Basic ... Timing

David Harry, West Midlands

Issue 4

Jul/Aug 83

Next Article >>

<< Prev Article

 

Basic programs, which are all I can write, sometimes run more slowly than I would like, so I try to incorporate any time-saving devices I can. Often there is more than one way to achieve the same effect-for example, GR.0 ? "ã" and ? CHR$(125) all clear the screen. This simple timing routine will enable you to find out which of the routines you can use is faster and whether there is any saving or penalty in memory usage. It uses a FOR-NEXT loop to execute your piece of code 1000 times but if that is too long, change the value of R in line 3000.

AtariLister - requires Java

The first RUN is a setting up run to isolate the effects of the program itself. Firstly enter a number of REM lines from line 20 onwards equivalent to the number of lines in the routine you are going to test as you are not testing the time taken to read line numbers. Next DIMension and initialise any strings or arrays from line 2000 on. Now RUN the program for the first time.

When the program asks you to change Line 1000, use the cursor up, enter the new line and enter your first routine for testing, starting from line 20. RUN the program and record the time and memory readings. Now change the routine to the one you want to compare and RUN again. Note the new readings and so on.

I originally wrote the routine to help with a program which used one long string to contain names and data for several individuals. Each name was allocated a 20 character section of string with blank characters (CHR$(32)) being used to extend names to the required length. This meant that where I wished the program to ask a question involving an individual's name, there was a variable length space before the next word in the question. One way round this would have been to use a FOR-NEXT loop to print the name one letter at a time, stopping at an end of name flag such as '*'. This seemed a little clumsy and slow. What I wanted was a character that would neither print anything on the screen nor move the cursor on. I found five: CHR$(158) - CLEAR TAB, CHR$(159) - SET TAB, CHR$(253) - SOUND BUZZER, CHR$(254) - DELETE CHARACTER, CHR$(255) - INSERT SPACE. The buzzer was obviously out - too noisy! Inserting or deleting characters would involve moving a number of characters (all blank) along the line which looked time consuming. However, it was still worth testing.

I first of all entered line 2000 DIM NAME$(20):NAME$= "ERMINTRUDE            ". This has ten letters and ten spaces. Then at line 20 ? NAME$;. Having RUN the program and entered line 1000 X=2728:M1=12370, I substituted CHR$(157) for each blank and re-ran. When I had timed each of the four characters I ran the FOR-NEXT loop with an end of name flag and "ERMINTRUDE" with no subsequent blanks for good measure. The initial run of 1000 repetitions took 54.56 secs. The other times were

CHR$	TIME	BYTES
158	-14.00 sec	0
159	-14.18 sec	0
254	137.58 sec	0
255	140.80 sec	0
FOR-NEXT loop	101.00 sec	76
“ERMINTRUDE”	-26.00 sec	-10

The saving in times, which came as a complete surprise to me, can only partly be explained by the extra lines and scrolling involved by printing the blanks. Ten CHR$(158) still print 0.005 seconds faster than ten CHR$(32). What surprising facts can you come up with?

top