Lock contention
Through tests on our new server I found that there is very high lock contention on the lock on the JumpFunctions table. This should be fixed. Somehow we need much more fine-grained locking. @StevenArzt something to think about... not sure how well this work with your "fast solver"
I just committed a lock-free version of the JumpFunctions data structure to the FastSolver in FlowDroid's unstable branch. It seems to work on my 8-core machine, but it would be great to also test on something which offers more parallelism.
The "incoming" and "endSummary" data structures still lock, this needs more work.
Great. Please email me your SSL public key (best through PGP), then I will add it to the server's authorized_hosts.