This commit moves all logging out of the standard library into an external
crate. This crate is the new crate which is responsible for all logging macros
and logging implementation. A few reasons for this change are:
* The crate map has always been a bit of a code smell among rust programs. It
has difficulty being loaded on almost all platforms, and it's used almost
exclusively for logging and only logging. Removing the crate map is one of the
end goals of this movement.
* The compiler has a fair bit of special support for logging. It has the
__log_level() expression as well as generating a global word per module
specifying the log level. This is unfairly favoring the built-in logging
system, and is much better done purely in libraries instead of the compiler
itself.
* Initialization of logging is much easier to do if there is no reliance on a
magical crate map being available to set module log levels.
* If the logging library can be written outside of the standard library, there's
no reason that it shouldn't be. It's likely that we're not going to build the
highest quality logging library of all time, so third-party libraries should
be able to provide just as high-quality logging systems as the default one
provided in the rust distribution.
With a migration such as this, the change does not come for free. There are some
subtle changes in the behavior of liblog vs the previous logging macros:
* The core change of this migration is that there is no longer a physical
log-level per module. This concept is still emulated (it is quite useful), but
there is now only a global log level, not a local one. This global log level
is a reflection of the maximum of all log levels specified. The previously
generated logging code looked like:
if specified_level <= __module_log_level() {
println!(...)
}
The newly generated code looks like:
if specified_level <= ::log::LOG_LEVEL {
if ::log::module_enabled(module_path!()) {
println!(...)
}
}
Notably, the first layer of checking is still intended to be "super fast" in
that it's just a load of a global word and a compare. The second layer of
checking is executed to determine if the current module does indeed have
logging turned on.
This means that if any module has a debug log level turned on, all modules
with debug log levels get a little bit slower (they all do more expensive
dynamic checks to determine if they're turned on or not).
Semantically, this migration brings no change in this respect, but
runtime-wise, this will have a perf impact on some code.
* A `RUST_LOG=::help` directive will no longer print out a list of all modules
that can be logged. This is because the crate map will no longer specify the
log levels of all modules, so the list of modules is not known. Additionally,
warnings can no longer be provided if a malformed logging directive was
supplied.
The new "hello world" for logging looks like:
#[phase(syntax, link)]
extern crate log;
fn main() {
debug!("Hello, world!");
}
This is mostly just an implementation detail, and anyone worried about the stack
bounds doesn't need to be bothered with the red zone because it's not usable
anyway.
Closes#12897
I ignored AtomicU64 methods on MIPS target
because libgcc doesn't implement MIPS32 64-bit atomic operations.
Otherwise it would cause link failure.
By the way, the patched LLVM doesn't have MIPS split stack anymore.
Should I file an issue about that?
lint: add lint for use of a `~[T]`.
This is useless at the moment (since pretty much every crate uses
`~[]`), but should help avoid regressions once completely removed from a
crate.
This commit fixes a small bug in the green scheduler where a scheduler task
calling `maybe_yield` would trip the assertion that `self.yield_check_count > 0`
This behavior was seen when a scheduler task was scheduled many times
successively, sending messages in a loop (via the channel `send` method), which
in turn invokes `maybe_yield`. Yielding on a sched task doesn't make sense
because as soon as it's done it will implicitly do a yield, and for this reason
the yield check is just skipped if it's a sched task.
I am unable to create a reliable test for this behavior, as there's no direct
way to have control over the scheduler tasks.
cc #12666, I discovered this when investigating that issue
- Added `TraitObject` representation to `std::raw`.
- Added doc to `std::raw`.
- Removed `Any::as_void_ptr()` and `Any::as_mut_void_ptr()`
methods as they are uneccessary now after the removal of
headers on owned boxes. This reduces the number of virtual calls needed from 2 to 1.
- Made the `..Ext` implementations work directly with the repr of
a trait object.
- Removed `Any`-related traits from the prelude.
- Added bench.
Bench before/after:
~~~
7 ns/iter (+/- 0)
4 ns/iter (+/- 0)
~~~
- Added `TraitObject` representation to `std::raw`.
- Added doc to `std::raw`.
- Removed `Any::as_void_ptr()` and `Any::as_mut_void_ptr()`
methods as they are uneccessary now after the removal of
headers on owned boxes. This reduces the number of virtual calls needed.
- Made the `..Ext` implementations work directly with the repr of
a trait object.
- Removed `Any`-related traits from the prelude.
- Added bench for `Any`
There's a lot of these types in the compiler libraries, and a few of the
older or private stdlib ones. Some types are obviously meant to be
public, others not so much.
This updates a number of ignore-test tests, and removes a few completely
outdated tests due to the feature being tested no longer being supported.
This brings a number of bench/shootout tests up to date so they're compiling
again. I make no claims to the performance of these benchmarks, it's just nice
to not have bitrotted code.
Closes#2604Closes#9407
Delete all the documentation from std::task that references linked
failure.
Tweak TaskBuilder to be more builder-like. .name() is now .named() and
.add_wrapper() is now .with_wrapper(). Remove .watched() and
.unwatched() as they didn't actually do anything.
- adds a `LockGuard` type returned by `.lock` and `.trylock` that unlocks the mutex in the destructor
- renames `mutex::Mutex` to `StaticNativeMutex`
- adds a `NativeMutex` type with a destructor
- removes `LittleLock`
- adds `#[must_use]` to `sync::mutex::Guard` to remind people to use it
Any single-threaded task benchmark will spend a good chunk of time in `kqueue()` on osx and `epoll()` on linux, and the reason for this is that each time a task is terminated it will hit the syscall. When a task terminates, it context switches back to the scheduler thread, and the scheduler thread falls out of `run_sched_once` whenever it figures out that it did some work.
If we know that `epoll()` will return nothing, then we can continue to do work locally (only while there's work to be done). We must fall back to `epoll()` whenever there's active I/O in order to check whether it's ready or not, but without that (which is largely the case in benchmarks), we can prevent the costly syscall and can get a nice speedup.
I've separated the commits into preparation for this change and then the change itself, the last commit message has more details.
Two unfortunate allocations were wrapping a proc() in a proc() with
GreenTask::build_start_wrapper, and then boxing this proc in a ~proc() inside of
Context::new(). Both of these allocations were a direct result from two
conditions:
1. The Context::new() function has a nice api of taking a procedure argument to
start up a new context with. This inherently required an allocation by
build_start_wrapper because extra code needed to be run around the edges of a
user-provided proc() for a new task.
2. The initial bootstrap code only understood how to pass one argument to the
next function. By modifying the assembly and entry points to understand more
than one argument, more information is passed through in registers instead of
allocating a pointer-sized context.
This is sadly where I end up throwing mips under a bus because I have no idea
what's going on in the mips context switching code and don't know how to modify
it.
Closes#7767
cc #11389
Instead, use an enum to allow running both a procedure and sending the task
result over a channel. I expect the common case to be sending on a channel (e.g.
task::try), so don't require an extra allocation in the common case.
cc #11389
The condition was the wrong direction and it also didn't take equality into
account. Tests were added for both cases.
For the small benchmark of `task::try(proc() {}).unwrap()`, this takes the
iteration time on OSX from 15119 ns/iter to 6179 ns/iter (timed with
RUST_THREADS=1)
cc #11389
Currently, a scheduler will hit epoll() or kqueue() at the end of *every task*.
The reason is that the scheduler will context switch back to the scheduler task,
terminate the previous task, and then return from run_sched_once. In doing so,
the scheduler will poll for any active I/O.
This shows up painfully in benchmarks that have no I/O at all. For example, this
benchmark:
for _ in range(0, 1000000) {
spawn(proc() {});
}
In this benchmark, the scheduler is currently wasting a good chunk of its time
hitting epoll() when there's always active work to be done (run with
RUST_THREADS=1).
This patch uses the previous two commits to alter the scheduler's behavior to
only return from run_sched_once if no work could be found when trying really
really hard. If there is active I/O, this commit will perform the same as
before, falling back to epoll() to check for I/O completion (to not starve I/O
tasks).
In the benchmark above, I got the following numbers:
12.554s on today's master
3.861s with #12172 applied
2.261s with both this and #12172 applied
cc #8341
This is in preparation for running do_work in a loop while there are no active
I/O handles. This changes the do_work and interpret_message_queue methods to
return a triple where the last element is a boolean flag as to whether work was
done or not.
This commit preserves the same behavior as before, it simply re-structures the
code in preparation for future work.
The green scheduler can optimize its runtime based on this by deciding to not go
to sleep in epoll() if there is no active I/O and there is a task to be stolen.
This is implemented for librustuv by keeping a count of the number of tasks
which are currently homed. If a task is homed, and then performs a blocking I/O
operation, the count will be nonzero while the task is blocked. The homing count
is intentionally 0 when there are I/O handles, but no handles currently blocked.
The reason for this is that epoll() would only be used to wake up the scheduler
anyway.
The crux of this change was to have a `HomingMissile` contain a mutable borrowed
reference back to the `HomeHandle`. The rest of the change was just dealing with
this fallout. This reference is used to decrement the homed handle count in a
HomingMissile's destructor.
Also note that the count maintained is not atomic because all of its
increments/decrements/reads are all on the same I/O thread.
This, the Nth rewrite of channels, is not a rewrite of the core logic behind
channels, but rather their API usage. In the past, we had the distinction
between oneshot, stream, and shared channels, but the most recent rewrite
dropped oneshots in favor of streams and shared channels.
This distinction of stream vs shared has shown that it's not quite what we'd
like either, and this moves the `std::comm` module in the direction of "one
channel to rule them all". There now remains only one Chan and one Port.
This new channel is actually a hybrid oneshot/stream/shared channel under the
hood in order to optimize for the use cases in question. Additionally, this also
reduces the cognitive burden of having to choose between a Chan or a SharedChan
in an API.
My simple benchmarks show no reduction in efficiency over the existing channels
today, and a 3x improvement in the oneshot case. I sadly don't have a
pre-last-rewrite compiler to test out the old old oneshots, but I would imagine
that the performance is comparable, but slightly slower (due to atomic reference
counting).
This commit also brings the bonus bugfix to channels that the pending queue of
messages are all dropped when a Port disappears rather then when both the Port
and the Chan disappear.
Beforehand, using a concurrent queue always mandated that the "shared state" be
stored internally to the queues in order to provide a safe interface. This isn't
quite as flexible as one would want in some circumstances, so instead this
commit moves the queues to not containing the shared state.
The queues no longer have a "default useful safe" interface, but rather a
"default safe" interface (minus the useful part). The queues have to be shared
manually through an Arc or some other means. This allows them to be a little
more flexible at the cost of a usability hindrance.
I plan on using this new flexibility to upgrade a channel to a shared channel
seamlessly.
Declare a `type SendStr = MaybeOwned<'static>` to ease readibility of
types that needed the old SendStr behavior.
Implement all the traits for MaybeOwned that SendStr used to implement.
This allows for easier static initialization of a pthread mutex, although the
windows mutexes still sadly suffer.
Note that this commit removes the clone() method from a mutex because it no
longer makes sense for pthreads mutexes. This also removes the Once type for
now, but it'll get added back shortly.
`Times::times` was always a second-class loop because it did not support the `break` and `continue` operations. Its playful appeal was then lost after `do` was disabled for closures. It's time to let this one go.
