With the many tweakable knobs in sparse, it is easy to have a sub-optimal setups. As sparse already have extensive logging of actions, it would be possible to write warnings in the log together with recommendations. Examples are:
If there are more counter allocations than pool size, it is probably because there are more threads than the pool can handle. Raising the pool size solved this.
If there are many attempts of doing sparse counting that blows the tracker, the cutoff-value is probably too high. Lowering the cutoff solves this.
With the many tweakable knobs in sparse, it is easy to have a sub-optimal setups. As sparse already have extensive logging of actions, it would be possible to write warnings in the log together with recommendations. Examples are:
If there are more counter allocations than pool size, it is probably because there are more threads than the pool can handle. Raising the pool size solved this.
If there are many attempts of doing sparse counting that blows the tracker, the cutoff-value is probably too high. Lowering the cutoff solves this.