RE: conduit load latency

[Vadim Strizhevsky <vadim optonline net>]

Carlos Morgado writes:
 > 
 > I just experienced a situation where high system load (actually everything was
 > running verrryyy slow due to interrupt flooding) caused gpilotd to abort just
 > after 
 > gpilotd-ERROR **: Unable to open conduit!
 > I believe this would be the backup conduit.
 > 
 > >From a user standpoint it looks like the delay in the load caused gpilotd to
 > assume it failed. From memory i'd say it uses dlopen() whom doesn't care less
 > about getting stoped in a mmap by a swapin. 
 > 
 > anyone has a light ? :)

I can't comment as to why it takes long to load the conduit, but it
certainly does. Even under normal conditions. Not sure why it failed
in your case though, perhaps malloc really ran out of memory? Anyway
just wanted to say that in the future we're planing to change the
loading strategy to load conduit only once in the begining. This
should improve the performance.

-Vadim