This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: incomplete ChangeLog entry for 31-Jul changes to sysv4.h?




On Tue, 14 Aug 2001, Geoffrey Keating wrote:

>
> Hi Dan,
>
> this change:
>
> 2001-07-31  Daniel Berlin  <dan@cgsoftware.com>
>
> 	PowerPC reorg and support for powerpc64-*-linux*.
>
> 	Also fixes emitting of constants on 32 bit and 64 bit
> 	platforms.
>
> 	* config.gcc: powerpc64-*-linux* is a new target.
> 	Things that needed aix.h now also include xcoff.h
>
> 	* config/rs6000/rs6000.h: Split XCOFF specific stuff into
> 	xcoff.h.
> 	Move AIX specific stuff into aix.h.
> 	(TARGET_AIX): Renamed to TARGET_XCOFF, since the AIX ABI is used
> 	with more than just XCOFF now.
> 	(SET_ASM_OP): Remove, now defined where needed.
> 	(FUNCTION_PROLOGUE): New macro definition.
> 	(FUNCTION_EPILOGUE): New macro definition.
> 	(ASM_OPEN_PAREN, ASM_CLOSE_PAREN): New macro definition.
>
> 	* config/rs6000/xcoff.h: New file.
>
> 	* config/rs6000/linux64.h: New file.
>
> 	* config/rs6000/darwin.h: Copy needed AIX alignment definitions.
>
> also defined DOUBLE_INT_ASM_OP in rs6000/sysv4.h, but there is no
> mention at all of sysv4.h in the ChangeLog.  Could you look at it?
I could have sworn i had written something about it.
I'll update the changelog entry
> Was this change supposed to be made?
Yes.
>  What is going on with
> DOUBLE_INT_ASM_OP, anyway?
>

Well, it's a .llong vs .quad issue. Before we just always used .llong for
AIX, and .quad for sysv4. However, since powerpc64 uses AIX ABI on ELF, it
should be using .quad. Since powerpc64 linux uses both aix.h and
sysv4.h, we'd get a mix of both depending on various factors (IE include
order/where a given macro came from).  You would also run into the issue
that some rs6000.h macros had chosen one format or the other, rather than
support both.
Since there was an INT_ASM_OP, and this is a "double int", it was called
DOUBLE_INT_ASM_OP.


> to CPP_ENDIAN_BIG_SPEC which I think are
> wrong, they revert an earlier patch which was a bugfix.
It's likely you are correct. I probably merged in the wrong direction
accidently.
I was trying to be very careful about them, but missed a few apparently.

>
> There is also a reformatted comment, which was reformatted to no
> longer fit in 80 columns; please don't do that.
>
Which comment?
I'll happily undo it.

> There was also a change to define ASM_CPU_SPEC and similar in sysv4.h,
> which I see was later backed out.
Yes.
>
> --
> Geoff Keating <geoffk@redhat.com>
>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]