Ch07-AdvCompArch-ManycoresAndGPUs-PaulKelly-V03

Shalf k yelick stencil computaon opmizaon and

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: hniques: –  Mul(core –  Simultaneous mul(threading (SMT) –  Vector instruc(ons –  Predica(on •  So basically a GPU core is a lot like the processor architectures we have studied! •  But the SIMT programming model makes it look different SIMT vs SIMD – GPUs without the hype •  SIMT: •  SIMD: •  one thread per lane •  SMT: a small number of threads run on the •  Adjacent threads same core to hide execute in lockstep memory latency (“warp”/”wavefront”) •  SMT: mul(ple “warps” •  Each thread may include...
View Full Document

This document was uploaded on 03/18/2014 for the course CO 332 at Imperial College.

Ask a homework question - tutors are online