Campus Event Calendar
Event Entry
What and Who
Efficient ray tracing on GPUs (Masters Seminar)
Stefan Popov
IMPRS
Talk
AG 1, AG 2, AG 3, AG 4, AG 5
MPI Audience
Note: We use this to send email in the morning.
Date, Time and Location
Tuesday, 24 January 2006
13:00
45 Minutes
46.1 - MPII
0.24
Saarbrücken
Abstract
Abstract:
Recursive ray tracing is a simple and powerful technique for rendering
high quality images. Even though recent algorithms have reached
real time frame rates, ray tracing is still very computationally
intensive. The computational power advantages of GPUs over CPUs
and their shading oriented architecture design make GPUs an attractive
platform for real-time ray tracing. However, up to this point, due to
architecture limits, no algorithm has been published that efficiently
exploits the power advantages of today's GPUs for ray-object
intersection phase.
This talk presents a new ray-object intersection algorithm based on the
recursive KD-tree traversal algorithm. It is expected to have the
same performance characteristics. However, it does not require a stack
and it is thus suitable for GPU implementations. Furthermore
it is a one-pass scalable algorithm, exploiting the GPU multi-threaded
architecture and GPU caches (in the case of coherent rays).
An algorithm designed for GPUs that can partially replace mail-boxing
will be presented as well.
Recursive ray tracing is a simple and powerful technique for rendering
high quality images. Even though recent algorithms have reached
real time frame rates, ray tracing is still very computationally
intensive. The computational power advantages of GPUs over CPUs
and their shading oriented architecture design make GPUs an attractive
platform for real-time ray tracing. However, up to this point, due to
architecture limits, no algorithm has been published that efficiently
exploits the power advantages of today's GPUs for ray-object
intersection phase.
This talk presents a new ray-object intersection algorithm based on the
recursive KD-tree traversal algorithm. It is expected to have the
same performance characteristics. However, it does not require a stack
and it is thus suitable for GPU implementations. Furthermore
it is a one-pass scalable algorithm, exploiting the GPU multi-threaded
architecture and GPU caches (in the case of coherent rays).
An algorithm designed for GPUs that can partially replace mail-boxing
will be presented as well.
Contact
Kerstin Kathy Meyer-Ross
9325-226
--email hidden
passcode not visible
logged in users only
Friederike Gerndt, 01/20/2006 13:09 -- Created document.