We now divide each scanline into prologue, aligned run, and epilogue.
Basically, we draw enough pixels one-by-one until we reach a grid
intersection. Then we draw full grid cell slices using fast memory
fills. Finally we go back to one-by-one for the epilogue.
This is roughly 2.5x faster in a microbenchmark and no longer dominates
the ImageViewer and PixelPaint resizing profiles.
The call to enumerable_own_property_names in the non-array case was
missing an exception check.
Fixes 1 test262 case (JSON/parse/reviver-object-own-keys-err.js)
This requires variables that should be exported to the script host
or to other scripts to be declared as var (such as in test-common.js),
since lexically-scoped variables won't be visible.
This was an old SerenityOS-specific syscall for donating the remainder
of the calling thread's time-slice to another thread within the same
process.
Now that Threading::Lock uses a pthread_mutex_t internally, we no
longer need this syscall, which allows us to get rid of a surprising
amount of unnecessary scheduler logic. :^)
This class was previously a spinlock that would call sys$donate()
to donate its timeslice to whichever thread was holding the lock.
Now that pthread_mutex_t has a fast path, let's implement Lock on top
of that instead and get rid of the last remaining user of sys$donate().
- Fix evaluation order: IsArray(O) should always be called and before
Get(O, @@toStringTag), previously it was the other way around and
IsArray would only be called if @@toStringTag is not a string
- Add missing exception checks to both function calls
- Add missing builtin tag for arguments object
Also, while we're here:
- Update variable names to match spec
- Add spec step comments
This implementation does not use locking or condition variables
internally; it's purely based on atomics and futexes.
Notably, concurrent sem_wait() and sem_post() calls can run *completely
in parallel* without slowing each other down, as long as there are empty
slots for them all to succeed without blocking.
Additionally, sem_wait() never executes an atomic operation with release
ordering, and sem_post() never executes an atomic operation with acquire
ordering (unless you count the syscall). This means the compiler and the
hardware are free to reorder code *into* the critical section.
This implementation features a fast path for pthread_cond_signal() and
pthread_cond_broadcast() for the case there's no thread waiting, and
does not exhibit the "thundering herd" issue in
pthread_cond_broadcast().
Fixes https://github.com/SerenityOS/serenity/issues/8432
This is a private function that locks the lock much like the regular
pthread_mutex_lock(), but causes the corresponding unlock operation to
always assume there may be other waiters. This is useful in case some
waiters are made to wait on the mutex's futex directly, without going
through pthread_mutex_lock(). This is going to be used by the condition
variable implementation in the next commit.
pthread_mutex is now an actual "sleeping" mutex, and not just a
spinlock! It still has a fast path that only uses atomics and (in the
successful case) returns immediately without sleeping. In case of
contention, it calls futex_wait(), which lets the kernel scheduler put
this thread to sleep, *and* lets it know exactly when to consider
scheduling it again.
These are convinient wrappers over the most used futex operations.
futex_wait() also does some smarts for timeout and clock handling.
Use the new futex_wait() instead of a similar private helper in
LibPthread.
async_enqueue() is a wrapper over the async_enqueue_buffer() call
to AudioServer. This allows users to asyncronously enqueue audio
samples, when the program requires non-blocking audio streaming.
This also makes ClientConnection use east-const everywhere.
We don't know what is a good time to wait after an audio buffer fails to
be processed by AudioServer. However, it seems like decreasing the wait
time to 10ms after such a failure should improve latency and has not
caused issues in my testing. After all, 10ms is quite some time in audio
sample magnitudes.
We need to move the result out of the BackgroundAction object before
posting the completion callback as there is a chance the
BackgroundAction instance gets freed before the event loop runs our
callback.
Fixes#7641
If calling the executor function throws an exception, the return value
of `vm.call()` will be an empty value, which we then passed as an
argument to the reject function, which is incorrect - what it actually
needs is the exception value. This stems from a misunderstanding of the
spec I had at the time of implementing this - in their case, the
exception value is part of the completion record returned by Call().
This error was previously masked as we would use a fallback
(`value_or(js_undefined())` for the empty value argument, but that was
removed in 57f7e6e.
Fixes#8447.
As the name implies (and the spec confirms), this is only ever going to
be an object or "nothing", or "undefined" in the spec. By taking this
literally and updating a check to check for `is_undefined()`, we
introduced a bug - the value was still initialized as an empty value.
Instead, use a pointer to an Object - either we have one, or we don't.
Fixes#8448.
Non-RangeError exceptions can be thrown by user implementations of
valueOf (which are called by to_index), and the specification
disallows changing the type of the thrown error.
