scsi rll trade off questions?

Ralf.Brown at B.GP.CS.CMU.EDU Ralf.Brown at B.GP.CS.CMU.EDU
Tue Jul 11 21:40:48 AEST 1989


In article <2966 at cadre.dsl.PITTSBURGH.EDU>, km at cadre.dsl.PITTSBURGH.EDU (Ken Mitchum) writes:
}>Item 5 in the list above is the current biggest concern to me.  If the
}>ST0x is not a DMA device what happens when I try to run software like
}>zmodem downloads that want to do serial and disk I/O simultaneously?
}
}The ST01 does support interrupts, so there is no need for the processor
}to "wait" for a transfer to take place. DMA was probably left off the
}card 1) to save money and 2) because there is no effective way in a DOS
}environment to take advantage of it.

Another reason for leaving off the DMA is that DMA is a big lose on AT-class
or 386 machines (at least as implemented by IBM PC compatibles--1 microsecond
per transfer, as opposed to 400 ns on even a 10/0 AT using programmed I/O
with the INSW and OUTSW instructions).  In addition to being slower than
INSW/OUTSW, DMA on a PC-compatible suffers from the problem of not being able
to cross 64K boundaries, which INSW/OUTSW have no problems with (though they
are, of course, limited to at most 64K at one time).

With a true multitasking OS, however, even slow DMA would be better than
programmed I/O, since another process can run while the transfer takes place.
 
--
UUCP: {ucbvax,harvard}!cs.cmu.edu!ralf -=-=-=- Voice: (412) 268-3053 (school)
ARPA: ralf at cs.cmu.edu  BIT: ralf%cs.cmu.edu at CMUCCVMA  FIDO: Ralf Brown 1:129/46
			Disclaimer? I claimed something?
"When things start going your way, it's usually because you stopped going the
 wrong way down a one-way street."



More information about the Comp.unix.wizards mailing list