Re: AW: AW: [avr-gcc-list] avr-gcc bug: global register variable is broken

[Klaus Rudolph]

Haase Bjoern (PT-BEU/MKP5) * schrieb:

> Hi Klaus,
>
>  
>
> > With register r10 for example the compile is totally corrupt. The 
> > register r10 is used in a function but not filled in by the caller! This
> > seems to be also a bug. And if lower registers work, why isn´t there any 
> > error message or warning?
> >    
>
> Because the praised individuals that have spend time on writing the free 
> compiler
> port probably didn't care about this point.
>
>  
Sorry, it is my bad english. I don´t want to say that there are bad 
programmers. I want to say that
we have a problem. It is not a question of fault anyway.

> > Because of undefined behaviour we must fix this bug. 
> > Also the libc doc says nothing against the use of global register vars.
> >    
>
> Yes, It is certainly a good idea to add some remarks to the documentation.
>
>  
I like more to fix the problem :-)

> > And in addition: I need more then one register above r16 if possible:-)
> > I have no idea how the internal machine description works, but I think 
> > there is something wrong in avr specific implementation.
> >    
>
> I will have a look at the code and tell you if I find out something on the constraints. 
> Meanwhile my guess is that the X-Register might work. BTW, in my opinion there is not something
> like "wrong" or "right". The compiler simply has been designed for most compact
> and efficient C-code. It has not been designed as a solution for your 
> non-standard problem. :-(.
>
>  
There is a general method how gnu handle a list of registers. The 
compiler framework knows really nothing about
the machine registers. This knowledge comes from the avr specific files 
(Traget Description macros & Functions)
I think that global registers variables work with other platforms so it 
is no question of compiler design I hope.
If the bug is general arround all targets, yes, then we have a bigger 
problem. But I think this is also solveable.
In general, the compiler itself is also only a bit software which could 
be made better :-))))

I currently digging into the GCC Internals manual and yes, there is a 
lot of text :-))

We will see who find the first solution.

Thanks for your ideas

   bye

       Klaus


> Yours,
>
> Björn
>
>
> Haase Bjoern (PT-BEU/MKP5) * schrieb:
>
>  
>
> > Hi,
> >
> > I have never tried it, but as much as I know, it should be possible to declare one of the lower 
> > registers for "fixed" purposes. I recommend you to try r2 and/or r3. The compiler will 
> > certainly not be able to work properly if you declare one of the registers usually being used for 
> > passing parameters to functions or one of the pointer registers to be "fixed". E.g. the Y 
> > register is required for the frame pointer.
> >
> > Yours,
> >
> > Björn Haase
> >
> > -----Ursprüngliche Nachricht-----
> > Von: address@hidden
> > [mailto:address@hidden Auftrag von Klaus Rudolph
> > Gesendet: Mittwoch, 1. Dezember 2004 15:49
> > An: address@hidden
> > Betreff: [avr-gcc-list] avr-gcc bug: global register variable is broken
> >
> >
> > Hi all,
> >
> > it is impossible to define a global register variable in avr-gcc.
> > The register is used by the compiler on some places and it is really not
> > preserved.
> >
> >
> > The actual gnu documentation says:
> > "Defining a global register variable in a certain register reserves that
> > register entirely for this use, at least within the current compilation. The
> > register will not be allocated for any other purpose in the functions in the
> > current compilation. The register will not be saved and restored by these
> > functions. Stores into this register are never deleted even if they would
> > appear to be dead, but references may be deleted or moved or simplified."
> >
> > I know that it is not ok to call to LIBRARY functions, but it is allowed to
> > use
> > all functions which are compiled with the knowledge of the reserved
> > register.
> > As a hint the documentation says that it is allowed to recompile all used
> > library functions with -ffixed-reg option to preserve the register for
> > use. 
> >
> > The bug is:
> > It is not possible to define a global register variable because it is used
> > from
> > ompiler.
> > It is impossible to preserve the register with -ffixed-reg option.
> >
> > Any idea for a workaround? I need some register vars for interrupt handler
> > :-(
> >
> > Is there allready a patch available or is this maybe fixed in current CVS
> > head of release 4.0 in gcc sources?
> >
> > Thanks
> >   Klaus
> >
> >
> >
> >
> >
> >
> >
> >
> >    
>
>
>
>