[power-gnu-discuss] Support for the MPC5554 (e200z6 core)

[David Brown]

Andrew Dennison wrote:
> On Tue, Jun 2, 2009 at 7:25 AM, David Brown <david at westcontrol.com> wrote:
>> Hi,
>>
>> I have a customer who is looking at using a Freescale MPC5554 in an
>> application, and is looking for appropriate tools.  I've been having a look
>> at the Lite edition of the CodeSourcery PPC gcc toolset. Unfortunately, I've
>> run into a big problem.
>>
>> The MPC5554 uses the e200z6 core, which is fairly similar to the e300c2 core
>> supported by the "-mcpu=e300c2" flag.  The problem with both these cores is
>> that they support hardware floating point for single precision arithmetic
>> using the "SPE" (Signal Processing Extension) instructions, with the
>> floating point data held in general purpose registers - they do not support
>> the Book E floating point registers and instructions.  My problem is that I
>> cannot get gcc to generate spe instructions for hardware floating point
>> code.
>>
>> Does anyone know if current gcc versions support using spe instructions for
>> hardware floating point, or if there is likely to be any support soon?  If I
>> can't find a way to generate the spe instructions from gcc, I'm going to
>> have to recommend my customer buys Code Worrier...
> 
> -fshort-double triggers gcc to emit the single precision floating
> point instructions.
> 

Yes, I've thought of that.  It could be useful if I were converting old 
code that used doubles, and wanted only single precision code, but it 
doesn't help here.

mvh.,

David