Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!rice!uw-beaver!ubc-cs!alberta!dvinci!weyr!p34.f200.n157.z1.FIDONET.ORG!Jim.Gilliland From: Jim.Gilliland@p34.f200.n157.z1.FIDONET.ORG (Jim Gilliland) Newsgroups: comp.os.os2 Subject: Re: Dos Mode Message-ID: <477.25CC1F81@weyr.FIDONET.ORG> Date: 2 Feb 90 12:00:47 GMT Sender: ufgate@weyr.FIDONET.ORG (newsout1.26) Organization: FidoNet node 1:157/200.34 - PC-OHIO II HST, Cleveland OH Lines: 35 PF> Do you have a 286? Switching from prot mode to real mode on a 286 PF> is a real scarey kludge. The keyboard controller resets the cpu PF> (!), part of the POST code checks some flag in memory, and skips PF> the RAM Test etc, and JMPs to some saved location. Unless you're PF> receiving characters at 2400 bps or less, you are almost certain to PF> loose characters. It might even happen at 2400. Wait a minute. I did some testing of comm overruns on an 80286 (for the OS/2 comm article I did for Byte a couple of years ago), and found that an 8MHz standard AT could keep up at 9600 bps with no loss of data - even with two other tasks running, including one in the DOS box. I found NO loss of data until I pushed the data rate to 19200, where the AT was no longer able to keep up. Certainly, it is possible to load up the machine enough to cause overruns, but you can do that on any computer. Even considering the mode-switch kludge, the 80286 (in its AT environment) does not have any innate trouble handling lower baud rates. Of course, it's also important that the communications code take proper advantage of large DosRead buffers (for the comm handle), multithreading, and appropriate use of priorities. It is quite possible for a communications program to compete with itself for resources if it is not carefully constructed. This could result in data overruns. Anyway, don't be too quick to blame the 80286. -- Jim Gilliland - via FidoNet node 1:140/22 UUCP: alberta!dvinci!weyr!157!200.34!Jim.Gilliland Internet: Jim.Gilliland@p34.f200.n157.z1.FIDONET.ORG Standard Disclaimers Apply...