If inline asm has a VGPR def, it must have come from a VGPR write
somewhere inside the asm. This should be further extended to all
read after write hazards.
Details
Details
Diff Detail
Diff Detail
Event Timeline
Comment Actions
LGTM, but it solves the problem just partially. One specific problem in one specific place. In general we cannot handle inline asm here.
Comment Actions
I'm working on a more general patch for this pattern but it seems we're missing a bit of test coverage for some of these hazards