Sparc PR: Is gnatsweb working?

[Jiri Gaisler]

Kenneth is right, cpukit/score/cpu/sparc/rtems/score/sparc.h
should be patched:

 -#if defined(_SOFT_FLOAT)
 -#define SPARC_HAS_FPU 0
 -#else
  #define SPARC_HAS_FPU 1
 -#endif

  #if SPARC_HAS_FPU
  #define CPU_MODEL_NAME "w/FPU"


The reason is (as we have dscussed on the list before) that the
FPU context save code is always present (in assembly) and the
prescense of an FPU is detrimed at run-time. The kernel is
always compiled with -msoft-float to avoid integer use of
FPU registers by gcc optimisations. This also leads to that
only one bsp is needed for leon's with and without FPU.

Jiri.


Joel Sherrill wrote:
> Kenneth J. Peters wrote:
> 
>> I have tried three times over the last few days to submit a PR, but
>> although the gnatsweb page acknowledges my submission, I do not get a
>> confirming email, and the PR does not appear. On  my first attempt I
>> also submitted a different PR, which did go into the system.
>>
>> The PR I am trying to submit applies to all sparc targets, the basic
>> info is:
>>
>> PR827 patch not fully applied to CVS tree (4.6.6 and CVS head)
>>
>> The PR827 patch for file cpukit/score/cpu/sparc/rtems/score/sparc.h
>> has not been applied in the CVS tree. The result is that sparc
>> applications unexpectedly do not use the FPU.
>>
>> I inspected the PR827 patches for all non-leon, non-network files, and
>> did not see any other significant missing pieces. I don't know if
>> there are any missing PR827 patches in the leon and network files.
>>
> 
> We won't be using GNATS forever but it does seem to still be working.  I
> saw the emails on
> the PRs.  I was watching them to see what you two decided. :)
> 
> --joel
> 
>> Ken Peters
>> Ken.Peters at jpl.nasa.gov
>>
> 
> .
>