RE: [sdcc-devel] [ sdcc-Bugs-479001 ] Code gen overwrites registers

[Sandeep D. <sa...@wi...>]

Alex,

This could happen when the register pressure is high.. I think
you need the complex example (if you simplify the problem will go away).
I sounds like two iTemps are sharing the same registers.

Sandeep

> -----Original Message-----
> From: sdc...@li...
> [mailto:sdc...@li...]On Behalf Of
> no...@so...
> Sent: Tuesday, November 06, 2001 10:12 PM
> To: no...@so...
> Subject: [sdcc-devel] [ sdcc-Bugs-479001 ] Code gen 
> overwrites registers
> 
> 
> Bugs item #479001, was opened at 2001-11-06 22:12
> You can respond by visiting: 
> http://sourceforge.net/tracker/?func=detail&atid=100599&aid=47
> 9001&group_id=599
> 
> Category: msc51(8051) target
> Group: None
> Status: Open
> Resolution: None
> Priority: 5
> Submitted By: Alex Karahalios (karahalios)
> Assigned to: Nobody/Anonymous (nobody)
> Summary: Code gen overwrites registers
> 
> Initial Comment:
> The following problem occurs with all versions of SDCC, 
> although this simply test case does not fail with the latest 
> version in CVS. I haven't been able to simplify my more 
> complex program yet.
> 
> The following C code:
> 
> extern char Data1[];
> 
> extern  int f1(void);
> extern void f2(char *,int);
> 
> void main(void) {
>     int l;
>     
>     l = f1();
>     f2(&Data1,l+10);
> }
> 
> generates the follow ASM code
> 
> ;       bug.c 9
>         lcall   _f1
>         mov     r2,dpl
>         mov     r3,dph
> ;       bug.c 10
>         mov     r2,#_Data1
> ; Peephole 3.c   changed mov to clr
>         clr  a
>         mov  r3,a
>         mov  r4,a
>         mov  a,#0x0A
>         add     a,r2
>         mov     _f2_PARM_2,a
> ; Peephole 180   changed mov to clr
>         clr  a
>         addc    a,r3
>         mov     (_f2_PARM_2 + 1),a
>         mov     dpl,r2
>         mov     dph,r3
>         mov     b,r4
>         lcall   _f2
> 
> Notice that after the call to _f1, the results are stored in 
> r2&r3. Immediately afterwards r2&r3 are overwritten. Later on 
> before the call to _f2, r2&r3 are  used, but have since been 
> destroyed.
> 
> As I mentioned, this simple test case does not currently fail 
> with the current CVS version, but my large program still does.
> 
> The current work around is to rewrite the C code so that 
> function parameters are precalculated before being used:
> 
>     l = f1()+10;
>     f2(&Data1,l);
> 
> Thanks,
> 
> Alex Karahalios
> 
> 
> 
> ----------------------------------------------------------------------
> 
> You can respond by visiting: 
> http://sourceforge.net/tracker/?func=detail&atid=100599&aid=47
9001&group_id=599

_______________________________________________
sdcc-devel mailing list
sdc...@li...
https://lists.sourceforge.net/lists/listinfo/sdcc-devel