Text archives Help
- From: "Biagio Cosenza" <cosenza@dia.unisa.it>
- To: manta@sci.utah.edu
- Subject: [Manta] About packetization performance
- Date: Thu, 15 May 2008 12:16:38 +0200
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:sender:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition:x-google-sender-auth; b=KKvd10hpayWPmWi/DemyM8kdaMcPO36/oVuluu3nq+lPoRLI7VfZcCvOQjg6xPqNxrQLWFZZJfenu7S2+yCantBjivYZmEnb9vgn0nxR5o3VPadKQIUVWKbED7Yjp3vABkwqw26vu7WaxiIUeIJV+r7j593L+j+t+4bDt6QvPms=
Hi,
thanks for your previous suggestions.
I've two question:
1# I need to use arbitrary size tiles and I'm working on my
ImageTraverser (...thanks Abe).
By using DynBVH you have better performance with 8x8 and 16x16 packets.
Supposing that I can use only square shaped packet of 8x8, then I'm
constrained to use tile of AxB where A and B are both multiple of 8,
further splitted in 8x8 packets.
Are square-shaped packet a constraint of Manta?
Can I create, for example, a packet of 4x8 rays? I suppose that
performance may have a degradation with less than 64 rays per packet
and using not square-shaped (i.e. line) packet.
2# I've used the .anim (...thanks Thiago).
It works but I got this problem with images:
"Could not load diffuse map: /Maps/FFfern.jpg: Unsupported image
format for image."
I'm sorry ... I'm wasting your time...
Thanks in advance for any suggestions and comments.
Cheers, Biagio
On Tue, May 13, 2008 at 5:42 AM, Thiago Ize <thiago@cs.utah.edu> wrote:
>
>
or pass it a .anim file that contains a list of all the obj files (give the
>
full path). A quick read of triangleSceneViewer.cc should explain all the
>
other features.
>
>
Thiago
>
>
>
>
Abe Stephens wrote:
>
[Adding the Manta mailing list to cc.]
>
>
Hi,
>
>
Obviously Manta targets multi-core/shared memory computers rather than
>
clusters-- every so often there is some interest in cluster support on the
>
mailing list, but I'm not aware of any high performance ray tracers based on
>
MPI today. Systems like OpenRT use their own communication protocols.
>
Manta's implementation can be roughly divided between the parallel pipeline
>
that coordinates individual threads (see RTRT.cc) and the modular rendering
>
stack that creates ray packets and then traces and shades the scene. You
>
might want to take a look at the image traversers and load balancers.
>
>
For animation using .obj key frames take a look at the triangleSceneViewer
>
scene: "bin/manta -scene lib/libscene_triangleSceneViewer.dylib" You can
>
specify several key frames with separate "-model" arguments on the command
>
line or in a shell script.
>
>
Others on this mailing list might have suggestions too.
>
>
Abe
>
>
>
On Fri, May 9, 2008 at 1:38 PM, Biagio Cosenza <cosenza@dia.unisa.it> wrote:
>
>
> Hello,
>
> I'm Biagio Cosenza, a Phd Student of the University of Salerno
>
> (Italy), actually involved in a E.C. project at the University of
>
> Stuttgart (Germany).
>
>
>
> I'm developing a parallel ray tracer working on a cluster of
>
> workstations (based on MPI).
>
> However, I'm going to change my serial implementation with Manta.
>
>
>
> I've two basic questions.
>
>
>
> 1 - I'm thinking to have something like a simple interface between the
>
> serial and the parallel stuffs, in order to easily change the serial
>
> implementation.
>
> In short, starting from the Manta-Skeleton, I want to implement a
>
> method (based on Manta) with a similar signature:
>
>
>
> # rendering of the tile in the given buffer img
>
> void renderTile(Image *img, int x1, int x2, int y1, int y2)
>
>
>
> For your experience and knowledge of Manta, is this the right approach?
>
> Where do I have to start looking for (classes)?
>
> Suggestions are welcomed...
>
>
>
> 2 - Test scenes. I need some animated test scenes. I've found in an
>
> old svn version (until version 1613) that there was a BART reader, but
>
> it is not more available. Does it works with i.e. BART robots scene?
>
> I saw and tested an obj importer, but is not clear how to import
>
> Wald's animated test scene with keyframes
>
> (http://www.sci.utah.edu/~wald/animrep). I suppose that I should
>
> implement a callback.
>
>
>
>
>
> I'm sorry for the very looooong email :(
>
>
>
> Thanks in advance for comments, suggestions and/or other.
>
> Regards, Biagio
>
>
>
> --
>
> Biagio Cosenza
>
> ISISLab, Dip. di Informatica ed Applicazioni "Renato M. Capocelli"
>
> Universita' degli Studi di Salerno
>
> http://www.dia.unisa.it/~cosenza
>
>
>
>
--
Biagio Cosenza
ISISLab, Dip. di Informatica ed Applicazioni "Renato M. Capocelli"
Universita' degli Studi di Salerno
http://www.dia.unisa.it/~cosenza
- [Manta] About packetization performance, Biagio Cosenza, 05/15/2008
Archive powered by MHonArc 2.6.16.