CSC 367: Cuda Development

This course discusses the principles, styles, and data-parallel patterns real-time software architecture using CUDA (Compute Unified Device Architecture). Trade-offs and ramifications of software architecture with respect to performance, maintainability, and reusability, will be explored. Topics include: Fermi and Kepler Architecture, Warps, Grid, Blocks, Thread coordination, Shared memory, synchronization and bulk texture memory. Focusing on CUDA Streams, scheduling, priorities and events using Atomic instructions with GPU and CPU interoperability.