Subject: Re: shark interrupt latency
To: None <augustss@cs.chalmers.se>
From: Lennart Augustsson <augustss@cs.chalmers.se>
List: port-arm32
Date: 08/04/1998 14:36:59
I wrote:
> So I wonder:  What on earth is blocking the interrupts for over 50ms?
> This must be a bug, there should be no legitimate reason for
> doing this.  It seems to work well when I play files from
> local disk or an mfs file system so I suspects the cs driver
> or the nfs code.

I've investigated some more now.  From what I can see now it is
actually not the interrupt that is delayed, but them DMA that is not
running at the proper speed (thus delaying the interrupt).  I checked
how far the DMA had come when the interrupt arrived and it seemed
right.

Possible explanations:  
 * The sound chip is not doing DMA as it should.  This seems unlikely
   as this is goverened by a xtal frequency divided by a set number.
 * DMA is not allowed as often as the sound chip wants it.  This
   also seems unlikely to me.  I know the ISA bus is a piece of crock,
   but surely it can sustain 2 * 180 Kb/s?
 * Something else?

Any suggestions?

        -- Lennart