AVR-Chat

John, Paul,

> Hi Paul,
> 
> This untried, but pretty sure that I would go for:
> 
> //assume that we have 16 blocks of 256 in the array...
> LDS r16,step  //this one 0-255?
> LDS r17,num   //this one 0-15?
> CLR r18	
> LDI r30,LOW(array)
> LDI r31,HIGH(array)
> ADD r30,r16
> ADC r31,r17
> LPM r18,r30
> OUT PORTD,r18
> 
> /*...and can be two cycles quicker if you dedicate a
>    permanent pointer to 'step' and 'num' sequentially in SRAM:
> 	LD  r16,X
> 	LDD r17,X+1
>   */

Actually, it can be faster still.  

 LDS  r30,step  //this one 0-255?
 LDS  r31,num   //this one 0-15?
 SUBI r30,LOW(-array)
 SBCI r31,HIGH(-array)
 LPM  r18,r30
 OUT  PORTD,r18

Assuming you align your array on a 256-byte boundary, we can drop one
instruction:

 LDS  r30,step  //this one 0-255?
 LDS  r31,num   //this one 0-15?
 SUBI r31,HIGH(-array)
 LPM  r18,r30
 OUT  PORTD,r18

If, however, you need a true C solution, you can use something like:

static const __code char *table[256] = (void *)0x4000;
// or whatever your compiler required for telling it about a pointer to
code space at 0x4000

Then you can write:

char c = table[y][x];

--
Paul Curtis, Rowley Associates Ltd  http://www.rowley.co.uk
CrossWorks for MSP430, ARM, and (soon) Atmel AVR processors