jonhoo / bustle Goto Github PK

Trying to run with some bigger table sizes. It looks like bustle is leaking memory and getting OOM killed. Any insight on this?

Worth noting is some implementations provide different features and guarantees such as being able to share reference guards across threads which is useful for async. I think that all maps included in the benchmark should be compared in features and guarantees. This can be a short table and doesn't have to be any sizable writeup. Just something to help a user choose an implementation that provides the features and guarantees they need.

I suggest adding an option to use a Zipf distribution to sample the keys. It might give a more realistic work load for heavy contention applications as many real word datasets follow it.

The rand_distr crate has an implementation that is not yet released here.

I think examples have some problems with allocation, currently, arc_rwlock_std example fails with the message below:

memory allocation of 1879048192 bytes failedAborted (core dumped)

Since the division is floor here, the for loop doesn't run for all the "ops" operations.

Ex: ops = 2048 and op_mix.len() = 100. This creates a slice of length = 2000 and 48 operations are left out. Should this be the case?

I feel like there was an assumption that the division is ceil because there is a clause in the for loop that breaks the loop prematurely.