Re: [KJ][RFC][PATCH] BIT macro cleanup

From: Richard Knutsson
Date: Fri Feb 23 2007 - 14:11:46 EST

Dmitry Torokhov wrote:
On 2/23/07, Richard Knutsson <ricknu-0@xxxxxxxxxxxxxx> wrote:
Dmitry Torokhov wrote:
> I was not talking about name (I hate BITWRAP) but behavior.
Oh, my bad :)
>> but mainly since it only enables wrapping of the long-type.
> I'd provde BIT and separate LLBIT for ones who really need long long.
> People who intereseted in smaller than BITS_PER_LONG bitmaps shoud use
> your proposal - BIT(x % DESIRED_WITH) and BIT should do modulo
> BITS_PER_LONG internally.
I agree that _if_ there is a "BITWRAP" then it should be long, but I
don't see the reason for it to be in bitops.h when it is only input.h
that uses it. + I find it different with BIT since it works as well with
'char' as 'long'.
Also, I think it would be best if the name indicated it is a 'long'.

Am a little bit curious why you would like it in bitops.h, but won't
complain if you do (think you have noticed my view of it ;))

Hm, I thought as was clear, but apparently I messed up explaining my position:

1. I don't like BITWRAP name at all and I don't want anything like
that near input code. I think BIT is just fine.
Oh, I think I understand now. So the (in input.h):
#undef BIT
#define BIT(...
business is what you want to do? Well, that I will not object to. Your patch with:
+#define BIT(nr) (1UL << (nr))
+#define LLBIT(nr) (1ULL << (nr))
+#define BITWRAP(nr) (1UL << ((nr) % BITS_PER_LONG))
in bitops.h made me believe the #undef in input.h was just a temporarily thing.

2. I don't want to use BIT(x % BITS_PER_BITLONG) as it will
significantly litter code in the input drivers. You want see whta bits
you are actually setting behind all these "% BITS_PER_BITLONG".
As I said before, I thought it should be defined as BITSWAP (or whatever) in input.h and then there is no more "% BITS_PER_LONG" litter. But redefining BIT seems like an equally good idea;
+ eas(y/ier) to understand and simple to implement
- another definition of BIT.

3. I think most of users could use input's implementation of BIT,
possibly using BIT(x % BM_WIDTH) format to further limit width of the
bitmap if needed.

4. LLBIT should be provided to users who really want long long.
Agreed. (As in the case of "BIT(x) (0x800...00ULL >> (x)) )

Richard Knutsson

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at