Solar Designer

I was wondering why on a system under other load, where I get a speedup at `--test --format=argon2 --tune=16`, I am only getting an OMP_SCALE of 1 (same as our...

Suggested auto-tune patch: ```diff +++ b/src/omp_autotune.c @@ -126,7 +126,7 @@ void omp_autotune_run(struct db_main *db) { struct db_salt *s = db->salts; - tune_cost = MIN(db->max_cost[0], options.loader.max_cost[0]); + tune_cost = MIN(db->real ?...

> ```diff > - tune_cost = MIN(db->max_cost[0], options.loader.max_cost[0]); > + tune_cost = MIN(db->real ? db->max_cost[0] : s->cost[0], options.loader.max_cost[0]); > ``` No, this doesn't always work right - seemed to work...

We've reset MKPC and OMP_SCALE to 1. As to benchmark hashes, there doesn't appear to be one common set of Argon2 parameters in use for password hashing. 4 MiB is...

Testing on https://3v4l.org I see that recent PHP uses: ```php

> BTW isn't the CPU Argon2 code capable of threading (for p > 1) by itself? Would it ever be beneficial to use that instead of (or together with, depending...

I'm sorry for the lack of review of this and your next PR so far. We do appreciate your contributions! @ldv-alt and I will need to discuss how we'd like...

> Looks like we'd need to use `__seccomp_filter_release` instead, or even the non-static `seccomp_filter_release` (for which we'd need a dummy task struct). Alternatively, since `__get_seccomp_filter` is currently just a `refcount_inc`,...

> Refcount overflow or hopefully saturation risk, too? I think the saturation should produce warnings in kernel messages. Just need to trigger over 2 billion credentials validations in one task...

> `struct seccomp_filter` is also not "public" We only need the `refs` field, which is the first: ``` e2cfabdfd0756 (Will Drewry 2012-04-12 16:47:57 -0500 226) struct seccomp_filter { b707ddee11d1d (Christian...