Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!MAIL.PHYSICS.LSA.UMICH.EDU!SEARS
From: SEARS@MAIL.PHYSICS.LSA.UMICH.EDU
Newsgroups: comp.sys.sgi
Subject: Interrupt response time for VME hardware interrupts.
Message-ID: <8912111100.aa22709@SMOKE.BRL.MIL>
Date: 11 Dec 89 13:24:00 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 18


I'm writing a VME hardware device driver, and have encountered the following
problem.  My hardware generates an interrupt, and the service routine must
send a short out to the device to stop it's operation as soon as possible.
It appears that there is a bit of overhead in processing an interrupt, and
my service routine may not be able to respond quickly enough.  I imagine 
that this is something intrinsic to a multitasking system, but I'd like
to know if there's any way to shorten the time between when a hardware
interrupt is generated and the interrupt handler code is actually
executed.

If it helps, the system is sleeping at splvme() when the interrupt comes
in; I'm doing data logging on existing equipment, so I have to do things
according to the timing constraints of the external eqpt.

Thanks,
Robert Sears
(sears@mich.physics.lsa.umich.edu)