The current implementation stores an array[2, float] for each param and a global params_lock.
In the perform block, if lock is acquired, the first entry of the param is used. Otherwise, the second (which is updated whenever a successful lock occurs).
Would an implementation with each param being an Atomic[float] be faster? This would need testing.
The current implementation stores an
array[2, float]
for each param and a globalparams_lock
.In the perform block, if lock is acquired, the first entry of the param is used. Otherwise, the second (which is updated whenever a successful lock occurs).
Would an implementation with each param being an
Atomic[float]
be faster? This would need testing.