Page MenuHomePhabricator

ARMEB: Vector narrow operations
AbandonedPublic

Authored by cpirker on Jun 6 2014, 8:20 AM.

Details

Reviewers
None
Summary

Hi all,

This patch implements correct vector narrow operations from <2 x 64> to <2 x 8>, <2 x 16>, and <2 x 32> and from <4 x 32 > to <4 x 8>, <4 x 16> vector type for big endian.
It applies to both integer and float type vectors. The CLANG vectorizer is likely to generate such vector extensions for array type casts.
You may also look into D4043, that covers the vector extension operations.

Please review.

Thanks,
Christian

Diff Detail

Event Timeline

cpirker updated this revision to Diff 10181.Jun 6 2014, 8:20 AM
cpirker retitled this revision from to ARMEB: Vector narrow operations.
cpirker updated this object.
cpirker edited the test plan for this revision. (Show Details)
cpirker added subscribers: Unknown Object (MLST), Konrad.
cpirker abandoned this revision.Jun 12 2014, 4:22 AM

Extensive testing showed that this was the wrong approach, as it has broken some tests.
I will issue another patch later this week.

Thanks,
Christian