This changes the way instruments and their related information is imported
from a Soundfont. Previously, each preset zone got a complete copy of the
instrument and related information, wasting a lot of precious memory.
After this change, an instrument and related information is only
imported once and then linked into all preset zones that use this
instrument.
This change separates the static instrument zone information read from
the Soundfont from the information that gets modified on import and
later via the legato handling. It opens opens up the possibility of
having unique instruments that only get imported once and then linked
into the individual preset zones.
This change adds a new feature that enables loading and unloading of
sample data on demand. As soon as a preset is selected for a channel, all
of it's samples are loaded into memory. When a preset is unselected, all
of it's samples are unloaded from memory (unless they are being used by
another selected preset).
This feature is disabled by default and can be switched on with the
"synth.dynamic-sample-loading" setting.
This change moves the Ogg Vorbis decompression to fluid_sffile, so that
this is the only place where we have to deal with compressed audio.
It also changes the way the samples are loaded for SF3 files: previously,
the compressed data was copied into memory, then the individual samples
were decompressed (resulting in both compressed and decompressed data to
stay in memory). Also, decompressed data wasn't cached, so previous loads
of the same file ran the decompressed again for each sample.
After this change, the vorbis decompression is changed so that it reads the
compressed data directly from the Soundfont file. And the resulting WAV
data is stored in the sample cache.
Removes the need for a pre-allocated stack of fluid_preset_t's. Upon
adding a loader specific preset, a fluid_preset_t is automatically
created. defsfont and ramsfont now keep track of the fluid_preset_t's,
not the loader specific ones.
Also changes the sfont::iteration_next signature to directly return
the next preset. This is possible as presets are now only created once
and returned as a pointer.
As fluid_defsfont contains code that deals with fluid_sfont_t and
fluid_defsfont_t, fluid_preset_t and fluid_defpreset_t, it sometimes gets
very confusion to know which type is currently being accessed by "sfont"
or "preset".
Also clarify some preset zone / inst zone ambiguities.
- move sample reading to fluid_sffile
- refactor sample cache to use fluid_list and separate long functions into
smaller ones
- include sample start and count in cache key, in preparation for lazy loading
- make defsfont use new sample cache loader interface
SFSample should provide the sample pointers as specified in the Soundfont
file. If any mangling of the pointers is required, it should happen in the
defsfont loader.
They were only used in soundfont loading code, so are probably a remnant
from Smurf. The rest of the FluidSynth code doesn't use underscore
functions, so remove them here as well for consistency.
Also use single quotes in double quoted string, to remove the need for
escaping chars.
