re: xft issues.

From: Martin Sevior (msevior@physics.unimelb.edu.au)
Date: Thu Jul 18 2002 - 15:15:25 EDT

  • Next message: Martin Sevior: "Revision Mark bugs."

    On Fri, 2002-07-19 at 04:37, Joaqun Cuenca Abela wrote:
    > On Thu, 2002-07-18 at 19:10, Martin Sevior wrote:
    > >
    > > Next when we actually load a *.abw file it is also very very slow.
    >
    > I'm done a lame benchmark. Loading a 205 pages document, with arial
    > bold, and Times New Roman normal, italic and bold, everything a bit
    > mixed in a usual arial for title times variants for body (a paragraph)
    > and everything repeated until I fill the 205 pages.
    >

    Actually we were getting terrible performance with 1 page docs. This is
    clearly different from what you're seeing. Unfortunately I did not keep
    a record of the pages of and pages of UT_DEBUG out put. Basically it w
    looping in xap_UnixFont.cpp and never getting a hit in the font cache.
    For some reason the font cache just wasn't working.

    > head abi with xft and profiling data activated load it in 2m 15s
    > stable abi (without xft nor profiling) load it in 1m 45s
    >
    > While I agree that it sucks, I just want to double check that the 28%
    > increment (I don't know if the profiling stuff is there for something,
    > but I don't think so) is that "very very slow" or if you're speaking of
    > something more visible.
    >

    This is not so bad. What we were seeing was about 100000% percent worse.

    > If it's something more visible, can you please send me a copy of the
    > file that it's giving you problems?
    >

    Sure here it is!

    Cheers

    Martin



    This archive was generated by hypermail 2.1.4 : Thu Jul 18 2002 - 15:20:49 EDT