Em 14/02/2014 00:23, William "Chops" Westfield escreveu:
> On Feb 11, 2014, at 1:24 AM, alan.b.pearce@stfc.ac.uk wrote:
>
>> I believe the free mode [of XC8] now does a minimal optimisation so that=
 any code generated will at least work in the way the code writer intended.
> I dunno.  I downloaded the latest, in hopes that this would be true.  But=
 it's still producing rather weird code:
>
> >From one of their "blinky" examples:
>
> void delay() {
>     int counter =3D 0;
>     for (counter =3D 0; counter<10000; counter++) {
>         ;
>     }
> }
>
> 7FD8	0E00	delay	MOVLW 0x0		;; 16 bits of 0 to counter.
> 7FDA	6E02		MOVWF 0x2, ACCESS
> 7FDC	0E00		MOVLW 0x0
> 7FDE	6E01		MOVWF counter, ACCESS
> 7FE0	0E00		MOVLW 0x0		;; and again (the source
> 7FE2	6E02		MOVWF 0x2, ACCESS	;;  does it twice, too, so
> 7FE4	0E00		MOVLW 0x0               ;;  I guess that's OK.)
> 7FE6	6E01		MOVWF counter, ACCESS


It is doing exactly what was requested.


> 7FE8	5002    lp:	MOVF 0x2, W, ACCESS
> 7FEA	0A80		XORLW 0x80		;; Strange compare to 10000
> 7FEC	0F59		ADDLW 0x59              ;;  I guess it's jumping through hoops=
 to wind up with a sequence
> 7FEE	0E10		MOVLW 0x10              ;;   that skips appropriately at the e=
nd, rather than resolving
> 7FF0	B4D8		BTFSC STATUS, 2, ACCESS  ;;   the branch in the middle.  But=
=85 Eww!
> 7FF2	5C01		SUBWF counter, W, ACCESS
> 7FF4	B0D8		BTFSC STATUS, 0, ACCESS
> 7FF6	D003		BRA 0x7FFE              ;; (exit) end of strange compare


It is calculating ( counter - 0x8000 + 0x5900 - 0x0010 ) that is exactly th=
e same as ( counter - 10000 )

1) Subtract (XORLW) 0x80 from the high byte. Note that the low byte isn't d=
isturbed.

2) Add 0x59 to the high byte. and test for zero. If the result is zero, cou=
nter is between 0x2700 (9984) and 0x27ff (10239). If the Z bit is clear and=
 the C bit is set then counter is above 10239 and it takes the BRA.

3) If the Z bit is set, counter is in the range between 9984 and 10239, it =
simply subtracts 0x10 (16) from the low byte. If the C bit gets set it is b=
ecause counter is above 9999 (0x270f) and it takes the BRA.


> 7FF8	4A01		INFSNZ counter, F, ACCESS
> 7FFA	2A02		INCF 0x2, F, ACCESS     ;; Pretty nice 16bit increment
> 7FFC	D7F5		BRA 0x7FE8              ;; (lp)
> 7FFE	0012	exit:	RETURN 0
>

--=20
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.