General Purpose Computation on Graphics （通用计算图形）.pdfVIP

General Purpose Computation on Graphics Processors (GPGPU) Mike Houston, Stanford University A little about me /~mhouston Education: – UC San Diego, Computer Science BS – Stanford University, Computer Science MS – Currently a PhD candidate at Stanford University Research – Parallel Rendering – High performance computing – Computation on graphics processors (GPGPU) 2 Mike Houston - Stanford University Graphics Lab What can you do on GPUs other than graphics? Large matrix/vector operations (BLAS) Protein Folding (Molecular Dynamics) FFT (SETI, signal processing) Ray Tracing Physics Simulation [cloth, fluid, collision] Sequence Matching (Hidden Markov Models) Speech Recognition (Hidden Markov Models, Neural nets) Databases Sort/Search Medical Imaging (image segmentation, processing) And many, many more… 3 Mike Houston - Stanford University Graphics Lab Why use GPUs? COTS – In every machine Performance – Intel 3.0 GHz Pentium 4 • 12 GFLOPs peak (MAD) • 5.96 GB/s to main memory – ATI Radeon X1800XT • 120 GFLOPs peak (fragment engine) • 42 GB/s to video memory 4 Mike Houston - Stanford University Graphics Lab Task vs. Data parallelism Task parallel – Independent processes with little communication – Easy to use • “Free” on modern operating systems with SMP Data parallel – Lots of data on which the same computation is being executed – No dependencies between data elements in each step in the