PerfTest V1.1

  This is version 1.1 of my hard disk PERFormance TEST.  It is a CLI based
program, so don't bother trying to use it from WorkBench.
  This program was written because I needed some way of evaluating hard
drive performance.  Basically, I wanted a program that would see just how
fast a hard disk & controller would transfer raw data, and I wanted to see
how much effect an overscan screen would have on the same system.  PerfTest
works by using the Read() and Write() functions to write & read a series of
files ranging from 64K to 512K.  It first determines the largest contiguous
chunk of free RAM (upto 512K) in your Amiga, then finds out how much free
space is available on your current drive.  With this info it determines how
large a file it can write, if any.
  All things considered, it is a simple program to use.  The user
indicates the number of times each chunk is written and read back.  A very
exact microsecond timer is used for accuracy, but the final result is
rounded to about .01 second.  Only the time involved in the actual read or
write is measured, not the time in opening the file, getting a lock or
deleting it when finished.  And the times reported are an average based on
the number of times each chunk size is tested.  Obviously, the more
iterations used, the more reliable the results are.  You can specify from
2-50 iterations.  More than 50 is probably useless, and 20-30 is fine. 
Keep in mind that for a loop of 50 (ranging from 64K to 512K), you end up
performing nearly one hundred megabytes of disk I/O. 
  The second parameter allows you to do the test while an overscan screen
is displayed.  The screen is 672x440 16 colors, and while not the most
severe overscan possible, it is enough for the test.  Some controllers,
especially SCSI, have their performance degraded by overscan displays.
(Early CBM A2090 controllers were reputed to actually stop working under
some overscan situations.)  What makes this a good test is the fact the
only difference is that the overscan screen is being displayed.  So run the
test once without overscan, then again on the same setup with overscan. 
Comparing the results will give you some idea of just how much your
controller is degraded when displaying overscan.

SYNTAX

  PerfTest n flag

           n=number of iterations 2-50
           flag=0 or 1, with 0 being no overscan and 1 means use overscan

   The program uses the current drive (it DOES NOT test a ram drive, as
this drive always has 0 blocks free).  Make sure you CD to the drive you
want to test before starting the program.  I suggest you also have you CLI
window fully open so you can more easily read the results when they are
printed to the screen.

NOTES

  A number of things can affect the results.  Obviously, the type of
controller and drive you are using will make a difference.  The amount of
data on the disk makes a difference, as does the degree of fragmentation.
Other factors such as interleave and number of buffers can change the
results.  The best data is generated with a freshly formatted, empty drive.
You can use either 'old' AmigaDOS or the 1.3 FastFileSystem, with the
latter being much faster of course.
  Also, the values reported here don't necessarily relate to those you will
get when loading or saving commercial programs.  But for those who need to
know, it does help compare one hard disk & controller to another.
  
WARNING

  I have TRIED to add some checks to the program so it is 'safe' to use,
and I have used it many, many times without problems. It does check to see
that there is enough free disk space, and it does check the users
parameters so they make some sort of sense.  HOWEVER, USE THIS AT YOUR OWN
RISK!  I CANNOT PROMISE IT WON'T BITE YOU, TRASH YOUR DISK, EAT YOUR
IRREPLACABLE FILES OR OTHERWISE MAKE YOU MISERABLE!
  I also suggest that once you start PerfTest, you let it finish.  Don't
reboot or otherwise interfere, 'cause it could cause you disk problems.

DISTRIBUTION

  While not public domain, you may copy this program and give it to others
for non-commercial use (I can't think of any commercial use anyway) as
long as this file is copied as well.  If you have any comments or bug
reports, let me know.  I can be reached on GENIE (LRWALLACE), PLINK
(LouWallace), COMPUSERVE (70616,1157), QUANTUMLINK (LRW or LouWallace) as
well as writing to me at Amiga World magazine.

                                              Lou Wallace