[U-Boot] [PATCH 2/8] armv7: cache maintenance operations for armv7

[Albert ARIBAUD]

Le 13/01/2011 13:05, Aneesh V a écrit :

>>> What I need is something like below:
>>>
>>> #define get_bit_field(nr, start, mask)\
>>> (((nr) & (mask)) >> (start))
>>>
>>> #define set_bit_field(nr, start, mask, val)\
>>> (nr) = ((nr) & ~(mask)) | (((val) << (start)) & (mask))
>>>
>>> Can these go in a generic header? If so, can I add them to
>>> "include/linux/bitops.h"
>>
>> After some more thought, I am wondering if a *generic* field setting and
>> getting macro is really useful. So far everyone is fine with at most
>> defining field-specific macros.
>
> Is it going to be easy if you have many fields to deal with?

I don't see how the generic macros ease anything. Instead of defining say

#define get_field_F(x)  ((x >> F_start) & F_mask)
#define set_field_F(x,v)  { x = (x ~ F_mask ) | (v << F_start) }

You'd have

#define get_field_F(x)  get_bit_field(x, F_start, F_mask)
#define set_field_F(x,v) set_bit_field(x, F_start, F_mask, v);

Which does not seem to bring any simplicity to me.

> However, I agree that the above may be specific to our needs.
>
> What may be of more generic interest may be something like this with
> the mask automatically generated:
> #define get_bit_field(nr, start, end)
> #define set_bit_field(nr, start, end, val)
>
> However, in our case I am already given the mask and start position for
> each field (automatically generated from hw database). So, I prefer the
> former versions.
>
> If that doesn't look useful for generic use I will put them in
> OMAP specific headers.
>
> Best regards,
> Aneesh

Amicalement,
-- 
Albert.