Disable peeling

Registered by Michael Hope

Peeling aligns memory pointers at run time both to give a performance boost where the CPU is faster with aligned accesses, and to support CPUs that can only do aligned vector accesses.

ARM supports unaligned vector access for no penalty over aligned and a small penalty over aligned-with-alignment hints. As a first pass, disable peeling completely.

Blueprint information

Status:
Complete
Approver:
Michael Hope
Priority:
High
Drafter:
Michael Hope
Direction:
Approved
Assignee:
Christophe Lyon
Definition:
Approved
Series goal:
Accepted for 4.7
Implementation:
Informational Informational
Milestone target:
milestone icon backlog
Started by
Christophe Lyon
Completed by
Matthew Gretton-Dann

Sprints

Whiteboard

[matthew-gretton-dann 201303020] Moved to Jira: http://cards.linaro.org/browse/TCWG-7
[matthew-gretton-dann 2013-02-03] Blocked on lack of vectorizer cost model.

Meta:
Headline: Disable peeling for alignment for ARM in GCC's auto-vectorizer
Acceptance: Patch accepted upstream and backported to GCC Linaro.
Roadmap ID: CARD-304

(?)

Work Items

Work items:
Add a target hook to say if peeling is beneficial: TODO
Add to the vectoriser: TODO
Fix any downstream assumptions: TODO
Add a target-supports to the testsuite and update test cases: TODO

Dependency tree

* Blueprints in grey have been implemented.

This blueprint contains Public information 
Everyone can see this information.

Subscribers

No subscribers.