This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU: Temporarily disable packed inlinable literals (v2f16, v2i16)
ClosedPublic

Authored by kzhuravl on Apr 21 2017, 11:11 AM.

Diff Detail

Repository
rL LLVM

Event Timeline

kzhuravl created this revision.Apr 21 2017, 11:11 AM
arsenm accepted this revision.Apr 21 2017, 11:24 AM

LGTM

This revision is now accepted and ready to land.Apr 21 2017, 11:24 AM
This revision was automatically updated to reflect the committed changes.

Any reason for this? This breaks instructions like

v_pk_fma_f16 v5, s101, 0, 0

in the assembler.

Any reason for this? This breaks instructions like

v_pk_fma_f16 v5, s101, 0, 0

in the assembler.

For now I am adding "-enable-packed-inlinable-literals" to the assembler tests.

Any reason for this? This breaks instructions like

v_pk_fma_f16 v5, s101, 0, 0

in the assembler.

We're not sure how the packed inline immediates are actually encoded