bnjbvr opened PR #3401 from remove-old-regalloc to main:
There's still a lot that can be done, but that'll be enough playing for tonight :-)
This removes the old register allocation, and the entire folder
regallocthat contained related code, like live range analysis, etc. and quite a lot in binemit too (maybe that can be entirely removed too with additional effort?).
cfallin submitted PR review.
bjorn3 submitted PR review.
bjorn3 created PR review comment:
Can you add a FIXME to make this non-optional?
bnjbvr updated PR #3401 from remove-old-regalloc to main.
bnjbvr submitted PR review.
bnjbvr created PR review comment:
// FIXME: make this non optional let backend = isa.get_mach_backend().expect("only mach backends nowadays");
bnjbvr updated PR #3401 from remove-old-regalloc to main.
cfallin submitted PR review.
cfallin created PR review comment:
More generally, I expect that we'll continue to tweak the data structures to remove
Option<>s on new bits and delete the old bits; the CLIF function representation still has some bits that can go (jt_offsetsfor example?) too.
bnjbvr submitted PR review.
bnjbvr created PR review comment:
I'd expect that too, yeah. The fact that so many things were marked
pub(and notpub(crate)) make it so the compiler is not seeing all the dead code out there.
bnjbvr merged PR #3401.
bjorn3 submitted PR review.
bjorn3 created PR review comment:
Can
EncCursorbe removed entirely? It doesn't update encodings anymore.
bnjbvr submitted PR review.
bnjbvr created PR review comment:
Sounds like yes!
Last updated: Jan 10 2026 at 02:36 UTC