2021-08-29 06:05:07

by Fengfei Xi

Subject: [PATCH] bcache: remove the redundant judgment on bi_size

The bi_size is unsigned int type data not less than 0,
so we can directly add bi_size without extra judgment

Signed-off-by: Fengfei Xi <[email protected]>
---
drivers/md/bcache/request.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
index 6d1de889b..2788eec3a 100644
--- a/drivers/md/bcache/request.c
+++ b/drivers/md/bcache/request.c
@@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
add_sequential(task);
i->sequential = 0;
found:
- if (i->sequential + bio->bi_iter.bi_size > i->sequential)
- i->sequential += bio->bi_iter.bi_size;
-
+ i->sequential += bio->bi_iter.bi_size;
i->last = bio_end_sector(bio);
i->jiffies = jiffies + msecs_to_jiffies(5000);
task->sequential_io = i->sequential;
--
2.17.1

2021-08-29 07:52:40

by Coly Li

Subject: Re: [PATCH] bcache: remove the redundant judgment on bi_size

On 8/29/21 12:49 PM, Fengfei Xi wrote:
> The bi_size is unsigned int type data not less than 0,
> so we can directly add bi_size without extra judgment
>
> Signed-off-by: Fengfei Xi <[email protected]>

NACK. The check is necessary to avoid redundant and unnecessary memory
write.

Coly Li

> ---
> drivers/md/bcache/request.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
> index 6d1de889b..2788eec3a 100644
> --- a/drivers/md/bcache/request.c
> +++ b/drivers/md/bcache/request.c
> @@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
> add_sequential(task);
> i->sequential = 0;
> found:
> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
> - i->sequential += bio->bi_iter.bi_size;
> -
> + i->sequential += bio->bi_iter.bi_size;
> i->last = bio_end_sector(bio);
> i->jiffies = jiffies + msecs_to_jiffies(5000);
> task->sequential_io = i->sequential;

2021-08-30 06:27:03

by Fengfei Xi

Subject: Re: [PATCH] bcache: remove the redundant judgment on bi_size

Thanks a lot. I understand the purpose.
So is the original judgment process too complicated? Can we judge bi_size directly? This will be more concise

@@ -423,7 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
add_sequential(task);
i->sequential = 0;
found:
- if (i->sequential + bio->bi_iter.bi_size > i->sequential)
+ if (!bio->bi_iter.bi_size)
i->sequential += bio->bi_iter.bi_size;

i->last = bio_end_sector(bio);

Thanks
Fengfei

-----邮件原件-----
发件人: Coly Li [mailto:[email protected]]
发送时间: 2021年8月29日 15:50
收件人: xifengfei (RD) <[email protected]>
抄送: [email protected]; [email protected]; [email protected]
主题: Re: [PATCH] bcache: remove the redundant judgment on bi_size

On 8/29/21 12:49 PM, Fengfei Xi wrote:
> The bi_size is unsigned int type data not less than 0, so we can
> directly add bi_size without extra judgment
>
> Signed-off-by: Fengfei Xi <[email protected]>

NACK. The check is necessary to avoid redundant and unnecessary memory write.

Coly Li

> ---
> drivers/md/bcache/request.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
> index 6d1de889b..2788eec3a 100644
> --- a/drivers/md/bcache/request.c
> +++ b/drivers/md/bcache/request.c
> @@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
> add_sequential(task);
> i->sequential = 0;
> found:
> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
> - i->sequential += bio->bi_iter.bi_size;
> -
> + i->sequential += bio->bi_iter.bi_size;
> i->last = bio_end_sector(bio);
> i->jiffies = jiffies + msecs_to_jiffies(5000);
> task->sequential_io = i->sequential;

2021-08-30 06:32:27

by Fengfei Xi

Subject: Re: [PATCH] bcache: remove the redundant judgment on bi_size

（Sorry, there was an obvious typo in the last email）
Thanks a lot. I understand the purpose.
So is the original judgment process too complicated? Can we judge bi_size directly? This will be more concise

@@ -423,7 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
add_sequential(task);
i->sequential = 0;
found:
- if (i->sequential + bio->bi_iter.bi_size > i->sequential)
+ if (bio->bi_iter.bi_size)
i->sequential += bio->bi_iter.bi_size;

i->last = bio_end_sector(bio);

Thanks
Fengfei

-----邮件原件-----
发件人: Coly Li [mailto:[email protected]]
发送时间: 2021年8月29日 15:50
收件人: xifengfei (RD) <[email protected]>
抄送: [email protected]; [email protected]; [email protected]
主题: Re: [PATCH] bcache: remove the redundant judgment on bi_size

On 8/29/21 12:49 PM, Fengfei Xi wrote:
> The bi_size is unsigned int type data not less than 0, so we can
> directly add bi_size without extra judgment
>
> Signed-off-by: Fengfei Xi <[email protected]>

NACK. The check is necessary to avoid redundant and unnecessary memory write.

Coly Li

> ---
> drivers/md/bcache/request.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
> index 6d1de889b..2788eec3a 100644
> --- a/drivers/md/bcache/request.c
> +++ b/drivers/md/bcache/request.c
> @@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
> add_sequential(task);
> i->sequential = 0;
> found:
> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
> - i->sequential += bio->bi_iter.bi_size;
> -
> + i->sequential += bio->bi_iter.bi_size;
> i->last = bio_end_sector(bio);
> i->jiffies = jiffies + msecs_to_jiffies(5000);
> task->sequential_io = i->sequential;

2021-08-31 01:37:40

by Coly Li

Subject: Re: [PATCH] bcache: remove the redundant judgment on bi_size

On 8/30/21 2:29 PM, Xifengfei wrote:
> （Sorry, there was an obvious typo in the last email）
> Thanks a lot. I understand the purpose.
> So is the original judgment process too complicated? Can we judge bi_size directly? This will be more concise
>
> @@ -423,7 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
> add_sequential(task);
> i->sequential = 0;
> found:
> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
> + if (bio->bi_iter.bi_size)
> i->sequential += bio->bi_iter.bi_size;
>
> i->last = bio_end_sector(bio);

The above change works, but the code readability decreased because
how/why i->sequential is maintained is not that directly visible.

This is a difference of coding styles. IMHO for this particular case,
the readability is more important than less CPU instructions.

Thanks.

Coly Li

> Thanks
> Fengfei
>
> -----邮件原件-----
> 发件人: Coly Li [mailto:[email protected]]
> 发送时间: 2021年8月29日 15:50
> 收件人: xifengfei (RD) <[email protected]>
> 抄送: [email protected]; [email protected]; [email protected]
> 主题: Re: [PATCH] bcache: remove the redundant judgment on bi_size
>
> On 8/29/21 12:49 PM, Fengfei Xi wrote:
>> The bi_size is unsigned int type data not less than 0, so we can
>> directly add bi_size without extra judgment
>>
>> Signed-off-by: Fengfei Xi <[email protected]>
> NACK. The check is necessary to avoid redundant and unnecessary memory write.
>
> Coly Li
>
>> ---
>> drivers/md/bcache/request.c | 4 +---
>> 1 file changed, 1 insertion(+), 3 deletions(-)
>>
>> diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
>> index 6d1de889b..2788eec3a 100644
>> --- a/drivers/md/bcache/request.c
>> +++ b/drivers/md/bcache/request.c
>> @@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
>> add_sequential(task);
>> i->sequential = 0;
>> found:
>> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
>> - i->sequential += bio->bi_iter.bi_size;
>> -
>> + i->sequential += bio->bi_iter.bi_size;
>> i->last = bio_end_sector(bio);
>> i->jiffies = jiffies + msecs_to_jiffies(5000);
>> task->sequential_io = i->sequential;

2021-08-31 02:07:58

by Fengfei Xi

Subject: Re: [PATCH] bcache: remove the redundant judgment on bi_size

OK. Thanks for our explanation. ^_^

Feng fei

-----邮件原件-----
发件人: Coly Li [mailto:[email protected]]
发送时间: 2021年8月31日 9:35
收件人: xifengfei (RD) <[email protected]>
抄送: [email protected]; [email protected]; [email protected]
主题: Re: [PATCH] bcache: remove the redundant judgment on bi_size

On 8/30/21 2:29 PM, Xifengfei wrote:
> （Sorry, there was an obvious typo in the last email）
> Thanks a lot. I understand the purpose.
> So is the original judgment process too complicated? Can we judge
> bi_size directly? This will be more concise
>
> @@ -423,7 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
> add_sequential(task);
> i->sequential = 0;
> found:
> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
> + if (bio->bi_iter.bi_size)
> i->sequential += bio->bi_iter.bi_size;
>
> i->last = bio_end_sector(bio);

The above change works, but the code readability decreased because how/why i->sequential is maintained is not that directly visible.

This is a difference of coding styles. IMHO for this particular case, the readability is more important than less CPU instructions.

Thanks.

Coly Li

> Thanks
> Fengfei
>
> -----邮件原件-----
> 发件人: Coly Li [mailto:[email protected]]
> 发送时间: 2021年8月29日 15:50
> 收件人: xifengfei (RD) <[email protected]>
> 抄送: [email protected]; [email protected];
> [email protected]
> 主题: Re: [PATCH] bcache: remove the redundant judgment on bi_size
>
> On 8/29/21 12:49 PM, Fengfei Xi wrote:
>> The bi_size is unsigned int type data not less than 0, so we can
>> directly add bi_size without extra judgment
>>
>> Signed-off-by: Fengfei Xi <[email protected]>
> NACK. The check is necessary to avoid redundant and unnecessary memory write.
>
> Coly Li
>
>> ---
>> drivers/md/bcache/request.c | 4 +---
>> 1 file changed, 1 insertion(+), 3 deletions(-)
>>
>> diff --git a/drivers/md/bcache/request.c
>> b/drivers/md/bcache/request.c index 6d1de889b..2788eec3a 100644
>> --- a/drivers/md/bcache/request.c
>> +++ b/drivers/md/bcache/request.c
>> @@ -423,9 +423,7 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
>> add_sequential(task);
>> i->sequential = 0;
>> found:
>> - if (i->sequential + bio->bi_iter.bi_size > i->sequential)
>> - i->sequential += bio->bi_iter.bi_size;
>> -
>> + i->sequential += bio->bi_iter.bi_size;
>> i->last = bio_end_sector(bio);
>> i->jiffies = jiffies + msecs_to_jiffies(5000);
>> task->sequential_io = i->sequential;