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.
v1: https://lore.kernel.org/all/TYCP286MB2323169D81A806A7C1F7FDF1CA309@TYCP286MB2323.JPNP286.PROD.OUTLOOK.COM
v2: Remove the ternary operator, based on Bart's suggestion
But this may lead to break for corner cases below:
BUILD_BUG_ON(blksize_bits(1025) != 11);
So make a minor modification by adding (SECTOR_SIZE - 1) before
shifting.
base-commit: 30209debe98b6f66b13591e59e5272cb65b3945e
Signed-off-by: Dawei Li <[email protected]>
---
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 57ed49f20d2e..7b537afe8b38 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 order_base_2((size + SECTOR_SIZE - 1) >> SECTOR_SHIFT) + SECTOR_SHIFT;
}
static inline unsigned int block_size(struct block_device *bdev)
--
2.25.1
On 10/29/22 19:17, Dawei Li wrote:
> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index 57ed49f20d2e..7b537afe8b38 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 order_base_2((size + SECTOR_SIZE - 1) >> SECTOR_SHIFT) + SECTOR_SHIFT;
> }
Why the rounding ("+ SECTOR_SIZE - 1")? The blksize_bits() argument
should be an argument of two.
Thanks,
Bart.
On Sat, Oct 29, 2022 at 08:00:58PM -0700, Bart Van Assche wrote:
> On 10/29/22 19:17, Dawei Li wrote:
> > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> > index 57ed49f20d2e..7b537afe8b38 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 order_base_2((size + SECTOR_SIZE - 1) >> SECTOR_SHIFT) + SECTOR_SHIFT;
> > }
>
> Why the rounding ("+ SECTOR_SIZE - 1")? The blksize_bits() argument should
> be an argument of two.
Yeah, that's what's supposed to be.
But I thought maybe a "just in case" is more robust?
Should we consider these corner cases(!is_power_of_2())?
Thanks.
>
> Thanks,
>
> Bart.
>
On Sat, Oct 29, 2022 at 08:33:22PM -0700, Bart Van Assche wrote:
> On 10/29/22 20:25, Dawei Li wrote:
> > On Sat, Oct 29, 2022 at 08:00:58PM -0700, Bart Van Assche wrote:
> > > On 10/29/22 19:17, Dawei Li wrote:
> > > > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> > > > index 57ed49f20d2e..7b537afe8b38 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 order_base_2((size + SECTOR_SIZE - 1) >> SECTOR_SHIFT) + SECTOR_SHIFT;
> > > > }
> > >
> > > Why the rounding ("+ SECTOR_SIZE - 1")? The blksize_bits() argument should
> > > be an argument of two.
> >
> > Yeah, that's what's supposed to be.
> > But I thought maybe a "just in case" is more robust?
> > Should we consider these corner cases(!is_power_of_2())?
>
> I don't think that the Linux kernel supports block sizes that are not a
> power of two. Hence my request to leave out the rounding code. Keeping that
> code would be misleading because it would suggest that the blksize_bits()
> argument can be something else than a power of two.
Thanks for the review, bart.
Will resend the updated patch.
>
> Thanks,
>
> Bart.
>
On 10/29/22 20:25, Dawei Li wrote:
> On Sat, Oct 29, 2022 at 08:00:58PM -0700, Bart Van Assche wrote:
>> On 10/29/22 19:17, Dawei Li wrote:
>>> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
>>> index 57ed49f20d2e..7b537afe8b38 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 order_base_2((size + SECTOR_SIZE - 1) >> SECTOR_SHIFT) + SECTOR_SHIFT;
>>> }
>>
>> Why the rounding ("+ SECTOR_SIZE - 1")? The blksize_bits() argument should
>> be an argument of two.
>
> Yeah, that's what's supposed to be.
> But I thought maybe a "just in case" is more robust?
> Should we consider these corner cases(!is_power_of_2())?
I don't think that the Linux kernel supports block sizes that are not a
power of two. Hence my request to leave out the rounding code. Keeping
that code would be misleading because it would suggest that the
blksize_bits() argument can be something else than a power of two.
Thanks,
Bart.