MicroBlaze port

Hi --

Attached are the final patches to the GCC trunk
and htdocs to add support for the Xilinx MicroBlaze
processor.  These patches have been updated to the
current head but are otherwise identical to the
patches which were previously approved.

The wwwdocs patches were approved by Gerald Pfeifer
off-list.

Rather than repost each patch individually with its
ChangeLog, the attached patch contains the ChangeLogs.
The ChangeLogs are the same as previously posted.

Please let me know ASAP if you see any problems
with checking these patches into GCC mainline.

On Tue, 28 Sep 2010, Michael Eager wrote:

> +extern char call_used_regs[];

Does this actually work now that call_used_regs is a macro in 
hard-reg-set.h?

> +extern int target_flags;

This will break when my patch 
<http://gcc.gnu.org/ml/gcc-patches/2010-09/msg02171.html> is in.  Don't 
explicitly declare any variables from .opt files.

> +/* This table translates options whose names would interfere
> +   with normal driver options and adds a 'Z' so that they can get to
> +   specs processing without interference.
> +
> +   The -xl-* options are deprecated and should be replace with their
> +   corresponding equivalent:
> +      -xl-mode-executable ==> -mxl-mode-executable
> +      -xl-mode-xmdstub    ==> -mxl-mode-xmdstub
> +      -xl-mode-bootstrap  ==> -mxl-mode-bootstrap
> +      -xl-mode-novectors  ==> -mxl-mode-novectors
> +*/
> +
> +#define TARGET_OPTION_TRANSLATE_TABLE \
> +  { "-xl-mode-executable", "-Zxl-mode-executable" }, \
> +  { "-xl-mode-xmdstub", "-Zxl-mode-xmdstub" },  \
> +  { "-xl-mode-bootstrap", "-Zxl-mode-bootstrap" }, \
> +  { "-xl-mode-novectors", "-Zxl-mode-novectors" }

Is it possible for you to use the .opt alias facility instead of 
TARGET_OPTION_TRANSLATE_TABLE?  If you make something an alias then you 
don't need to handle it in specs - you could alias these options directly 
to the -m versions, I'd have thought.  I'd like to get rid of 
TARGET_OPTION_TRANSLATE_TABLE.

> +#undef SWITCH_TAKES_ARG
> +#define SWITCH_TAKES_ARG(CHAR)						\
> +  (DEFAULT_SWITCH_TAKES_ARG (CHAR) || (CHAR) == 'G')

You're missing two other pieces of -G support: handling it in the 
handle_option hook, and use of g.opt in config.gcc.  See 
<http://gcc.gnu.org/ml/gcc-patches/2010-07/msg02340.html> and 
<http://gcc.gnu.org/ml/gcc-patches/2010-08/msg00066.html>.

Michael Eager wrote:
> Hi --
> 
> Attached are the final patches to the GCC trunk
> and htdocs to add support for the Xilinx MicroBlaze
> processor.  These patches have been updated to the
> current head but are otherwise identical to the
> patches which were previously approved.
> 
> The wwwdocs patches were approved by Gerald Pfeifer
> off-list.
> 
> Rather than repost each patch individually with its
> ChangeLog, the attached patch contains the ChangeLogs.
> The ChangeLogs are the same as previously posted.
> 
> Please let me know ASAP if you see any problems
> with checking these patches into GCC mainline.

Patches to support MicroBlaze have been checked in.

Michael,

My platform doesn't have FLOAT_MIN or FLOAT_MAX,  aren't the C/C++
standard names FLT_MIN and FLT_MAX?

I am also seeing cmpsi-2.c fail on execution but I haven't tracked
down the reason for that yet.

Steve Ellcey
sje@cup.hp.com


--- gcc/testsuite/gcc.c-torture/execute/cmpsf-1.c       (revision 0)
+++ gcc/testsuite/gcc.c-torture/execute/cmpsf-1.c       (revision 0)

+float args[] =3D
+{
+  0.0F,
+  1.0F,
+  -1.0F,=20
+  FLOAT_MAX,
+  FLOAT_MIN,
+  0.0000000000001F,
+  123456789.0F,
+  -987654321.0F
+};

On Tue, Sep 28, 2010 at 11:52 AM, Michael Eager <eager@eagerm.com> wrote:
> Hi --
>
> Attached are the final patches to the GCC trunk
> and htdocs to add support for the Xilinx MicroBlaze
> processor.  These patches have been updated to the
> current head but are otherwise identical to the
> patches which were previously approved.
>
> The wwwdocs patches were approved by Gerald Pfeifer
> off-list.
>
> Rather than repost each patch individually with its
> ChangeLog, the attached patch contains the ChangeLogs.
> The ChangeLogs are the same as previously posted.
>
> Please let me know ASAP if you see any problems
> with checking these patches into GCC mainline.
>

It caused:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45856

H.J. Lu wrote:

> 
> It caused:
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45856

I've updated the bug report with patches for the
test cases.  Tested on x86 and x86-64.