This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU/R600: Convert buffer_id to VTX_READ input
ClosedPublic

Authored by jvesely on Jul 21 2016, 2:22 PM.

Details

Reviewers
tstellarAMD
Summary

Use Patterns to match buffer ids to vtx_reads
Add buffer id to asm string

Diff Detail

Repository
rL LLVM

Event Timeline

jvesely updated this revision to Diff 64964.Jul 21 2016, 2:22 PM
jvesely retitled this revision from to AMDGPU/R600: Convert buffer_id to VTX_READ input.
jvesely updated this object.
jvesely added a reviewer: tstellarAMD.
jvesely set the repository for this revision to rL LLVM.
jvesely added subscribers: llvm-commits, arsenm.

What's the reason for moving to separate patterns?

lib/Target/AMDGPU/EvergreenInstructions.td
143–144

Capitalize, indent comment and remove commented out line. The comment should probably go on the encoding bit definition

What's the reason for moving to separate patterns?

'twas the easiest way to add it to the asm string.

jvesely updated this revision to Diff 65078.Jul 22 2016, 8:02 AM

remove dead code and comments
change buffer_id to lower case to be consistent with other input field names

jvesely marked an inline comment as done.Jul 28 2016, 7:20 AM
tstellarAMD accepted this revision.Aug 5 2016, 4:29 PM
tstellarAMD edited edge metadata.

LGTM.

This revision is now accepted and ready to land.Aug 5 2016, 4:29 PM
jvesely closed this revision.Aug 16 2016, 8:36 AM

committed as r278749