Re: [PATCH 1/4] lib: vsprintf: Optimize division by 10 for small integers.

[Denys Vlasenko]

On Fri, Aug 3, 2012 at 7:21 AM, George Spelvin <linux@xxxxxxxxxxx> wrote:
> Shrink the reciprocal approximations used in put_dec_full4
> based on the comments in put_dec_full9.
>
> Signed-off-by: George Spelvin <linux@xxxxxxxxxxx>
> Cc: Denys Vlasenko <vda.linux@xxxxxxxxxxxxxx>
> Cc: Michal Nazarewicz <mina86@xxxxxxxxxx>
> ---
>  lib/vsprintf.c |    5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> I was looking over the code and noticed that the constants could be smaller.
>
> diff --git a/lib/vsprintf.c b/lib/vsprintf.c
> index c3f36d41..2f32fe8 100644
> --- a/lib/vsprintf.c
> +++ b/lib/vsprintf.c
> @@ -243,13 +243,14 @@ char *put_dec(char *buf, unsigned long long n)
>
>  /* Second algorithm: valid only for 64-bit long longs */
>
> +/* See comment in put_dec_full9 for choice of constants */
>  static noinline_for_stack
>  char *put_dec_full4(char *buf, unsigned q)
>  {
>         unsigned r;
> -       r      = (q * 0xcccd) >> 19;
> +       r      = (q * 0xccd) >> 15;
>         *buf++ = (q - 10 * r) + '0';
> -       q      = (r * 0x199a) >> 16;
> +       q      = (r * 0xcd) >> 11;

I would use 16-bit shifts instead of smaller ones.
There may be CPUs on which "get upper half of 32-bit reg"
operation is cheaper or smaller than a shift.

-- 
vda
--
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/