#249 Performance Profiling in Fedora
Closed: published by rlengland. Opened by glb.

Article Summary:

Learn how to use the profiling tools for GNOME to discover performance issues and fix them.

Article Description:

The GNOME community has wanted frame-pointers for a decade on x86_64 so that we could improve system-wide performance. With Fedora 39 not only do we have frame-pointers but also a new version of the Sysprof profiler which can take advantage of them.

Learn how profilers work and why frame-pointers are currently necessary to have a good experience with system-wide profiling.

Fedora helping to lead this change has already impacted other distributions to follow suit.

Note to Editors:

I have this article mostly finished at this point. So turn around can be fairly quick.

https://discussion.fedoraproject.org/t/article-proposal-performance-profiling-in-fedora/101654/1


Metadata Update from @glb:
- Issue assigned to hergertme
- Issue tagged with: article, needs-image

Metadata Update from @steiner:
- Custom field image-editor adjusted to steiner

Article banner up for review.

performance-profiling.png

Metadata Update from @steiner:
- Issue untagged with: needs-image

I have a rough draft at https://fedoramagazine.org/wp-admin/post.php?post=39673&action=edit which could use a look over.

Especially hopeful to improve the writing style.

It's quite long, but hopefully that is okay as I split it into two basic parts:

  1. Some simple steps to get you started playing with it
  2. The nitty-gritty details of how/why it works

Metadata Update from @rlengland:
- Custom field preview-link adjusted to https://fedoramagazine.org/?p=39673&preview=true&preview_id=39673

@hergertme We'll take a look. Thanks for the link.

Just realized after seeing the Plasma 6 Test Days issue that, for some reason, the banners I was exporting came out in half the intended resolution, so I'm fixing those now:

performance-profiling.jpg

Hello @hergertme I've started to review your article and the first thing I notice is that the images are all "embedded" in the document. I'm not certain if that is the correct term, but they have not been uploaded to WordPress and included from there. It is not possible to create links to the images in WordPress (something the editors usually check for and set up).

The problem with this situation is that users can not select the images and have them viewable in a separate window in an enlarged view. The size of the images are small enough that much of the detail is lost and I feel that the images add significantly to the value of your article. I would request that you upload your images to WordPress Media and include them from there.

I will continue to review but I wanted to let you know about this so you can work on it when you can.

Thanks.

@hergertme Investigating further I see that you have uploaded the images but the insertion was the issue. I have replaced the images and they are not presenting in a better form and linking now works.

@hergertme I completed an edit pass on your article and a couple items could be addressed.

  • The following paragraph in the Container Namespaces section is rather unclear to me.
    Specifically the "To make that more difficult..." sentence. Is it possible to reword that a bit for more clarity?

The symbolization process needs to resolve what /usr/lib/libglib-2.so inside the application mount namespace is on the host system. To make that more difficult it needs to be able to resolve potential .gnu_debuglink ELF section headers to find locate the debuginfo version of the binary. That will again reference a file within the mount namespace of the process which could look quite different from the host.

  • I added links to articles on "ORC", "DWARF", "eBPF", and "SFrame". Please check them and make certain they are appropriate for those (like me) that may be unfamiliar with the terms. If there is something available for "SHSTK" and "LBR" their addition might be good.

  • I made some sentence simplifications for readability and adjusted the paragraph splits as well, but nothing major there.

It would be very good if you would review the article to make certain I have not accidentally changed something.

Let us know when you feel it is ready to publish or if we can assist in any way.

Do note that added edits/corrections on your part are totally acceptable. Over all the article was very well done. I know I learned a lot.

Fedora users computers are our production.

should probably be

Fedora users' computers are our production.

and

The Linux kernel removed it’s support

should be

The Linux kernel removed its support

and

Each user-space process has it’s own memory layout

should be

Each user-space process has its own memory layout


Also, "Fedora 39" --> "Fedora Linux 39"

@glb ...sigh... Apostrophes always elude me, even when I know where they are suppose to go. Thanks for checking up on me.

I wasn't "checking up on you" so much as I was "scoping out the article". I just happened to notice a few things as I was perusing it. 🙂

That was only humour. I lost the smiley. :-)
Added the "Linux" in several places as well.

Metadata Update from @rlengland:
- Custom field editor adjusted to rlengland

Thanks everyone! Double thanks for fixing up my poor , and ' usage and providing the helpful links!

I did another pass and cleaned up some sentences here and there to be more understandable.

From my perspective I think it's good to go!

@hergertme Great. Thank you. I'm going to read through it one more time and then schedule it for publication tomorrow (31 Jan).

Thanks, again, for contributing.

Metadata Update from @rlengland:
- Custom field publish adjusted to 2024-01-31

Issue status updated to: Closed (was: Open)
Issue close_status updated to: published

Metadata