[PARPORT] low performance

From: Stephan Wefing (wefing@gmx.de)
Date: Sun Jul 02 2000 - 20:51:25 EDT

  • Next message: dmaul683@juno.com: "Re: [PARPORT] low performance"

    hi,

    i'm using an smp 2.2.10 kernel (suse 6.2) on an intel box (2x PII 266mhz, tyan
    tiger s1832dl, 512mb) with a ps printer (kyocera fs-1750) attached to the
    parport (5m cable, otherwise it is ieee-1284 compatible). to communicate with
    the parport i use the parport, parport_pc, and the lp modules. in the bios i
    selected the ecp protocol for the parport and the printer uses autodetection.
    no irq and no dma selected in /etc/conf.modules. no pnp selected, neither in
    the bios ("pnp aware os") nor in the kernel.

    everything works well for small ps files, however, when i send an 8mb ps file
    to the printer via "cat poster.ps > /dev/lp0" it takes about 90s elapsed time
    (80s cpu time) to transfer the data to the printer. thus the transfer rate is
    about 100kb/s, which is far less than expected for a "high speed" parport.
    the resulting printout is perfect, nothing is missing.

    monitoring the system during the transfer with "top" and "xosview" shows that
    after a short initial phase the "cat" needs a complete cpu. even more strange:
    other running processes start to consume an abnormal amount of cpu time as
    well (i observed a maximum of 75% for the monitoring xosview). same for the X
    server so that the mouse pointer does not readily follow movements of the
    mouse.

    i experimented a lot, but nothing changed the situation:
    - using irq=7 and/or dma=3 (in accordance with the bios). as revealed by
      /proc/parport/0/hardware and xosview, the irq is actually used during the
      transfer.
    - any reasonable combination of the possible bios parport protocol types
      (ecp, epp1.7, epp1.9, bidirectional, normal) and the printer's parport
      protocol types (auto, high speed, nibble (high), normal). for the normal
      protocol the transfer time is about 150s.
    - pnp activated in the bios and in the kernel (-> parport_probe module). no
      diagnostics in /var/log/{boot.msg,messages,warn} with "klogd -c 8".
    - using the standard suse kernel with only one cpu activated.

    i observed the same phenomenon with kernel 2.0.36 on the same box and a
    lexmark k1220 attached to it.

    any idea what's happening, what to do? is this a "normal" behavior?

    THANKS!

    stephan

    +--------------------------------------------------------------------+
    | Stephan Wefing |
    | Stiftung caesar email : wefing@gmx.de |
    | Friedensplatz 16 http : www.caesar.de |
    | D-53111 Bonn phone : +49 +228 9656 277 |
    | Germany fax : +49 +228 9656 111 |
    +--------------------------------------------------------------------+

    -- To unsubscribe, send mail to: linux-parport-request@torque.net --
    -- with the single word "unsubscribe" in the body of the message. --



    This archive was generated by hypermail 2b29 : Sun Jul 02 2000 - 20:42:57 EDT