This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU: Add support for MCExpr to instruction printer
ClosedPublic

Authored by jvesely on May 1 2016, 9:54 PM.

Diff Detail

Repository
rL LLVM

Event Timeline

jvesely updated this revision to Diff 55779.May 1 2016, 9:54 PM
jvesely retitled this revision from to AMDGPU: Add support for MCExpr to instruction printer.
jvesely updated this object.
jvesely set the repository for this revision to rL LLVM.

Needs a test

lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.cpp
608–609 ↗(On Diff #55779)

Why change this? It's already being truncated. Maybe add an explicit static_cast to the Imm assignment?

612 ↗(On Diff #55779)

Single quotes

jvesely updated this revision to Diff 55949.May 2 2016, 11:04 PM
jvesely edited edge metadata.
jvesely marked an inline comment as done.
jvesely added a subscriber: llvm-commits.

Use single quotes

lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.cpp
608–609 ↗(On Diff #55779)

Since there should be only 32 bit values, the idea was to print all 64 bits and catch any bugs/corruption in asm output.

Needs a test

I extended tests in D19792 to cover this

tstellarAMD accepted this revision.May 6 2016, 5:11 PM
tstellarAMD edited edge metadata.

LGTM.

This revision is now accepted and ready to land.May 6 2016, 5:11 PM
arsenm added inline comments.May 6 2016, 5:13 PM
lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.cpp
608–609 ↗(On Diff #55949)

That's not happening anyway, because there is the implicit cast in the BitsToFloat call

This revision was automatically updated to reflect the committed changes.