Re: [avr-libc-dev] [bug #12941] Table of Signal Names in "Interrupts and Signals" is wrong/incomplete

[Rolf Ebert]

On Thu, 5 May 2005 08:33:45 +1100, Dmitry K. <address@hidden>  
wrote:

> On Wednesday 04 May 2005 08:26, Rolf Ebert wrote:
>
> > * the interrupt short names in the XML files sometimes contain funny
> > characters that cannot be part of an identifier.  I replace all of  
> > comma,
> > blank, or slash to underscores in the Python script.  I don't know why
> > Atmel created interrupt names containing these characters in the first
> > place...
> >
> >      Rolf
>
>
> First, the speech is about INTERRUPT SERVICE ROUTINES names. Not about of
> electrical signals or human readable designations.

Yes, that's what I was talking about. Excuse my sloppy wording.

> Let us look ATtiny2313 datasheet (next page after Table 21).
> ISR names differ from specified above:
>     ...
>     rjmp   INT0
>     ...
>     rjmp   TIM1_CAPT
>     ...
>     rjmp   TIM1_OVF

This is sample assembler code.  In assembler you can use whatever name  
pleases you.  The rjmp labels are not related in any way to the names of  
the interrupts.  Perhaps that is what you wanted to say above.

gcc doesn't have the same freedom for choosing names of the interrupt  
service routines.  Their internal names are fixed to __vector_<xx>  
anyway.  The corresponding io header file predefines the names that are to  
be used in "SIGNAL(<name>)".  Certainly it makes sense to call the ISR the  
same as the triggering interrupt. AVR libc prepends SIG_ to all names, it  
could as well append _HANDLER, that's not the point.

I still think the ISR name should be based on the documented (in the Atmel  
data sheets) interrupt names.  The XML for the three mentioned interrupts  
of attiny2313 reads:

<VECTOR2>
 <PROGRAM_ADDRESS>$001</PROGRAM_ADDRESS>
 <SOURCE>INT0</SOURCE>
 <DEFINITION>External Interrupt Request 0</DEFINITION>
</VECTOR2>
...
<VECTOR4>
 <PROGRAM_ADDRESS>$003</PROGRAM_ADDRESS>
 <SOURCE>TIMER1 CAPT</SOURCE>
 <DEFINITION>Timer/Counter1 Capture Event</DEFINITION>
</VECTOR4>
...
<VECTOR6>
 <PROGRAM_ADDRESS>$005</PROGRAM_ADDRESS>
 <SOURCE>TIMER1 OVF</SOURCE>
 <DEFINITION>Timer/Counter1 Overflow</DEFINITION>
</VECTOR6>

Note the blanks in the values of SOURCE of vector4 and vector6.


The Atmel assembler include file (AvrAssembler/Appnotes/tn2313def.inc)  
reads:

.equ    INT0addr        = 0x001     ;External Interrupt0
...
.equ    ICP1addr        = 0x003     ;Input capture interrupt 1
...
.equ    OVF1addr        = 0x005     ;Overflow1 Interrupt

It claims to be autogenerated from the XML file. Although, I don't know  
which algorithm generates "OVF1addr" from the XML info above...


> Certainly, it would be good to use names from Atmel's datasheets from the
> first. But, unfortunately, Atmel is very inaccurate in ISR's names.
> Identical subroutines have different names for different chips (for  
> example,
> EE_RDY in AT90S4433 and EE_READY in ATtiny2313). Sometimes the name  
> clashes
> with other identificators (INT0 above).
>
> Dmitry.

It won't clash with a prefix like SIG_ to all names, as it is right now.

Different names for different chips triggered by the same interrupt is  
another topic.  Personally, as a programmer I'd rather look up the name of  
an identifier in the doc, i.e. data sheet, than to guess the name from  
another chip.  But then, there are projects that want to use a single  
piece of code for several chips and it isn't beautiful to have #ifdefs all  
over the place.

     Rolf