This commit implements a workaround for an [upstream LLVM bug][1] where custom
sections were accidentally duplicated amongst codegen units when ThinLTO passes
were performed. This is due to the fact that custom sections for wasm are stored
as metadata nodes which are automatically imported into modules when ThinLTO
happens. The fix here is to forcibly delete the metadata node from imported
modules before LLVM has a chance to try to copy it over.
[1]: https://bugs.llvm.org/show_bug.cgi?id=38184
This commit removes a hack in our ThinLTO passes which removes available
externally functions manually. The [upstream bug][1] has long since been fixed,
so we should be able to rely on LLVM natively for this now!
[1]: https://bugs.llvm.org/show_bug.cgi?id=35736
use checked write in `LineWriter` example
The example was wrong because it didn't check the return value of
`write()`, and it didn't flush the buffer before comparing the contents
of the file.
Fixes#51621.
Don't build twice the sanitizers on Linux
This commit is an attempted fix at #50887. It was noticed that on that issue
we're building both x86_64 and i386 versions of libraries, but we only actually
need the x86_64 versions! This hopes that the build race condition exhibited
in #50887 is connected to building both architectures and/or building a lot of
libraries, so this should help us build precisely what we need and no more.
This commit stabilizes the `#[wasm_import_module]` attribute as
`#[link(wasm_import_module = "...")]`. Tracked by #52090 this new directive in
the `#[link]` attribute is used to configured the module name that the imports
are listed with. The WebAssembly specification indicates two utf-8 names are
associated with all imported items, one for the module the item comes from and
one for the item itself. The item itself is configurable in Rust via its
identifier or `#[link_name = "..."]`, but the module name was previously not
configurable and defaulted to `"env"`. This commit ensures that this is also
configurable.
Closes#52090
This commit is an attempted fix at #50887. It was noticed that on that issue
we're building both x86_64 and i386 versions of libraries, but we only actually
need the x86_64 versions! This hopes that the build race condition exhibited
in #50887 is connected to building both architectures and/or building a lot of
libraries, so this should help us build precisely what we need and no more.
rustc: Enable `use_extern_macros` in 2018 edition
This was previously enabled via `proc_macro`, but since `proc_macro` is now
stable this is no longer the case. Explicitly include it in the 2018 edition
here.
Allow clippy to be installed with make install
After #51122 clippy is available as a component but doesn't install when building from source.
This PR allows to install clippy with extended tools.
rustc: Fix a suggestion for the `proc_macro` feature
This feature is stable, we shouldn't suggest it any more! Instead suggest the
real feature, `use_extern_macros`.
Amend option.take examples
It wasn't abundantly clear to me what `.take` returned. Perhaps this is a slightly frivolous change, but I think it's an improvement. =)
Apologies if I'm not following proper procedures.
Handle array manually in str case conversion methods
Avoiding the overhead incurred from `String.extend(char.to_lowercase())` showed a notable performance improvement when I benchmarked it.
I tested on these strings:
```rust
ALL_LOWER: "loremipsumdolorsitametduosensibusmnesarchumabcdefgh"
ALL_UPPER: "LOREMIPSUMDOLORSITAMETDUOSENSIBUSMNESARCHUMABCDEFGH"
REALISTIC_UPPER: "LOREM IPSUM DOLOR SIT AMET, DUO SENSIBUS MNESARCHUM"
SIGMAS: "ΣΣΣΣΣ ΣΣΣΣΣ ΣΣΣΣΣ ΣΣΣ ΣΣΣΣ, ΣΣΣ ΣΣΣΣΣΣΣΣ ΣΣΣΣΣΣΣΣΣΣ"
WORD_UPPER: "Lorem Ipsum Dolor Sit Amet, Duo Sensibus Mnesarchum"
```
the performance improvements of `to_lowercase()` were
```
running 10 tests
test tests::all_lower ... bench: 1,752 ns/iter (+/- 49)
test tests::all_lower_new ... bench: 1,266 ns/iter (+/- 15) -28%
test tests::all_upper ... bench: 1,832 ns/iter (+/- 39)
test tests::all_upper_new ... bench: 1,337 ns/iter (+/- 18) -27%
test tests::realistic_upper ... bench: 1,993 ns/iter (+/- 14)
test tests::realistic_upper_new ... bench: 1,445 ns/iter (+/- 22) -27%
test tests::sigmas ... bench: 1,342 ns/iter (+/- 39)
test tests::sigmas_new ... bench: 1,226 ns/iter (+/- 16) -9%
test tests::word_upper ... bench: 1,899 ns/iter (+/- 12)
test tests::word_upper_new ... bench: 1,381 ns/iter (+/- 26) -27%
```
and of `to_uppercase()`
```
running 10 tests
test tests::all_lower ... bench: 1,813 ns/iter (+/- 20)
test tests::all_lower_new ... bench: 1,321 ns/iter (+/- 16) -27%
test tests::all_upper ... bench: 1,629 ns/iter (+/- 22)
test tests::all_upper_new ... bench: 1,241 ns/iter (+/- 9) -24%
test tests::realistic_upper ... bench: 1,670 ns/iter (+/- 24)
test tests::realistic_upper_new ... bench: 1,241 ns/iter (+/- 17) -26%
test tests::sigmas ... bench: 2,053 ns/iter (+/- 20)
test tests::sigmas_new ... bench: 1,753 ns/iter (+/- 23) -15%
test tests::word_upper ... bench: 1,873 ns/iter (+/- 30)
test tests::word_upper_new ... bench: 1,412 ns/iter (+/- 25) -25%
```
I gave up on the more advanced method from #52061 as it wasn't always a clear improvement and would help in even less cases if this PR was merged.
Enable default inlining in platform intrinsics
Since [#28273](https://github.com/rust-lang/rust/issues/28273) has been fixed for quite some time, it might be a good idea to return to default inlining in platform intrinsics.
This commit changes the exit status of rustc to 1 in the presence of
compilation errors. In the event of an unexpected panic (ICE) the
standard panic error exit status of 101 remains.
A run-make test is added to ensure that the exit code does not regress,
and compiletest is updated to check for an exit status of 1 or 101,
depending on the mode and suite.
This is a breaking change for custom drivers.
Fixes#51971.
rustc: Use link_section, not wasm_custom_section
This commit transitions definitions of custom sections on the wasm target from
the unstable `#[wasm_custom_section]` attribute to the
already-stable-for-other-targets `#[link_section]` attribute. Mostly the same
restrictions apply as before, except that this now applies only to statics.
Closes#51088
Avoid most allocations in `Canonicalizer`.
Extra allocations are a significant cost of NLL, and the most common
ones come from within `Canonicalizer`. In particular, `canonical_var()`
contains this code:
indices
.entry(kind)
.or_insert_with(|| {
let cvar1 = variables.push(info);
let cvar2 = var_values.push(kind);
assert_eq!(cvar1, cvar2);
cvar1
})
.clone()
`variables` and `var_values` are `Vec`s. `indices` is a `HashMap` used
to track what elements have been inserted into `var_values`. If `kind`
hasn't been seen before, `indices`, `variables` and `var_values` all get
a new element. (The number of elements in each container is always the
same.) This results in lots of allocations.
In practice, most of the time these containers only end up holding a few
elements. This PR changes them to avoid heap allocations in the common
case, by changing the `Vec`s to `SmallVec`s and only using `indices`
once enough elements are present. (When the number of elements is small,
a direct linear search of `var_values` is as good or better than a
hashmap lookup.)
The changes to `variables` are straightforward and contained within
`Canonicalizer`. The changes to `indices` are more complex but also
contained within `Canonicalizer`. The changes to `var_values` are more
intrusive because they require defining a new type
`SmallCanonicalVarValues` -- which is to `CanonicalVarValues` as
`SmallVec` is to `Vec -- and passing stack-allocated values of that type
in from outside.
All this speeds up a number of NLL "check" builds, the best by 2%.
r? @nikomatsakis
This was previously enabled via `proc_macro`, but since `proc_macro` is now
stable this is no longer the case. Explicitly include it in the 2018 edition
here.
