Ch07-AdvCompArch-ManycoresAndGPUs-PaulKelly-V03

Fatahalian m houston a closer look at gpus commun acm

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: m of locality) 31 GT200 streaming mul(processor hSp://www.realworldtech.com/page.cfm?Ar(cleID=RWT090808195242 GT200 streaming mul(processor •  Mul(threaded instruc(on fetch and issue unit •  8 streaming processors (SP) –  32 ­bit ALU and FPU (e.g. bitwise, min, max; FP add, mul, mad) –  Branch unit (e.g. cmp) •  2 special ­func(on units (SFU) –  Transcendental func(ons (e.g. reciprocal, sine), interpola(on –  Interpola(on hardware can be used as 4 FP mul(pliers •  16KB shared memory (as fast as registers, if no conflicts) •  64 ­bit FP and integer unit (not in G80) •  16K 32 ­bit registers (only 8K in G80) SM mul(threaded instruc(on unit •  Instruc(on unit creates, manages, schedules and executes threads in groups of 32 threads called warps •  Warp instruc(ons are fetched into an instruc(on buffer •  Scoreboard qualifies warp instruc(ons for issue •  Scheduler priori(ses ready to issue warp instruc(ons ba...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online