A GPU interval library based on Boost.Interval

Caroline Collange (1) , Jorge Flórez (2) , David Defour (1)

1 LP2A - Laboratoire de Physique Appliquée et d'Automatique
2 IIiA - Institut d'informàtica i aplicacions

Interval arithmetic is widely used in numerical algorithms requiring reliability. Ray tracing of implicit surface is one of these applications that use interval arithmetic to increase the quality of a produced image. However these applications are computationally demanding. One solution is to use graphics processing unit (GPU) in order to take advantage of its computational power. We describe in this paper a GPU implementation of interval operators based on the Boost library. We tested these operators on a ray tracing algorithms and observe several order of execution speed improvements over the CPU version with the same image quality.

Mots clés: GPU; boost; interval arithmetic; cuda; cg

Domaines: Architectures Matérielles [cs.AR]; Arithmétique des ordinateurs

Caroline Collange, Jorge Flórez, David Defour. A GPU interval library based on Boost.Interval. 8th Conference on Real Numbers and Computers, Jul 2008, Santiago de Compostela, Spain. pp.61-71. ⟨hal-00263670v2⟩