Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sample.eng.ohio-state.edu!purdue!ccncsu!lamar!mglacy
From: mglacy@lamar.ColoState.EDU
Newsgroups: comp.lang.pascal
Subject: Sorting with 80x87
Message-ID: <14826@ccncsu.ColoState.EDU>
Date: 10 May 91 13:17:19 GMT
Sender: news@ccncsu.ColoState.EDU
Lines: 31

Sender: 
Reply-To: mglacy@lamar.ColoState.EDU ()
Followup-To: 
Distribution: 
Organization: Colorado State University, Fort Collins, CO  80523
Keywords:  Sorting, 80287, 80387 

I revised a program to take advantage of my 80287 co-processor.
The program generates some 10,000 values of a random variable, then sorts
them and outputs a frequency distribution for the variable. I find
that, in the {$N+} mode, the program is only marginally faster in overall
exec. time.  Everything except the sort seems to run MUCH faster, but
the sort is now a real dog.  Sorting in {N+} mode is so slow that it negates 
the time saved elsewhere in the program.

I'm using a plain-vanilla Quick Sort, which in the {$N-} state sorted
the array[1..10000] of real in a couple of seconds; when I changed to
the {$N+} state (and changed all the reals to  extended),
the sort portion of the program was MUCH slower.

I'm using TP4.0, but writing essentially standard Pascal code.

1.  Why is the sort so slow?    I wouldn't think that sorting extended values,
even though they are 10 bytes long (vs. 6 bytes for ordinary software reals),
would make so much difference.

2.  Can anyone think of a way to get around this slowdown?

Thanks,

Mike Lacy