c31 lecture 16

For cpus gpus fpgas with complete step code

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: CPUs, GPUs, FPGAs, with complete step code generation in some cases 44 Customizable Heterogeneous Platform (CHP) Creation Cache parameters Cache parameters Cache size & Cache size & configuration configuration Cache vs. SPM Cache vs. SPM … … NoC parameters NoC parameters Interconnect Interconnect topology topology # of virtual channels # of virtual channels Routing policy Routing policy Link bandwidth Link bandwidth Router pipeline Router pipeline depth depth Number of RF-I Number of RF-I enabled routers enabled routers RF-I channel and RF-I channel and bandwidth allocation bandwidth allocation … … Customizable Heterogeneous Platform (CHP) Core parameters Core parameters Frequency & voltage Frequency & voltage Datapath bit width Datapath bit width Instruction window Instruction window size size Issue width Issue width Cache size & Cache size & configuration configuration Register file Register file organization organization # of thread contexts # of thread contexts … … Custom instructions & Custom instructions & accelerators accelerators Shared vs. private accelerators Shared vs. private accelerators Choice of accelerators Choice of accelerators Custom instruction selection Custom instruction selection Amount of programmable fabric Amount of programmable fabric … … $ $ $ $ $ $ $ $ Fixed Fixed Core Core Fixed Fixed Core Core Fixed Fixed Core Core Fixed Fixed Core Core Custom Custom Core Core Custom Custom Core Core Custom Custom Core Core Custom Custom Core Core Prog Prog Fabric Fabric Prog Prog Fabric Fabric accelerator accelerator accelerator accelerator Reconfigurable RF-I bus Reconfigurable optical bus Transceiver/receiver Optical interface Key questions: Optimal trade-off between efficiency & customizability Which options to fix at CHP creation? Which to be set by CHP 45 Experimental Platform Thrust: Highlights Server-class platform Mobile platform Server/client model Computation vs. display CPU, GPU, FPGA, et...
View Full Document

This note was uploaded on 04/03/2014 for the course CS 31 taught by Professor Melkanoff during the Fall '00 term at UCLA.

Ask a homework question - tutors are online