Project

General

Profile

Actions

Bug #228

closed

Subbuffer sizes and powers of 2

Added by Matthew Khouzam over 12 years ago. Updated over 11 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Target version:
Start date:
04/26/2012
Due date:
% Done:

0%

Estimated time:

Description

A subbuffer should be able to take any value, it is unreasonable to ask a user in userland to memorize the powers of 2 greater than 16.

A good compromise would be to cause lttng tool to create subbuffers that are rounded up to the nearest power of 2. I can personnaly garanty that a user asking for 1000000 bytes per subbuffer will not be outraged if the buffer is in reality 1024^2.

Another useful thing would be to have human friendly input like 4k 16k 1m 2g...


Files


Related issues 1 (0 open1 closed)

Related to LTTng-tools - Bug #235: lttng_enable_channel with too little subbuffers size returns an error that isn't explicitResolved05/08/2012

Actions
Actions #1

Updated by Mathieu Desnoyers over 12 years ago

Issue #228 has been reported by Matthew Khouzam.

----------------------------------------
Bug #228: Subbuffer sizes and powers of 2
https://bugs.lttng.org/issues/228

Author: Matthew Khouzam
Status: New
Priority: Normal
Assignee:
Category:
Target version:

A subbuffer should be able to take any value, it is unreasonable to ask a user in userland to memorize the powers of 2 greater than 16.

A good compromise would be to cause lttng tool to create subbuffers that are rounded up to the nearest power of 2. I can personnaly garanty that a user asking for 1000000 bytes per subbuffer will not be outraged if the buffer is in reality 1024^2.

Another useful thing would be to have human friendly input like 4k 16k 1m 2g...

Agreed, this should be done in lttng cmd line tool. Patch is welcome,

Thanks,

Mathieu

--
You have received this notification because you have either subscribed to it, or are involved in it.
To change your notification preferences, please click here: http://bugs.lttng.org/my/account

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Actions #2

Updated by Yannick Brosseau over 12 years ago

  • Target version set to 2.1 pre
Actions #3

Updated by Yannick Brosseau over 12 years ago

  • Target version deleted (2.1 pre)

It might also be an helper function (in the lib) that do the round up

-- edited to add (in the lib)

Actions #5

Updated by Mathieu Desnoyers about 12 years ago

Please implement by pulling lttng-ust libringbuffer/backend_internal.h get_count_order() into lttng-tools. This code is already tested, and FWIW more efficient than the patch you proposed. Moreover, it follows the project coding style.

Thanks,

Mathieu

Actions #6

Updated by Matthew Khouzam about 12 years ago

That's a pretty cool function. What is a FLS? I can tell the lttng_ust_fls() function is more desirable for this than get_count_order since if we give ti a subbuffer of 2^n, we don't need to allocate 2^n+1... or do we?

Also, the only comments are "arch-agnostic implementation" event though that is great to know, I think we really need to know what is a fls even more and what get_count_order does.

Actions #7

Updated by Mathieu Desnoyers about 12 years ago

Issue #228 has been updated by Matthew Khouzam.

That's a pretty cool function. What is a FLS? I can tell the lttng_ust_fls() function is more desirable for this than get_count_order since if we give ti a subbuffer of 2^n, we don't need to allocate 2^n+1... or do we?

Also, the only comments are "arch-agnostic implementation" event though that is great to know, I think we really need to know what is a fls even more and what get_count_order does.

fls = find last bit set. There is a similar implementation in the Linux
kernel, so you can find its documentation there.

fls has corner-cases with the exact power values: it's not the semantic
we need. We want get_count_order. get_count_order, if given 2^n, returns
2^n, not 2^(n+1) like you assume.

Thanks,

Mathieu

----------------------------------------
Bug #228: Subbuffer sizes and powers of 2
https://bugs.lttng.org/issues/228#change-775

Author: Matthew Khouzam
Status: New
Priority: Normal
Assignee:
Category:
Target version:

A subbuffer should be able to take any value, it is unreasonable to ask a user in userland to memorize the powers of 2 greater than 16.

A good compromise would be to cause lttng tool to create subbuffers that are rounded up to the nearest power of 2. I can personnaly garanty that a user asking for 1000000 bytes per subbuffer will not be outraged if the buffer is in reality 1024^2.

Another useful thing would be to have human friendly input like 4k 16k 1m 2g...

--
You have received this notification because you have either subscribed to it, or are involved in it.
To change your notification preferences, please click here: http://bugs.lttng.org/my/account

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Actions #8

Updated by David Goulet about 12 years ago

  • Status changed from New to Confirmed
  • Target version set to 2.2

I'm +1 to adding a helper call in liblttng-ctl rather than hacking the values behind the user back.

Flagging this for 2.2. Anyone want to contribute, feel free! :)

Actions #9

Updated by Mathieu Desnoyers over 11 years ago

  • Status changed from Confirmed to Resolved
commit 16068db5dda7041571a5520dc69c96af43b799a2
Author: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Date:   Thu May 2 11:53:08 2013 -0400

    Fix: check errors in lttng command argument values

    Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

commit 8ce58badbfbff17e050194e01a9120ac6058bc1c
Author: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Date:   Thu May 2 11:33:46 2013 -0400

    Implement get_count_order in lttng utils

    Imported from Userspace RCU rculfhash.c.

    Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Actions

Also available in: Atom PDF