Message-ID: <20070720072522.GB27772@wotan.suse.de>
Date:	Fri, 20 Jul 2007 09:25:22 +0200
From:	Nick Piggin <npiggin@...e.de>
To:	Joachim Deguara <joachim.deguara@....com>
Cc:	ricklind@...ibm.com, lkml List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Documentation update sched-stat.txt

On Wed, Jul 18, 2007 at 11:11:30AM +0200, Joachim Deguara wrote:
> While learning about schedstats I found that the documentation in the tree is 
> old.  I updated it and found some interesting stuff like schedstats version 
> 14 is the same as version and version 13 never saw a kernel release!  Also 
> there are 6 fields in the current schedstats that are not used anymore.  Nick 
> had made them irrelevant in commit 476d139c218e44e045e4bc6d4cc02b010b343939 
> but never removed them.
> 
> Thanks to Rick's perl script who I borrowed some of the updated descriptions 
> from.

Ah, thanks, I actually didn't realise there was such good documentation
there. Patch looks good.

BTW. I have a simple program to do a basic statistical summary of the
multiprocessor balancing if you are interested and haven't seen it.

Acked-by: Nick Piggin <npiggin@...e.de>

> 
> -Joachim
> 
> --
> Updating schedstats documentation from version 10 to 14.
> 
> Signed-off-by: Joachim Deguara <joachim.deguara@....com>
> 
> Index: kernel/Documentation/sched-stats.txt
> ===================================================================
> --- kernel.orig/Documentation/sched-stats.txt
> +++ kernel/Documentation/sched-stats.txt
> @@ -1,10 +1,11 @@
> -Version 10 of schedstats includes support for sched_domains, which
> -hit the mainline kernel in 2.6.7.  Some counters make more sense to be
> -per-runqueue; other to be per-domain.  Note that domains (and their 
> associated
> -information) will only be pertinent and available on machines utilizing
> -CONFIG_SMP.
> +Version 14 of schedstats includes support for sched_domains, which hit the
> +mainline kernel in 2.6.20 although it is identical to the stats from version
> +12 which was in the kernel from 2.6.13-2.6.19 (version 13 never saw a kernel
> +release).  Some counters make more sense to be per-runqueue; other to be
> +per-domain.  Note that domains (and their associated information) will only
> +be pertinent and available on machines utilizing CONFIG_SMP.
>  
> -In version 10 of schedstat, there is at least one level of domain
> +In version 14 of schedstat, there is at least one level of domain
>  statistics for each cpu listed, and there may well be more than one
>  domain.  Domains have no particular names in this implementation, but
>  the highest numbered one typically arbitrates balancing across all the
> @@ -27,7 +28,7 @@ to write their own scripts, the fields a
>  
>  CPU statistics
>  --------------
> -cpu<N> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 
> 27 28
> +cpu<N> 1 2 3 4 5 6 7 8 9 10 11 12
>  
>  NOTE: In the sched_yield() statistics, the active queue is considered empty
>      if it has only one process in it, since obviously the process calling
> @@ -39,48 +40,20 @@ First four fields are sched_yield() stat
>       3) # of times just the expired queue was empty
>       4) # of times sched_yield() was called
>  
> -Next four are schedule() statistics:
> -     5) # of times the active queue had at least one other process on it
> -     6) # of times we switched to the expired queue and reused it
> -     7) # of times schedule() was called
> -     8) # of times schedule() left the processor idle
> -
> -Next four are active_load_balance() statistics:
> -     9) # of times active_load_balance() was called
> -    10) # of times active_load_balance() caused this cpu to gain a task
> -    11) # of times active_load_balance() caused this cpu to lose a task
> -    12) # of times active_load_balance() tried to move a task and failed
> -
> -Next three are try_to_wake_up() statistics:
> -    13) # of times try_to_wake_up() was called
> -    14) # of times try_to_wake_up() successfully moved the awakening task
> -    15) # of times try_to_wake_up() attempted to move the awakening task
> -
> -Next two are wake_up_new_task() statistics:
> -    16) # of times wake_up_new_task() was called
> -    17) # of times wake_up_new_task() successfully moved the new task
> +Next three are schedule() statistics:
> +     5) # of times we switched to the expired queue and reused it
> +     6) # of times schedule() was called
> +     7) # of times schedule() left the processor idle
>  
> -Next one is a sched_migrate_task() statistic:
> -    18) # of times sched_migrate_task() was called
> -
> -Next one is a sched_balance_exec() statistic:
> -    19) # of times sched_balance_exec() was called
> +Next two are try_to_wake_up() statistics:
> +     8) # of times try_to_wake_up() was called
> +     9) # of times try_to_wake_up() was called to wake up the local cpu
>  
>  Next three are statistics describing scheduling latency:
> -    20) sum of all time spent running by tasks on this processor (in ms)
> -    21) sum of all time spent waiting to run by tasks on this processor (in 
> ms)
> -    22) # of tasks (not necessarily unique) given to the processor
> -
> -The last six are statistics dealing with pull_task():
> -    23) # of times pull_task() moved a task to this cpu when newly idle
> -    24) # of times pull_task() stole a task from this cpu when another cpu
> -	was newly idle
> -    25) # of times pull_task() moved a task to this cpu when idle
> -    26) # of times pull_task() stole a task from this cpu when another cpu
> -	was idle
> -    27) # of times pull_task() moved a task to this cpu when busy
> -    28) # of times pull_task() stole a task from this cpu when another cpu
> -	was busy
> +    10) sum of all time spent running by tasks on this processor (in jiffies)
> +    11) sum of all time spent waiting to run by tasks on this processor (in
> +        jiffies)
> +    12) # of timeslices run on this cpu
>  
>  
>  Domain statistics
> @@ -89,65 +62,95 @@ One of these is produced per domain for 
>  CONFIG_SMP is not defined, *no* domains are utilized and these lines
>  will not appear in the output.)
>  
> -domain<N> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
> +domain<N> <cpumask> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 
> 23 24 25 26 27 28 29 30 31 32 33 34 35 36
>  
>  The first field is a bit mask indicating what cpus this domain operates over.
>  
> -The next fifteen are a variety of load_balance() statistics:
> -
> -     1) # of times in this domain load_balance() was called when the cpu
> -	was idle
> -     2) # of times in this domain load_balance() was called when the cpu
> -	was busy
> -     3) # of times in this domain load_balance() was called when the cpu
> -	was just becoming idle
> -     4) # of times in this domain load_balance() tried to move one or more
> -	tasks and failed, when the cpu was idle
> -     5) # of times in this domain load_balance() tried to move one or more
> -	tasks and failed, when the cpu was busy
> -     6) # of times in this domain load_balance() tried to move one or more
> -	tasks and failed, when the cpu was just becoming idle
> -     7) sum of imbalances discovered (if any) with each call to
> -	load_balance() in this domain when the cpu was idle
> -     8) sum of imbalances discovered (if any) with each call to
> -	load_balance() in this domain when the cpu was busy
> -     9) sum of imbalances discovered (if any) with each call to
> -	load_balance() in this domain when the cpu was just becoming idle
> -    10) # of times in this domain load_balance() was called but did not find
> -	a busier queue while the cpu was idle
> -    11) # of times in this domain load_balance() was called but did not find
> -	a busier queue while the cpu was busy
> -    12) # of times in this domain load_balance() was called but did not find
> -	a busier queue while the cpu was just becoming idle
> -    13) # of times in this domain a busier queue was found while the cpu was
> -	idle but no busier group was found
> -    14) # of times in this domain a busier queue was found while the cpu was
> -	busy but no busier group was found
> -    15) # of times in this domain a busier queue was found while the cpu was
> -	just becoming idle but no busier group was found
> -
> -Next two are sched_balance_exec() statistics:
> -    17) # of times in this domain sched_balance_exec() successfully pushed
> -	a task to a new cpu
> -    18) # of times in this domain sched_balance_exec() tried but failed to
> -	push a task to a new cpu
> -
> -Next two are try_to_wake_up() statistics:
> -    19) # of times in this domain try_to_wake_up() tried to move a task based
> -	on affinity and cache warmth
> -    20) # of times in this domain try_to_wake_up() tried to move a task based
> -	on load balancing
> +The next 24 are a variety of load_balance() statistics in grouped into types
> +of idleness (idle, busy, and newly idle):
>  
> +     1) # of times in this domain load_balance() was called when the
> +        cpu was idle
> +     2) # of times in this domain load_balance() checked but found
> +        the load did not require balancing when the cpu was idle
> +     3) # of times in this domain load_balance() tried to move one or
> +        more tasks and failed, when the cpu was idle
> +     4) sum of imbalances discovered (if any) with each call to
> +        load_balance() in this domain when the cpu was idle
> +     5) # of times in this domain pull_task() was called when the cpu
> +        was idle
> +     6) # of times in this domain pull_task() was called even though
> +        the target task was cache-hot when idle
> +     7) # of times in this domain load_balance() was called but did
> +        not find a busier queue while the cpu was idle
> +     8) # of times in this domain a busier queue was found while the
> +        cpu was idle but no busier group was found
> +
> +     9) # of times in this domain load_balance() was called when the
> +        cpu was busy
> +    10) # of times in this domain load_balance() checked but found the
> +        load did not require balancing when busy
> +    11) # of times in this domain load_balance() tried to move one or
> +        more tasks and failed, when the cpu was busy
> +    12) sum of imbalances discovered (if any) with each call to
> +        load_balance() in this domain when the cpu was busy
> +    13) # of times in this domain pull_task() was called when busy
> +    14) # of times in this domain pull_task() was called even though the
> +        target task was cache-hot when busy
> +    15) # of times in this domain load_balance() was called but did not
> +        find a busier queue while the cpu was busy
> +    16) # of times in this domain a busier queue was found while the cpu
> +        was busy but no busier group was found
> +
> +    17) # of times in this domain load_balance() was called when the
> +        cpu was just becoming idle
> +    18) # of times in this domain load_balance() checked but found the
> +        load did not require balancing when the cpu was just becoming idle
> +    19) # of times in this domain load_balance() tried to move one or more
> +        tasks and failed, when the cpu was just becoming idle
> +    20) sum of imbalances discovered (if any) with each call to
> +        load_balance() in this domain when the cpu was just becoming idle
> +    21) # of times in this domain pull_task() was called when newly idle
> +    22) # of times in this domain pull_task() was called even though the
> +        target task was cache-hot when just becoming idle
> +    23) # of times in this domain load_balance() was called but did not
> +        find a busier queue while the cpu was just becoming idle
> +    24) # of times in this domain a busier queue was found while the cpu
> +        was just becoming idle but no busier group was found
> +
> +   Next three are active_load_balance() statistics:
> +    25) # of times active_load_balance() was called
> +    26) # of times active_load_balance() tried to move a task and failed
> +    27) # of times active_load_balance() successfully moved a task
> +
> +   Next three are sched_balance_exec() statistics:
> +    28) sbe_cnt is not used
> +    29) sbe_balanced is not used
> +    30) sbe_pushed is not used
> +
> +   Next three are sched_balance_fork() statistics:
> +    31) sbf_cnt is not used
> +    32) sbf_balanced is not used
> +    33) sbf_pushed is not used
> +
> +   Next three are try_to_wake_up() statistics:
> +    34) # of times in this domain try_to_wake_up() awoke a task that
> +        last ran on a different cpu in this domain
> +    35) # of times in this domain try_to_wake_up() moved a task to the
> +        waking cpu because it was cache-cold on its own cpu anyway
> +    36) # of times in this domain try_to_wake_up() started passive balancing
>  
>  /proc/<pid>/schedstat
>  ----------------
>  schedstats also adds a new /proc/<pid/schedstat file to include some of
>  the same information on a per-process level.  There are three fields in
> -this file correlating to fields 20, 21, and 22 in the CPU fields, but
> -they only apply for that process.
> +this file correlating for that process to:
> +     1) time spent on the cpu
> +     2) time spent waiting on a runqueue
> +     3) # of timeslices run on this cpu
>  
>  A program could be easily written to make use of these extra fields to
>  report on how well a particular process or set of processes is faring
>  under the scheduler's policies.  A simple version of such a program is
>  available at
> -    http://eaglet.rain.com/rick/linux/schedstat/v10/latency.c
> +    http://eaglet.rain.com/rick/linux/schedstat/v12/latency.c
> 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists