Manta Interactive Ray Tracer Development Mailing List

Text archives Help


Re: [MANTA] rendering vs. displaying


Chronological Thread 
  • From: James Bigler <bigler@cs.utah.edu>
  • To: Abe Stephens <abe@sci.utah.edu>
  • Cc: "'manta@sci.utah.edu'" <manta@sci.utah.edu>
  • Subject: Re: [MANTA] rendering vs. displaying
  • Date: Thu, 16 Jun 2005 20:13:19 -0600
Very cool!

With the amount of time it took to send the image compared to the amount of time to render at 1 spp, I would definitely kick up the number of samples a bit. ;)

Also, I don't know what method you used to schlep the data around, but if you use ssh definitely turn on compression "ssh -C" if you haven't already.

I would also use "unu resample -s = x0.2 x0.2 " to make a 1/5 smaller version. It will give the effect of having multiple samples per pixel and will give it a really nice look.

James

Abe Stephens wrote:
Hey,

Today we rendered the Boeing777 cross section of the entire plane at 46080x14400 1spp.

I was pleased to see:

Benchmark completed in 7.42894 seconds (1 frames, 0.134609 frames per second)

This was run on an Altix using 508 processors.

This was only the beginning of the frame's journey. The output frame was approximately 2.47 GB and was located on a computer in the mid-west. It took use several hours to copy the file to a file server in Mountain View, and then another half hour or so to move the file to a Prism workstation. Next we copied the file into memory (several moments) and then displayed it. (And we actually got to finally see the image....)

Anyone know where we can get something this size printed? (We will probably render another image with more spp.)

Abe








Archive powered by MHonArc 2.6.16.

Top of page