[comp.windows.x] Sun Benchmarks

lewin@savax.UUCP (Stuart Lewin) (12/21/88)

Back around Beta-X time when we were doing some performance
measurements on X (for the Sun server), I too noticed that the
server was going in to read the i/o devices every time around the
loop. It turned out that the Sun ddx code was setting the flags
to determine if input exists and then never bothering to clear it
after the first input arrived - hence once the first device input
came along, it went and looked for input every time after that.
If the code hasn't been fixed since then, it's probably still the
case...
Stu
<lewin@savax.com or ...!decvax!savax!lewin>

jdi@sparky.UUCP (John Irwin) (12/23/88)

Your message:

    Back around Beta-X time when we were doing some performance
    measurements on X (for the Sun server), I too noticed that the
    server was going in to read the i/o devices every time around the
    loop. It turned out that the Sun ddx code was setting the flags
    to determine if input exists and then never bothering to clear it
    after the first input arrived - hence once the first device input
    came along, it went and looked for input every time after that.
    If the code hasn't been fixed since then, it's probably still the
    case...
    Stu
    <lewin@savax.com or ...!decvax!savax!lewin>
--------

This appears to work correctly in Release 3.  The SigIOHandler increments
isItTimeToYield, which causes ProcessInputEvents to run.  ProcessInputEvents
then sets isItTimeToYield back to 0.

BTW -- WhatEver happened to nIcE, EASy to TYpe lOwErCaSe nAMEs?

	-- John