On Thu, Nov 29, 2018 at 12:34:12PM -0700, Jens Axboe wrote:
> If we aren't forced to do round robin tag allocation, just use the
> allocation hint to find the index for the tag word, don't use it for the
> offset inside the word.

Maybe also add "We're already fetching that cache line, so we might as
well check the whole word."

> This avoids a potential extra round trip in the
> bit looping.
> 
> Signed-off-by: Jens Axboe <ax...@kernel.dk>
> 
> ---
> 
> diff --git a/lib/sbitmap.c b/lib/sbitmap.c
> index fdd1b8aa8ac6..2987b2ac8ed7 100644
> --- a/lib/sbitmap.c
> +++ b/lib/sbitmap.c
> @@ -118,10 +118,19 @@ int sbitmap_get(struct sbitmap *sb, unsigned int 
> alloc_hint, bool round_robin)
>  
>       index = SB_NR_TO_INDEX(sb, alloc_hint);
>  
> +     /*
> +      * Unless we're doing round robin tag allocation, just use the
> +      * alloc_hint to find the right word index. No point in looping
> +      * twice in find_next_zero_bit() for that case.
> +      */
> +     if (round_robin)
> +             alloc_hint = SB_NR_TO_BIT(sb, alloc_hint);
> +     else
> +             alloc_hint = 0;
> +
>       for (i = 0; i < sb->map_nr; i++) {
>               nr = __sbitmap_get_word(&sb->map[index].word,
> -                                     sb->map[index].depth,
> -                                     SB_NR_TO_BIT(sb, alloc_hint),
> +                                     sb->map[index].depth, alloc_hint,
>                                       !round_robin);
>               if (nr != -1) {
>                       nr += index << sb->shift;
> @@ -130,12 +139,8 @@ int sbitmap_get(struct sbitmap *sb, unsigned int 
> alloc_hint, bool round_robin)
>  
>               /* Jump to next index. */
>               index++;
> -             alloc_hint = index << sb->shift;
> -
> -             if (index >= sb->map_nr) {
> +             if (index >= sb->map_nr)
>                       index = 0;
> -                     alloc_hint = 0;
> -             }

We need to set alloc_hint = 0 here for the round_robin case.

>       }
>  
>       return nr;
> 
> -- 
> Jens Axboe
> 

Reply via email to