Hi all
The cpu-time is in tens of seconds
and the run-time is in thousands of secounds.
but the source code doesn't follow it.
I fixed it and also some white-spaces.
Could you please check this patch.
Signed-off-by: Naohiro Ooiwa <[email protected]>
---
mm/oom_kill.c | 10 +++++-----
1 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 64e5b4b..bddab74 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -100,14 +100,14 @@ unsigned long badness(struct task_struct *p, unsigned long uptime)
/*
* CPU time is in tens of seconds and run time is in thousands
- * of seconds. There is no particular reason for this other than
- * that it turned out to work very well in practice.
+ * of seconds. There is no particular reason for this other than
+ * that it turned out to work very well in practice.
*/
- cpu_time = (cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
- >> (SHIFT_HZ + 3);
+ cpu_time = ((cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
+ >> SHIFT_HZ) / 10UL;
if (uptime >= p->start_time.tv_sec)
- run_time = (uptime - p->start_time.tv_sec) >> 10;
+ run_time = (uptime - p->start_time.tv_sec) / 1000UL;
else
run_time = 0;
--
1.5.4.1
> Hi all
>
> The cpu-time is in tens of seconds
> and the run-time is in thousands of secounds.
>
> but the source code doesn't follow it.
Have you seen any trouble?
>
> I fixed it and also some white-spaces.
> Could you please check this patch.
>
>
> Signed-off-by: Naohiro Ooiwa <[email protected]>
> ---
> mm/oom_kill.c | 10 +++++-----
> 1 files changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index 64e5b4b..bddab74 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -100,14 +100,14 @@ unsigned long badness(struct task_struct *p, unsigned long uptime)
>
> /*
> * CPU time is in tens of seconds and run time is in thousands
> - * of seconds. There is no particular reason for this other than
> - * that it turned out to work very well in practice.
> + * of seconds. There is no particular reason for this other than
> + * that it turned out to work very well in practice.
> */
> - cpu_time = (cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
> - >> (SHIFT_HZ + 3);
> + cpu_time = ((cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
> + >> SHIFT_HZ) / 10UL;
>
> if (uptime >= p->start_time.tv_sec)
> - run_time = (uptime - p->start_time.tv_sec) >> 10;
> + run_time = (uptime - p->start_time.tv_sec) / 1000UL;
> else
> run_time = 0;
in some architecture, shift-op outperfom divide-op largely.
why do you need this change?
Hi KOSAKI-san
Thank you for quick reply and checking my patch.
> Have you seen any trouble?
No I haven't.
> in some architecture, shift-op outperfom divide-op largely.
Of course, but I think that the oom-killer doesn't need high performance.
Do you think oom-killer needs it ?
> why do you need this change?
Nothing special,
but I write a tips about oom-killer now.
The comments and the source code don't match.
so I think how to write about badness point now.
Therefore, I only think the source code should conform to the comments.
Regards,
Naohiro Ooiwa.
KOSAKI Motohiro さんは書きました:
>> Hi all
>>
>> The cpu-time is in tens of seconds
>> and the run-time is in thousands of secounds.
>>
>> but the source code doesn't follow it.
>
> Have you seen any trouble?
>
>> I fixed it and also some white-spaces.
>> Could you please check this patch.
>>
>>
>> Signed-off-by: Naohiro Ooiwa <[email protected]>
>> ---
>> mm/oom_kill.c | 10 +++++-----
>> 1 files changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
>> index 64e5b4b..bddab74 100644
>> --- a/mm/oom_kill.c
>> +++ b/mm/oom_kill.c
>> @@ -100,14 +100,14 @@ unsigned long badness(struct task_struct *p, unsigned long uptime)
>>
>> /*
>> * CPU time is in tens of seconds and run time is in thousands
>> - * of seconds. There is no particular reason for this other than
>> - * that it turned out to work very well in practice.
>> + * of seconds. There is no particular reason for this other than
>> + * that it turned out to work very well in practice.
>> */
>> - cpu_time = (cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
>> - >> (SHIFT_HZ + 3);
>> + cpu_time = ((cputime_to_jiffies(p->utime) + cputime_to_jiffies(p->stime))
>> + >> SHIFT_HZ) / 10UL;
>>
>> if (uptime >= p->start_time.tv_sec)
>> - run_time = (uptime - p->start_time.tv_sec) >> 10;
>> + run_time = (uptime - p->start_time.tv_sec) / 1000UL;
>> else
>> run_time = 0;
>
> in some architecture, shift-op outperfom divide-op largely.
> why do you need this change?
>
>
>