SignalP 6 works well enough on cloud instances and laptops etc, but it seems to take forever on Zeus.
I suspect it has something to do with the network file-system.
See if we can do something about that.
They also create a lot of files and plots we don't need without any way to turn it off.
That will be killing performance, but shouldn't be as much as what I'm seeing.
Should email them and ask for an option to disable per-protein plots etc.
It looks like the OMP implementation creates some sort of Lock, which doesn't play nicely with external parallelisation.
Reverted to OMP based threading with f9108b6 .
SignalP 6 works well enough on cloud instances and laptops etc, but it seems to take forever on Zeus. I suspect it has something to do with the network file-system.
See if we can do something about that. They also create a lot of files and plots we don't need without any way to turn it off. That will be killing performance, but shouldn't be as much as what I'm seeing. Should email them and ask for an option to disable per-protein plots etc.