> -----Original Message-----
> From: Hulatt, Jon [spam_OUTjhulattTakeThisOuTMONSTEREUROPE.COM]
> Sent: 14 October 2003 16:08
> To: .....PICLISTKILLspam@spam@MITVMA.MIT.EDU
> Subject: [PIC:] Efficiency of C?
>
>
> void OutputByte( unsigned char w)
> {
>         unsigned char nibble;
>
>         nibble = w >> 4;
>         OutputNibble(nibble);
>         nibble = w & 0xF;
>         OutputNibble(nibble);
>
>         return;
> }
>
> That seems all well and good in C. But will it compile to anything as
> effient as an ASM ecquivalent of:-
>
> cblock
>         temp
> endc
>
> outputbyte:                                             ;
> byte to output is
> in W
>
>       movwf  temp
>     swapf  temp,W  ; get the hi nibble in the lo 4 bits
>     andlw  0xF
>     call            outputnibble
>
>     movf   temp,W  ; recover the lo nibble & send
>     andlw  0xF
>     call   outputnibble
>
>     return

>Hi,
>
>I've been using MPASM up to now, but I am considering the benefits of
>developing in C, since I'm a C++ programmer by trade, and a PIC hobbyist.
>It'll make development much simpler for me.
>
>C being a higher level language obviously introduces some potential issues
>for efficiency. But how severe are these? Some of the features of the PIC16
>instruction set allow some neat coding techniques that create fast, small
>code.
>
>
>

> Not bad, the first two instructions are for bank setting.  I chose a 16F877
> so a smaller device may require less instructions.  Note that the compiler
> has optimised the last call to OutputNibble by converting it to a jump.
> However also note that the nibble variable is not actually required, a
> better implementation would be:
>
> void OutputByte( unsigned char w )
> {
>         OutputNibble((unsigned char)(w >> 4));
>         OutputNibble((unsigned char)(w & 0xF));
> }
>
> giving:
>
>     33  07EE                     _OutputByte
>     34                           ;      _w assigned to ?a_OutputByte+0
>     35  0000                     _OutputByte$w  set     ?a_OutputByte
>     36                           ;_w stored from w
>     37  07EE  1283                      bcf     3,5
>     38  07EF  1303                      bcf     3,6
>     39  07F0  00A0                      movwf   ?a_OutputByte
>     41  07F1  0E20                      swapf   ?a_OutputByte,w
>     42  07F2  390F                      andlw   15
>     43  07F3  27E9                      fcall   _OutputNibble
>     45  07F4  0820                      movf    ?a_OutputByte,w
>     46  07F5  390F                      andlw   15
>     47  07F6  2FE9                      ljmp    _OutputNibble
>
> Now apart from the bank setting bits which weren't included in your assembly
> example, this has exactly the same number of instructions, but actually
> executes more quickly due to the jump optimisation.  The ljmp macro
> degenerates to a single goto if the OutputNibble resides in the same program
> memory page as the OutputByte function.
>
> Hope that calms your fears over the code generation quality of HiTech :o)
>
> Mike