chapter3-m2-ziavras

Ziavras loop processingunrolling 4 observation 36

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: er makes address change 6 clock cycles per iteration clock cycles per iteration (Complete 1 iteration & store back 1 array element) S. Ziavras Loop Processing/Unrolling (4) • Observation: 3/6 cycles process array elements 3/6 • Objective: include more array operations in 6 cycles Loop unrolling unrolling Replicates loop body many times Uses Uses different regs. for different iterations Adjusts loop termination code Example: Apply to previous example loop. Assume: • 4 copies of loop body • R1 initially contains a multiple of 32 • Eliminate obviously redundant operations • Do not reuse any of the regs. not reuse any of the regs S. Ziavras Loop Processing/Unrolling (5) Loop: L.D ADD.D Merge S.D DADDUI L.D instrs. ADD.D S.D L.D Drop ADD.D unneeded S.D BNE operations L.D ADD.D S.D DADDUI BNE F0,0(R1) F4,F0,F2 F4,0(R1) ;drop DADDUI/BNE F6,-8(R1) F8,F6,F2 F8,-8(R1) ;drop DADDUI/BNE F10,-16(R1) F12,F10,F2 F12,-16(R1) ;drop DADDUI/BNE F14,-24(R1) Not easy: easy F16,F14,F2 Requires symboli...
View Full Document

This document was uploaded on 02/09/2014.

Ask a homework question - tutors are online