PLATFORM GamSimAnImageTechMeth

Platform Performance

Project Leader Project Co-leader
Alexandra Fedorova
Simon Fraser University
David Fleet
University of Toronto

Harnessing the parallelism of new processors (such as heterogeneous multicore processors) to perform complex, real-time computations for games requires new parallel algorithms for surface representation and generation in geometric modeling and real-time graphics. 

New methods for triangulating surfaces to produce better tessellations and fast polygonization of mathematical surfaces are being developed. The new CASCADE parallel programming framework automatically detects implicit dependencies among tasks using compile-time analyses and runtime information. A method for avoiding race conditions in parallel games by comparing compile-time analyses of memory locations used by each task has the potential to achieve greater parallelism than a human programmer would using traditional tools.

PLATFORM will relieve game developers from thread and shared state management and will provide faster parallel algorithms for geometry processing.