Summary: 4-port cards

Uwe Doering gemini at geminix.in-berlin.de
Thu Apr 25 20:18:12 AEST 1991


davidg%aegis.or.jp at kyoto-u.ac.jp (Dave McLane) writes:

>gemini at geminix.in-berlin.de (Uwe Doering) writes:
>
>> I disagree. I developed FAS for NS16550A chips, and have six of them on my
>> own serial card. And obviously, FAS uses the 16 byte FIFOs. The `AF' version
>> has a faster timing, but is functionally equivalent to the `A' version. The
>> trailing `N' indicates the packaging technique used for this chip, in
>> this case a plastic body in DIL form.
>
>I'm not sure how you know that "it's obvious that FAS uses the 16 byte
>FIFOs." What indication did you get from FAS? 
>
>The reason that I ask I am running FAS 2.0.8 with both NS16550AFN 
>chips and the UARTs that are integrated with other functions on my 
>Dell 333D; FAS accepts both and works with both ... until you see 
>how much CPU time is being used. The significant reduction in time
>would seem to indicate FAS is using the FIFO buffers in the
>NS16550AFNs but not in the others....

Be assured that I did lots of performance tests during the development
of FAS. Besides the six NS16550A chips I have two NS16450 UARTs on
my 8-port card, so I could verify that with NS16550A chips FAS needs
only a small fraction of the CPU time needed for NS16450 chips. At the
time I released the first FAS version, there where no `AF' chips
on the market. So I used the `A' chips and developed FAS with the
help of the NS16550A data sheet. If there wouldn't have been a
significant performance gain with the NS16550A chips I would never
have released FAS.

If you feel that the UARTs on your main board consume too much CPU
time, then they are no NS16550A chips. At least I don't know of
any NS16550A chips that are integrated with other functions, so
I think that there are rather NS16450 or 8250 chips on the main
board. But then I don't quite understand how this led you to the
assumption that the `AF' type has another functionallity than the
`A' type of the NS16550.

      Uwe
-- 
Uwe Doering  |  INET : gemini at geminix.in-berlin.de
Berlin       |----------------------------------------------------------------
Germany      |  UUCP : ...!unido!fub!geminix.in-berlin.de!gemini



More information about the Comp.unix.sysv386 mailing list