I didn't say I would convert it to GPGPU. I said prepare for converting to
gpgpu, What that means is, If there is recursion inside code, convert it to
iterative code. If there is, a data structure that doesn't work well on
GPU, extract it and replace it with one that works better. This is
something that would help should someone decides its time to port the code.
I agree, porting the entire raytracer to GPU is very hard work, and
definitelly one that I am not suited for. Even if there was porting
included, it would be to a small part of the program say couple of
functions that do lots of calculations and are suited.
Here is what I actually aim to do:
*The first half of the project will be dedicated to determining
optimization opportunities and exposing them through documenting and aiming
tests at them, to help others involved in the project notice them.*
The second half will be aimed at implementing some of these optimizations
and testing their feasibility and success.
Or for short:
Analyze
Benchmark
Optimize
Check
I will try to make the proposal easier to read.
Thank you for the input, it is appreciated
On Wed, Mar 19, 2014 at 5:19 PM, Christopher Sean Morrison
<brl...@mac.com>wrote:
>
> On Mar 19, 2014, at 11:44 AM, Tose Nikolov wrote:
>
> Here is my google summer of code proposal, please share your toughts.
> http://brlcad.org/wiki/User:Todor_nikolov/proposal
>
>
> Todor, this looks like a good start, but I hope to see some more detail on
> what specifically you aim to accomplish. You speak about the merits of
> GPGPU ray tracing, but don't really say what you're going to do or how
> you're going to do it.
>
> For example, we have a dozen existing ray tracers in BRL-CAD all build on
> our LIBRT ray trace library. That library has two dozen primitive types
> and there are separate intersection functions for each of them. There's no
> way you'd possibly convert that entire infrastructure to GPGPU within the
> timeframe of GSoC, so you need to do some homework and figure out exactly
> what you're going to do. You could maybe focus on the implicit primitives
> (the easiest) or on polygonal meshes (second easiest) or on NURBS (hardest
> but most valuable), ...
>
> On that same note, your timeline needs improvement. You currently propose
> spend nearly month getting yourself familiar with our code. That's not
> going to work. That should all happen before the coding window even
> begins. That's why the timeline schedules five weeks before coding begins
> on May 19th.
>
> From now until then is the time to get intimately familiar with our code.
> Especially for a project as hard as the one you're suggesting, this is not
> just helpful but requisite. Moreover, as you don't seem to have any
> experience with GPGPU code (you don't mention any other than saying you
> have deep knowledge -- what does that mean?), I'd hope to see a good patch
> from you that demonstrates your ability to read and write code quickly,
> perhaps some modification to our "rt" raytracer program or librt library.
>
> Keep at it. ;-)
>
> Cheers!
> Sean
>
>
>
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/13534_NeoTech
> _______________________________________________
> BRL-CAD Developer mailing list
> brlcad-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/brlcad-devel
>
>
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
_______________________________________________
BRL-CAD Developer mailing list
brlcad-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/brlcad-devel
