 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hello,
I'm trying to locate portions of POV-ray that bottleneck performance so that I
can try and offload some of POV-ray's workload onto a hardware accelerator. I'm
just trying to see how fast I can get POV-ray to run by programming an FPGA to
accelerate certain portions of the code, but I'm finding it hard to locate in
the source.
Thanks!
-Michael Poon
Post a reply to this message
|
|
| |
| |
|
|
From: Thorsten Froehlich
Subject: Re: Computationally intensive parts of POV-ray?
Date: 17 Mar 2010 03:39:40
Message: <4ba0873c$1@news.povray.org>
|
|
|
| |
| |
|
|
On 17.03.10 00:29, Michael wrote:
> I'm trying to locate portions of POV-ray that bottleneck performance so that I
> can try and offload some of POV-ray's workload onto a hardware accelerator. I'm
> just trying to see how fast I can get POV-ray to run by programming an FPGA to
> accelerate certain portions of the code, but I'm finding it hard to locate in
> the source.
The bottlenecks for POV-Ray that best fit into hardware are the same as for
any other ray-tracer: The bounding. It is uniform and simple (for hardware)
and can easily be solved in parallel. You can find various SIGGRAPH papers
(early to mid 2000s) on real-time ray-tracing that also covered or
referenced hardware BSP algorithm implementations.
Thorsten, POV-Team
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thorsten Froehlich <tho### [at] trf de> wrote:
> On 17.03.10 00:29, Michael wrote:
> > I'm trying to locate portions of POV-ray that bottleneck performance so that I
> > can try and offload some of POV-ray's workload onto a hardware accelerator. I'm
> > just trying to see how fast I can get POV-ray to run by programming an FPGA to
> > accelerate certain portions of the code, but I'm finding it hard to locate in
> > the source.
>
> The bottlenecks for POV-Ray that best fit into hardware are the same as for
> any other ray-tracer: The bounding. It is uniform and simple (for hardware)
> and can easily be solved in parallel. You can find various SIGGRAPH papers
> (early to mid 2000s) on real-time ray-tracing that also covered or
> referenced hardware BSP algorithm implementations.
>
> Thorsten, POV-Team
Hmm, ok, I figured it would be that or calculating intersections. The thing I
need help on though is figuring out where to plug in my hardware into the
source. So looking through the source, I'd take a guess that most of the
computation is done in render.cpp, but a lot of it seems boilerplate and I'm
having a difficult time trying to find where all of the computation is being
done. If you could point me to certain functions, I'd appreciate it greatly.
Thanks!
Michael
Post a reply to this message
|
|
| |
| |
|
|
From: Thorsten Froehlich
Subject: Re: Computationally intensive parts of POV-ray?
Date: 26 Mar 2010 02:59:39
Message: <4bac5b5b@news.povray.org>
|
|
|
| |
| |
|
|
On 25.03.10 22:45, Michael wrote:
> Hmm, ok, I figured it would be that or calculating intersections. The thing I
> need help on though is figuring out where to plug in my hardware into the
> source. So looking through the source, I'd take a guess that most of the
> computation is done in render.cpp, but a lot of it seems boilerplate and I'm
> having a difficult time trying to find where all of the computation is being
> done. If you could point me to certain functions, I'd appreciate it greatly.
If you need a well documented source code, you might be better off doing
your project as an extension to the PBRT Ray-Tracer (www.pbrt.org), which is
actually part of the book with the same name.
Otherwise you will need to take a look at the
source/backend/support/bsptree.cpp source file and how it is tied into
POV-Ray 3.7.
Thorsten
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
