very basic porting

very basic porting

[Gabriele Caracausi]

Hello All,

i'm working to Microchip PIC18 porting for university purposes...
I've written the four files (.h, .md, .c and protos), studied for 9 months
and created cc1 for pic18.

But when I try to assign a value to a variable the cc1 sends me an error in
prepare_cmp_insn - optabs.c (line 3505).
I've setted cmpqi e cmphi RTL patterns... but in your opinion, the problem
is in .md or in .h file? Or what can be the problem?

Someone has a very basic porting of other 8bit microcontrollers (like ip2k
or avr) to take as example to write the minimal set of code in .md and .h
files?


Thanks a lot.
Write me for all types of questions!
Gabriele Caracausi
Italy

Re: very basic porting

[Jan Hubicka]

> Hello All,
> 
> i'm working to Microchip PIC18 porting for university purposes...
> I've written the four files (.h, .md, .c and protos), studied for 9 months
> and created cc1 for pic18.
> 
> But when I try to assign a value to a variable the cc1 sends me an error in
> prepare_cmp_insn - optabs.c (line 3505).

Can you, for start, show some context of line 3505 in your optabs.c?  In
current mainline, prepare_cmp_insn has 3 different aborts in it none at
line 3505.  The line numbers depend on version of compiler you do use...

Honza
> I've setted cmpqi e cmphi RTL patterns... but in your opinion, the problem
> is in .md or in .h file? Or what can be the problem?
> 
> Someone has a very basic porting of other 8bit microcontrollers (like ip2k
> or avr) to take as example to write the minimal set of code in .md and .h
> files?
> 
> 
> Thanks a lot.
> Write me for all types of questions!
> Gabriele Caracausi
> Italy
> 

R: very basic porting

[Gabriele Caracausi]

> > Hello All,
> >
> > i'm working to Microchip PIC18 porting for university purposes...
> > I've written the four files (.h, .md, .c and protos), studied
> for 9 months
> > and created cc1 for pic18.
> >
> > But when I try to assign a value to a variable the cc1 sends me
> an error in
> > prepare_cmp_insn - optabs.c (line 3505).
>
> Can you, for start, show some context of line 3505 in your optabs.c?  In
> current mainline, prepare_cmp_insn has 3 different aborts in it none at
> line 3505.  The line numbers depend on version of compiler you do use...

The version in 3.3. The line is

  if (class == MODE_FLOAT)
    prepare_float_lib_cmp (px, py, pcomparison, pmode, punsignedp);

  else
    abort ();

... for me the problem is that in previous lines the function doesn't return
because not one of conditions listed occurs.
What do you think? For me the problem is in my .h configuration... what do
you think? Have you an idea?

Thanks. Gab.

Re: R: very basic porting

[Jim Wilson]

Gabriele Caracausi wrote:
> The version in 3.3. The line is
> 
>   if (class == MODE_FLOAT)
>     prepare_float_lib_cmp (px, py, pcomparison, pmode, punsignedp);
> 
>   else
>     abort ();

This isn't enough information.  We need to know why none of the previous 
conditions matched.  For instance, what is the value of mode?  What are 
the values of the cmp_optab->handlers[] array?
-- 
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com

Re: R: very basic porting

[Gabriele Caracausi]

Jim Wilson writes: 

> Gabriele Caracausi wrote:
>> The version in 3.3. The line is 
>> 
>>   if (class == MODE_FLOAT)
>>     prepare_float_lib_cmp (px, py, pcomparison, pmode, punsignedp); 
>> 
>>   else
>>     abort ();
> 
> This isn't enough information.  We need to know why none of the previous 
> conditions matched.  For instance, what is the value of mode?  What are 
> the values of the cmp_optab->handlers[] array?

You are right but fortunately yesterday afternoon i've resolved the 
problem!!! :) infact in my .md file I haven't declared all conditional 
branch insns. 

Thank you all.
Gabriele.