busy with peach - Brecht Van Lommel
>>> <brechtvanlommel AT pandora DOT be>  ) to see if he plans to pick it
>>> back up or if his todo list is long enough that he is unlikely to
>>> return to it in the near future.
>>>
>>>>  Also, I'm not quite sure what are the custom for such a project. Do I
>>>> need
>>>>  to submit something and/or does such a project needs to be approved in
>>>> some
>>>>  way?
>>>
>>> Letting us know of your plans on this list is an excellent start.  If
>>> you don't hear from brecht soon, I'd dive in.  Once you have it either
>>> finished or good preliminary results please place a patch in the patch
>>> tracker, and email a notice to the list.  Also you might want to
>>> provide a testing build for users to play with so that bugs are less
>>> likely to be encountered when the patch is reviewed.
>>>
>>> You might also want to stop by #blendercoders on irc.freenode.net
>>>
>>> Also we have a weekly meeting on sundays for developer
>>> discussion/decisions - see the wiki for details and times.
>>> (Discussion can take place at any time of the week and on the mailing
>>> list, so the meeting is primarily for decisions and status updates -
>>> also immediately after the meeting the majority of developers are
>>> present so asking questions and discussions have the highest chance of
>>> someone knowledgable in that area being present ).
>>>
>>> Glad to have you join us, good luck,
>>>
>>> LetterRip
>>
>>_______________________________________________
>>Bf-committers mailing list
>>Bf-committers at blender.org
>>http://lists.blender.org/mailman/listinfo/bf-committers
>>
>>
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers