New M68K target for GCC

[Chris Johns]

Joel Sherrill wrote:
>   
> Email Peter Barada and cc me.  I see patches from him on the gcc list 
> which are
> not in 4.1.1 which appear to be want you want.  Google "gcc coldfire 
> 528x" and
> you should get at least an ffs pattern.  I am curious what has become of his
> effort.  Maybe it is already on the head.

I thought the latest work is being done by Richard Sandiford. An example 
of a number of patches submitted this month and maybe of interest here is:

  http://gcc.gnu.org/ml/gcc-patches/2007-01/msg00764.html

Is Peter still active with Coldfires ?

> 
> Anyway gcc 4.1.1 already has a -m528x flag but I don't see much magic 
> based on that yet.
> 

I thought we have talked about this in the past when Eric was adding a 
special instruction for the 5282 processor. I cannot find any references 
to that discussion and cannot remember the detail.

If we start adding support for the possible Coldfire options as compiler 
flags for RTEMS we would have an explosion of libraries in a multilib 
build with little if any difference. Describing a device as a gcc option 
for a device's on-chip functions is a mistake. GCC should only deal with 
instruction set difference that the compiler uses including FPU. In the 
case of a special instruction for Eric it was a specific instruction for 
a specific function used in one place in RTEMS. It did not warrant a 
whole extra set of libraries.

Regards
Chris