gcc 3.3.5 arm-elf .. which multilib for soft-float?

[Joel Sherrill <joel@OARcorp.com>]

Ralf Corsepius wrote:
> On Tue, 2004-11-02 at 13:33, Richard Earnshaw wrote:
> 
>>On Tue, 2004-11-02 at 12:24, Joel Sherrill  wrote:
>>
>>>Richard Earnshaw wrote:
>>>
>>>>On Tue, 2004-11-02 at 00:21, Joel Sherrill  wrote:
>>>>
>>>>
>>>>>Hi,
>>>>>
>>>>>It looks like btween gcc 3.3.3 and 3.3.5, the file config/arm/t-arm-elf
>>>>>was reworked.  I am getting linking errors which indicate that the
>>>>>default libc.a is hard-float.
>>>>>
>>>>>The gcc documentation states that -mhard-float is the default for the
>>>>>compiler and I only see these libc.a's
>>>>>
>>>>
>>>>
>>>>-mhard-float has *never* been the default for the arm-elf
>>>>configuration.  In fact, AFAIK the only supported configuration of gcc
>>>>on ARM that still uses the FPA hard-float format is Linux.
>>>
>>>I rather suspected this because I didn't recall any RTEMS user CPU model
>>>I had ever looked at having hard float.
>>>
>>>But the gcc documentation states that hard-float is ALWAYS the default:
>>
>>The documentation is out of date.  There is a time when that was correct
>>for all the (then) supported configurations, but it's now far from the
>>truth.
> 
> 
> OK, so I understand (and from what I see in gcc-3.4.x's source code),
> the default for the arm code generation now is soft-float?
> 
> 
> However, gcc-3.4.x seems to miss to pass this on to the assembler:
> 
> Some experiments with an arm-rtems-gcc using this multilib setup:
> MULTILIB_OPTIONS     = marm/mthumb msoft-float/mhard-float
> MULTILIB_DIRNAMES    = arm thumb soft fpu
> MULTILIB_EXCEPTIONS  = *mthumb/*mhard-float*
> MULTILIB_MATCHES     = marm=mno-thumb
> 
> shows this:
> 1.
> # arm-rtems4.7-gcc -v -msoft-float ...
> ...
> /opt/rtems-4.7/arm-rtems4.7/bin/as --gdwarf2 -mfpu=softfpa ...
> 
> 2.
> # arm-rtems4.7-gcc -v -mhard-float ...
> ...
> /opt/rtems-4.7/arm-rtems4.7/bin/as --gdwarf2 ...
> 
> 3.
> # arm-rtems4.7-gcc -v ...
> ...
> /opt/rtems-4.7/arm-rtems4.7/bin/as --gdwarf2 ...
> 
> => arm-elf/gcc misses to pass -mfpu=softfpa for the "compiler's
> default".
> 
> I.e. IMO, the actual cause for our problems is an inconsistency between
> the GCC's defaults and "as"'s default.
> 
> IMO, this can be regarded from two POVs:
> a) either arm/GCC's ASM_SPECS should be changed to reflect GCC other
> flags having been changed
> b) or binutils's default has to be changed to "-mfpu=softfpa" 
> 
> I am currently trying to implement a) into arm/rtems-elf.h for
> gcc-3.4.x.

Personally I think that is preferable because it fixes the problem
without depending on having a specific binutils version/behavior.

I know a port may have depend on having a binutils of a specific vintage
to get a new instruction and that seems unavoidable.  But in this case
it is safer to pass flags and force a behavior than depending on
default.  If you changed binutils to match gcc's assumption, gcc would 
be depending on the behavior of an unreleased
binutils and be broken with all released versions.

> Ralf

--joel