[PATCH] perf tools: Fix an error on perf-bench document

[PATCH] perf tools: Fix an error on perf-bench document

[Namhyung Kim]

The pipe should be messaging.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/Documentation/perf-bench.txt |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/perf/Documentation/perf-bench.txt b/tools/perf/Documentation/perf-bench.txt
index a3dbadb26ef5..d53d22f0c6ee 100644
--- a/tools/perf/Documentation/perf-bench.txt
+++ b/tools/perf/Documentation/perf-bench.txt
@@ -51,8 +51,8 @@ SUITES FOR 'sched'
 Suite for evaluating performance of scheduler and IPC mechanisms.
 Based on hackbench by Rusty Russell.
 
-Options of *pipe*
-^^^^^^^^^^^^^^^^^
+Options of *messaging*
+^^^^^^^^^^^^^^^^^^^^^^
 -p::
 --pipe::
 Use pipe() instead of socketpair()
-- 
1.7.9.2

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Ingo Molnar]

* Namhyung Kim <namhyung@kernel.org> wrote:

> The pipe should be messaging.
> 
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> ---
>  tools/perf/Documentation/perf-bench.txt |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/perf/Documentation/perf-bench.txt b/tools/perf/Documentation/perf-bench.txt
> index a3dbadb26ef5..d53d22f0c6ee 100644
> --- a/tools/perf/Documentation/perf-bench.txt
> +++ b/tools/perf/Documentation/perf-bench.txt
> @@ -51,8 +51,8 @@ SUITES FOR 'sched'
>  Suite for evaluating performance of scheduler and IPC mechanisms.
>  Based on hackbench by Rusty Russell.
>  
> -Options of *pipe*
> -^^^^^^^^^^^^^^^^^
> +Options of *messaging*
> +^^^^^^^^^^^^^^^^^^^^^^
>  -p::
>  --pipe::
>  Use pipe() instead of socketpair()
> -- 

There's several typos all across the perf bench documentation:

1)

       This perf bench command is general framework for benchmark suites.

s/is general/is a general

2)

The output of 'perf bench --help' outputs weird looking 
formatting characters like:

           .ft C
           % perf bench sched pipe                      # with no style specified
           (executing 1000000 pipe operations between two tasks)
                   Total time:5.855 sec
                           5.855061 usecs/op
                           170792 ops/sec
           .ft

3)

This:

           all: test all suite (pseudo suite)

s/all suite (psudo suite)/all benchmark suites

(this should be fixed for perf bench mem as well.)

4)

perf bench mem memcpy -h outputs:

    -l, --length <1MB>    Specify length of memory to copy. available unit: B, MB, GB (upper and lower)

bad capitalization: s/. a/. A

s/unit/units

5)

This:

    -c, --clock           Use CPU clock for measuring

should probably clarify why the CPU clock matters and what 
measurement is done otherwise.

Thanks,

	Ingo

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Namhyung Kim]

Hi, Ingo

On Mon, 18 Jun 2012 11:05:01 +0200, Ingo Molnar wrote:
> There's several typos all across the perf bench documentation:
>
> 1)
>
>        This perf bench command is general framework for benchmark suites.
>
> s/is general/is a general
>

Will fix.


> 2)
>
> The output of 'perf bench --help' outputs weird looking 
> formatting characters like:
>
>            .ft C
>            % perf bench sched pipe                      # with no style specified
>            (executing 1000000 pipe operations between two tasks)
>                    Total time:5.855 sec
>                            5.855061 usecs/op
>                            170792 ops/sec
>            .ft
>

I have no idea on this, maybe a problem of asciidoc?


> 3)
>
> This:
>
>            all: test all suite (pseudo suite)
>
> s/all suite (psudo suite)/all benchmark suites
>
> (this should be fixed for perf bench mem as well.)
>

Will fix.


> 4)
>
> perf bench mem memcpy -h outputs:
>
>     -l, --length <1MB>    Specify length of memory to copy. available unit: B, MB, GB (upper and lower)
>
> bad capitalization: s/. a/. A
>
> s/unit/units
>

Will fix too.


> 5)
>
> This:
>
>     -c, --clock           Use CPU clock for measuring
>
> should probably clarify why the CPU clock matters and what 
> measurement is done otherwise.
>

This looks a bit weird. If -c switch was given, it'd use the cycles
event not the cpu-clock event. If omitted, it'd use the gtod().

Thanks,
Namhyung

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Hitoshi Mitake]

On Tue, Jun 19, 2012 at 5:48 PM, Namhyung Kim <namhyung@kernel.org> wrote:
> Hi, Ingo
>
> On Mon, 18 Jun 2012 11:05:01 +0200, Ingo Molnar wrote:
>> There's several typos all across the perf bench documentation:
>>
>> 1)
>>
>>        This perf bench command is general framework for benchmark suites.
>>
>> s/is general/is a general
>>
>
> Will fix.
>
>
>> 2)
>>
>> The output of 'perf bench --help' outputs weird looking
>> formatting characters like:
>>
>>            .ft C
>>            % perf bench sched pipe                      # with no style specified
>>            (executing 1000000 pipe operations between two tasks)
>>                    Total time:5.855 sec
>>                            5.855061 usecs/op
>>                            170792 ops/sec
>>            .ft
>>
>
> I have no idea on this, maybe a problem of asciidoc?
>
>
>> 3)
>>
>> This:
>>
>>            all: test all suite (pseudo suite)
>>
>> s/all suite (psudo suite)/all benchmark suites
>>
>> (this should be fixed for perf bench mem as well.)
>>
>
> Will fix.
>
>
>> 4)
>>
>> perf bench mem memcpy -h outputs:
>>
>>     -l, --length <1MB>    Specify length of memory to copy. available unit: B, MB, GB (upper and lower)
>>
>> bad capitalization: s/. a/. A
>>
>> s/unit/units
>>
>
> Will fix too.
>
>
>> 5)
>>
>> This:
>>
>>     -c, --clock           Use CPU clock for measuring
>>
>> should probably clarify why the CPU clock matters and what
>> measurement is done otherwise.
>>
>
> This looks a bit weird. If -c switch was given, it'd use the cycles
> event not the cpu-clock event. If omitted, it'd use the gtod().
>

Very sorry, I misused the terms clock and cycles. The memset uses
hardware cycles event, the above naming and description with "clock"
is very bad...
I will write and send a patch for fix the problem.

-- 
Hitoshi Mitake
h.mitake@gmail.com

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Hitoshi Mitake]

On Mon, Jun 25, 2012 at 12:59 AM, Hitoshi Mitake <h.mitake@gmail.com> wrote:
> On Tue, Jun 19, 2012 at 5:48 PM, Namhyung Kim <namhyung@kernel.org> wrote:
>> Hi, Ingo
>>
>> On Mon, 18 Jun 2012 11:05:01 +0200, Ingo Molnar wrote:
>>> There's several typos all across the perf bench documentation:
>>>
>>> 1)
>>>
>>>        This perf bench command is general framework for benchmark suites.
>>>
>>> s/is general/is a general
>>>
>>
>> Will fix.
>>
>>
>>> 2)
>>>
>>> The output of 'perf bench --help' outputs weird looking
>>> formatting characters like:
>>>
>>>            .ft C
>>>            % perf bench sched pipe                      # with no style specified
>>>            (executing 1000000 pipe operations between two tasks)
>>>                    Total time:5.855 sec
>>>                            5.855061 usecs/op
>>>                            170792 ops/sec
>>>            .ft
>>>
>>
>> I have no idea on this, maybe a problem of asciidoc?
>>
>>
>>> 3)
>>>
>>> This:
>>>
>>>            all: test all suite (pseudo suite)
>>>
>>> s/all suite (psudo suite)/all benchmark suites
>>>
>>> (this should be fixed for perf bench mem as well.)
>>>
>>
>> Will fix.
>>
>>
>>> 4)
>>>
>>> perf bench mem memcpy -h outputs:
>>>
>>>     -l, --length <1MB>    Specify length of memory to copy. available unit: B, MB, GB (upper and lower)
>>>
>>> bad capitalization: s/. a/. A
>>>
>>> s/unit/units
>>>
>>
>> Will fix too.
>>
>>
>>> 5)
>>>
>>> This:
>>>
>>>     -c, --clock           Use CPU clock for measuring
>>>
>>> should probably clarify why the CPU clock matters and what
>>> measurement is done otherwise.
>>>
>>
>> This looks a bit weird. If -c switch was given, it'd use the cycles
>> event not the cpu-clock event. If omitted, it'd use the gtod().
>>
>
> Very sorry, I misused the terms clock and cycles. The memset uses
> hardware cycles event, the above naming and description with "clock"
> is very bad...
> I will write and send a patch for fix the problem.

I wrote the simple fix for eliminatng wrong usage of "clock":
https://github.com/mitake/linux/commit/b7836d41deff430c16058ca729120dccfa41e74e

This patch assumes the change made by the Namhyung's patch. So I'm
planning to send this after the Namhyung's patch is applied to tip
tree. Or should the patches be unified before applied?


-- 
Hitoshi Mitake
h.mitake@gmail.com

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Namhyung Kim]

On Wed, 27 Jun 2012 23:28:22 +0900, Hitoshi Mitake wrote:
> I wrote the simple fix for eliminatng wrong usage of "clock":
> https://github.com/mitake/linux/commit/b7836d41deff430c16058ca729120dccfa41e74e
>
> This patch assumes the change made by the Namhyung's patch. So I'm
> planning to send this after the Namhyung's patch is applied to tip
> tree. Or should the patches be unified before applied?

The patch is now in Arnaldo's tree. So you can see it in tip tree soon
unless something bad happens.

Anyway, looking at your commit (I think it's better just to send it via
email with mentioning the dependency), it seems only addresses memset
part but I think memcpy has the same problem, right?

Thanks,
Namhyung

Re: [PATCH] perf tools: Fix an error on perf-bench document

[Hitoshi Mitake]

On Thu, Jun 28, 2012 at 10:56 AM, Namhyung Kim <namhyung@kernel.org> wrote:
> On Wed, 27 Jun 2012 23:28:22 +0900, Hitoshi Mitake wrote:
>> I wrote the simple fix for eliminatng wrong usage of "clock":
>> https://github.com/mitake/linux/commit/b7836d41deff430c16058ca729120dccfa41e74e
>>
>> This patch assumes the change made by the Namhyung's patch. So I'm
>> planning to send this after the Namhyung's patch is applied to tip
>> tree. Or should the patches be unified before applied?
>
> The patch is now in Arnaldo's tree. So you can see it in tip tree soon
> unless something bad happens.
>
> Anyway, looking at your commit (I think it's better just to send it via
> email with mentioning the dependency), it seems only addresses memset
> part but I think memcpy has the same problem, right?

Thanks for your pointing. I had to fix the problem of memcpy too.
I'll send the new patch later.


-- 
Hitoshi Mitake
h.mitake@gmail.com