chapter1-AppendixA-m2-ziavras

Chapter1-AppendixA-m2-ziavras

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: execution of any instruction to be driven by data (operand) availability. High-degree of parallelism at the fine-grain instructional level Some material is adapted from D. Culler & D. Patterson (UCB) S. Ziavras Control Flow Vs. Data Flow Loop Fetch Fetch PC = PC + 1 Decode Operand Fetch Fetch Execute Write Back Loop Some material is adapted from D. Culler & D. Patterson (UCB) Dataflow Execution • Not possible to describe the operation into neat, discrete steps • Results obtained from obtained from executed instruction are called data tokens • Data tokens are passed directly between instructions • Receiving instructions check conditions and fire if conditions are met are met S. Ziavras Control Flow Vs. Data Flow (2) Control-flow example 1. C=A+B 2. D=C*A 3. E=C-B 4. F=C+D Dataflow execution A B + C E Execution follows operations in data in data-flow graph graph Some material is adapted from D. Culler & D. Patterson (UCB) * + D F S. Ziavras Control Flow Vs. Data Flow (3) • Dataflow computer: instruction execution driven by data availability ♦ In theory, an instruction is ready for theory an instruction is ready for execution when all its operands become available ♦ Instructions are not ordered ♦ Data directly held inside instructions di ♦ Data generated by an instruction are replicated into many copies & forwarded directly to instructions that need it, in the form of data tokens th it th ♦ Data tokens, once consume...
View Full Document

This document was uploaded on 02/09/2014.

Ask a homework question - tutors are online