home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.3b1
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!rpi!psinntp!psinntp!monymsys!sonyd1.Broadcast.Sony.COM!hico2!kak
- From: kak@hico2.westmark.com (Kris A. Kugel)
- Subject: Disableing the phone to gain speed (was: Re: 3b1 Magic POKE?)
- Message-ID: <Bxrr1w.BM8@hico2.westmark.com>
- Summary: Do we have to wait for the source?
- Keywords: phone device driver 3B1 unixpc att7300 on-board modem
- Reply-To: kak@hico2.westmark.com
- Organization: High Country Software
- X-Newsreader: Tin 1.1 PL5
- References: <1992Oct16.221844.14874@cbnewsc.cb.att.com>
- Distribution: usa
- Date: Sun, 15 Nov 1992 17:53:54 GMT
- Lines: 42
-
- Craig M. Votava (cmv@looney.att.com) wrote:
- : |> My motivation for asking this is to inquire whether you
- : |> have found any "magic pokes" that can enhance the present system in
- : |> the tradition of the "serial.pat" adb script. As I recall you
- : |> folks mentioned that the 3b1 polled the status of the /dev/ph
- : |> devices every clock tick (60Hz on the 3b1), and you theorized that
- : |> a gain could be gotten by disabling this polling.
- :
- : Yes, in fact one of the guys who was on customer support for the unix-pc
- : kernel told me that a total rewrite of the phone polling code to make it
- : interrupt driven would save quite a bit of processing time.
-
- If we could get good speedups if the phone code was
- that interrupt-driven rather than polled, shouldn't
- we be able to do something even before the source code
- becomes available? (which, after all, may be never)
-
- What I THINK the questioner was asking, (and what I'd
- be interested in hearing) was if a kernel patch could
- be developed that would disable the phone handling altogether.
-
- Personally, having invested in a worldblazer, I'd rather
- my worldblazer be able to process at 1600cps, rather
- than have a second phone available at 120cps. Given that
- tradeoff, I'd just as soon disable anything having to do
- with that on-board modem, or at least prior to a new kernel
- with a less-costly modem processing scheme.
- And I'd guess the phone polling slows EVERYTHING down,
- not just the serial port.
-
- It seems to me like a patch to replace the (hypothical)
- call to the phone-status checking routine with a no-op
- would do the trick. At least it SOUNDS easy.
-
- So let me add some additional encouragement for somebody (Craig?)
- to look into this. The first step would be to see what the entry
- points into the phone polling code are, and if they are separate
- enough to disable easily. I'd bet on some (a) easily disabled
- subroutine call(s).
-
- Kris A. Kugel 908-842-2707
- hico2!kak kak@hico2.westmark.com
-