[net.micro.pc] Weirdness in the PCjr - faster than normal 'bug'.

iverson@cory.BERKELEY.EDU (Tim Iverson) (01/25/86)

	Recently, a friend of mine said that by firing up SuperCalc and then
exiting his PCjr would perform nearly twice as fast as it did before using
SuperCalc.  Thinking this behavior to be rather odd, I ran a bench mark
program (the public domain dhrystone program distributed over net.sources)
before and after the alleged speed increase.  Well, it was faster: 195
dhrystones/sec before, and 396 dhrystones/sec after (the same as a PC).

	Needless to say, I am curious about the cause of this serendipitous
event, and if any of you happen to know the answer I would like to know why
this happens.  Please post to the net (no e-mail), as I am sure others are
interested as well.

Thanks,
Tim Iverson.

dyer@harvard.UUCP (Steve Dyer) (01/25/86)

This is an educated guess, but I assume that the Supercalc package may
have a 'terminate and stay resident' portion which then causes any
further programs to be loaded above the lower 128K of memory which
must be shared with the video display circuitry (this only holds true
if your friend's PCJr has additional memory above 128K.)  Memory in
the 0 - 128K range is much 'slower' due to the display circuitry's need
for the addition of 2 extra wait states per memory access.
-- 
/Steve Dyer
dyer@harvard.harvard.edu
harvard!dyer

revatt@aluxz.UUCP (EVATT) (01/30/86)

> 
> 	Recently, a friend of mine said that by firing up SuperCalc and then
> exiting his PCjr would perform nearly twice as fast as it did before using
> SuperCalc.  Thinking this behavior to be rather odd, I ran a bench mark
> program (the public domain dhrystone program distributed over net.sources)
> before and after the alleged speed increase.  Well, it was faster: 195
> dhrystones/sec before, and 396 dhrystones/sec after (the same as a PC).
> 
> 	Needless to say, I am curious about the cause of this serendipitous
> event, and if any of you happen to know the answer I would like to know why
> this happens.  Please post to the net (no e-mail), as I am sure others are
> interested as well.
> 
> Thanks,
> Tim Iverson.

On the PCjr with programs that run in the bottom 128K, the CPU and the
screen display must compete for memory cycle since there is no
separate screen memory.   Programs running above 128K run at same
speed as the PC (as long as they are not doing i/o like writing to
disk or screen).  The program he's running may reallocate the memory
and leave the new allocation when it exits.  Most memory expansion
units provide drivers to allocate the memory in various ways.


Robert Evatt                
AT&T Bell Labs, Allentown, PA
{ihnp4!aluxz!revatt}