ppigazzini

@MJZ1977 > So basically, fishtest always tries to change all variables at the same time. SPSA = **Simultaneous** perturbation stochastic approximation > One improvement can be to take flip values...

> I don't think we're typically stuck in local minima. At least, I have never seen evidence of that. In that case (a proper implemented) SPSA should be able to...

@MJZ1977 try with a blatant wrong value eg KnightSafeCheck = 790000. If the SPSA is not able to recover a value that makes sense (eg 2000) then: - that parameter...

> that might not be such a good test... this could be so far off that e.g. the local gradient is zero, so progress won't be made. @vondele you know...

> Similarly, imagine a patch that makes an engine win more often, but needing 1000move games. These results would take a long time to arrive, and especially with many workers,...

Tested with `mongosh` and a script to reset the password to plain text. Tests passed: - password hashed with web login - password hashed with worker login - password hashed...

What about this code refactoring? ```diff 38,39c38,40 < def hash_password(self, password): < return PasswordHasher().hash(password) --- > def hash_password(self, user): > user["password"] = PasswordHasher().hash(user["password"]) > self.save_user(user) 57,58c58 < user["password"] = self.hash_password(user["password"])...

> I think 64 MB memory cost per hash is reasonable given the number of workers. If you wish, you can also check the timing performance of argon2-cffi for the...

We have a problem. During that simple test on PROD the CPU usage was huge, I reverted the code otherwise the provider will shutdown the VPS in a couple of...

Better CPU usage (half usage wrt the default parameters, see https://argon2-cffi.readthedocs.io/en/stable/api.html): ```python #default from https://argon2-cffi.readthedocs.io/en/stable/api.html #ph = PasswordHasher(time_cost: int = 3, memory_cost: int = 65536, parallelism: int = 4, hash_len:...