Suzy B Software 2/Suzy B Software CD-ROM 2 (1994).iso/new_file/utilitys/speedom2/speedom.doc

Text File | 1995-04-25 | 9.9 KB | 194 lines

                               Speed-o-meter
                               =============
                      "The multitasker's speed gauge"

                       ╜1994 DataBasement Software
                           All Rights Reserved

                         By Erin Matthew Monaco
                            Programmed using
                       Lattice C 5.60 + DevPac III.


     This is another S.A.S (Short And Simple) ShareWare product from
DataBasement Software. If you find it useful and use it I'd appreciate
it if you could send the $5 registration fee to me at:

  Erin Monaco
  35244 Chestnut St.
  Wayne, MI  48184

I also accept GEnie Gifts of Time page M080. My GEnie Address is:
  E.MONACO

     You have the right to copy and distribute this program for a
reasonable PD disk fee. You must insure that distribution is complete
with this document as its the only identification the program carries.

History:
Version 2:
- EVT_TIMER vector now installs using XBRA protocol. Tested appears
  to work fine. Now SPEEDOM will unlink safely when run with
  other XBRA compliant programs which redirect EVT_TIMER.
  (This includes running multiple copies of SPEEDOM).

- To keep window size to a minimum I removed the relatively worthless
  SPEEDOM window title. Now when SPEEDOM is first run the title will
  display "Wait..." this will remain till SPEEDOM has read its .CNF
  file, or while it calculates your computers optimal speed and then
  creates the .CNF file.  After that the title will display the
  applications speed percentage in decimal (in addition to the bar
  graph already displayed in the window) Now you can see at a glance
  what the actual percentage is. Since its possible (as explained
  below) a plus "+" sign will appear to the right of the percent symbol
  when you happen to exceed 100% performance. ('+' actually added after
  this document was written.)

- Changed the second line from "|  5|0  |" to just "|   |   |" but
  the ticks still represent 25%, 50% and 75% of optimal speed. This
  is fairly accurate but not lined up "precisely". Which is why I added
  the decimal % display.

- Changed the code a little but it still works as explained for version
  one. Which is what this doc file talks about.

Version 2: Added Tidbits.
- Once I added the accurate decimal percentage display (and no longer
  had to guess at speeds) I found that my initially reported slowdowns
  on mouse movements and running Geneva to be on the low side.  In
  other words, the slowdown was even more dramatic then I thought. To
  see an excellent example of this configure SPEEDOM on a completely
  bare system as suggested.  Then when your speed starts getting reported
  move your mouse as quickly as you can across the screen.  Do not be
  surprised to see a 25% slowdown in application execution.  If you have
  warp_9 (or possibly some other mouse/screen accelerator run it and
  try SPEEDOM again)  With warp_9 3.75 Moving the mouse on my AdSpeed
  STE (16Mhz) accelerated machine I started reading 100%... it could
  have been even more but the program caps at 100 for other technical
  reasons.

- Running Geneva and setting the default flags to single-tasking Speedom
  reported about 89%.  This is with neodesk (which gets suspended),
  Genevas Task Manager and Warp_9 CP.  I found that if I put the acc's
  (Task Manager and Warp_9 CP to sleep, by shift clicking on their
  menu entries.. yes you CAN do that! Now running SPEEDOM I was back up
  to 100%!!!  (This could be useful if you plan on doing something where
  every ounce of speed would be nice. The acc's can be reactivated by
  simply clicking on their entries, when they pop up just close them...
  so they are STILL available from the application if you need them!!!)
  WOW!  This method works fairly well when multitasking.  Doing the same
  thing in multitasking speedom was running at about 44%.  By manually
  (shift-clicking) putting neodesk and the acc's to sleep I managed to
  get up around 98%! Very interesting and useful indeed.

- I have noticed that when I configure SPEEDOM at 8Mhz on a bare system...
  except for the Adspeed acc I used to drop down to 8Mhz before running
  SPEEDOM, that I'd tipically get 98% efficency right off the bat. When
  Warp_9 is added, it jumps to 100 (again possibly beyond). I attribute
  this to the fact that I poll the machine for 5 seconds taking 5-
  1 second counts and then I take the highest (not average) optimal
  ratio of instructions executed divided by time taken to execute them.
  (thats why I need the EVT_TIMER vector!) Not to mention I cheated
  on the instruction count "system" which does not actually represent
  instructions executed, but hopefully a rough estimate of this anyway!
  What I'm saying is "results will very from machine to machine, system
  to system, and most accurate results will be gotten from configuring
  SPEEDOM on your bare system.

- If you think about it since I use a constant for executed instructions
  that increments a counter each pass through the loop, addind software
  that accelerates such things as mouse and screen would be reflected
  in a better performance percentage. Say each pass through the loop
  increments the instruction counter by 100. Now say because of various
  "accelerator" programs we manage to get through the loop in only 80
  instructions. The instruction counter is still incremented by 100
  so we manage to get through the loop a few more times then we did
  before we used the accelerators. Thus we will have a higher
  instruction to time ratio... giving us a reported percentage that was
  better then the original. (Don't worry if you don't get it)

 Version 1:
- Initial release.

                            Version 1 Docs

     This program is a breeze to use. Just run it using your favorite
method and watch it go! Currently it only operates as a program- it can
not be run as an ACC... yet!

     When you run the program you'll be presented with a very small
window. The top line of the window (small text size) is a graphic bar
indicator. The next line simply has three "|"'s and 50 to mark the
half way point. The Graph Bar is a simple percentage indicator. Going
from 1 to 100. With "|" at 25, 50 and 75 Percent marks.  This is a
completely visual program with no interaction required. To quit just
close the window. The window can be moved anywhere on screen. This
program gives a visual indicator of how fast your programs are running,
as compared to the optimal speed they could run on your machine.

     "OK, the program's under 6K with embedded RSC and all... how does
it know my machine's optimal speed?"

     Great question. It doesn't so we need to teach it first. When you
run the program it checks for a file SPEEDOM.CNF. If it finds it it will
load in the 4 Byte value contained in that file. If it doesn't it will
Test your machine's speed 5 times (takes under 10 seconds) and then it
will compute this value and create a SPEEDOM.CNF to the current directory
for you! (Simple) After this is done it will proceed with rating the
application's speed.  This was the easiest way I could think of to do
this accurately (mostly) on any machine from 8Mhz, 16Mhz, 32... 64...
a zillion or whatever.

     The best way to configure the program is to run it on a BARE system
with nothing else running, no accessories, etc. If you ever choose to
re-calibrate the program, just delete the SPEEDOM.CNF file and run
SPEEDOM to create a new one.  You could if you choose configure it from
within Geneva or MultiTos or whatever. But for the most accurate system
performance readings you should configure it with a bare system at the
highest CPU speed (in the case of accelerators, etc) that you can.

INTERESTING TIDBITS:

     In testing/debugging this program I saw a few neat (disturbing)
things. Configured on a bare system (no warp_9, mouse accelerators etc)
Moving the mouse in a quick circle around the screen seemed to slow
the application by about 20% on my TOS 1.62 ROMS... thats scary! With
WARP_9 I could not detect a slowdown at all!

     Running SPEEDOM as a single-task application I averaged around
90% of my normal speed with mouse movements slowing it down by another
5% or so. (Faster you move the mouse, greater the slowdown... this was
with warp_9 on). Defaulting back to multitasking having NeoDesk and
a few accessories I ran two copies of SPEEDOM and got a speed just under
50%. For every copy run after that system speed dropped by only about 5%
or so.


QUICK DISCLAIMER:

     By making this program and everything I am not trying to promote
any products (like warp_9, although its well worth it <g>) or pound
any programs (like Geneva... I LOVE Geneva and respect Dan alot for
his accomplishments and GRIBNIF support? Is superb) but no endorsements
from me! <g>

     Programs that do not like to surrender control to Geneva/MultiTOS
will not be reflected very well. (If SPEEDOM doesn't get time to work...
then it won't be updated at all!)  Also as sorta mentioned above things
such as moving the mouse, clicking on objects and using menus and stuff
will temporarily reduce the application's percentage of code executed.
The graphic bar display is normally updated every second or so. Unless
one of the above mentioned factors prevents this.

     I'm no genius. I make mistakes and believe it or not I don't always
know WHAT I'M DOING!!!  This program is a perfect example of my
"Winging It". While I believe this program does represent an applications
speed in terms of a percentage of optimal speed fairly accurately, I
could be seriously wrong! In any event it is nice to look at and watch :)

     Comments and suggestions are welcomed and appreciated, as are
registrations (hint!) feel free to leave E-mail on GEnie to me at
E.MONACO or send a letter to the above address.

  Enjoy! Erin @DataBasement Software.