Moving empty destructors out of the .cpp files and into headers
allows them to be devirtualized in certain cases.
(When the compiler can't "see" a function in a header, it must largely
assume it's some black box that the linker will resolve.)
While we're at it, use C++11's `= default` to define empty virtual
desturctors for us.
For some classes (e.g., Piano), nothing is derived from it, so we can
mark the class as final and remove any explicit virtual dtor.
There are many other places where this can be done, but this is a large
enough patch as-is.
* ThreadableJob: Move from AtomicInt to std::atomic
This is the first in a series of commits that migrates away from
AtomicInt towards the C++11 standard library types.
This would allow the removal of AtomicInt and related Qt
version-specific code.
While we're at it, also make ProcessingState an `enum class` so that
it's not implicitly convertible to integers.
* LocklessAllocator: Switch from AtomicInt to std::atomic_int
If it looks like some assignments around the Compare & Swap loops went
missing, it's because compare_exchange_weak() updates the first argument
to the current value when it fails (i.e., returns false).
* BufferManager: Remove extra AtomicInt include
* MixerWorkerThread: AtomicInt to std::atomic_int
* NotePlayHandle: AtomicInt to std::atomic_int
* Remove AtomicInt.h
* Move from QAtomicPointer to std::atomic<T*>
This removes some #ifdef trickery that was being used to get
load-acquire and store-release from Qt5 and "plain" (presumably
sequentially-consistent) loads and stores from Qt4.
C++11 (and subsequent C++ standards) provide portable ways to issue
atomic hardware instructions, which allow multiple threads to load,
store, and modify integers without taking a lock. The standard also
defines a memory model that lets you express the ordering guarantees
around these atomic operations. (x86 is relatively strongly-ordered, but
many other common architectures, such as ARM, are free to reorder loads
and stores unless told not to.)
This patch removes the lock from shared_object and replaces it with the
standard thread-safe reference counting implementation used in
C++'s std::shared_ptr, Rust's std::sync::Arc, and many others.
Additional resources on the topic:
https://assets.bitbashing.io/papers/concurrency-primer.pdfhttps://www.youtube.com/watch?v=ZQFzMfHIxng
Previously lmms used themed dialogs for project saving/opening, but not
when editing settings (edit -> settings).
With this change, the settings editor also uses themed dialogs.
When exporting a project lmms performs extra tests for bad data.
The tests are for infs and nans. Switching these tests on for all
occasions as the extra performance hit would be in the order of
only ~2% and the problems, when it hits the end user, are hard to
debug and/or work around.
After testing for inf/nan we clamp the sound to +/-4.0f as sometimes
you will get large transients passing through (an issue that is currently
only present when exporting).
Fixes: #1048
Adds the function randsv, which gives you persistent upon note plays and waveforms transit in the gui.
Moves lmms exprtk submodule back to latest upstream
Instead of calling MouseMoveEvent(), the TCO's "text float" text and position are updated.
This prevents left-clicking the right edge of a resizable TCO from decreasing its size.
Also, removed an unused variable: m_oldTime
Add `PerfTime` class representing a point in CPU time and `PerfLogTimer`, used for measuring and logging a time period. Used in `ProjectRenderer::run()
* Remove DataType.
An AutomatableModel should not need to know what concrete type it is.
Use virtual methods and implement them in the derived class instead of
testing the type in the base class.
* Remove unused method
* Remove m_hasLinkedModels
We can compute it on-the-fly with very little cost, and doing so
simplifies the code.
* Remove extra 'public:'
Probably a remnant of merging master
