Optimizing CRC16 routine.

[Mike, K8LH]

Mike,
So obviously BoostC compiles this to faster code but doesn't the BoostC compiled code you posted use GOTO as well?

I posted 16F1827 code. A slightly unfair comparison because 18F devices don't have a "lsrf" instruction.

I didn't realize you were using an 18F until you posted your assembler output, and so I tested the following code which is similar to yours, for an 18F in BoostC;

    MB_CRC ^= work;             //
    for(MB_j=0; MB_j<8; MB_j++) //                    
    { if(MB_CRC.0)              //
      { MB_CRC >>= 1;           //
        MB_CRC ^= 0xA001;       //
      }
      else                      //
      { MB_CRC >>= 1;           //
      }
    }                           // Next bit in the byte

Here's the BoostC output which uses "bra" instructions and runs in about 115 cycles (worst case);

0008            label1
0008  500A      	MOVF gbl_work, W
000A  1A05      	XORWF gbl_MB_CRC, F
000C  6A09      	CLRF gbl_MB_j
000E            label2
000E  0E08      	MOVLW 0x08
0010  6009      	CPFSLT gbl_MB_j
0012  D7FA      	BRA	label1
0014  A005      	BTFSS gbl_MB_CRC,0
0016  D008      	BRA	label3
0018  90D8      	BCF STATUS,C
001A  3206      	RRCF gbl_MB_CRC+D'1', F
001C  3205      	RRCF gbl_MB_CRC, F
001E  0E01      	MOVLW 0x01
0020  1A05      	XORWF gbl_MB_CRC, F
0022  0EA0      	MOVLW 0xA0
0024  1A06      	XORWF gbl_MB_CRC+D'1', F
0026  D003      	BRA	label4
0028            label3
0028  90D8      	BCF STATUS,C
002A  3206      	RRCF gbl_MB_CRC+D'1', F
002C  3205      	RRCF gbl_MB_CRC, F
002E            label4
002E  2A09      	INCF gbl_MB_j, F
0030  D7EE      	BRA	label2