seq_put_decimal_ull() et al

seq_put_decimal_ull() et al

[Alexey Dobriyan]

> Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree

I think the fix is to create proper binary interface for exporting
interrupt counts.
Programs can parse raw data equally well if programming language is
even minimally reasonable
(POSIX shell may not qualify, though).

Re: seq_put_decimal_ull() et al

[Alexey Dobriyan]

On Tue, Mar 13, 2012 at 4:35 PM, Alexey Dobriyan <adobriyan@gmail.com> wrote:
>> Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
>
> I think the fix is to create proper binary interface for exporting
> interrupt counts.
> Programs can parse raw data equally well if programming language is
> even minimally reasonable
> (POSIX shell may not qualify, though).

Ugh, I meant a file with binary data, not complete fileless interface.

Re: seq_put_decimal_ull() et al

[KAMEZAWA Hiroyuki]

On Tue, 13 Mar 2012 15:35:54 +0200
Alexey Dobriyan <adobriyan@gmail.com> wrote:

> > Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
> 
> I think the fix is to create proper binary interface for exporting
> interrupt counts.
> Programs can parse raw data equally well if programming language is
> even minimally reasonable
> (POSIX shell may not qualify, though).

Ok, please try with all tool-chain updates.
procps, top, sar, and more ?


Thanks,
-Kame

Re: seq_put_decimal_ull() et al

[KAMEZAWA Hiroyuki]

On Wed, 14 Mar 2012 09:31:56 +0900
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:

> On Tue, 13 Mar 2012 15:35:54 +0200
> Alexey Dobriyan <adobriyan@gmail.com> wrote:
> 
> > > Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
> > 
> > I think the fix is to create proper binary interface for exporting
> > interrupt counts.
> > Programs can parse raw data equally well if programming language is
> > even minimally reasonable
> > (POSIX shell may not qualify, though).
> 
> Ok, please try with all tool-chain updates.
> procps, top, sar, and more ?
> 
One concern is maintainance of 2 files of the same data.

Thanks,
-Kame

Re: seq_put_decimal_ull() et al

[Alexey Dobriyan]

On Wed, Mar 14, 2012 at 5:08 AM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Wed, 14 Mar 2012 09:31:56 +0900
> KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>
>> On Tue, 13 Mar 2012 15:35:54 +0200
>> Alexey Dobriyan <adobriyan@gmail.com> wrote:
>>
>> > > Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
>> >
>> > I think the fix is to create proper binary interface for exporting
>> > interrupt counts.
>> > Programs can parse raw data equally well if programming language is
>> > even minimally reasonable
>> > (POSIX shell may not qualify, though).
>>
>> Ok, please try with all tool-chain updates.
>> procps, top, sar, and more ?
>>

Well, why not?
We can tune print code or we can create maximally fast interface and forget.

> One concern is maintainance of 2 files of the same data.

The other is all those multiplying print functions.

Re: seq_put_decimal_ull() et al

[KAMEZAWA Hiroyuki]

(2012/03/16 20:16), Alexey Dobriyan wrote:

> On Wed, Mar 14, 2012 at 5:08 AM, KAMEZAWA Hiroyuki
> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>> On Wed, 14 Mar 2012 09:31:56 +0900
>> KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>>
>>> On Tue, 13 Mar 2012 15:35:54 +0200
>>> Alexey Dobriyan <adobriyan@gmail.com> wrote:
>>>
>>>>> Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
>>>>
>>>> I think the fix is to create proper binary interface for exporting
>>>> interrupt counts.
>>>> Programs can parse raw data equally well if programming language is
>>>> even minimally reasonable
>>>> (POSIX shell may not qualify, though).
>>>
>>> Ok, please try with all tool-chain updates.
>>> procps, top, sar, and more ?
>>>
> 
> Well, why not?
> We can tune print code or we can create maximally fast interface and forget.
> 

Do you have an idea where to place it ? 
/proc/stat_binary ?
/proc/binary/stat ?

Hm, to parse /proc/stat binary format, the user has to know the format of the
kernel version.

Do we show a format file as
/proc/binary/stat_format and show
# cat /proc/binary/stat_format
  =IHHHIHIQQQIHQQ...

as python's pack/unpack ? or
export the format via /usr/include/linux/proc_format.h ? as a structure ?

any better idea ?

Thanks,
-Kame

Re: seq_put_decimal_ull() et al

[Alexey Dobriyan]

On Mon, Mar 19, 2012 at 3:29 AM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> (2012/03/16 20:16), Alexey Dobriyan wrote:
>
>> On Wed, Mar 14, 2012 at 5:08 AM, KAMEZAWA Hiroyuki
>> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>>> On Wed, 14 Mar 2012 09:31:56 +0900
>>> KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>>>
>>>> On Tue, 13 Mar 2012 15:35:54 +0200
>>>> Alexey Dobriyan <adobriyan@gmail.com> wrote:
>>>>
>>>>>> Subject:    + procfs-add-num_to_str-to-speed-up-proc-stat-fix-2.patch added to -mm tree
>>>>>
>>>>> I think the fix is to create proper binary interface for exporting
>>>>> interrupt counts.
>>>>> Programs can parse raw data equally well if programming language is
>>>>> even minimally reasonable
>>>>> (POSIX shell may not qualify, though).
>>>>
>>>> Ok, please try with all tool-chain updates.
>>>> procps, top, sar, and more ?
>>>>
>>
>> Well, why not?
>> We can tune print code or we can create maximally fast interface and forget.
>>
>
> Do you have an idea where to place it ?
> /proc/stat_binary ?
> /proc/binary/stat ?
>
> Hm, to parse /proc/stat binary format, the user has to know the format of the
> kernel version.
>
> Do we show a format file as
> /proc/binary/stat_format and show
> # cat /proc/binary/stat_format
>  =IHHHIHIQQQIHQQ...
>
> as python's pack/unpack ? or
> export the format via /usr/include/linux/proc_format.h ? as a structure ?
>
> any better idea ?

I'm thinking of separate file with interrupts statistics. Let's not
mix everything.
Hmm, let me see how top(1) parses /proc/interrupts.