Newsgroups: comp.sys.transputer
From: dbc@sybus.com (David Cameron)
Subject: Re: One dumb question about DMA
Organization: Sybus Corporation, Largo, FL
Date: Wed, 6 Jul 1994 11:21:39 GMT
Message-ID: <1994Jul6.112139.23371@sybus.com>

In article <Cs9v8K.3IK@wizzy.com> andyr@wizzy.com (Andy Rabagliati) writes:
>Jan Vorbrueggen <jan@neuroinformatik.ruhr-uni-bochum.de> wrote:
>
>
>>Ah, but Andy, isn't it the case that the DMA machines indeed do stop
>>accessing memory when a hi-pri job is executing, although this isn't
>>documented?
>
>Indeed, Roger Shepherd corrected me in another post.
>
>>Anyway, the solution is to have the processes serving the hard links
>>run at hi-pri as well, which is better for the latency as well, and to
>>do the number-crunchin at lo-pri.
>>

Does this also apply to the VPC on the T9000, or will that function
independently of the main processor ?
[I know what the book says, but I become concerned when people start
talking about undocumented "features".]

Dave

