100/ naming

[jkotlinski]

I don't anymore particularly like the naming of 100/ (which shifts top of stack right by 8, inserting 0)

• 100 no longer makes sense, because durexForth now uses DECIMAL base by default
• / does not make total sense, because it implies arithmetic right shift.

Any ideas? Are there any known precedents?

Right now, I'm leaning towards 256/...

[ekipan]

I was actually thinking the same while writing my own multiply and divide words! I usually name them with decimal. It's just a style thing so it doesn't actually matter but my preference is for 256/. I guess $100/ or #256/ would be clearer but with 40 columns for source code space is at a premium.

[ekipan]

I'm poking at a block editor proof of concept. It uses these words:

: 32/  2/ 2/ 2/ 2/ 2/ ;
: 40*  dup 2* 2* + 2* 2* 2* ;
: 1024* [ lsb lda,x asl,a asl,a
  msb sta,x 0 lda,# lsb sta,x ] ;

[Whammo]

It's strange, that an operation so ubiquitous be not standardized. Perhaps, simply documenting which way of going about it is fastest, and then leaving to the user.
Though, it did occur to me that a series of operations could be created to treat the top of stack as if it were a bit stack.

[ekipan]

Something like this?

: popbit  dup 1 and swap 2/ ;
: pushbit  2* or ;
T{ %101 popbit -> 1 %10 }T
T{ 1 %10 pushbit -> %101 }T

[Whammo]

Careful, or we could introduce bitrot

😄

[Whammo]

So many ways to accomplish the operation.

code 256/
msb lda, lsb sta, 0 lda,# msb sta,
;code

Edit: oh.

[ekipan]

The builtin one is only four instructions.

Actually that stack effect comment is wrong! Should be ( u1 -- u2 ) @jkotlinski

Edit: @Whammo it won't work because you have to index by the X register. You could do this:

: 256/ [ 0 ldy,# msb lda,x ] pushya nip ;

But that's just wasting a bunch of loads, stores, and an inx and dex pair.

[jkotlinski]

How about HIBYTE?

[jkotlinski]

In present-day 8-bit assemblers and compilers, HIGH seems to be the most common choice.

[ekipan]

code >lohi ( u1 -- u2 u3 )
msb lda,x dex, lsb sta,x 
0 lda,# msb sta,x msb 1+ sta,x
;code

code lohi> ( u1 u2 -- u3 )
lsb lda,x inx, msb sta,x
;code

Re: your other comment as well. This is with >lo and >hi coupled together. If you need both bytes I guess this is fewer instructions.

[jkotlinski]

>lohi could be named split or possibly bytesplit. Seems like a very useful word.

[jkotlinski]

Yeah... I think split fits perfectly. Best option so far...

[jkotlinski]

I really love split :) Super useful.

[jkotlinski]

Another idea: Replace 100/ with

code byteswap
lda msb,x ldy lsb,x
sty msb,x sta lsb,x
;code

[jkotlinski]

Good:

• obvious naming (or?)
• is more generally useful than 100/
• usually works as drop-in replacement for 100/

Bad:

• is not always drop-in replacement for 100/

[Whammo]

c! and the like ignores the high byte.
Perhaps, c>! and the like ignores the low.

[Whammo]

: 256/ w c>! w c@ ;

[jkotlinski]

How it turned out: #546