Josh Stone [Thu, 28 Mar 2019 01:15:25 +0000 (18:15 -0700)]
Rollup merge of #59283 - SimonSapin:branchless-ascii-case, r=joshtriplett
Make ASCII case conversions more than 4× faster
Reformatted output of `./x.py bench src/libcore --test-args ascii` below. The `libcore` benchmark calls `[u8]::make_ascii_lowercase`. `lookup` has code (effectively) identical to that before this PR, and ~~`branchless`~~ `mask_shifted_bool_match_range` after this PR.
~~See [code comments](https://github.com/rust-lang/rust/pull/59283/commits/ce933f77c865a15670855ac5941fe200752b739f#diff-01076f91a26400b2db49663d787c2576R3796) in `u8::to_ascii_uppercase` in `src/libcore/num/mod.rs` for an explanation of the branchless algorithm.~~
**Update:** the algorithm was simplified while keeping the performance. See `branchless` v.s. `mask_shifted_bool_match_range` benchmarks.
Credits to @raphlinus for the idea in https://twitter.com/raphlinus/status/1107654782544736261, which extends this algorithm to “fake SIMD” on `u32` to convert four bytes at a time. The `fake_simd_u32` benchmarks implements this with [`let (before, aligned, after) = bytes.align_to_mut::<u32>()`](https://doc.rust-lang.org/std/primitive.slice.html#method.align_to_mut). Note however that this is buggy when addition carries/overflows into the next byte (which does not happen if the input is known to be ASCII).
This could be fixed (to optimize `[u8]::make_ascii_lowercase` and `[u8]::make_ascii_uppercase` in `src/libcore/slice/mod.rs`) either with some more bitwise trickery that I didn’t quite figure out, or by using “real” SIMD intrinsics for byte-wise addition. I did not pursue this however because the current (incorrect) fake SIMD algorithm is only marginally faster than the one-byte-at-a-time branchless algorithm. This is because LLVM auto-vectorizes the latter, as can be seen on https://rust.godbolt.org/z/anKtbR.
Benchmark results on Linux x64 with Intel i7-7700K: (updated from https://github.com/rust-lang/rust/pull/59283#issuecomment-474146863)
Josh Stone [Thu, 28 Mar 2019 01:15:19 +0000 (18:15 -0700)]
Rollup merge of #57565 - petrochenkov:turbowarn, r=Centril
syntax: Remove warning for unnecessary path disambiguators
`rustfmt` is now stable and it removes unnecessary turbofishes, so removing the warning as discussed in https://github.com/rust-lang/rust/pull/43540 (where it was introduced).
One hardcoded warning less.
bors [Wed, 27 Mar 2019 13:20:16 +0000 (13:20 +0000)]
Auto merge of #59415 - varkor:values_since_snapshot, r=eddyb
Refactor InferenceFudger (née RegionFudger)
- Rename `RegionFudger` (and related methods) to `InferenceFudger`.
- Take integer and float inference variables into account.
- Refactor `types_created_since_snapshot` and `vars_created_since_snapshot` with the [new version of ena](https://github.com/rust-lang-nursery/ena/pull/21).
- Some other refactoring in the area.
bors [Wed, 27 Mar 2019 08:58:40 +0000 (08:58 +0000)]
Auto merge of #55780 - ogoffart:span_source_text, r=petrochenkov
Introduce proc_macro::Span::source_text
A function to extract the actual source behind a Span.
Background: I would like to use `syn` in a `build.rs` script to parse the rust code, and extract part of the source code. However, `syn` only gives access to proc_macro2::Span, and i would like to get the source code behind that.
I opened an issue on proc_macro2 bug tracker for this feature https://github.com/alexcrichton/proc-macro2/issues/110 and @alexcrichton said the feature should first go upstream in proc_macro. So there it is!
Since most of the Span API is unstable anyway, this is guarded by the same `proc_macro_span` feature as everything else.
bors [Wed, 27 Mar 2019 05:25:50 +0000 (05:25 +0000)]
Auto merge of #59447 - GuillaumeGomez:rollup, r=GuillaumeGomez
Rollup of 7 pull requests
Successful merges:
- #59004 ([rustdoc] Improve "in parameters" search and search more generally)
- #59026 (Fix moving text in search tabs headers)
- #59197 (Exclude old book redirect stubs from search engines)
- #59330 (Improve the documentation for std::convert (From, Into, AsRef and AsMut))
- #59424 (Fix code block display in portability element in dark theme)
- #59427 (Link to PhantomData in NonNull documentation)
- #59432 (Improve some compiletest documentation)
Guillaume Gomez [Tue, 26 Mar 2019 21:26:42 +0000 (22:26 +0100)]
Rollup merge of #59330 - DevQps:improve-std-convert-documentation, r=steveklabnik
Improve the documentation for std::convert (From, Into, AsRef and AsMut)
# Description
In this PR I updated the documentation of From, Into, AsRef and AsMut, as well as the general std::convert module documentation. The discussion in #59163 provided information that was not yet present in the docs, or was not expressed clearly enough. I tried to clarify the examples that were already present in the docs as well as add more information about considered best-practices that came out of the discussion in #59163
@steveklabnik I hope I didn't change too much. This is an initial version! I will scan through everything tomorrow as well again to see if I made any typo's or errors, and maybe make some small changes here and there.
bors [Tue, 26 Mar 2019 13:58:30 +0000 (13:58 +0000)]
Auto merge of #59434 - Centril:bootstrap-to-2019-03-20, r=Mark-Simulacrum
Bump bootstrap compiler to 2019-03-20
Includes https://github.com/rust-lang/rust/pull/59295 and by extension https://github.com/rust-lang/rust/pull/59047, which unblocks https://github.com/rust-lang/rust/pull/58253, https://github.com/rust-lang/rust/pull/58837, and possibly https://github.com/rust-lang/rust/pull/59336, and so therefore:
Rollup merge of #59374 - faern:simplify-checked-duration-since, r=shepmaster
Simplify checked_duration_since
This follows the same design as we updated to in #56490. Internally, all the system specific time implementations are checked, no panics. Then the panicking publicly exported API can just call the checked version of itself and make do with a single panic (`expect`) at the top.
Since the internal sys implementations are now checked, this gets rid of the extra `if self >= &earlier` check in `checked_duration_since`. Except likely making the generated machine code simpler, it also reduces the algorithm from "Check panic condition -> call possibly panicking method" to just "call non panicking method".
Added two test cases:
* Edge case: Make sure `checked_duration_since` on two equal `Instant`s produce a zero duration, not a `None`.
* Most common/intended usage: Make sure `later.checked_duration_since(earlier)`, returns an expected value.
Rollup merge of #59362 - pnkfelix:demo-from-iterator-short-circuiting, r=Centril
Demo `FromIterator` short-circuiting
while looking at a FIXME in `FromIterator for Option` and `FromIterator for Result`, I realized that the current documentation does not have example code showing exactly what is meant by "no further elements are taken."
The code snippets provided here are meant to correct that.
bors [Sun, 24 Mar 2019 21:26:57 +0000 (21:26 +0000)]
Auto merge of #59382 - davidtwco:rfc-2008-refactoring, r=petrochenkov
Separate `DefId`s for variants and their constructors
Part of #44109. Split off from #59376. See [Zulip topic](https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/rfc-2008/near/132663140) for previous discussion.
bors [Sun, 24 Mar 2019 17:53:45 +0000 (17:53 +0000)]
Auto merge of #59397 - kennytm:rollup, r=kennytm
Rollup of 7 pull requests
Successful merges:
- #59213 (Track changes to robots.txt)
- #59239 (Remove inline assembly from hint::spin_loop)
- #59251 (Use a valid name for graphviz graphs)
- #59296 (Do not encode gensymed imports in metadata)
- #59328 (Implement specialized nth_back() for Box and Windows.)
- #59355 (Fix ICE with const generic param in struct)
- #59377 (Correct minimum system LLVM version in tests)
David Wood [Sun, 24 Mar 2019 17:21:59 +0000 (18:21 +0100)]
Remove `CtorOf` from `Node::Ctor`.
This commit removes `CtorOf` from `Node::Ctor` as the parent of the
constructor can be determined by looking at the node's parent in the few
places where knowing this is necessary.
kennytm [Sun, 24 Mar 2019 07:32:57 +0000 (15:32 +0800)]
Rollup merge of #59377 - smaeul:patch/system-llvm, r=nikic
Correct minimum system LLVM version in tests
Since commit 9452a8dfa3ba, the new debug info format is only generated
for LLVM 8 and newer versions. However, the tests still assume that LLVM
7 will use the new debug info format. Fix the tests (and a comment in
the code) to match the actual version check.
bors [Sun, 24 Mar 2019 14:26:11 +0000 (14:26 +0000)]
Auto merge of #58305 - scalexm:chalk-continued, r=nikomatsakis
(WIP) Small fixes in chalkification
Small fixes around region constraints and builtin impls. There are still some type inference errors, for example the following code errors out:
```rust
fn main() {
let mut x: Vec<i32> = Vec::new();
// ^^^^^^^^ cannot infer type for `std::vec::Vec<_>`
}
```
but explicitly specifying `Vec::<i32>::new` works.
With these few fixes, the following code now passes type-checking:
```rust
fn main() {
let mut x: Vec<i32> = Vec::<i32>::new();
x.push(5);
println!("{:?}", x);
}
```
I also fixed the implied bounds bug as discussed on Zulip and in https://github.com/rust-lang-nursery/chalk/pull/206
kennytm [Sun, 24 Mar 2019 07:32:52 +0000 (15:32 +0800)]
Rollup merge of #59328 - koalatux:iter-nth-back, r=scottmcm
Implement specialized nth_back() for Box and Windows.
Hi there, this is my first pull request to rust :-)
I started implementing some specializations for DoubleEndedIterator::nth_back() and these are the first two. The problem has been discussed in #54054 and nth_back() is tracked in #56995.
I'm stuck with the next implementation so I though I do a PR for the ones I'm confident with to get some feedback.
kennytm [Sun, 24 Mar 2019 07:32:49 +0000 (15:32 +0800)]
Rollup merge of #59296 - petrochenkov:stdup, r=estebank
Do not encode gensymed imports in metadata
(Unless they are underscore `_` imports which are re-gensymed on crate loading, see https://github.com/rust-lang/rust/pull/56392.)
We cannot encode gensymed imports properly in metadata and if we encode them improperly, we can get erroneous name conflicts downstream.
Gensymed imports are produced by the compiler, so we control their set, and can be sure that none of them needs being encoded for use from other crates.
A workaround that fixes https://github.com/rust-lang/rust/issues/59243.
kennytm [Sun, 24 Mar 2019 07:32:48 +0000 (15:32 +0800)]
Rollup merge of #59251 - matthewjasper:fix-graphviz, r=petrochenkov
Use a valid name for graphviz graphs
Hiridification has broken graphviz output because `HirId` has a more complex display implemetation than `NodeId`. Since the id was just used to generate a distinct identifier, we just pull out the various constituent indexed.