This commit adds a normalization for line and column numbers in stderr
files where the line/col is from the source directory rather than
the test itself - thereby removing the need to update tests as
compiler source changes.
expand the documentation on the `Unpin` trait
provides an overview of the Pin API which the trait is for,
and show how it can be used in making self referencial structs
part of #49150
syntax: Optimize some literal parsing
Currently in the `wasm-bindgen` project we have a very very large crate that's
procedurally generated, `web-sys`. To generate this crate we parse all of a
browser's WebIDL and we then generate bindings for all of the APIs contained
within.
The resulting Rust file is 18MB large (wow!) and currently takes a very long
time to compile in debug mode. On the nightly compiler a *debug* build takes 90s
for the crate to finish. I was curious what was taking so long and upon
investigating a *massive* portion of the time was spent in the `lit_token`
method of the compiler, primarily formatting strings via `format!`.
Upon some more investigation it looks like the `byte_str_lit` was allocating an
error message once per byte, causing a very large number of allocations to
happen for large literals, of which wasm-bindgen generates quite a few (some are
MB large).
This commit fixes the issue by lazily allocating the error message, only doing
so if the error message is actually needed (which should be never). As a result,
the debug mode compilation time for our `web-sys` crate decreased from 90s to
20s, a very nice improvement! (although we've still got some work to do).
update lld submodule to include RISCV patch
This pulls in one new commit, to add support for linking static RISCV
binaries, suitable for the new riscv32imac-unknown-none-elf target.
See: https://github.com/rust-lang/lld/pull/1
Change target triple used to check for lldb in build-manifest
The wrong target triple was used for lldb in build-manifest. lldb is
only built for macOS, so update the triple to reflect that.
This is an attempt to fix bug#48168.
restore the page title after escaping out of a search
Currently if I start a search in the docs, but then hit ESC, the "Results for..." title is still there in my browser tab. This is a simple attempt to fix that. I see that there's a separate `var previousTitle = document.title` thing happening in `startSearch()`, but as far as I can tell that's only related to the back stack? I'd also appreciate feedback on the right place to declare the `titleBeforeSearch` variable.
Testing-wise, I've confirmed by hand that the tab title restores correctly after building with `./x.py doc --stage 1 src/libstd`, but nothing more involved than that. What else should I test?
Mark libserialize functions as inline
Got to thinking: "what if that big pile of tiny functions isn't inlining as it should?"
So a few `replace-regex` later the local perf run says this:
<details>
![](https://i.imgur.com/gvdJEgG.png)
</details>
Not huge, but still a win, which is interesting. Want to verify with the real perf run, but I understand there's a backlog.
I didn't notice any increase in compile time or binary sizes for rustc/libs.
Change `Rc::inc_{weak,strong}` to better hint optimization to LLVM
As discussed in #13018, `Rc::inc_strong` and `Rc::inc_weak` are changed to allow compositions of `clone` and `drop` to be better optimized. Almost entirely as in [this comment](https://github.com/rust-lang/rust/issues/13018#issuecomment-408642184), except that `abort` on zero is added so that a `drop(t.clone())` does not produce a zero check followed by conditional deallocation.
This is different from #21418 in that it doesn't rely on `assume`, avoiding the prohibitive compilation slowdown.
[Before and after IR](https://gist.github.com/hermord/266e55451b7fe0bb8caa6e35d17c86e1).
`fn resolve_legacy_scope` can now resolve only to `macro_rules!` items,
`fn resolve_lexical_macro_path_segment` is for everything else - modularized macros, preludes
Currently in the `wasm-bindgen` project we have a very very large crate that's
procedurally generated, `web-sys`. To generate this crate we parse all of a
browser's WebIDL and we then generate bindings for all of the APIs contained
within.
The resulting Rust file is 18MB large (wow!) and currently takes a very long
time to compile in debug mode. On the nightly compiler a *debug* build takes 90s
for the crate to finish. I was curious what was taking so long and upon
investigating a *massive* portion of the time was spent in the `lit_token`
method of the compiler, primarily formatting strings via `format!`.
Upon some more investigation it looks like the `byte_str_lit` was allocating an
error message once per byte, causing a very large number of allocations to
happen for large literals, of which wasm-bindgen generates quite a few (some are
MB large).
This commit fixes the issue by lazily allocating the error message, only doing
so if the error message is actually needed (which should be never). As a result,
the debug mode compilation time for our `web-sys` crate decreased from 90s to
20s, a very nice improvement! (although we've still got some work to do).
The Great Generics Generalisation: HIR Followup
Addresses the final comments in #48149.
r? @eddyb, but there are a few things I have yet to clean up. Making the PR now to more easily see when things break.
cc @yodaldevoid
Add the identity function as core::convert::identity
## New notes
This implements rust-lang/rfcs#2306 (see https://github.com/rust-lang/rust/issues/53500).
## Old notes (ignore this in new reviews)
Adds the identity function `fn id<T>(x: T) -> T { x }` to core::convert and the prelude.
Some motivations for why this is useful are explained in the doc tests.
Another is that using the identity function instead of `{ x }` or `|x| x` makes it clear that you intended to use an identity conversion on purpose.
The reasoning:
+ behind adding this to `convert` and not `mem` is that this is an identity *conversion*.
+ for adding this to the prelude is that it should be easy enough to use that the ease of writing your own identity function or using a closure `|x| x` doesn't overtake that.
I've separated this out into two feature gates so that the addition to the prelude can be considered and stabilized separately.
cc @bluss
In some profiling on OSX I saw the `write` syscall as quite high up on
the profiling graph, which is definitely not good! It looks like we're
setting the output stream of an object file as directly to a file
descriptor which means that we run the risk of doing lots of little
writes rather than a few large writes.
This commit fixes this issue by adding a buffered stream on the output,
causing the `write` syscall to disappear from the profiles on OSX.
The `Buf` vs. non-`Buf` distinction is no longer necessary, and the
nastiest code in this file can be removed.
To minimize this patch, `IdxSet` is made a typedef of `IdxSetBuf`. The
next patch will remove this typedef.
This makes it more like `AllSets::{gen,kill}_set`, removes the need for
a bunch of bitset range computations, and removes the need for `Bits`.
It's marginally less efficient, because we have to allocate one bitset
per basic block instead of one large shared bitset, but the difference
is negligible in practice.