Add small-copy optimization for copy_from_slice
## Summary
During benchmarking, I found that one of my programs spent between 5 and 10 percent of the time doing memmoves. Ultimately I tracked these down to single-byte slices being copied with a memcopy. Doing a manual copy if the slice contains only one element can speed things up significantly. For my program, this reduced the running time by 20%.
## Background
I am optimizing a program that relies heavily on reading a single byte at a time. To avoid IO overhead, I read all data into a vector once, and then I use a `Cursor` around that vector to read from. During profiling, I noticed that `__memmove_avx_unaligned_erms` was hot, taking up 7.3% of the running time. It turns out that these were caused by calls to `Cursor::read()`, which calls `<&[u8] as Read>::read()`, which calls `&[T]::copy_from_slice()`, which calls `ptr::copy_nonoverlapping()`. This one is implemented as a memcopy. Copying a single byte with a memcopy is very wasteful, because (at least on my platform) it involves calling `memcpy` in libc. This is an indirect call when libc is linked dynamically, and furthermore `memcpy` is optimized for copying large amounts of data at the cost of a bit of overhead for small copies.
## Benchmarks
Before I made this change, `perf` reported the following for my program. I only included the relevant functions, and how they rank. (This is on a different machine than where I ran the original benchmarks. It has an older CPU, so `__memmove_sse2_unaligned_erms` is called instead of `__memmove_avx_unaligned_erms`.)
```
#3 5.47% bench_decode libc-2.24.so [.] __memmove_sse2_unaligned_erms
#5 1.67% bench_decode libc-2.24.so [.] memcpy@GLIBC_2.2.5
#6 1.51% bench_decode bench_decode [.] memcpy@plt
```
`memcpy` is eating up 8.65% of the total running time, and the overhead of dispatching to a specialized fast copy function (`memcpy@GLIBC` showing up) is clearly visible. The price of dynamic linking (`memcpy@plt` showing up) is visible too.
After this change, this is what `perf` reports:
```
#5 0.33% bench_decode libc-2.24.so [.] __memmove_sse2_unaligned_erms
#14 0.01% bench_decode libc-2.24.so [.] memcpy@GLIBC_2.2.5
```
Now only 0.34% of the running time is spent on memcopies. The dynamic linking overhead is not significant at all any more.
To add some more data, my program generates timing results for the operation in its main loop. These are the timings before and after the change:
| Time before | Time after | After/Before |
|---------------|---------------|--------------|
| 29.8 ± 0.8 ns | 23.6 ± 0.5 ns | 0.79 ± 0.03 |
The time is basically the total running time divided by a constant; the actual numbers are not important. This change reduced the total running time by 21% (much more than the original 9% spent on memmoves, likely because the CPU is stalling a lot less because data dependencies are more transparent). Of course YMMV and for most programs this will not matter at all. But when it does, the gains can be significant!
## Alternatives
* At first I implemented this in `io::Cursor`. I moved it to `&[T]::copy_from_slice()` instead, but this might be too intrusive, especially because it applies to all `T`, not just `u8`. To restrict this to `io::Read`, `<&[u8] as Read>::read()` is probably the best place.
* I tried copying bytes in a loop up to 64 or 8 bytes before calling `Read::read`, but both resulted in about a 20% slowdown instead of speedup.
Make core::fmt::Void a non-empty type.
Adding back this change that was removed from PR #36449 because it's a fix and because I immediately hit a problem with it again when I started implementing my fix for #12609.
Based on the discussion in https://github.com/rust-lang/rust/pull/37573,
it is likely better to keep this limited to std::io, instead of
modifying a function which users expect to be a memcpy.
Ultimately copy_from_slice is being a bottleneck, not io::Cursor::read.
It might be worthwhile to move the check here, so more places can
benefit from it.
During benchmarking, I found that one of my programs spent between 5 and
10 percent of the time doing memmoves. Ultimately I tracked these down
to single-byte slices being copied with a memcopy in io::Cursor::read().
Doing a manual copy if only one byte is requested can speed things up
significantly. For my program, this reduced the running time by 20%.
Why special-case only a single byte, and not a "small" slice in general?
I tried doing this for slices of at most 64 bytes and of at most 8
bytes. In both cases my test program was significantly slower.
rustdoc: link to cross-crate sources directly.
Fixes#37684 by implementing proper support for getting the `Span` of definitions across crates.
In rustdoc this is used to generate direct links to the original source instead of fragile redirects.
This functionality could be expanded further for making error reporting code more uniform and seamless across crates, although at the moment there is no actual source to print, only file/line/column information.
Closes#37870 which is also "fixes" #37684 by throwing away the builtin macro docs from libcore.
After this lands, #37727 could be reverted, although it doesn't matter much either way.
Refactor TraitObject to Slice<ExistentialPredicate>
For reference, the primary types changes in this PR are shown below. They may add in the understanding of what is discussed below, though they should not be required.
We change `TraitObject` into a list of `ExistentialPredicate`s to allow for a couple of things:
- Principal (ExistentialPredicate::Trait) is now optional.
- Region bounds are moved out of `TraitObject` into `TyDynamic`. This permits wrapping only the `ExistentialPredicate` list in `Binder`.
- `BuiltinBounds` and `BuiltinBound` are removed entirely from the codebase, to permit future non-constrained auto traits. These are replaced with `ExistentialPredicate::AutoTrait`, which only requires a `DefId`. For the time being, only `Send` and `Sync` are supported; this constraint can be lifted in a future pull request.
- Binder-related logic is extracted from `ExistentialPredicate` into the parent (`Binder<Slice<EP>>`), so `PolyX`s are inside `TraitObject` are replaced with `X`.
The code requires a sorting order for `ExistentialPredicate`s in the interned `Slice`. The sort order is asserted to be correct during interning, but the slices are not sorted at that point.
1. `ExistentialPredicate::Trait` are defined as always equal; **This may be wrong; should we be comparing them and sorting them in some way?**
1. `ExistentialPredicate::Projection`: Compared by `ExistentialProjection::sort_key`.
1. `ExistentialPredicate::AutoTrait`: Compared by `TraitDef.def_path_hash`.
Construction of `ExistentialPredicate`s is conducted through `TyCtxt::mk_existential_predicates`, which interns a passed iterator as a `Slice`. There are no convenience functions to construct from a set of separate iterators; callers must pass an iterator chain. The lack of convenience functions is primarily due to few uses and the relative difficulty in defining a nice API due to optional parts and difficulty in recognizing which argument goes where. It is also true that the current situation isn't significantly better than 4 arguments to a constructor function; but the extra work is deemed unnecessary as of this time.
```rust
// before this PR
struct TraitObject<'tcx> {
pub principal: PolyExistentialTraitRef<'tcx>,
pub region_bound: &'tcx ty::Region,
pub builtin_bounds: BuiltinBounds,
pub projection_bounds: Vec<PolyExistentialProjection<'tcx>>,
}
// after
pub enum ExistentialPredicate<'tcx> {
// e.g. Iterator
Trait(ExistentialTraitRef<'tcx>),
// e.g. Iterator::Item = T
Projection(ExistentialProjection<'tcx>),
// e.g. Send
AutoTrait(DefId),
}
```
This commit adds a new attribute that instructs the compiler to emit
target specific code for a single function. For example, the following
function is permitted to use instructions that are part of SSE 4.2:
#[target_feature = "+sse4.2"]
fn foo() { ... }
In particular, use of this attribute does not require setting the
-C target-feature or -C target-cpu options on rustc.
This attribute does not have any protections built into it. For example,
nothing stops one from calling the above `foo` function on hosts without
SSE 4.2 support. Doing so may result in a SIGILL.
This commit also expands the target feature whitelist to include lzcnt,
popcnt and sse4a. Namely, lzcnt and popcnt have their own CPUID bits,
but were introduced with SSE4.
Show multiline spans in full if short enough
When dealing with multiline spans that span few lines, show the complete span instead of restricting to the first character of the first line.
For example, instead of:
```
% ./rustc file2.rs
error[E0277]: the trait bound `{integer}: std::ops::Add<()>` is not satisfied
--> file2.rs:13:9
|
13 | foo(1 + bar(x,
| ^ trait `{integer}: std::ops::Add<()>` not satisfied
|
```
show
```
% ./rustc file2.rs
error[E0277]: the trait bound `{integer}: std::ops::Add<()>` is not satisfied
--> file2.rs:13:9
|
13 | foo(1 + bar(x,
| ________^ starting here...
14 | | y),
| |_____________^ ...ending here: trait `{integer}: std::ops::Add<()>` not satisfied
|
```
The [proposal in internals](https://internals.rust-lang.org/t/proposal-for-multiline-span-comments/4242/6) outlines the reasoning behind this.
Separate function bodies from their signatures in HIR
Also give them their own dep map node.
I'm still unhappy with the handling of inlined items (1452edc1), but maybe you have a suggestion how to improve it.
Fixes#35078.
r? @nikomatsakis
They don't implement FnLikeNode anymore, instead are handled differently
further up in the call tree. Also, keep less information (just def ids
for the args).
Setup two tasks, one of which only processes the signatures, in order to
isolate the typeck entries for signatures from those for bodies.
Fixes#36078Fixes#37720
This used to work with the rustc_clean attribute, but doesn't anymore
since my rebase; but I don't know enough about the type checking to find
out what's wrong. The dep graph looks like this:
ItemSignature(xxxx) -> CollectItem(xxxx)
CollectItem(xxxx) -> ItemSignature(xxxx)
ItemSignature(xxxx) -> TypeckItemBody(yyyy)
HirBody(xxxx) -> CollectItem(xxxx)
The cycle between CollectItem and ItemSignature looks wrong, and my
guess is the CollectItem -> ItemSignature edge shouldn't be there, but
I'm not sure how to prevent it.
bors
Martin Pool
Jeffrey Seyfried
Michael Woerister
Theodore DeRego
Alex Crichton
Guillaume Gomez
Mark-Simulacrum
Sébastien Marie
Ruud van Asseldonk
Eduard-Mihai Burtescu
Liigo Zhuang
Andrew Gallant
Doug Goldstein
Florian Diebold
Niko Matsakis