| |
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):
11001101
01011000
10100111
11111111
00101000
01010101
11011010
00100110 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....
|
| |
Perplex
Registered: Feb 2009
Posts: 255 |
Quoting Rudicurrent techniques just seem too slow for many chars. wished someone could do some magic tricks :)
Magic sometimes happens when you ditch working on a generic solution and instead concentrate on one that fits your particular data set or use case.
Not implying that a generic solution is less interesting, mind. |
| |
Rudi
Account closed
Registered: May 2010
Posts: 125 |
Some notes to self. altho, i dont know if it will do any miracles. maybe one can reduce it to Axiss solution, but I have no idea:
Basically what happens from a layman perspective (like me), bits are copied like this:
where each number are the bit-indices. consider the second number (after the arrow) to be the destination (of zero memory), and source (infront of arrow):
copy from (bit-index) -> to:0->7
1->15
2->23
3->31
4->39
.
.(and so on..)
.
63->56 looking at those destination values in binary produce a pattern like this:07: 00000|111|
15: 00001|111|
23: 00010|111|
31: 00011|111|
39: 00100|111|
47: 00101|111|
55: 00110|111|
63: 00111|111| and06: 00000|110|
14: 00001|110|
22: 00010|110|
30: 00011|110|
38: 00100|110|
46: 00101|110|
54: 00110|110|
62: 00111|110| and so on...
notice last column (trits) are same, but decrease, and msbs are increasing. wether or not one can derive something out of this i dont know. of course these are just index-values and dont have any sensible meaning for copying from and to memory. cycles would be massive using this technique, but my thought was to get something sensible out of this (which i did not yet)... |
| |
Slammer
Registered: Feb 2004
Posts: 416 |
I think its a shame that your cyclecounts for this problem is so low allready. I would like to see the spritecollision solution. Something like storing the 8 bytes on row 7 of succeding chars. Stretch the line and use the LFT trick to calculate the solution. But that would take 8 lines. |
| |
tlr
Registered: Sep 2003
Posts: 1790 |
Quote: I think its a shame that your cyclecounts for this problem is so low allready. I would like to see the spritecollision solution. Something like storing the 8 bytes on row 7 of succeding chars. Stretch the line and use the LFT trick to calculate the solution. But that would take 8 lines.
Interesting idea! It would not necessarily take 8 lines. IIRC collisions can be retriggered during the line. You will be limited by the amount of pixels a sprite covers though. |
| |
Rudi
Account closed
Registered: May 2010
Posts: 125 |
I dont see how sprites will solve the C2P-problem (with as few cycles as possible)? but then again, Im not too experienced with sprites on the C64. A goal would be to calculate as many chars (per frame) as possible for C2P. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11386 |
tlr: wasnt there some VIC that only clears (or retriggers?) the collisions once? |
| |
tlr
Registered: Sep 2003
Posts: 1790 |
Quote: tlr: wasnt there some VIC that only clears (or retriggers?) the collisions once?
That was lightpen triggering. It's working very differently on 6569R1. |
| |
Skate
Registered: Jul 2003
Posts: 494 |
I just made a rough calculation. Using 16k of precalc data (which is already terrible), each char takes 500-600 cycles to rotate. So, c2p routine looks like the best solution in every aspect. I'm still thinking an alternative approach though. |
| |
Flavioweb
Registered: Nov 2011
Posts: 463 |
Quoting SkateI just made a rough calculation. Using 16k of precalc data (which is already terrible), each char takes 500-600 cycles to rotate.
Just to say (...maybe i'm off-topic...) this code took 544 cicles rotate a char counterclockwise.
*=$C000
.FOR S=0, S<8, S=S+1
LDA SOURCE+S
.FOR D=0, D<8, D=D+1
ASL
ROL DEST+D
.NEXT
.NEXT
RTS
;-------------------
SOURCE
.BYTE %11111111
.BYTE %10000000
.BYTE %01000000
.BYTE %00100000
.BYTE %00010000
.BYTE %00001000
.BYTE %00000100
.BYTE %11111111
DEST
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
.BYTE %00000000
;--------------------------------------- |
| |
Rudi
Account closed
Registered: May 2010
Posts: 125 |
This takes ~472 cycles, and is basically similar to Oswalds approach: ;counter-clockwise rotation by 90 degrees.
;column 1
ror $74
rol
ror $75
rol
ror $76
rol
ror $77
rol
ror $78
rol
ror $79
rol
ror $7a
rol
ror $7b
rol
sta $7c
;tot = 59 cycles
;column 2
ror $74
rol
ror $75
etc..
ror $7b
rol
sta $83
;tot = 59*8 = 472 cycles did a test in zp. so rotated (destination) bytes are in $7c to $83. this req eight bytes to be copied to zeropage ($74 to $7b) (as an example). |
| Previous - 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 - Next |
