This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU][MC][GFX10] Improved diagnistic message for image instructions w/o DIM modifier
AbandonedPublic

Authored by dp on Nov 25 2020, 2:48 AM.

Details

Reviewers
rampitec
Summary

See bug 48241

Diff Detail

Event Timeline

dp created this revision.Nov 25 2020, 2:48 AM
Herald added a project: Restricted Project. · View Herald TranscriptNov 25 2020, 2:48 AM
dp requested review of this revision.Nov 25 2020, 2:48 AM
rampitec added inline comments.Nov 25 2020, 10:14 AM
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
4032

It is not necessarily "dim". For example image_bvh* do not have dim modifier.

dp abandoned this revision.Nov 25 2020, 10:58 AM
dp added inline comments.
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
4032

Thanks. I did not know about these instructions. I thought that at least in gfx1010 all MIMG opcodes require DIM.

rampitec added inline comments.Nov 25 2020, 10:59 AM
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
4032

You can query if opcode has dim operand.

foad added a subscriber: foad.Nov 26 2020, 1:17 AM
foad added inline comments.
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
4031

Please use isGFX10Plus, which I just added in D92092.

dp added inline comments.Nov 26 2020, 2:13 AM
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
4031

Sure, thanks.