Remove uniform array move MIR passes
This PR fixes a number of bugs caused by limitations of this pass
* Projections from constant indexes weren't being canonicalized
* Constant indexes from the start weren't being canonicalized (they could have different min_lengths)
* It didn't apply to non-moves
This PR makes the following changes to support removing this pass:
* ConstantIndex of arrays are now generated in a canonical form (from the start, min_length is the actual length).
* Subslices are now split when generating move paths and when checking subslices have been moved.
Additionally
* The parent move path of a projection from an array element is now calculated correctly
closes#66502
The comparison can be performed on the raw bytes, as the chars can
only match if their UTF8 encoding matches.
This avoids the `is_char_boundary` checks and translates to a straight
`u8` slice comparison which is optimized to a memcmp or inline
comparison where appropriate.
async/await: improve not-send errors, part 2
Part of #64130. Fixes#65667.
This PR improves the errors introduced in #64895 so that they have specialized messages for `Send` and `Sync`.
r? @nikomatsakis
The reason we were invoking `builtin_deref` was to enable comparisons
when the type was `&T`. For the reasons outlined in the comment, those
comparisons failed because the regions disagreed.
When we dynamically link against libLLVM.so (as opposed to statically
linking LLVM), we need libLLVM.so to be present in the stage0 sysroot,
so that stage1 tools (which are built against the stage0
compiler+sysroot) can see it at build time (when the linker is run)
See the comment in the commit for more details
This commit builds on #65501 continue to simplify the build system and
compiler now that we no longer have multiple LLVM backends to ship by
default. Here this switches the compiler back to what it once was long
long ago, which is linking LLVM directly to the compiler rather than
dynamically loading it at runtime. The `codegen-backends` directory of
the sysroot no longer exists and all relevant support in the build
system is removed. Note that `rustc` still supports a dynamically loaded
codegen backend as it did previously, it just no longer supports
dynamically loaded codegen backends in its own sysroot.
Additionally as part of this the `librustc_codegen_llvm` crate now once
again explicitly depends on all of its crates instead of implicitly
loading them through the sysroot. This involved filling out its
`Cargo.toml` and deleting all the now-unnecessary `extern crate`
annotations in the header of the crate. (this in turn required adding a
number of imports for names of macros too).
The end results of this change are:
* Rustbuild's build process for the compiler as all the "oh don't forget
the codegen backend" checks can be easily removed.
* Building `rustc_codegen_llvm` is much simpler since it's simply
another compiler crate.
* Managing the dependencies of `rustc_codegen_llvm` is much simpler since
it's "just another `Cargo.toml` to edit"
* The build process should be a smidge faster because there's more
parallelism in the main rustc build step rather than splitting
`librustc_codegen_llvm` out to its own step.
* The compiler is expected to be slightly faster by default because the
codegen backend does not need to be dynamically loaded.
* Disabling LLVM as part of rustbuild is still supported, supporting
multiple codegen backends is still supported, and dynamic loading of a
codegen backend is still supported.
rustc: allow non-empty ParamEnv's in global trait select/eval caches.
*Based on #66963*
This appears to alleviate the symptoms of #65510 locally (without fixing WF directly), and is potentially easier to validate as sound (since it's a more ad-hoc version of queries we already have).
I'm opening this PR primarily to test the effects on perf.
r? @nikomatsakis cc @rust-lang/wg-traits
Add options to --extern flag.
This changes the `--extern` flag so that it can take a series of options that changes its behavior. The general syntax is `[opts ':'] name ['=' path]` where `opts` is a comma separated list of options. Two options are supported, `priv` which replaces `--extern-private` and `noprelude` which avoids adding the crate to the extern prelude.
```text
--extern priv:mylib=/path/to/libmylib.rlib
--extern noprelude:alloc=/path/to/liballoc.rlib
```
`noprelude` is to be used by Cargo's build-std feature in order to use `--extern` to reference standard library crates.
This also includes a second commit which adds the `aux-crate` directive to compiletest. I can split this off into a separate PR if desired, but it helps with defining these kinds of tests. It is based on #54020, and can be used in the future to replace and simplify some of the Makefile tests.
Fix constant propagation for scalar pairs
We now only propagate a scalar pair if the Rvalue is a tuple with two scalars. This for example avoids propagating a (u8, u8) value when Rvalue has type `((), u8, u8)` (see the regression test). While this is a correct thing to do, implementation is tricky and will be done later.
Fixes#66971Fixes#66339Fixes#67019
Optimize Ord trait implementation for bool
Casting the booleans to `i8`s and converting their difference into `Ordering` generates better assembly than casting them to `u8`s and comparing them.
Fixes#66780
#### Comparison([Godbolt link](https://rust.godbolt.org/z/PjBpvF))
##### Old assembly:
```asm
example::boolean_cmp:
mov ecx, edi
xor ecx, esi
test esi, esi
mov eax, 255
cmove eax, ecx
test edi, edi
cmovne eax, ecx
ret
```
##### New assembly:
```asm
example::boolean_cmp:
mov eax, edi
sub al, sil
ret
```
##### Old LLVM-MCA statistics:
```
Iterations: 100
Instructions: 800
Total Cycles: 234
Total uOps: 1000
Dispatch Width: 6
uOps Per Cycle: 4.27
IPC: 3.42
Block RThroughput: 1.7
```
##### New LLVM-MCA statistics:
```
Iterations: 100
Instructions: 300
Total Cycles: 110
Total uOps: 500
Dispatch Width: 6
uOps Per Cycle: 4.55
IPC: 2.73
Block RThroughput: 1.0
```
Rollup of 9 pull requests
Successful merges:
- #66377 (Update RELEASES.md for 1.40.0)
- #67134 (Ensure that we get a hard error on generic ZST constants if their bod…)
- #67152 (Sort auto trait and blanket implementations display)
- #67154 (Fix typos in src/libcore/alloc.rs docs)
- #67168 (corrected comment in E0478)
- #67178 (Move non clean impls items)
- #67180 (doc: Use .copied() instead of .cloned() in Vec example)
- #67181 (Update hashmap doc)
- #67193 (In which we start tracking polonius in `-Z self-profile`)
Failed merges:
r? @ghost
Currently, the type `struct S { x: u32, y: u32, tag: () }` is
incorrectly described like this:
```
print-type-size type: `S`: 8 bytes, alignment: 4 bytes
print-type-size field `.x`: 4 bytes
print-type-size field `.tag`: 0 bytes, offset: 0 bytes, alignment: 1 bytes
print-type-size padding: 4 bytes
print-type-size field `.y`: 4 bytes, alignment: 4 bytes
```
Specifically:
- The `padding` line is wrong. (There is no padding.)
- The `offset` and `alignment` on the `.tag` line shouldn't be printed.
The problem is that multiple fields can end up with the same offset, and
the printing code doesn't handle this correctly.
This commit fixes it by adjusting the field sorting so that zero-sized fields
are dealt with before non-zero-sized fields. With that in place, the
printing code works correctly.
The commit also corrects the "something is very wrong" comment.
The new output looks like this:
```
print-type-size type: `S`: 8 bytes, alignment: 4 bytes
print-type-size field `.tag`: 0 bytes
print-type-size field `.x`: 4 bytes
print-type-size field `.y`: 4 bytes
```
Update RLS and Rustfmt
Fixes#66885
Rustfmt is bumped here to version 1.4.11
~For now we pull two versions of rustfmt in Cargo.toml (RLS uses 1.4.11, rustfmt source only has 1.4.10), so~
1. ~it'd be great to upload the source used to publish rustfmt 1.4.11 @topecongiro~
2. ~I can downgrade the rustfmt used in RLS (it was an oversight on my part when merging https://github.com/rust-lang/rls/pull/1598)~
From a `Vec<Ty>` to a `Vec<InferTy>`, because that's a more restrictive
type. This is a perf win because the ultra-hot function
`shallow_resolve_changed` has less pattern-matching to do.
