dale@wucs1.wustl.edu (Dale Frye) (12/22/89)
I was given a program to install on an ATT 6386 WGS. the program
was written by a professor (Mech. Eng.) using Quick Basic. The
program does a great deal of computations. Running in native DOS
the program will invert and do some calculations on a test matrix
in 20 seconds. Under Simultask (VP/ix) it takes almost 5 minutes.
We have concluded that QB is checking for and not finding the 80837.
Turbo Pascal, AutoCad and Ansys run fine. (Ansys won't run without
a math coprocessor)
Any ideas or suggestions?
P.S. I expect a slowdown of 30% running under SimulTask. This isn't
even close.
P.P.S. SimulTask allows DOS to run as a task under UNIX on 386 boxes.
Dale Frye
Washington University in St. Louisfredex@cg-atla.UUCP (Fred Smith) (12/23/89)
In article <1989Dec21.234703.15536@cec1.wustl.edu> dale@wucs1.wustl.edu (Dale Frye) writes: >I was given a program to install on an ATT 6386 WGS. the program >was written by a professor (Mech. Eng.) using Quick Basic. The >program does a great deal of computations. Running in native DOS >the program will invert and do some calculations on a test matrix >in 20 seconds. Under Simultask (VP/ix) it takes almost 5 minutes. >We have concluded that QB is checking for and not finding the 80837. > >Any ideas or suggestions? > >Dale Frye >Washington University in St. Louis Dale et al: You can attempt to verify your conclusion by setting an environment variable named NO87=(any value you like) then running your program. Do this OUTSIDE the SimulTask enviornment. The presence of this variable should cause the qB runtime to assume that there is no math coprocessor availalbe, even if there is one. You should be able then to approximately duplicate the performance measurements you got while using SimulTask, if in fact the problem IS as you guess. As to how to fixit, I am afraid I can't help there! Good luck! Fred