Re: Uniform Driver Interface (UDI)

Tim Smith (tzs@pop.nwnexus.com)
Tue, 1 Sep 1998 07:23:17 -0700 (PDT)


On Mon, 31 Aug 1998, Erik Andersen wrote:
> without the usual pain. I am a bit concerned though. One of the major areas
> where Linux shines is speed. If every device is to be run in:
> "an encapsulating environment for drivers with well-defined
> interfaces which isolate drivers from OS policies and from
> platform and I/O bus dependencies"
> then isn't this equivalent to saying that performace will suck?

Not really, in general (assuming they do UDI right...I haven't looked at the
details), because the OS side of things can be optimized to work with UDI
drivers. E.g., if UDI SCSI drivers expect ASPI (remember, I haven't read
the spec...I'm just making this up for an example) and the OS generates
CAM requests for SCSI drivers, instead of a performance-sapping CAM<-->ASPI
translation layer, one would change the OS to just generate ASPI instead.
At some level, something is going from OS-level stuff (read this file)
to driver-level stuff (fill this buffer from this disk block). As long
as the thing that turns OS-level things to driver-level things knows about
UDI, performance should be fine.

--Tim Smith

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html