Re: lshift bug

[Anton Ertl]

On Sun, May 17, 2020 at 10:22:03PM +0200, Klaus Schleisiek wrote:
> Am 17.05.2020 um 18:10 schrieb Anton Ertl:
> 
> > and this should work as you expect.  But I would not call a result 0
> > "mathematically correct".
> 
> Aber ja doch! Lshift ist doch so definiert, dass die niederwrtigen Bits mit
> Nullen gefüllt werden.

For lshift the shift count results in an ambuguous condition.

For 2**, 0 is just as wrong as any other number that fits into a cell.

> > : 2** 1 over lshift swap 8 cells < and ;
> 
> Das funktioniert dann aber nur bei 64 Bit.

This works on every system with 8-bit address units (and Bernd
discussed how to eliminate that dependency), except for very large
numbers.  Fix:

: 2** ( u1 -- u2 ) 1 over lshift swap 8 cells u< and ;

> Und bedenke bitte, das ich bei uCore auch mit 16, 18, 24, 27 und 32 Bits
> Datenwortbreite arbeite und keine "Bytes" habe.

Ok, for systems with arbitrary address units, use:

s" address-unit-bits" environment? [if]
  constant address-unit-bits [then]
: 2** ( u1 -- u2 ) 1 over lshift swap address-unit-bits cells u< and ;

- anton