[CSDb] - User Forums - Fast way to rotate a char?

You are not logged in - nap

CSDb User Forums

Forums > C64 Coding > Fast way to rotate a char?

2017-01-04 08:32

Rudi
Account closed

Registered: May 2010
Posts: 125

Fast way to rotate a char?

Im not talking about rol or ror, but swap bits so that they are rotated 90 degrees:

Example:

a char (and the bits can be random):

10110010 byte 1..
11010110 byte 2.. etc..
00111001
01010110
11011010
10110101
00110011
10110100

after "rotation" (rows and columns are swapped):

is it possible to use lookup tables for this or would that lookup table be too big?
or other lookuptable for getting and setting bits?

-Rudi

... 105 posts hidden. Click here to view all posts....

2017-01-12 09:31

Rastah Bar
Account closed

Registered: Oct 2012
Posts: 336

Quoting Rudi

2. swap lower nybbles of byte0->byte4, byte1->byte5 etc.

That helped. If all bits from block 'a' in my post above are put entirely in destination0, and the bits of the 4x2 block above that in destination1, nybbles have to be swapped at the end, but the code

sta selfmod:+1
and #$0f
sta destination0
selfmod:
lda moveHighNybbleToLowNybble,x
sta destination1

simplifies to just 'STA destination' and with this I can reduce the cycle count to 312, I think.

2017-01-13 09:05

ChristopherJam

Registered: Aug 2004
Posts: 1409

I'm not managing to get below 308 cycles (source and destination non zero page).

However, my second stage is only 23 cycles, less for when stage chaining is done. Perhaps this can help optimise one of the routines above?

#define swap1(s1,s2,m,d1,d2) \
    .(        :\
    lda s1    :\
    eor s2    :\
    sta ztb   :\
    and#m     :\
    eor s2    :\
    sta d1    :\
    eor ztb   :\
    sta d2    :\
    .)

(also - I could get down to 300 by moving the code to zero page, but then you'd have to jmp there and back, and couldn't unroll for multiple source/destinations - one char takes most of the page)

2017-01-13 10:32

Rastah Bar
Account closed

Registered: Oct 2012
Posts: 336

Quoting ChristopherJam

I'm not managing to get below 308 cycles (source and destination non zero page).

However, my second stage is only 23 cycles.

Yes, thanks! That brings my lookup table approach down to 301 cycles when source and dest are in ZP and to 329 cycles when they are not.

2017-01-13 13:14

Axis/Oxyron
Account closed

Registered: Apr 2007
Posts: 91

Christopher: That 3x EOR thing is exactly what we did on Amiga back in the days. Didnt expect this to have an advantage on 6502. But where is the shifting taking place? Or is this only in 1 of the 3 passes, and the other pathes correct the bitorder with a table lookup?

2017-01-13 17:10

Rudi
Account closed

Registered: May 2010
Posts: 125

Color Bar: 301 cycles is nice, just 2 cycles more than what my code does.

If anyone can get below 299 cycles then pls explain what method you use :p

2017-01-13 18:17

ChristopherJam

Registered: Aug 2004
Posts: 1409

Quoting Axis/Oxyron

Christopher: That 3x EOR thing is exactly what we did on Amiga back in the days. Didnt expect this to have an advantage on 6502. But where is the shifting taking place? Or is this only in 1 of the 3 passes, and the other pathes correct the bitorder with a table lookup?

Sweet. Yes, the code above is only used in the second of the three passes; first and third are very similar to yours, only with a bit shuffle on one of the input bytes in each pair on the third pass.

2017-01-13 18:22

ChristopherJam

Registered: Aug 2004
Posts: 1409

Quoting Rudi

If anyone can get below 299 cycles then pls explain what method you use :p

Could we make the rules the same as for Axis' contribution before we compare cycle counts?

-neither input nor output bytes on zero page
-zero page intermediate results are fine
-code is not relocated to zero page either.

In practice it's unlikely there'd be enough space in ZP for source and destination charsets, and copying the data in and out would add at least an extra 102 cycles per char.

2017-01-13 18:32

Rastah Bar
Account closed

Registered: Oct 2012
Posts: 336

Good idea. I am down to 327 cycles now according to these rules.

2017-01-13 20:41

Rudi
Account closed

Registered: May 2010
Posts: 125

No, I dont think thats a good idea. But feel free to restrict yourselves to your own rules. I totally dont know what you are talking about anyway...

2017-01-13 21:31

Rastah Bar
Account closed

Registered: Oct 2012
Posts: 336

Quote: No, I dont think thats a good idea. But feel free to restrict yourselves to your own rules. I totally dont know what you are talking about anyway...

OK, I will keep mentioning both cases (all input and output bytes either in ZP or in Mem).

One version uses 301 cycles if all are in ZP, another version takes 327 cycles when all are in Mem.