Re: [PACH] smp: uninline num_online_cpus() & num_possible_cpus()

[Andrew Morton <akpm@google.com>]

On Fri, 05 Dec 2008 18:33:44 +0100
Eric Dumazet <dada1@cosmosbay.com> wrote:

> num_online_cpus() and num_possible_cpus() are not performance
> critical and are quite large.
> 
> Unlining them shrinks kernel text size by 7523 bytes on x86,
> if NR_CPUS>32
> 
> Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
> ---
>  include/linux/cpumask.h |    4 ++--
>  init/main.c             |   12 ++++++++++++
>  2 files changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
> index 21e1dd4..f9b2b51 100644
> --- a/include/linux/cpumask.h
> +++ b/include/linux/cpumask.h
> @@ -507,8 +507,8 @@ extern cpumask_t cpu_present_map;
>  extern cpumask_t cpu_active_map;
>  
>  #if NR_CPUS > 1
> -#define num_online_cpus()	cpus_weight_nr(cpu_online_map)
> -#define num_possible_cpus()	cpus_weight_nr(cpu_possible_map)
> +extern int num_online_cpus(void);
> +extern int num_possible_cpus(void);
>  #define num_present_cpus()	cpus_weight_nr(cpu_present_map)
>  #define cpu_online(cpu)		cpu_isset((cpu), cpu_online_map)
>  #define cpu_possible(cpu)	cpu_isset((cpu), cpu_possible_map)
> diff --git a/init/main.c b/init/main.c
> index 7e117a2..a1a3e55 100644
> --- a/init/main.c
> +++ b/init/main.c
> @@ -376,6 +376,18 @@ EXPORT_SYMBOL(cpu_mask_all);
>  int nr_cpu_ids __read_mostly = NR_CPUS;
>  EXPORT_SYMBOL(nr_cpu_ids);
>  
> +int num_online_cpus(void)
> +{
> +	return cpus_weight_nr(cpu_online_map);
> +}
> +EXPORT_SYMBOL(num_online_cpus);
> +
> +int num_possible_cpus(void)
> +{
> +	return cpus_weight_nr(cpu_possible_map);
> +}
> +EXPORT_SYMBOL(num_possible_cpus);
> +
>  /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
>  static void __init setup_nr_cpu_ids(void)
>  {
	
Looks OK.

That area in init/main.c is horrid - it took quite some staring through
the ifdef tangle for me to convince myself that the code you added was
reliably SMP-only.

Perhaps sometime a lot of this cpu masky code should be moved over to
kernel/cpu.c and cleaned up.