971023 CPU32 patch

[Robin Kirkham]

Herewith a patch for impoving egcs's external behaviour with respect to the
CPU32, in generic cross-compiler configurations (e.g., m68k-coff):

    o the flag -mcpu32 by default no longer emits 68881 floating-point
      instructions
    o -mcpu32 defines preprocessor symbols __mcpu32__, __mcpu32 (and mcpu32,
      without -ansi, as per normal)
    o -m68332 still works, but is now redundant. It defines __m68332__ and so
      on, along with __mcpu32__ etc but is otherwise the same as -mcpu32 
    o the CPU32 now has its own multilib
    o a redundant m68000/msoft-float multilib is no longer built

I think the redundant -m68332 and -m68302 flags could be removed (-m68302 is
already identical to -m68000, except for a preprocessor symbol). I was talked 
out of providing -m switches for the whole 683xx family, and I think it
should be "all or none".

This patch does not alter the internal gcc code-generator behavour, in so
far as it still regards the CPU32 as a 68020 variant, equivalent to by
default -m68020 -mno-bitfields -msoft-float.

For this patch to work, you need to use a very recent binutils snapshot, or
apply the binutils-2.8.1 patch below. This is to prevent gas complaining
about certain "tst.l %a0" instructions.

I've tested this patch on egcs 971023 and 971016.

Robin Kirkham			CSIRO Manufacturing Science and Technology
Project Engineer		Locked Bag 9, Preston 3072, Australia
robin.kirkham@mlb.dmt.csiro.au	Phone: +61 3 9662-7756  Fax: +61 3 9662-7851



-------------- next part --------------
A non-text attachment was scrubbed...
Name: bin00000.bin
Type: text/x-diff
Size: 16006 bytes
Desc: not available
URL: <https://gcc.gnu.org/pipermail/gcc/attachments/19971024/cbaa9d66/attachment.bin>