Repository navigation
v0.0.4
A third full review of the code, and a lighter audio thread (no crash, memory or audio-thread problem
was found; each fix below has a test that fails without it).
- Q-Links and automation cost a fraction of the CPU they did: every knob MPC wrote (a Q-Link
turned, automation playing back) made EffectForce work out every parameter again, and every write,
the crossfader's too, made it look through all of them for values to show MPC; with every module
on, automation playing cost 12% more than the same sound standing still. Now a knob costs about
what it changes. On the Force, every module on at its heaviest with automation playing went past the
CPU budget in 0.0.3 (p99 15.9% of a block) and stays within it now (13.9%); every module on alone
14.5% -> 13.7%, the looper rolling under the fader 7.5% -> 6.7%, the reverb a fifth lighter. The
sound is the same, bit for bit, but for the looper's float rounding (-116 dB). - Grain: after a jump to shorter or denser grains (a preset, a scene, a Q-Link) the cloud could
fall silent for up to 1.3 s, waiting out the old grain interval. - Presets: a preset file that won't load (empty, or half-copied onto the card) no longer stops
PREV / NEXT and the stepper at it: they step over it. - Filter: its tail was reckoned far too short at very high cutoffs, so a resonant filter a scene
switches out could be stopped while it still rang. - The crossfader's bar is redrawn when the fader goes back to where a loaded project had left it.