Re: RFA: MN10300: Fix frame related notes and unwind info

[Nick Clifton <nickc at redhat dot com>]

Hi Richard,

  Thanks for the patch review.  I have attached a revised patch which 
covers all of the points you raised except for:

> Consider marking the epilogue as well, though usually that's
> not as simple as setting RTX_FRAME_RELATED_P.  You generally
> have to use most of the REG_CFA_* specific notes too.

I was not confident that I would get this right, and my goal at the 
moment is to get the MN10300 port to build and run the gcc/g++ 
testsuites.  I have some major changes to submit to the MN10300 backend, 
and I would like to get those in before I look at perfecting the debug 
information generation.

The revised patch also includes one small change that was not in the 
initial version - it disables the combine-stack-adjust pass.  This pass 
is breaking the cc0-setter/cc0-user relationship (for the MN10300), but 
since I have a future patch to submit that completely removes the use of 
cc0, I have not investigated any further.

OK to apply ?

Cheers
  Nick

gcc/ChangeLog
2010-10-05  Nick Clifton  <nickc@redhat.com>

	* config/mn10300/mn10300.h (FIRST_PSEUDO_REGISTER): Increment by
        one.
        (MDR_REGNUM): Define.
        (FIXED_REGISTERS, CALL_USED_REGISTERS): Add MDR as a fixed
        register.
        (REG_CLASS_CONTENTS): Add MDR to ALL_REGS.
        (INCOMING_RETURN_ADDR_RTX): Define in terms of MDR.
        (REGISTER_NAMES): Add MDR.
        (DWARF2_DEBUGGING_INFO): Define to 1.
        * config/mn10300/mn10300.c (TARGET_EXCEPT_UNWIND_INFO): Define.
        (F): New function.  Sets RTX_FRAME_RELATED_P.
        (mn10300_gen_multiple_store): Use F.
        (expand_prologue): Use F.  Use gen_movsf() to push floating
        point registers.
        (expand_epilogue): Use gen_movsf() to pop floating point
        registers.
	(mn10300_option_override): Disable combine stack adjust pass.

Attachment: mn10300.fsf.patch.2.bz2
Description: application/bzip