[PATCH] block: simplify blksize_bits() implementation

[PATCH] block: simplify blksize_bits() implementation

[Dawei Li]

Convert current looping-based implementation into bit operation,
which can bring improvement for:

1) bitops is more efficient for its arch-level optimization.

2) Given that blksize_bits() is inline, _if_ @size is compile-time
constant, it's possible that order_base_2() _may_ make output
compile-time evaluated, depending on code context and compiler behavior.

Signed-off-by: Dawei Li <set_pte_at@outlook.com>
---
 include/linux/blkdev.h | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 50e358a19d98..117061c8b9a1 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1349,12 +1349,7 @@ static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
 /* assumes size > 256 */
 static inline unsigned int blksize_bits(unsigned int size)
 {
-	unsigned int bits = 8;
-	do {
-		bits++;
-		size >>= 1;
-	} while (size > 256);
-	return bits;
+	return size > 512 ? order_base_2(size) : 9;
 }
 
 static inline unsigned int block_size(struct block_device *bdev)
-- 
2.25.1

Re: [PATCH] block: simplify blksize_bits() implementation

[Bart Van Assche]

On 10/26/22 08:14, Dawei Li wrote:
> Convert current looping-based implementation into bit operation,
> which can bring improvement for:
> 
> 1) bitops is more efficient for its arch-level optimization.

As far as I know blksize_bits() is not used in the hot path so 
performance of this function is not critical.

> 2) Given that blksize_bits() is inline, _if_ @size is compile-time
> constant, it's possible that order_base_2() _may_ make output
> compile-time evaluated, depending on code context and compiler behavior.
> 
> Signed-off-by: Dawei Li <set_pte_at@outlook.com>
> ---
>   include/linux/blkdev.h | 7 +------
>   1 file changed, 1 insertion(+), 6 deletions(-)
> 
> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index 50e358a19d98..117061c8b9a1 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -1349,12 +1349,7 @@ static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
>   /* assumes size > 256 */
>   static inline unsigned int blksize_bits(unsigned int size)
>   {
> -	unsigned int bits = 8;
> -	do {
> -		bits++;
> -		size >>= 1;
> -	} while (size > 256);
> -	return bits;
> +	return size > 512 ? order_base_2(size) : 9;
>   }

How about optimizing this function even further by eliminating the 
ternary operator, e.g. as follows (untested)?

         return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT;

Thanks,

Bart.

Re: [PATCH] block: simplify blksize_bits() implementation

[Keith Busch]

On Wed, Oct 26, 2022 at 09:29:21AM -0700, Bart Van Assche wrote:
> On 10/26/22 08:14, Dawei Li wrote:
> > Convert current looping-based implementation into bit operation,
> > which can bring improvement for:
> > 
> > 1) bitops is more efficient for its arch-level optimization.
> 
> As far as I know blksize_bits() is not used in the hot path so performance
> of this function is not critical.

blksize_bits() is used on every IO going through iomap_dio_bio_iter(),
though the usage there is completely unnecessary and can be removed.