[PATCH] D99587: [AMDGPU] Deduce sign from address space

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D99587#2658501 <https://reviews.llvm.org/D99587#2658501>, @sebastian-ne wrote:

> We already use this connection in `SIInstrInfo::isLegalFLATOffset`, there is a check for `ST.hasFlatSegmentOffsetBug() && AddrSpace == AMDGPUAS::FLAT_ADDRESS`. I guess this is already broken then.

Yeah, that's wrong

> Is there a good way to find the instruction variant (global, scratch or flat) from the instruction (or intrinsic) or do we need to pass this through everywhere?

Something directly from the instruction should work, but that won't help you in the pattern matching contexts


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99587/new/

https://reviews.llvm.org/D99587