svn commit: r221614 - projects/largeSMP/sys/powerpc/include

[Attilio Rao]

2011/5/12 Artem Belevich <art at freebsd.org>:
> On Thu, May 12, 2011 at 3:22 PM, Attilio Rao <attilio at freebsd.org> wrote:
>> 2011/5/12 Artem Belevich <art at freebsd.org>:
>>> On Thu, May 12, 2011 at 5:15 AM, Attilio Rao <attilio at freebsd.org> wrote:
>>>> 2011/5/8 Attilio Rao <attilio at freebsd.org>:
>>>>> Author: attilio
>>>>> Date: Sun May  8 00:39:49 2011
>>>>> New Revision: 221614
>>>>> URL: http://svn.freebsd.org/changeset/base/221614
>>>>>
>>>>> Log:
>>>>>  All architectures define the size-bounded types (uint32_t, uint64_t, etc.)
>>>>>  starting from base C types (int, long, etc).
>>>>
>>>> mips seems having the same issue, so here is my patch:
>>>> http://www.freebsd.org/~attilio/largeSMP/mips-atomic.diff
>>>
>>> I see at least one problem. N32 ABI is ILP32 even though it can use
>>> 64-bit registers for "long long".
>>
>> I'm sorry but this _longlong is non standard for our atomic scheme,
>> nothing in the kernel uses it and in general I'd say to not add them
>> if not strictly necessary.
>> Said that, in the -CURRENT code it doesn't seem present, or I'm missing it?
>
> The key is that N32 ABI does use 64-bit registers and can perform
> 64-bit atomic ops.
> Unfortunately, 64-bit type is long long or [u]int64_t. The point is
> that it's not *long*.
>>
>>>> #if defined(__mips_n64) || defined(__mips_n32)
>>>> static __inline void
>>>>-atomic_set_64(__volatile uint64_t *p, uint64_t v)
>>>>+atomic_set_long(__volatile u_long *p, u_long v)
>>>
>>> Using _long here would not match the assembly on N32.
>>>
>>> If you stick with your changes, then you should drop __mips_n32 from
>>> the ifdef above.
>>> You may also want to add atomic_*_longlong for n32. Actually, for o64 as well.
>>
>> I'm not entirely sure in which way the above is different... or at
>> least, I may have overlooked the _long support for __mips_n32... can
>> you please elaborate a bit more?
>
> The inline assembly for atomic_set_long above uses 64-bit load/store.
> This is not going to work for 32-bit long in N32 ABI.

That cames from my mips assembly ignorance, I apologize.
I just gave a look at types.h and what it just I basically see is:
- __mips_n32 is just the same as plain !__mips_n32 && !__mips_n64
- __mips_n64 just addes the operations for 64 bits operands

Thus what I need to make sure is:
- __mips_n32 (and !__mips_n32 && !__mips_n64) case work on 32 bits
operand on long too
- __mips_n64 works on 64 bits long operands

Do you have a specifi opinion on _char/_short operands?
Do you have strong feeling on the _longlong stuff? Do you think it
will be useful in mips?

About the last question I think that it would be more fair to use _64
directly, when possible.

Thanks,
Attilio


-- 
Peace can only be achieved by understanding - A. Einstein