[PATCH] D83561: [ScheduleDAG] Move DBG_VALUEs after first term forward.

Tue Jul 14 11:31:04 PDT 2020

jmorse added a comment.

(Drive-by review) Thanks for the patch -- it  looks good, and generating legal machine code is definitely better than illegal,

> it seems like currently there is no way to insert a proper DBG_VALUEs for such registers anyways.

I agree with this -- IMHO, this patch should also make any moved DBG_VALUE an undef / $noreg DBG_VALUE. If the DBG_VALUE somehow refers to a register defined by the first terminator, then it'll definitely be wrong after being moved, and will be clobbered on the next instruction anyway. All other DBG_VALUEs can't have been attached to the terminator node, and will have been emitted before the first terminator (lines 986 of this patch to 996).

(Shameless plug -- using instruction referencing instead as per  http://lists.llvm.org/pipermail/llvm-dev/2020-February/139440.html was largely inspired by this EmitSchedule method, which is tricky to get right).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D83561/new/

https://reviews.llvm.org/D83561