NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/arch/11947

Text File | 1992-12-26 | 1.6 KB | 32 lines

Newsgroups: comp.arch
Path: sparky!uunet!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!ira.uka.de!Sirius.dfn.de!rz.ruhr-uni-bochum.de!jan
From: [e-mail] (Jan Vorbrueggen)
Subject: Re: question about DMA devices
Message-ID: <[e-mail]>
In-reply-to: [e-mail]'s message of 24 Dec 92 01:37:25 GMT
Organization: Inst. f. Neuroinformatik, Ruhr-Universitaet Bochum, FRG
References: <[e-mail]>
    <[e-mail]>
Date: 26 Dec 92 12:58:08
Lines: 19

In article <[e-mail]> [e-mail] (Don Stokes) writes:

   The MicroVAX I addressed memory on the Qbus, in the same way it addressed
   peripherals.  DMA is no different on these systems, except that there is no
   need for mapping registers because the memory is in the same address space
   as the peripheral and can be addressed directly.

Na, Don, the mapping registers not only serve to map (eg) Unibus addresses
into (physical) system addresses, they're also doing the pagewise scatter/
gather for you. If you don't have them, you i) don't do DMA, ii) do DMA only
a page at a time (seems hardly worth the effort), iii) or have to make sure
that virtually contiguous address space is also physically contiguous, eg use
a base&bound scheme like the first PDP10s did (and some Crays do?). I think
at some point VMS added a pool allocation routine called something like
EXE$ALOPHYCNTG to allocate virtually _and_ physically contiguous memory
from nonpaged pool, in order to support the type of VAx which didn't have
mapping registers.

    Jan