Translate shims using MIR #39628
Translate shims using MIR #39628
Conversation
|
r? @pnkfelix (rust_highfive has picked a reviewer for you, use r? to override) |
|
☔ The latest upstream changes (presumably #39586) made this pull request unmergeable. Please resolve the merge conflicts. |
|
Hi @arielb1, |
|
Some functions in Rust, such as enum variant constructions (e.g. |
|
cc @rust-lang/compiler |
|
@arielb1 to what end? what you're doing and why you're doing it are equally important. Your commit messages don't provide information on your rationale 🙁 |
|
@mrhota less special-casing and less convoluted code in the compiler |
|
It also pulls trans-specific code out into MIR in a way that will work for alternative backends (e.g. miri, mir2wasm), reducing duplication between them. |
src/librustc/ty/instance.rs
Outdated
| #[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)] | ||
| pub enum InstanceDef<'tcx> { | ||
| Item(DefId), | ||
| // <fn() as FnTrait>::call_* |
There was a problem hiding this comment.
A more explicit comment would be nice -- It's not clear to me what the Ty<'tcx> represents here.
|
☔ The latest upstream changes (presumably #40091) made this pull request unmergeable. Please resolve the merge conflicts. |
arielb1
force-pushed
the
shimmir
branch
2 times, most recently
from
f52b583
to
6f607d6
Compare
| Deep | ||
| } | ||
|
|
||
| pub trait DropElaborator<'a, 'tcx: 'a> : fmt::Debug { |
There was a problem hiding this comment.
Why is this file not in librustc_mir/transform?
There was a problem hiding this comment.
Because it's not a MIR pass.
src/librustc/ty/context.rs
Outdated
| pub shim_mir_cache: RefCell<FxHashMap<ty::InstanceDef<'tcx>, | ||
| &'tcx RefCell<Mir<'tcx>>>>, | ||
| pub mir_shimmer: RefCell<Option<Box<ty::MirShimmer<'tcx>>>>, | ||
|
|
There was a problem hiding this comment.
Can you put a FIXME on these to eventually integrate them better with the rest of on-demand?
| @@ -2302,6 +2312,22 @@ impl<'a, 'gcx, 'tcx> TyCtxt<'a, 'gcx, 'tcx> { | |||
| queries::mir::get(self, DUMMY_SP, did).borrow() | |||
| } | |||
|
|
|||
| /// Return the possibly-auto-generated MIR of a (DefId, Subst) pair. | |||
| pub fn instance_mir(self, instance: ty::InstanceDef<'gcx>) | |||
src/librustc/ty/context.rs
Outdated
| @@ -423,6 +423,8 @@ pub struct GlobalCtxt<'tcx> { | |||
| pub hir: hir_map::Map<'tcx>, | |||
| pub maps: maps::Maps<'tcx>, | |||
|
|
|||
| // This is not a ty::maps cache because it contains Tys, instead | |||
| // it is structured similarly to selection_cache. | |||
There was a problem hiding this comment.
But if you have a dep_node method, you can use it to move this to on-demand.
arielb1
force-pushed
the
shimmir
branch
2 times, most recently
from
7d069ec
to
bd92573
Compare
|
☔ The latest upstream changes (presumably #39648) made this pull request unmergeable. Please resolve the merge conflicts. |
|
Now ready (except for cleanup, and I prefer to land it now given that this PR is rather big enough already). r? @eddyb |
arielb1
force-pushed
the
shimmir
branch
3 times, most recently
from
040f333
to
9f5def1
Compare
|
I can't see the end of that appveyor log. |
|
sccache failure @bors retry |
|
⌛ Testing commit 5dc8548 with merge 35617fc... |
|
💔 Test failed - status-appveyor |
|
⌛ Testing commit 5dc8548 with merge 0b1ddef... |
|
Mac builders offline @bors retry |
|
⌛ Testing commit 5dc8548 with merge 5e30f81... |
|
💔 Test failed - status-travis |
Translate shims using MIR This removes one large remaining part of old trans.
|
Mac test timeout? @bors retry |
|
☀️ Test successful - status-appveyor, status-travis |
This removes one large remaining part of old trans.