[avr] Patch for -mint8 option

[avr] Patch for -mint8 option

[Svein E. Seldal]

[-- Attachment #1: Type: text/plain, Size: 626 bytes --]

Hello,

I have put together a little patch for setting the size of the long long 
type when the -mint8 option is used. As there are no 32-bits type in 
-mint8 mode, and the fact that none of the other sizeof's are compliant 
to the C-specification, I think that it would be appropriate to do this.

I have also modified the specs to define the symbol __AVR_INT8__, making 
it possible for applications to differentiate when -mint8 are used or not.


Svein

gcc/ChangeLog:
2002-11-25  Svein E. Seldal  <Svein.Seldal@solidas.com>

	* config/avr/avr.h: Reduce the long long type to 32-bit in
	-mint8 mode and define __AVR_INT8__

[-- Attachment #2: avr-mint8.diff --]
[-- Type: text/plain, Size: 2036 bytes --]

Index: gcc/config/avr/avr.h
===================================================================
RCS file: /prosjekt/gnu/gcc/gcc/gcc/config/avr/avr.h,v
retrieving revision 1.71
diff -c -3 -p -r1.71 avr.h
*** gcc/config/avr/avr.h	24 Sep 2002 12:48:52 -0000	1.71
--- gcc/config/avr/avr.h	20 Nov 2002 10:40:46 -0000
*************** Boston, MA 02111-1307, USA.  */
*** 38,43 ****
--- 38,45 ----
  	builtin_define ("__AVR_MEGA__");	\
        if (TARGET_NO_INTERRUPTS)			\
  	builtin_define ("__NO_INTERRUPTS__");	\
+       if (TARGET_INT8)                          \
+         builtin_define ("__AVR_INT8__");        \
      }						\
    while (0)
  
*************** extern int avr_asm_only_p;
*** 204,210 ****
     used in `cpp'.  */
  
  
! #define LONG_LONG_TYPE_SIZE 64
  /* A C expression for the size in bits of the type `long long' on the
     target machine.  If you don't define this, the default is two
     words.  If you want to support GNU Ada on your machine, the value
--- 206,212 ----
     used in `cpp'.  */
  
  
! #define LONG_LONG_TYPE_SIZE ( INT_TYPE_SIZE == 8 ? 32 : 64 )
  /* A C expression for the size in bits of the type `long long' on the
     target machine.  If you don't define this, the default is two
     words.  If you want to support GNU Ada on your machine, the value
*************** extern int avr_case_values_threshold;
*** 2508,2514 ****
     (and ANSI C) library functions `memcpy' and `memset' rather than
     the BSD functions `bcopy' and `bzero'.  */
  
! #define CPP_SPEC "%{posix:-D_POSIX_SOURCE}"
  
  /* A C string constant that tells the GNU CC driver program options to
     pass to CPP.  It can also specify how to translate options you
--- 2510,2516 ----
     (and ANSI C) library functions `memcpy' and `memset' rather than
     the BSD functions `bcopy' and `bzero'.  */
  
! #define CPP_SPEC "%{posix:-D_POSIX_SOURCE} %{mint8:-D__AVR_INT8__}"
  
  /* A C string constant that tells the GNU CC driver program options to
     pass to CPP.  It can also specify how to translate options you

unreviewed patch

[Svein E. Seldal]

Hi,

This patch has not been reviewed yet. Any news?

http://gcc.gnu.org/ml/gcc-patches/2002-11/msg01502.html


Svein

[avr] Patch for -mint8 option -- unreviewed patch

[Svein E. Seldal]

Hi all,

I'm still waiting for approval to check this in:

http://gcc.gnu.org/ml/gcc-patches/2002-11/msg01502.html

I was hoping on getting this into 3.3, but I totally forgot about it...


Regards,
Svein

unreviwed patch -- 3rd attempt (since 2002-11)

[Svein E. Seldal]

Svein E. Seldal wrote:
> 
> Hi all,
> 
> I'm still waiting for approval to check this in:
> 
> http://gcc.gnu.org/ml/gcc-patches/2002-11/msg01502.html

Regards,
Svein

Re: [avr] Patch for -mint8 option -- unreviewed patch

[Jim Wilson]

Using 32-bit long long with an 8-bit target is sensible, and is already
done for some other 8-bit targets.  So the concept is OK.

You are defining __AVR_INT8__ in two places, which seems redundant.  You
only need it in one place.  The correct place would be the first one in
TARGET_CPU_CPP_BUILTINS.

The ChangeLog entry isn't correct.  You need to list every macro
modified, and briefly describe the change.  You haven't done this.  So
it should look something like this
* config/avr/avr.h (TARGET_CPU_CPP_BUILTINS): Define __AVR_INT8__
(LONG_LONG_TYPE_SIZE): Check INT_TYPE_SIZE.

The patch is OK with these changes.

The avr has two listed maintainers, so I included them in the cc's in
case they want to comment.

I was going to suggest that a doc change was missing, but then I noticed
that the -mint8 option isn't documented.  Perhaps you could add the
missing documentation?

Jim