Re: [Qemu-devel] [4932] Preliminary PPC64/Linux host support

[malc <av1474@comtv.ru>]

On Sat, 26 Jul 2008, Andreas F?rber wrote:

>
> Am 25.07.2008 um 20:18 schrieb malc:
>
>> On Thu, 24 Jul 2008, C.W. Betts wrote:
>> 
>>> Building it under OS X 10.5.4 doesn't work.  After editing the config-host 
>>> files, I get the following:
>>> In file included from 
>>> /Users/cwbetts/makestuff/qemu-allmac/src/exec-all.h:320,
>>>              from /Users/cwbetts/makestuff/qemu-allmac/src/vl.c:136:
>>> /Users/cwbetts/makestuff/qemu-allmac/src/qemu-lock.h:216:2: error: #error 
>>> unimplemented CPU support
>>> 
>>> After adding
>>> #define __powerpc__ 1
>>> #define __powerpc64__ 1
>
> I don't think it's is a good idea to define these as such, it may confuse 
> other software. You should check the places where they are used and add 
> appropriate checks for either __ppc__ or __ppc64__ respectively (they are 
> never defined at the same time, contrary to Linux).

You are replying to both me and C.W. Betts this is somewhat confusing.

> In TCG Spotlight came up only with this:
>
> diff --git a/tcg/tcg.h b/tcg/tcg.h
> index bc5b902..860bce9 100644
> --- a/tcg/tcg.h
> +++ b/tcg/tcg.h
> @@ -413,7 +413,7 @@ uint64_t tcg_helper_divu_i64(uint64_t arg1, uint64_t 
> arg2);
> uint64_t tcg_helper_remu_i64(uint64_t arg1, uint64_t arg2);
>
> extern uint8_t code_gen_prologue[];
> -#if defined(__powerpc__) && !defined(__powerpc64__)
> +#if (defined(__powerpc__) || defined(__ppc__)) && !defined(__powerpc64__)
> #define tcg_qemu_tb_exec(tb_ptr) \
>    ((long REGPARM __attribute__ ((longcall)) (*)(void 
> *))code_gen_prologue)(tb
> #else
>
> and it shouldn't affect ppc64. Not sure if it's needed at all for OSX? What's 
> it supposed to change?
>
>>> It chokes up under
>>> In file included from /usr/include/inttypes.h:255,
>>>              from /Users/cwbetts/makestuff/qemu-allmac/src/cpu-defs.h:29,
>>>              from 
>>> /Users/cwbetts/makestuff/qemu-allmac/src/target-i386/exec.h:30,
>>>              from /Users/cwbetts/makestuff/qemu-allmac/src/cpu-exec.c:22:
>>> /usr/lib/gcc/powerpc-apple-darwin9/4.2.1/include/stdint.h:35: error: 
>>> conflicting types for ?int64_t?
>>> /Users/cwbetts/makestuff/qemu-allmac/src/dyngen-exec.h:59: error: previous 
>>> declaration of ?int64_t? was here
>>> /usr/lib/gcc/powerpc-apple-darwin9/4.2.1/include/stdint.h:55: error: 
>>> conflicting types for ?uint64_t?
>>> /Users/cwbetts/makestuff/qemu-allmac/src/dyngen-exec.h:42: error: previous 
>>> declaration of ?uint64_t? was here
>>> make: *** [cpu-exec.o] Error 1
>> 
>> Uh, weird,
>
> Well, afaik dyngen didn't have support for ppc64...

No that's not what i find weird, the weird part is darwin elects to use
(unsigned) long long for (u)int64_t while Linux uses plain longs.

>> what are [u]int64_t are defined to on darwin?
>
> uint64_t is defined to u_int64_t (stdint.h), which is defined to unsigned 
> long long (ppc/types.h).
> int64_t is defined to long long (ppc/types.h).
>
>>> I know that this isn't exactly designed for PPC Macs, but the tcg files 
>>> compiled without a problem.
>> 
>> What gives you that idea?
>
> He probably meant designed for PPC Mac OS X - you said so, and it doesn't 
> work yet. See my patch, there are currently numbers hardcoded for Linux.

It's not just about numbers, last time i looked (cursory) the ABIs where
a lot different (PPC32 case) just changing the numbers wont get one very
far i think.

-- 
mailto:av1474@comtv.ru