rand: inform the optimiser that indexing is never out-of-bounds.
This uses a bitwise mask to ensure that there's no bounds checking for
the array accesses when generating the next random number. This isn't
costless, but the single instruction is nothing compared to the branch.
A `debug_assert` for "bounds check" is preserved to ensure that
refactoring doesn't accidentally break it (i.e. create values of `cnt`
that are out of bounds with the masking causing it to silently wrap-
around).
Before:
test test::rand_isaac ... bench: 990 ns/iter (+/- 24) = 808 MB/s
test test::rand_isaac64 ... bench: 614 ns/iter (+/- 25) = 1302 MB/s
After:
test test::rand_isaac ... bench: 877 ns/iter (+/- 134) = 912 MB/s
test test::rand_isaac64 ... bench: 470 ns/iter (+/- 30) = 1702 MB/s
(It also removes the unsafe code in Isaac64Rng.next_u64, with a *gain*
in performance; today is a good day.)
The performance hit from these checks is significant, but unoptimized
builds are already incredibly slow. Enabling these checks results in
better test coverage since there are bots doing unoptimized builds, and
the cost is relatively small in the context of an unoptimized build.
This also allows using `JEMALLOC_FLAGS` to override the default
configure flags.
I've found that 64k is still too much and continue to see the errors as reported
in #14940. I've locally found that 32k fails, and 24k succeeds, so I've trimmed
the size down to 10000 which the included links in the added comment end up
recommending.
It sounds like the limit can still be hit with many threads in play, but I have
yet to reproduce this, so I figure we can wait until that's hit (if it's
possible) and then take action.
I've found that 64k is still too much and continue to see the errors as reported
in #14940. I've locally found that 32k fails, and 24k succeeds, so I've trimmed
the size down to 8192 which libuv happens to use as well.
It sounds like the limit can still be hit with many threads in play, but I have
yet to reproduce this, so I figure we can wait until that's hit (if it's
possible) and then take action.
This was inspired by seeing a LLVM flatline of **~600MB** when running rustc with jemalloc (each type's `t_box_` is allocated on the heap, creating a lot of fragmentation, which jemalloc can deal with, unlike glibc).
By default, 32-bit Windows executables are restricted to 2GiB of address
space even when running on 64-bit Windows when 4GiB is available.
Closes#17043
This uses a bitwise mask to ensure that there's no bounds checking for
the array accesses when generating the next random number. This isn't
costless, but the single instruction is nothing compared to the branch.
A `debug_assert` for "bounds check" is preserved to ensure that
refactoring doesn't accidentally break it (i.e. create values of `cnt`
that are out of bounds with the masking causing it to silently wrap-
around).
Before:
test test::rand_isaac ... bench: 990 ns/iter (+/- 24) = 808 MB/s
test test::rand_isaac64 ... bench: 614 ns/iter (+/- 25) = 1302 MB/s
After:
test test::rand_isaac ... bench: 877 ns/iter (+/- 134) = 912 MB/s
test test::rand_isaac64 ... bench: 470 ns/iter (+/- 30) = 1702 MB/s
(It also removes the unsafe code in Isaac64Rng.next_u64, with a *gain*
in performance; today is a good day.)
A match in callee.rs was recognizing some foreign fns as named tuple constructors. A reproducible test case for this is nearly impossible since it depends on the way NodeIds happen to be assigned in different crates.
Fixes#15913
Closes#16813
r? @nikomatsakis I feel like I should be checking more things in check_rvalues, but not sure what - I don't properly understand expr_use_visitor
This assert was likely inherited from some point, but it's not quite valid as a
no-timeout read may enter this loop, but data could be stolen by any other read
after the socket is deemed readable.
I saw this fail in a recent bors run where the assertion was tripped.
Sometimes (e.g. on Rust CI) the "expand description" text of the
collapse toggle was displayed by default, when a page is first
loaded (even though the description is expanded), because some
Content-Security-Policy settings disable inline CSS.
Setting it the style with the `.css` method allows the output to be used
in more places.
Issue can be reproduced by the following:
```
$ cat main.rs
fn main() {}
$ rustc -Z print-link-args -Lfoo -Lbar main.rs
```
Run the rustc command a few times and observe that the order of the '-L' 'foo' '-L' 'bar' options randomly changes.
Actually hit this issue in practice on Windows when specifying two -L directories to rustc, one with rust-sdl2 in it and one with the C SDL2.dll. Since Windows file systems aren't case-sensitive, gcc randomly attempted to link against the rust sdl2.dll instead of SDL2.dll if that -L directory happened to come first.
The randomness was due to addl_lib_search_paths being a HashSet. Changed it to a Vec instead which maintains the ordering.
Unsure how to test this though since it is random by nature; suggestions very welcome.
The performance hit from these checks is significant, but unoptimized
builds are already incredibly slow. Enabling these checks results in
better test coverage since there are bots doing unoptimized builds, and
the cost is relatively small in the context of an unoptimized build.
This also allows using `JEMALLOC_FLAGS` to override the default
configure flags.
Its arguments were inverted.
For instance it displayed this message:
```
warning: this extern crate syntax is deprecated. Use: extern create "foobar" as foo;
```
Instead of:
```
warning: this extern crate syntax is deprecated. Use: extern create "foo" as foobar;
```
By default, 32-bit Windows executables are restricted to 2GiB of address
space even when running on 64-bit Windows when 4GiB is available.
Closes#17043
If you browse to, say, http://doc.rust-lang.org/libc/types/os/common/posix01/struct.timeval.html , you will see the "location" window showing `libc::types::os::common::posix01`. The first element points to a crate and others point modules. This patch adds the bold attribute to the first (ie. crate) element so that it stands out more.
