Implements a ring buffer class for LMMS, which is designed to be flexible, efficient and thread-safe.
Due to flexible design, it supports various methods of operation:
- set delays/sizes in absolute frame values, ignoring samplerate
- set delays/sizes in milliseconds with samplerate-awareness
- multiple inputs -> single output
- single input -> multiple outputs
Efficiency is achieved by working in buffers and using memcpy/memset for audio operations, except when additive mixing is needed: then MixHelpers are used
Thread-safety is guaranteed with QMutex
Also featuring a very efficient buffer-based system for transporting sample-exact control data
Also interpolation for automations
The native Amplifier is a reference implementation for taking advantage of sample-exact data and is currently
the only one that does so, it can be used to test things out, and as documentation/example for implementing the
same elsewhere
Basically, this works as such:
- if you click shift *after* starting a note move OR after creating a new note, the note move action is switched into resize mode, so you can quickly resize the note you just created, or the note you just moved. This saves time and improves workflow - at least based on my own experience: I've always wished I could do this, this is a huge time saving when you want to quickly jot down notes of differing lengths.
- if shift is already pressed when you click, the above will not happen, because that would mess with the note copy function. Copying notes with shift-dragging works the same as before.
- note test playback is halted when you click shift while moving. This is purely because it was causing some crackling noise, probably because of the changing length of a note that is currently playing. Maybe that can be fixed later, although it's arguably better not to hear the note while resizing - it's consistent with the other resize.
- works on group of notes as well, if you start moving a group of notes and then click shift, it will go into resize. Exception is notes copied with shift-drag... for obvious reasons.
- that should be all. Testing appreciated.
We don't want to loose the settings of an effect plugin even if it's not
available and thus can't be instantiated. Therefore remember original
settings data and save them back properly.
Partly closes#733.
Since we now provide the wavetables as pre-generated files, there's no delay caused by their initialization
so we can move it to the startup of the software. I thought engine.cpp is the best place for this, it makes
conceptually more sense than main.cpp IMO.
This way each instrument that wants to use them in the future won't have to call the initialization function
separately, making things a bit easier.
These methods are used to fetch the automated value of a model at a given MidiTime
These are still untested but that shouldn't be a problem since they aren't actually used by anything yet... but I'll be doing some testing and bugfixing (if needed) for them later.
These will be an important step in making sampletracks eventually be reliably playable from any position, and more generically, being able to reliably convert MidiTime to real time. Of course they can be useful for other things too (not sure what though, yet).
- Uses existing functionality in FxMixer & FxMixerView to manipulate channels
- Instruments sending to the manipulated channels get automatically updated
- In the future I hope to implement a drag/drop functionality instead of the clunky context menu but this is a good first step until then
- Also added in a little QWhatsThis help message for the FX line, also accessible from context menu
That lmms_math thing got mixed in accidentally, but it's also a good change: always include math.h in lmms_math - that way, other parts of the software can just #include lmms_math, and won't have to #include both math.h and lmms_math, also the yet unused sinc function in it seems to need it so this prevents problems down the line
An end frame variable always has to be greater than a start frame in
order to prevent crashes due to negative frame counts being calculated
in getSampleFragment() otherwise.
Closes#629.
In order to get a uniform behaviour when starting with a new project or
opening another project, reset volume and panning of last edited note so
new placed notes have default volume and panning.
Closes#644.
The runtime-generation is still there as a fallback, and the file generation code is left in as commented-out, because
it might be needed in the future
There was some memory alignment logic inside the Mixer cpp file. To break down
the code and separate things into smaller modules, the aligned memory stuff
was crammed into a new class called MemoryHelper.
The MemoryHelper can be reused for any other aligned memory that may be needed
by another component.
Samplebuffer: use qualitysettings for interpolation (currently defaults to SINC_FASTEST on playback), also: in visualize, draw both channels instead of averaging them into one graph (otherwise, samplebuffers with counter-phase content show up as flatline)
