Thoughts on password hashing: use parallel threads w\ @solardiz-style large immutable ROM /cc @stouset @veroq @CodesInChaos
@bascule @solardiz @stouset @VeroQ @CodesInChaos the goal: maximize both circuit size and "good-guy" parallelism (with smaller PHF loads)
-
-
@bascule@CodesInChaos What I am currently planning to provide multi-threaded code for is ROM initialization. Faster auth service startup.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos ... where would the pre-initialized ROM come from on the user's PC? Speeding up its initialization is more relevant.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos Another use case is in a KDF on a user's own computer, where multi-threaded KDF could run much faster. However ...Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos The question is then: is reducing the latency (when the server is mostly idle) e.g. from 10 ms to 5 ms worth it?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos Sure, full server capacity is normally not reached. Extra parallelism could reduce typical latency by a few ms.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos On large auth servers, parallelism is available from concurrent requests, so little need for threads at lower levelThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@bascule@CodesInChaos That's good thinking, but I think there are currently relatively few use cases where threads+ROM would be appropriateThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.