Optimizing CRC16 routine.

[Mike, K8LH]

... <snip> ... I'm using TMR1 to measure the execution time of the routine ... <snip> ...

Could that be a problem? Do you have anything similar to the MPLAB Simulator Stopwatch?

... <snip>... This time I got 1533 cycles ... <snip> ...

My goodness, that doesn't sound quite right. That's close to what you got when you were processing nine bytes. Could there be a problem with the method you're using to count cycles?

I would expect you to get code similar to the assembler code generated from my BoostC "Do/While" loop;

  crcword ^= work; i = 8;   //
0040  0845  	MOVF crc16_00000_arg_work, W
0041  06B7  	XORWF gbl_crcword, F
0042  3008  	MOVLW 0x08
0043  00C6  	MOVWF crc16_00000_1_i

  do {                      //
0044        label5

    crcword >>= 1;          //
0044  36B8  	LSRF gbl_crcword+D'1', F
0045  0CB7  	RRF gbl_crcword, F

    if(status.C)            //
0046  1C03  	BTFSS gbl_status,0
0047  284C  	GOTO	label6
004C        label6

      crcword ^= 0x0A01;    //
0048  3001  	MOVLW 0x01
0049  06B7  	XORWF gbl_crcword, F
004A  300A  	MOVLW 0x0A
004B  06B8  	XORWF gbl_crcword+D'1', F

  } while(--i);             //
004C  03C6  	DECF crc16_00000_1_i, F
004D  1D03  	BTFSS STATUS,Z
004E  2844  	GOTO	label5
}