ARM Improve 64-bit division routines in libgcc

Registered by Matthew Gretton-Dann on 2012-11-01

libgcc's 64-bit division routines assume that if you don't have native 64x64 division you do have native 64x32. No ARM cores have such a divide instruction. We should investigate improving this.

Blueprint information

Status:
Not started
Approver:
Matthew Gretton-Dann
Priority:
Medium
Drafter:
Matthew Gretton-Dann
Direction:
Approved
Assignee:
Mans Rullgard
Definition:
Discussion
Series goal:
Accepted for 4.8
Implementation:
Informational Informational
Milestone target:
milestone icon 4.8-2013.09

Related branches

Sprints

Whiteboard

[matthew-gretton-dann 2013-03-21] This is now tracked in TCWG-24

Meta:
Headline: Improve performance of 64-bit divides in ARMv7-A
Acceptance criteria: Patches landed upstream
Roadmap id: CARD-304

(?)

Work Items

Work items:
Write efficient 64x64 divide for cores with UDIV/SDIV support: TODO
Write efficient 64x64 divide for cores without any divide instruction: TODO
Benchmark: TODO
Document: TODO
Land patches upstream: TODO

This blueprint contains Public information 
Everyone can see this information.