home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Power-Programmierung
/
CD1.mdf
/
pascal
/
library
/
dos
/
rtl
/
tp60
/
rate60.pas
< prev
next >
Wrap
Pascal/Delphi Source File
|
1991-02-20
|
6KB
|
149 lines
{ ========================================================================== }
{ Rate60.pas - Rate timer using only low memory clock. ver 6.0, 02-20-91 }
{ }
{ This rate timer is a highly accurate timer based on the low memory clock. }
{ It is device independent by not accessing any special timer hardware and }
{ does not use any interrupts. Accuracy is +-1 repetition. }
{ }
{ This timer is great for timing short events (1 microsecond to .1 second) }
{ to get a Reps-per-second result. The secret to such great accuracy is }
{ keying in on the fact that the low memory clock is very accurate and on }
{ schedule every time it ticks. So, just make the event work by the clock }
{ rather than vice-versa. }
{ }
{ The timer is set for a 1 second test. If you can't get enough repetitions }
{ (>100) to get any resolution, you are free to change TestTime up to 3600 }
{ seconds. }
{ }
{ The timer even accounts for the overhead of the timer itself and any other }
{ overhead that is a part of your event. Just insert your code and go. You }
{ may experience a little jitter while running in the integrated environment,}
{ but is solid when run as an executable file. }
{ }
{ For further explanation of the high-resolution feature of this timer, see }
{ comments at the end of this file. }
{ }
{ Public Domain by James H. LeMay, Eagle Performance Software }
{ ========================================================================== }
{$A+,D+,E-,L-,N-,R-,S-,V-}
program RateTimer;
uses
Crt;
const
TestTime = 1; { seconds }
TicksPerDay = 1573040.0; { DOS timer ticks/day. }
TicksPerSec = TicksPerDay/86400.0;
var
LowClock: word absolute $0000:$046C;
Tick1,Ticks: word;
Reps,OverheadReps: longint;
RepsPerSec,TickFactor: real;
R: real absolute RepsPerSec; { Variable name is easier to Watch }
{ -- Place your variables here: -- }
S1,S2: string;
b: byte;
L1,L2,L3: longint;
procedure WaitForTick;
begin
Tick1 := LowClock;
repeat
until LowClock<>Tick1;
Tick1 := LowClock;
end;
procedure RunTime (Seconds: word);
begin
if Seconds>3600
then Seconds:=3600;
Ticks := trunc(Seconds*TicksPerSec);
TickFactor := Seconds*TicksPerSec/Ticks; { Accounts for trunc of seconds }
WaitForTick;
end;
procedure Init;
begin
{ ** Initialize any of your variables here: ** }
S1 := 'Now is the time to check out Eagle Performance Software Products'+
'for speed tests';
S2 := S1;
L1 := 12345678;
L2 := 100;
end;
procedure ResetVariables;
begin
{ ** Place any variables that need to be reset after each test here: ** }
{ StrMove (S2,S1); }
end;
procedure Test;
begin
Reps := 0;
RunTime (TestTime);
repeat
{ ** Insert your test routine here ** }
L3 := L1 div L2;
{ ** End of test routine ** }
ResetVariables;
inc (Reps);
until LowClock-Tick1>=Ticks;
end;
procedure OverHead;
begin
OverHeadReps := 0;
RunTime (TestTime);
repeat
ResetVariables;
inc (OverHeadReps);
until LowClock-Tick1>=Ticks;
end;
procedure CalcRate;
begin
RepsPerSec := TickFactor/((TestTime)*(1.0/Reps-1.0/OverHeadReps));
WriteLn ('Strings/sec = ',RepsPerSec:8:0);
end;
begin
Init;
Test;
OverHead;
CalcRate;
end.
{ ========================================================================== }
{ The simplicity of the rate timer is deceptive. Don't let it fool you. If }
{ you take a just a cursory look, you may think that you can only get 1/18.2 }
{ second resolution out this timer. But think again. }
{ }
{ There are two clocks in your computer - (1) the 8253/8254 timer chip, and }
{ (2) the CPU. (One doesn't usually think of the CPU as a clock!) Remember,}
{ the low-memory output for the clock comes from the 8253/8254 chip which is }
{ accurate to 1-microsecond, even though it doesn't "tick" but every }
{ 1/18.2... seconds. }
{ }
{ Since we can't get high resolution from the low-memory clock, the program }
{ does the reverse and gets the high resolution from the CPU instead. There }
{ are two types of applications for a timer - (1) EVENT TIMER: duration of a }
{ a single event, and (2) RATE TIMER: the number of events in a single }
{ duration. Our application in STRG is the latter. Where resolution (not }
{ accuracy) is low in the timer, it is gained in the CPU by making sure }
{ there are a LARGE number of events. }
{ }
{ Notice that this timer is accurate to +/- 1 repetition as opposed to some }
{ value in "seconds". That's because it measures EVENTS, not seconds. So, }
{ a test of a given duration producing 100,000 events will certainly have }
{ higher resolution than the same duration producing only 100 events. The }
{ former would have 0.001% resolution while the latter only 1%. If you }
{ increase the duration of the test, you can get even higher resolution. }
{ }
{ RATE is meant to be a test tool for SYS to give a rough idea of }
{ comparative performance and works quite well. }
{ ========================================================================== }