Changes the type `mir::Mir` into `mir::Body`
Fixes part 1 of #60229 (previously attempted in #60242).
I stumbled upon the issue and it seems that the previous attempt at solving it was not merged. This is a second try more up-to-date.
The commit should have changed comments as well.
At the time of writting, it passes the tidy and check tool.
The commit should have changed comments as well.
At the time of writting, it passes the tidy and check tool.
Revisions asked by eddyb :
- Renamed of all the occurences of {visit/super}_mir
- Renamed test structures `CachedMir` to `Cached`
Fixing the missing import on `AggregateKind`
Once upon a time (commit 9bd35c07c2) there were two kinds of
projection: one for places, and one for constants. It therefore made
sense to share the `Projection` struct for both. Although the different
use-cases used different concrete types, sharing was made possible by
type-parameterisation of `Projection`.
Since then, however, the usage of projections in constants has
disappeared, meaning that (forgetting lifetimes for a moment) the
parameterised type is only every instantiated under one guise. So it may
as well be a concrete type.
Place related refactors
Meanwhile I was working on Place 2 I'm finding some little things that I had on my branch but preferred to land a separate PR for things that are simpler to merge.
r? @oli-obk
Multi-variant layouts for generators
This allows generators to overlap fields using variants, but doesn't do any such overlapping yet. It creates one variant for every state of the generator (unresumed, returned, panicked, plus one for every yield), and puts every stored local in each of the yield-point variants.
Required for optimizing generator layouts (#52924).
There was quite a lot of refactoring needed for this change. I've done my best in later commits to eliminate assumptions in the code that only certain kinds of types are multi-variant, and to centralize knowledge of the inner mechanics of generators in as few places as possible.
This change also emits debuginfo about the fields contained in each variant, as well as preserving debuginfo about stored locals while running in the generator.
Also, fixes#59972.
Future work:
- Use this change for an optimization pass that actually overlaps locals within the generator struct (#52924)
- In the type layout fields, don't include locals that are uninitialized for a particular variant, so miri and UB sanitizers can check our memory (see https://github.com/rust-lang/rust/issues/59972#issuecomment-483058172)
- Preserve debuginfo scopes across generator yield points
Add a tidy check for files with over 3,000 lines
Files with a large number of lines can cause issues in GitHub (e.g. https://github.com/rust-lang/rust/issues/60015) and also tend to be indicative of opportunities to refactor into less monolithic structures.
This adds a new check to tidy to warn against files that have more than 3,000 lines, as suggested in https://github.com/rust-lang/rust/issues/60015#issuecomment-483868594. (This number was chosen fairly arbitrarily as a reasonable indicator of size.) This check can be ignored with `// ignore-tidy-filelength`.
Existing files with greater than 3,000 lines currently ignore the check, but this helps us spot when files are getting too large. (We might try to split up all files larger than this in the future, as in https://github.com/rust-lang/rust/issues/60015).
This increases the size of some important types, such as `ast::Expr` and
`mir::Statement`. However, it drastically reduces how much the interner
is used, and the fields are more natural sizes that don't require bit
operations to extract.
As a result, instruction counts drop across a range of workloads, by as
much as 12% for incremental "check" builds of `script-servo`.
Peak memory usage goes up a little for some cases, but down by more for
some other cases -- as much as 18% for non-incremental builds of
`packed-simd`.
The commit also:
- removes the `repr(packed)`, because it has negligible effect, but can
cause undefined behaviour;
- replaces explicit impls of common traits (`Copy`, `PartialEq`, etc.)
with derived ones.
Remove adt_def from projections and downcasts in MIR
As part of optimizing generator layouts in MIR, we'd like to allow downcasting generators to variants which do not have a corresponding `def_id`, since they are created by the compiler.
This refactor hopes to allow that, without regressing perf.
r? @eddyb
This commit makes two changes - separating the `NodeId` that identifies
an enum variant from the `NodeId` that identifies the variant's
constructor; and no longer creating a `NodeId` for `Struct`-style enum
variants and structs.
Separation of the variant id and variant constructor id will allow the
rest of RFC 2008 to be implemented by lowering the visibility of the
variant's constructor without lowering the visbility of the variant
itself.
No longer creating a `NodeId` for `Struct`-style enum variants and
structs mostly simplifies logic as previously this `NodeId` wasn't used.
There were various cases where the `NodeId` wouldn't be used unless
there was an unit or tuple struct or enum variant but not all uses of
this `NodeId` had that condition, by removing this `NodeId`, this must
be explicitly dealt with. This change mostly applied cleanly, but there
were one or two cases in name resolution and one case in type check
where the existing logic required a id for `Struct`-style enum variants
and structs.