I can´t explain this (math precision)

[bola_dor]

I know that roughly the mathematical precision of the ZX81 starts to lose after 9 digits, one would expect rounding and exponential expressions.
What I can't explain is why the same statement "LET a=-0.999999999999999999999999" gives two different results in this case.
I would expect -1 as rounding, or anything else, what puzzles me is that the result is not consistent.
I tried the same listing on Spectrum BASIC and the result is always 2... why not -1 ?

• 10 LET A=-0.9999999999999999
  20 PRINT -0.9999999999999999
  30 PRINT A
  40 POKE 16436,255
  50 POKE 16437,255
  60 LET SS=PEEK 16436+256*PEEK 16437
  70 FOR F=1 TO 100
  80 LET A=-0.9999999999999999
  90 NEXT F
  100 LET EU=PEEK 16436+256*PEEK 16437
  110 LET E1=ABS SS-EU
  120 PRINT A;":",E1/50;""""

test.jpg (1.43 KiB) Viewed 4599 times

[1024MAK]

Which ROM is your machine using?

Using the last and latest Sinclair official ZX81 ROM I get:

Maths - rounding

The above was done on an emulator (the excellent one for iOS) using 16K RAM and the 649 ROM.

I also tried on a emulator for a ZX Spectrum 128, but without the timing parts of the program (so only lines 10 to 30, 70 to 90 and line 120.

This printed -1 three times.

Mark

[bola_dor]

Which ROM is your machine using?

Using the last and latest Sinclair official ZX81 ROM I get:
IMG_6117.jpeg
The above was done on an emulator (the excellent one for iOS) using 16K RAM and the 649 ROM.

I also tried on a emulator for a ZX Spectrum 128, but without the timing parts of the program (so only lines 10 to 30, 70 to 90 and line 120.

This printed -1 three times.

Mark

i used Eighyone v1.29 zx81.version2.rom
and spec48.rom and spec128.rom

Now I tried zx81.version3.rom and results are al 2...
anyway I woul spect -1

[XavSnap]

Hi bola_dor,

Try to edit the line 80 using shifted 1 and press New/Line without changes.