Re: [PARPORT] Low ECP speed

From: Fred Barnes (frmb2@ukc.ac.uk)
Date: Mon May 28 2001 - 05:27:00 EDT

  • Next message: Fernando Ruza: "[PARPORT] Module parport_lowlevel."

    You write:
    >
    > I have the same problem!?
    > The write speed is about 400 KB/s but read speed is very low (only 70 KB/s).

    Yup. Writes will be done using DMA or programmed I/O if it can. Reads revert
    to the software-emulated mode at the moment (due to the brokenness of the
    hardware ECP reading code -- this is something I'm attempting to fix, albeit
    slowly atm..).

    > But how can I found out if I am using the ECP hardware or only software emulation of ECP?

    The difference in speed is usually the best indication. I've only ever
    seen around 70 k/s for software-emulated stuff. The best hardware speed
    I've seen so far was around 800 k/s in EPP mode (both directions, using the
    PARPORT_EPP_FAST flag (PP_SETFLAGS/PP_GETFLAGS for ppdev).

    > I heard about a speed of 1 MB/s, how can I reach it?

    Dunno. In some cases (zip drive for example) the driver can rely on certain
    assumptions, such as the device always being ready to accept data, and a known
    amount of data always being available in the reverse direction. The avoidance
    of error-checking (timeouts, etc.) really speeds things up in this case.

    After getting a digital logic analyser on the parport lines once, it was
    discovered that the device was willing to go *much* faster (probably up to
    2.5 m/s in ECP), but the PC was restricting it.. We never conclusively
    discovered what the cause was (could be any number of things, but probably
    related to a crappy (standard) parport). I would hope that some PCI cards
    might be able to deliver faster data rates (iirc, warp9 (www.fapo.com) do
    some PCI cards, and their chips are really rather good).

    -- Fred

    -- 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 : Mon May 28 2001 - 06:07:57 EDT