Deprecate crate_type and crate_name nested inside #![cfg_attr]
This implements the proposal in https://github.com/rust-lang/rust/pull/83676#issuecomment-811213956, with a future compatibility lint imposed on usage of crate_type/crate_name inside cfg's.
This is a compromise between removing `#![crate_type]` and `#![crate_name]` completely and keeping them as a whole, which requires somewhat of a hack in rustc and is impossible to support by gcc-rust. By only removing `#![crate_type]` and `#![crate_name]` nested inside `#![cfg_attr]` it becomes possible to parse them before a big chunk of the compiler has started.
Replaces https://github.com/rust-lang/rust/pull/83676
```rust
#![crate_type = "lib"] // remains working
#![cfg_attr(foo, crate_type = "bin")] // will stop working
```
# Rationale
As it currently is it is possible to try to access the stable crate id before it is actually set, which will panic. The fact that the Session contains mutable state beyond debugging things also doesn't completely sit well with me. Especially once parallel rustc becomes the default.
I think there is currently also a cyclic dependency where you need to set the stable crate id to be able to load crates, but you need to load crates to expand proc macro attributes that may define #![crate_name] or #![crate_type]. Currently crate level proc macro attributes are unstable or completely unsupported (can't remember which), so this is not a problem, but it may become an issue in the future.
Finally if we want to add incremental compilation to macro expansion or even parsing, we need the StableCrateId to be created together with the Session or even earlier as incremental compilation determines the incremental compilation session dir based on the StableCrateId.
In LLVM 14, our current method of setting section flags to avoid
embedding the `.llvmbc` section into final compilation artifacts
will no longer work, see issue #90326. The upstream recommendation
is to instead embed the entire bitcode using module-level inline
assembly, which is what this change does.
I've kept the existing code for platforms where we do not need to
set section flags, but possibly we should always be using the
inline asm approach.
This one is a heavy `'tcx` user.
Two interesting ones:
This one had the `'tcx` declared on the function, despite the trait taking a `'tcx`:
```diff
-impl Visitor<'_> for UsedLocals {
+impl<'tcx> Visitor<'tcx> for UsedLocals {
fn visit_statement(&mut self, statement: &Statement<'tcx>, location: Location) {
```
This one use in-band for one, and underscore for the other:
```diff
-pub fn remove_dead_blocks(tcx: TyCtxt<'tcx>, body: &mut Body<'_>) {
+pub fn remove_dead_blocks<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
```
Avoid string validation in rustc_serialize, check a marker byte instead
Since the serialization format isn't self-describing we need a way to detect when encoder and decoder don't match up. But for strings it doesn't have to be utf8 validation, which currently does cost a few percent of performance.
Instead we can use a marker byte at the end to be reasonably sure that we're dealing with a string and it wasn't overwritten in some way.
Support AVR for inline asm!
A first pass at support for the AVR platform in inline `asm!`. Passes the initial compiler tests, have not yet done more complete verification.
In particular, the register classes could use a lot more fleshing out, this draft PR so far only includes the most basic.
cc `@Amanieu` `@dylanmckay`
Remove a dead code path.
It is neither documented nor can I see any way it could ever be reached.
Also, no tests fail when turning that arm into an ICE
Fix AnonConst ICE
I am not sure if this is even the correct place to fix this issue, but i went down the path where the generic args came from and i wasn't able to find a clear cause for this down there. But if anybody has a suggestion what i should do, just tell me.
This fixes: https://github.com/rust-lang/rust/issues/91267
We already use the object crate for generating uncompressed .rmeta
metadata object files. This switches the generation of compressed
.rustc object files to use the object crate as well. These have
slightly different requirements in that .rmeta should be completely
excluded from any final compilation artifacts, while .rustc should
be part of shared objects, but not loaded into memory.
The primary motivation for this change is #90326: In LLVM 14, the
current way of setting section flags (and in particular, preventing
the setting of SHF_ALLOC) will no longer work. There are other ways
we could work around this, but switching to the object crate seems
like the most elegant, as we already use it for .rmeta, and as it
makes this independent of the codegen backend. In particular, we
don't need separate handling in codegen_llvm and codegen_gcc.
codegen_cranelift should be able to reuse the implementation as
well, though I have omitted that here, as it is not based on
codegen_ssa.
This change mostly extracts the existing code for .rmeta handling
to allow using it for .rustc as well, and adjust the codegen
infrastructure to handle the metadata object file separately: We
no longer create a backend-specific module for it, and directly
produce the compiled module instead.
This does not fix#90326 by itself yet, as .llvmbc will need to be
handled separately.
When we point at a binding to suggest giving it a type, erase all the
type for ADTs that have been resolved, leaving only the ones that could
not be inferred. For small shallow types this is not a problem, but for
big nested types with lots of params, this can otherwise cause a lot of
unnecessary visual output.
This largely avoids remapping from and to the 'real' indices, with the exception
of predecessor lookup and the final merge back, and is conceptually better.
As the paper indicates, the unprocessed vertices in the DFS tree and processed
vertices are disjoint, and we can use them in the same space, tracking only the index
of the split.
This replaces the previous implementation with the simple variant of
Lengauer-Tarjan, which performs better in the general case. Performance on the
keccak benchmark is about equivalent between the two, but we don't see
regressions (and indeed see improvements) on other benchmarks, even on a
partially optimized implementation.
The implementation here follows that of the pseudocode in "Linear-Time
Algorithms for Dominators and Related Problems" thesis by Loukas Georgiadis. The
next few commits will optimize the implementation as suggested in the thesis.
Several related works are cited in the comments within the implementation, as
well.
Implement the simple Lengauer-Tarjan algorithm
This replaces the previous implementation (from #34169), which has not been
optimized since, with the simple variant of Lengauer-Tarjan which performs
better in the general case. A previous attempt -- not kept in commit history --
attempted a replacement with a bitset-based implementation, but this led to
regressions on perf.rust-lang.org benchmarks and equivalent wins for the keccak
benchmark, so was rejected.
The implementation here follows that of the pseudocode in "Linear-Time
Algorithms for Dominators and Related Problems" thesis by Loukas Georgiadis. The
next few commits will optimize the implementation as suggested in the thesis.
Several related works are cited in the comments within the implementation, as
well.
On the keccak benchmark, we were previously spending 15% of our cycles computing
the NCA / intersect function; this function is quite expensive, especially on
modern CPUs, as it chases pointers on every iteration in a tight loop. With this
commit, we spend ~0.05% of our time in dominator computation.
Also add a test case for inserting a semicolon on extern fns.
Without this fix, we got an error like this:
error: expected one of `->`, `where`, or `{`, found `}`
--> chk.rs:3:1
|
2 | fn foo()
| --- - expected one of `->`, `where`, or `{`
| |
| while parsing this `fn`
3 | }
| ^ unexpected token
Since this is inside an extern block, you're required to write function
prototypes with no body. This fixes a regression, and adds a test case
for it.
since the serialization format isn't self-describing we need a way to detect
when encoder and decoder don't match up. but that doesn't have to
be utf8 validation for strings, which does cost a few % of performance.
Instead we can use a marker byte at the end to be reasonably
sure that we're dealing with a string and it wasn't overwritten in some
way.
Stop enabling `in_band_lifetimes` in rustc_data_structures
There's a conversation started in the tracking issue about possibly unaccepting `in_band_lifetimes`, but it's used heavily in the compiler, and thus there'd need to be a bunch of PRs like this if that were to happen.
So here's one to see how much of an impact it has. For this crate, at least, it doesn't seem like in-band was a big win -- about half the places that were using it didn't even need a named lifetime.
(Oh, and I removed `nll` while I was here too, since it didn't seem needed. Let me know if I should put that back.)
r? `@petrochenkov`
