On 8 September 2015 at 00:07, Maciej W. Rozycki <macro@xxxxxxxxxxxxxx> wrote:
> On Mon, 7 Sep 2015, Yousong Zhou wrote:
>
>> diff --git a/arch/mips/include/uapi/asm/swab.h b/arch/mips/include/uapi/asm/swab.h
>> index 8f2d184..8b9a390 100644
>> --- a/arch/mips/include/uapi/asm/swab.h
>> +++ b/arch/mips/include/uapi/asm/swab.h
>> @@ -8,6 +8,11 @@
>> #ifndef _ASM_SWAB_H
>> #define _ASM_SWAB_H
>>
>> +/*
>> + * Enable the optimized version only when compiling without MIPS16.
>> + */
>> +#ifndef __mips16
>> +
>> #include <linux/compiler.h>
>> #include <linux/types.h>
>>
>> @@ -66,4 +71,5 @@ static inline __attribute_const__ __u64 __arch_swab64(__u64 x)
>> #define __arch_swab64 __arch_swab64
>> #endif /* __mips64 */
>> #endif /* MIPS R2 or newer or Loongson 3A */
>> +#endif /* ifndef __mips16 */
>> #endif /* _ASM_SWAB_H */
>
> I think it will best go with the main #if which checks the conditions
> that have to be met for this optimisation to be possible; there is no gain
> from nesting the conditions here.
>
Yes, you are right, and I should not wrap that __SWAB_64_THRU_32__
inside the #if check.
> Also you need a second patch paired with this to undo your previous
> `nomips16' change which will be no longer needed (and in the case of the
> `.set' part not wanted either).
Wow, I did not know that my previous patch was merged, also into all
those stable branches, and some with kind minor modifications... Is
it the case that I only need to firstly revert that commit in
mips-upstream-sfr/mips-for-linux-next, then followed by refreshed
version of this one?
Regards,
yousong
