Investigate putting Cortex-Strings memcpy() and friends into the kernel

Registered by Bernhard Rosenkraenzer on 2012-06-26

Why?
The kernel may benefit from our memcpy(). We need to investigate this to prove or disprove it this.

Context?
General Android optimization work.

What gets produced?
Android builds with the changes or a Wiki page with the results of the experiment.

Where will the work get put?
In Android builds or on a Wiki.

Blueprint information

Status:
Started
Approver:
Zach Pfeffer
Priority:
Low
Drafter:
Bernhard Rosenkraenzer
Direction:
Approved
Assignee:
Bernhard Rosenkraenzer
Definition:
Approved
Series goal:
Accepted for future
Implementation:
Blocked
Milestone target:
milestone icon backlog
Started by
Bernhard Rosenkraenzer on 2012-07-02

Related branches

Sprints

Whiteboard

Notes:
[2012/7/2 pfefferz] Put notes here.
Blocked waiting for completion of:
https://blueprints.launchpad.net/linaro-android/+spec/android-faster-memcpy
to decide which memcpy() we should investigate inside the kernel
[2012/7/23 pfefferz] Since JB landed mid-cycle all of Bero's work was put on hold

Meta:
Roadmap id: ANDROID-OPTIMIZATION
Headline: New memcpy() etc. routines have been put into the kernel, or the lack of performance improvements from that have been documented.
Acceptance: New memcpy() etc. routines have been put into the kernel, or the lack of performance improvements from that have been documented.

(?)

Work Items

Work items:
Build a kernel with the memcpy routine replaced with the one put into Linaro's Bionic: TODO
Benchmark it against normal kernel: TODO
If it performs better, get the patch into at least one Linaro kernel: TODO
If things perform better, push to kernel.org: TODO
If it doesn't perform better, document that fact on the wiki: TODO

This blueprint contains Public information 
Everyone can see this information.