Re: [Patch] Scale pidhash_shift/pidhash_size up based on num_possible_cpus().

[ebiederm@xmission.com (Eric W. Biederman)]

Robin Holt <holt@sgi.com> writes:

> For large cpu configurations, we find the number of pids in a pidhash
> bucket cause things like 'ps' to perform slowly.  Raising pidhash_shift
> from 12 to 16 cut the time for 'ps' in half on a 2048 cpu machine.
>
> This patch makes the upper limit scale based upon num_possible_cpus().
> For machines 128 cpus or less, the current upper limit of 12 is
> maintained.

It looks like there is a magic limit we are dancing around.

Can we please make the maximum for the hash table size be based
on the maximum number of pids.  That is fls(PID_MAX_LIMIT) - 6?

12 actually looks like it was fls(PID_MAX_LIMIT) - 3 at one point in
time.

Basing the hash table size on PID_MAX_LIMIT seems to say interesting
about the maximum number of hash chain entries we will tolerate in
practice.

Eric

> Signed-off-by: Robin Holt <holt@sgi.com>
>
>
> Index: contention_unroll/kernel/pid.c
> ===================================================================
> --- contention_unroll.orig/kernel/pid.c	2008-07-31 11:59:21.154284073 -0500
> +++ contention_unroll/kernel/pid.c	2008-07-31 11:59:22.862497720 -0500
> @@ -502,9 +502,10 @@ void __init pidhash_init(void)
>  {
>  	int i, pidhash_size;
>  	unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT);
> +	int pidhash_shift_ul = max(12, fls(num_possible_cpus() - 1) + 5);
>  
>  	pidhash_shift = max(4, fls(megabytes * 4));
> -	pidhash_shift = min(12, pidhash_shift);
> +	pidhash_shift = min(pidhash_shift_ul, pidhash_shift);
>  	pidhash_size = 1 << pidhash_shift;
>  
>  	printk("PID hash table entries: %d (order: %d, %Zd bytes)\n",