Execution speed, faster at beginning versus end of long program

[stroebeljc]

I have noticed a significant difference in the execution speed of BASIC lines that are at the beginning of a program versus those at the end of a large program. I do not have a "timing" analysis, but I do notice significant impacts of speed when looping to produce speech sounds when the loops are in these different locations.

I believe it is due to the way line addresses are determined during run-time. The LINE-ADDR ROM routine always starts at the beginning of the program space and checks each line until it gets to the one it wants. Obviously, this can take a lot longer if those lines are at the tail end of a large program.

Has anyone ever noticed this? Is there a discussion about this on this forum that I missed?

[dr beep]

Yes this is known issue.

Your games best have initialize at then end (only called once) and on a ZX Spectrum DEF FN is better place at the start (scanned first)
Often used code is best placed high as possible in the listings.

Machinecode will just execute what is at the Program Counter and always points there. No problem in MC.

[1024MAK]

Yes, this has been known about since the 1980s and as dr beep says, affects both the ZX81 and the ZX Spectrum.

Hence your first line should be:
10 GOTO 8000 : REM Jump to set-up - note not needed if the program auto-runs to a different line number after loading
20 REM any timing critical code
...

200 REM main loop
...

1000 REM less often used code
...

8000 REM set-up code

9000 REM on ZX Spectrum, any data
9010 DATA
...

Mark