SharpEye: A parallel, portable, selective rendering system based on RADIANCE
SharpEye:
A parallel, portable, selective rendering system based on RADIANCE
Kurt Debattista
Francisco
Pereira (speaker)
Bristol University
ABSTRACT
Using RADIANCE as the core of our renderer we are in the process of developing a
rendering architecture which will speed-up rendering time for animations
through the use of message passing parallelism, perceptual based selective
rendering and by efficiently exploiting the underlying hardware. Our
parallel system TPL (Thread Parallel Library) is a portable parallel library
which lies on top of any message system providing thread-safe mechanisms for
any vendor-specific MPI or PVM implementation. Furthermore, we are
investigating how to best make use of temporal and spatial coherence in a
parallel environment. The perceptual based selective rendering exploits
limitations in the human visual system to only render salient parts of an image
or animation at high quality while the rest of the image, not noticed by the
human visual system, can be rendered at significantly lower quality.
Finally we use a hardware resource allocater (HRA) to ensure we distribute
tasks amongst available hadware resources efficiently, ensuring we exploit the
full spectrum of today s hardware SMPs, streaming instructions, cache
exploitation, GPUs and possibly FPGAs. The talk will focus on the three
aspects of our system and the interaction amongst them.