[PATCH] D82763: MIR: Infer not-SSA for subregister defs

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/CodeGen/MIRParser/MIRParser.cpp:330
+    // FIXME: def_empty should count as an SSA violation.
+    // Subregister defs are invalid in SSA.
+    const MachineOperand *RegDef = MRI.getOneDef(Reg);
----------------
qcolombet wrote:
> Replying to the FIXME: Unclear, if Reg is a reserved reg, maybe that's fine.
> I could also be that we removed the implicit_defs.
I was working on making it illegal to remove implicit_defs in ssa at the time I did this; however the complication is implicit_defs are folded into undef operands just before phi elimination so we end up in a weird half-SSA state


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82763/new/

https://reviews.llvm.org/D82763