[sdcc-devel] [ sdcc-Bugs-1572431 ] yy -=x*6: wrong assembler code

[SourceForge.net <no...@so...>]

Bugs item #1572431, was opened at 2006-10-07 00:08
Message generated for change (Comment added) made by frief
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1572431&group_id=599

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: C-Front End
Group: None
>Status: Closed
Resolution: None
Priority: 5
Submitted By: Wolfgang Heidrich (wolfgang75)
Assigned to: Nobody/Anonymous (nobody)
Summary: yy -=x*6: wrong assembler code

Initial Comment:
I'm using:
;--------------------------------------------------------
; File Created by SDCC : FreeWare ANSI-C Compiler
; Version 2.5.0 #1020 (May  8 2005)
; This file generated Fri Oct  6 23:10:33 2006
;--------------------------------------------------------
        .module bug
        .optsdcc -mmcs51 --model-small

The following is the sample code that produces, in my
opinion,
the wrong code:
--------------------------------------------
#include <stdio.h>

void putchar( char a )
{
        a=0;
}

void calc( unsigned int yy, unsigned char x)
{
        yy -= x*6; // <<<<<<<<<<<<<<<<<<<<<<<<<<<<< The
problem is here.
        printf("Result: %02u ", yy );
}

void main( void )
{
        unsigned char i;

        for( i=0; i<10; i++ ) calc( 280, i );
}
-------------------------------------------

The following is the produced assembler code ( sdcc bug.c )
of function calc(...):

;------------------------------------------------------------
;Allocation info for local variables in function 'calc'
;------------------------------------------------------------
;x                         Allocated with name
'_calc_PARM_2'
;yy                        Allocated to registers r2 r3
;------------------------------------------------------------
;bug.c:8: void calc( unsigned int yy, unsigned char x)
;       -----------------------------------------
;        function calc
;       -----------------------------------------
_calc:
;     genReceive
        mov     r2,dpl
        mov     r3,dph
;bug.c:10: yy -= x*6;
;     genMult
;     genMultOneByte
        mov     a,_calc_PARM_2
        mov     b,#0x06
        mul     ab
;     genMinus
        setb    c
;       Peephole 236.l  used r2 instead of ar2
        subb    a,r2
        cpl     a
        mov     r2,a
        mov     a,r3
        subb    a,b
        mov     r3,a
...
...
...

As far as I understood there is missing a "cpl c" just
after the "cpl a".

Detailed explanation:
The compiler tries to implement "yy -= x*6;" (at least
concerning the low byte) similar to following c code:

p = x*6

yy = p - yy

yy = -yy

For the negation "yy = -yy" the compiler needs to
substract 1 and calculate the complement.
Substracting 1 is done indirectly with "setb c" before
"subb a,r2". The complement is done with "cpl a".
But the compiler forgot to complement the carry bit.
The high byte is substracted differently
as "yy = yy - p", but now using the wrong
(uncomplemented) carry bit.

Example: yy = 280 = 0x118 and x = 46 = 0x2E
=> r2=0x18 r3=0x01

mul ab ; => b=0x01 a=0x14
setb c ; => c=1
subb a,r2 ; => a = a-r2-c = 0x14-0x18-1 = 0xFB = -5 and
c=1 because "borrow" happened.
cpl a ; => a = 0x04
mov r2,a ; => r2 = 0x04 (this is correct)
mov a,r3 ; => a = 0x01
subb a,b ; a = a-b-c = 0x01-0x01-1 = 0xff (carry bit is
set, see above, this is wrong)
mov r3,a ; r3 = 0xff

So the 8051 gets as result 0xff04 instead of the
correct result 0x04.

Concerning everything else of sdcc I'm highly pleased,
thank you very much for the great work!
Many greetings from Munich!

Wolfgang Heidrich

----------------------------------------------------------------------

>Comment By: Frieder Ferlemann (frief)
Date: 2006-10-16 10:56

Message:
Logged In: YES 
user_id=589052

Please get a more recent version of SDCC. This was fixed on
2005-10-26 by Bernhard Held. 

(This report is a duplicate of bug report #1270906)

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1572431&group_id=599