We don't actually need this for now, but eventually it would be nice to
run icount benchmarks on multiple targets. Start tagging artifact names
with the architecture, and allow passing `--tag` to `ci-util.py` in
order to retrieve the correct one.
Currently we run logspace tests for extensive tests, but there isn't any
reason we couldn't also run more kinds of tests more extensively (e.g.
more edge cases, combine edge cases with logspace for multi-input
functions, etc). As a first step toward making this possible, make
`extensive` a new field in `CheckCtx`, and rename `QuickSpaced` to
`Spaced`.
We may soon want to use some new nightly features in `compiler-builtins`
and `libm`, specifically `cfg_target_has_reliable_f16_f128` which was
added in the past few weeks. This will mean we need a newer toolchain
for benchmarks to continue building.
Bump to the current latest nightly so we are not blocked on this down
the line.
When multiple merges to `master` happen before a CI run completes, the
in-progress job is getting canceled. Fix this by using the commit sha
for the group key if a pull request number is not available, rather than
`github.ref` (which is always `refs/head/master` after merge). This
should prevent jobs running on previous commits from getting cancelled,
while still ensuring there is only ever one active run per pull request.
These are now provided by `compiler-builtins`, so there is no need to
also build the C versions. This was detected by checking for duplicate
symbols and not excluding weak symbols (like CI currently does).
As seen at [1], LLVM uses `long long` on LLP64 (to get a 64-bit integer
matching pointer size) and `long` on everything else, with exceptions
for AArch64 and AVR. Our current logic always uses an `i32`. This
happens to work because LLVM uses 32-bit instructions to check the
output on x86-64, but the GCC checks the full 64-bit register so garbage
in the upper half leads to incorrect results.
Update our return type to be `isize`, with exceptions for AArch64 and
AVR.
Fixes: https://github.com/rust-lang/compiler-builtins/issues/919
[1]: 0cf3c437c1/compiler-rt/lib/builtins/fp_compare_impl.inc (L11-L27)
These were deleted during refactoring in 0a2dc5d9 ("Combine the source
files for more generic implementations") but got added back by accident
in 54bac411 ("refactor: Move the libm crate to a subdirectory"). Remove
them again here.
This will get checked in the compiler-builtins repository. The rules are
slightly different from rust-lang/rust since `use_small_heuristics =
"Max"` tends to create difficult to read output with math-heavy code
(e.g. multiple `if` statements with numbers are small enough that they
sometimes get collapsed into a single line).
Use the Josh [1] utility to add `compiler-builtins` as a subtree, which
will allow us to stop using crates.io for updates. This is intended to
help resolve some problems when unstable features change and require
code changes in `compiler-builtins`, which sometimes gets trapped in a
bootstrap cycle.
This was done using `josh-filter` built from the r24.10.04 tag:
git fetch https://github.com/rust-lang/compiler-builtins.git 233434412fe7eced8f1ddbfeddabef1d55e493bd
josh-filter ":prefix=library/compiler-builtins" FETCH_HEAD
git merge --allow-unrelated FILTERED_HEAD
The HEAD in the `compiler-builtins` repository is 233434412f ("fix an if
statement that can be collapsed").
[1]: https://github.com/josh-project/josh
Fast path for `register_region_obligation`
If a type has no params, infer, placeholder, or non-`'static` free regions, then we can skip registering outlives obligations since the type has no components which affect lifetime checking in an interesting way.
move expensive layout sanity check to debug assertions
It is [hard to fix](https://github.com/rust-lang/rust/issues/141006#issuecomment-2883415000) the slowness in the uninhabitedness computation for very big types but we can fix the very specific case of them being called during the layout sanity checks, as described in #140944.
So this PR moves this uninhabitedness check to the other expensive layout sanity checks that are ran under `debug_assertions`.
It makes building the `lemmy_api_routes` crate's self-profile `layout_of` query go from
```
+--------------------------------------------------------+-----------+-----------------+----------+------------+---------------------------------+
| Item | Self time | % of total time | Time | Item count | Incremental result hashing time |
+--------------------------------------------------------+-----------+-----------------+----------+------------+---------------------------------+
| layout_of | 63.02s | 41.895 | 244.26s | 123703 | 50.30ms |
+--------------------------------------------------------+-----------+-----------------+----------+------------+---------------------------------+
```
on master (2m17s total), to
```
| layout_of | 330.21ms | 0.372 | 26.90s | 123703 | 53.19ms |
```
with this PR (1m15s total).
(Note that the [perf run results](https://github.com/rust-lang/rust/pull/141039#issuecomment-2884688756) below look a bit better than [an earlier run](https://perf.rust-lang.org/compare.html?start=4eca99a18eab3d4e28ed1ce3ee620d442955a470&end=c4a00993f8ee02c7565e7be652608817ea2fb97d&stat=instructions:u) I did in another PR. There may be some positive noise there, or post-merge results could differ a bit)
Since we discussed this today, r? `@compiler-errors` — and cc `@lcnr` and `@RalfJung.`
Merge mir query analysis invocations
r? `@ghost`
same thing as https://github.com/rust-lang/rust/pull/140854 just a different set of queries
Doing this in general has some bad cache coherence issues because the query caches are laid out in Vec<QueryResult> lists per query where each index refers to a DefId in the same order as we're iterating. Iterating two or more lists at the same time does have cache issues, so I want to poke a bit at it to see if we can't merge just a few of them at a time.
Initial implementation of `core_float_math`
Since [1], `compiler-builtins` makes a certain set of math symbols
weakly available on all platforms. This means we can begin exposing some
of the related functions in `core`, so begin this process here.
It is not possible to provide inherent methods in both `core` and `std`
while giving them different stability gates, so standalone functions are
added instead. This provides a way to experiment with the functionality
while unstable; once it is time to stabilize, they can be converted to
inherent.
For `f16` and `f128`, everything is unstable so we can move the inherent
methods.
The following are included to start:
* floor
* ceil
* round
* round_ties_even
* trunc
* fract
* mul_add
* div_euclid
* rem_euclid
* powi
* sqrt
* abs_sub
* cbrt
These mirror the set of functions that we have in `compiler-builtins`
since [1], with the exception of `powi` that has been there longer.
Details for each of the changes is in the commit messages.
Tracking issue: https://github.com/rust-lang/rust/issues/137578
[1]: https://github.com/rust-lang/compiler-builtins/pull/763
try-job: aarch64-gnu
tru-job: armhf-gnu
try-job: i686-msvc-1
try-job: test-various
try-job: x86_64-mingw-1
try-job: x86_64-mingw-2
Rollup of 9 pull requests
Successful merges:
- #135808 (Implement Display for ``rustc_target::callconv::Conv``)
- #137432 (Add as_ascii_unchecked() methods to char, u8, and str)
- #139103 (deduplicate abort implementations)
- #140917 (checktools.sh: fix bashism)
- #141035 (turn lld warning on old gccs into info log)
- #141118 (Enable rust-analyzer to go from query definition to the corresponding provider field)
- #141121 (Only select true errors in `impossible_predicates`)
- #141125 (check coroutines with `TypingMode::Borrowck` to avoid cyclic reasoning)
- #141131 (Make some `match`es slightly more ergonomic in `librustdoc`)
r? `@ghost`
`@rustbot` modify labels: rollup
Make some `match`es slightly more ergonomic in `librustdoc`
Bunch of small cleanups I found while working on other stuff, mostly getting rid of superfluous `*`s and `ref [mut]`s in `match`es,
makes the code less sigil/keyword-heavy, and slightly improves readability IMHO.
Also flattens a few nested `match`es.
check coroutines with `TypingMode::Borrowck` to avoid cyclic reasoning
MIR borrowck taints its output if an obligation fails. This could then cause `check_coroutine_obligations` to silence its error, causing us to not emit and actual error and ICE.
Fixes the ICE in https://github.com/rust-lang/trait-system-refactor-initiative/issues/199. It is unfortunately still a regression.
r? compiler-errors
turn lld warning on old gccs into info log
As discussed in #140964 and IRL, this PR switches the spammy warning shown unconditionally when an old gcc doesn't support `-fuse-ld=lld` and we retry linking without it, to an info debug log so we don't lose it.
r? `@Mark-Simulacrum`
Fixes#140964
checktools.sh: fix bashism
Follow-up to https://github.com/rust-lang/rust/pull/140903. Turns out `tests/{pass,panic}` only properly expands in bash, not in dash. :/
r? `@WaffleLapkin`
deduplicate abort implementations
Currently, the code for process aborts is duplicated across `panic_abort` and `std`. This PR uses `#[rustc_std_internal_symbol]` to make the `std` implementation available to `panic_abort` via the linker, thereby deduplicating the code.
Add as_ascii_unchecked() methods to char, u8, and str
This PR adds the `as_ascii_unchecked()` method to `char`, `u8`, and `str`, allowing users to convert these types to `ascii::Char`s (see #110998) in an `unsafe` context without first checking for validity. This method was already available for `[u8]`, so this PR makes the API more consistent across other types.
MIR borrowck taints its output if an obligation fails. This could then cause
`check_coroutine_obligations` to silence its error, causing us to not emit
and actual error and ICE.
