Re: [PATCH] lib/int_sqrt.c: Optimize square root function

From: Davidlohr Bueso
Date: Mon Feb 02 2015 - 23:41:57 EST


On Mon, 2015-02-02 at 11:13 -0800, Linus Torvalds wrote:
> IOW, instead of
>
> m = 1UL << (BITS_PER_LONG - 2);
>
> perhaps something like
>
> m = 1UL << (BITS_PER_LONG/2- 2);
> if (m < x)
> m <<= BITS_PER_LONG/2;
>
> (assuming gcc can change that code into a "cmov") might cut down the
> "lots of empty loops" case in half for small values of 'x'.

Makes a lot of sense.

> There's probably some other better cheap initial guess value estimator.

Just to get a feeling for the normal arg range in the non-driver parts
that use this thing:

fs/ceph/super.h: congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
fs/nfs/write.c: nfs_congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
fs/nfsd/nfscache.c: limit = (16 * int_sqrt(low_pages)) << (PAGE_SHIFT-10);
kernel/rcu/tree_plugin.h: ls = int_sqrt(nr_cpu_ids);
mm/memcontrol.c: inactive_ratio = int_sqrt(10 * gb);
mm/page_alloc.c: ratio = int_sqrt(10 * gb);
mm/page_alloc.c: new_min_free_kbytes = int_sqrt(lowmem_kbytes * 16);

So mostly values that scale according to mem or cpu.



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/