Re: [PATCH] mask ADT: new mask.h file [2/22]

[Paul Jackson <pj@sgi.com>]

In my previous reply to Rusty, I wrote:
> 	    struct cpumap { DECLARE_BITMAP(bits, NR_CPUS); };
> 	    struct cpumask s, d1, d2;
> 	    bitmap_or(s.bits, d1.bits, d2.bits);

Brain dead code alert (as well as a typo alert for the 'cpumap') - that
last line needs to be:

	    bitmap_or(s.bits, d1.bits, d2.bits, NR_CPUS);

Which is why the 60 odd cpumask and nodemask specific operation macros
exist, to avoid having to explicitly specify the bitsize on each call

In other words, I understand that the following three possibilities
exist for coding these masks:

    /* specify bitsize both on declarations and operations */
    struct cpumask { DECLARE_BITMAP(bits, NR_CPUS); };
    struct cpumask s, d1, d2;
    bitmap_or(s.bits, d1.bits, d2.bits, NR_CPUS);	/* explicit bitsize */

or:

    /* specify bitsize on declaration; use specialized operations */
    struct cpumask { DECLARE_BITMAP(bits, NR_CPUS); };
    struct cpumask s, d1, d2;
    cpus_or(s.bits, d1.bits, d2.bits);			/* 'cpu' implies NR_CPUS */

or:

    /* carry the bitsize in the structure [pseudo C alert] */
    struct mask { int nbits = NR_CPUS; DECLARE_BITMAP(bits, NR_CPUS); };
    struct mask s, d1, d2;
    mask_or(s.bits, d1.bits, d2.bits);			/* mask_* ops get size from struct */

Am I missing any choices?  Which do you prefer?

I understand that the kernel currently does the 2nd choice, encoding the
bitsize in the operation name.

My personal preference is to continue doing this 2nd choice.

-- 
                          I won't rest till it's the best ...
                          Programmer, Linux Scalability
                          Paul Jackson <pj@sgi.com> 1.650.933.1373