Poking At A Big NUMA Benchmark Regression In Linux 5.18 Git

[phoronix]

Phoronix: Poking At A Big NUMA Benchmark Regression In Linux 5.18 Git

There still is a few days left to the Linux 5.18 merge window but already I've started firing up benchmarks of this new kernel on a handful of desktops and servers so far. One benchmark though in particular has been showing a staggering performance drop on Linux 5.18 on multiple systems but overall Linux 5.18 in my testing thus far has been working out well...

https://www.phoronix.com/scan.php?page=news_item&px=Linux-518-Stress-NUMA-Goes-Boom

[Linuxxx]

Interesting to note that the patch notes talk about CONFIG_PREEMPT_RT, while the actual patches to turn Linux into a "hard" real-time OS haven't been merged yet...

[F.Ultra]

Interesting to note that the patch notes talk about CONFIG_PREEMPT_RT, while the actual patches to turn Linux into a "hard" real-time OS haven't been merged yet...

While technically true, there have been merges of parts of the RT patches here and there. The CONFIG_PREEMPT_RT setting to kconfig was added back in 2019 to make it easier to merge more of the RT patches upstream. And it does change things in the current kernel even though it still doesn't make it fully hard realtime.

[Danny3]

OMG, the regression is huge!

Hope to see it fixed soon.

I still don't understand what is the Linux foundation doing with all that money, while not full time employing Michael or making available for him a few systems where he can automate building the Linux kernel for each commit and do performance testing so that performance regressions are noticed as soon as possible.

[onlyLinuxLuvUBack]

OMG, the regression is huge!

Hope to see it fixed soon.

I still don't understand what is the Linux foundation doing with all that money, while not full time employing Michael or making available for him a few systems where he can automate building the Linux kernel for each commit and do performance testing so that performance regressions are noticed as soon as possible.

All Michael has to do is code they will do the rest , this is from their web....



SUPPORT PROGRAMS



You code, we’ll do the rest

[skeevy420]

All Michael has to do is code they will do the rest , this is from their web....



SUPPORT PROGRAMS



You code, we’ll do the rest

#!/usr/bin/env zsh

# Send $1,000,000 to skeevy420@Phoronix

[coder]

OMG, the regression is huge!

It's a corner case that synthetic tests like that are designed to tease out. Otherwise, it'd be showing up a lot more broadly and probably wouldn't have gotten this far.

I still don't understand what is the Linux foundation doing with all that money,

Are you trying to understand, or you just see a big pot of money with the name "Linux" written somewhere on it, and assume that most of it should go towards Linux kernel development?

while not full time employing Michael or making available for him a few systems where he can automate building the Linux kernel for each commit and do performance testing so that performance regressions are noticed as soon as possible.

Again that's not how these sorts of things usually work. For a project to get funded from the foundation, there's probably some kind of grant process, where applicants would have to submit a proposal for funding. If Michael wanted to jump through those hoops and formalize what he's doing, I'm sure he could try for it.

The other possibility is that the kernel project decides to step up its own regression testing effort and creates a position, whereupon maybe he can apply for a full-time position. That would probably mean no more Phoronix, as we've known it, unless he wants to hand off reins to someone else.

I'm all for better regression testing. I don't really know anything about what the kernel project does, so maybe someone here could help enlighten us. However, I think you can be assured that many downstream users of the Kernel are testing release candidates on their own workloads and filing bugs when they spot significant regressions. This latter point could have created a sense that the kernel project doesn't need to worry so much about doing its own performance tests. Also, consider that you'd need a potentially phenomenal amount of hardware to do a decent job of it.

[coder]

All Michael has to do is code they will do the rest , this is from their web....



SUPPORT PROGRAMS



You code, we’ll do the rest

It's important to read the web page, in order to understand what they mean by "the rest". And you could've included a link, so we can at least be sure we're looking at the same thing you quoted from.

https://linuxfoundation.org/projects-support-programs/

In a phrase, they're offering: "Expert resources and useful tools to help make your project a success, from ideation to enterprise adoption."

What constitutes resources & tools seems to be:

• Developer Enablement
• Business and IT Operations
• Ecosystem Development
• Project Operations
• LFX Tools
• Event Management
• Training and Certification
• Infrastructure Management
• Legal and IP Management
• Finance and Operations
• HR Support
• Marketing Operations
• Community Growth
• Creative Support

In other words, all of the non-technical or non-core stuff needed to make a project succeed and scale. It is not simply getting some of that sweet, sweet cash for writing a few (or even quite a lot of) shell scripts and PHP or whatever.

The trend, in philanthropy, seems to be about putting resources where they can have a multiplicative effect. Sometimes, barriers take only a little of the right resources to remove, and part of those resources come in the form of expertise. Otherwise, you could end up pouring lots and lots of cash into a relatively small number of projects with little to show for it, because the project leaders quite likely don't even have all the expertise, time, attention, and interest to use it most effectively.

In the startup business, it's typical for a tech person to partner up with a business person and co-found a company. That's because most techies don't have the interest or expertise to deal with all the business stuff. However, most open source projects are started by one or more techies, without the sort of expertise to deal with all the other stuff. So, their focus seems to be about providing a lot of that other stuff + the hosting infrastructure and some of the other dev ops.

[onlyLinuxLuvUBack]

It's important to read the web page, in order to understand what they mean by "the rest". And you could've included a link, so we can at least be sure we're looking at the same thing you quoted from.

https://linuxfoundation.org/projects-support-programs/

In a phrase, they're offering: "Expert resources and useful tools to help make your project a success, from ideation to enterprise adoption."

What constitutes resources & tools seems to be:

• Developer Enablement
• Business and IT Operations
• Ecosystem Development
• Project Operations
• LFX Tools
• Event Management
• Training and Certification
• Infrastructure Management
• Legal and IP Management
• Finance and Operations
• HR Support
• Marketing Operations
• Community Growth
• Creative Support

In other words, all of the non-technical or non-core stuff needed to make a project succeed and scale. It is not simply getting some of that sweet, sweet cash for writing a few (or even quite a lot of) shell scripts and PHP or whatever.

The trend, in philanthropy, seems to be about putting resources where they can have a multiplicative effect. Sometimes, barriers take only a little of the right resources to remove, and part of those resources come in the form of expertise. Otherwise, you could end up pouring lots and lots of cash into a relatively small number of projects with little to show for it, because the project leaders quite likely don't even have all the expertise, time, attention, and interest to use it most effectively.

In the startup business, it's typical for a tech person to partner up with a business person and co-found a company. That's because most techies don't have the interest or expertise to deal with all the business stuff. However, most open source projects are started by one or more techies, without the sort of expertise to deal with all the other stuff. So, their focus seems to be about providing a lot of that other stuff + the hosting infrastructure and some of the other dev ops.

>> Otherwise, you could end up pouring lots and lots of cash into a relatively small number of projects with little to show for it,

Right right, like skype wasnt a waste of money ?

These big corps throw tons of cash around with little to show for it

notice the top bar has big corp icons to their stuff but no codeberg?

No agendas. No politics. Just stuff developers need to build and maintain projects.


Linux foundation has big rich cats as platinums but can only afford to have conference in austin,tx?

If you are trying to help grow why isn't even one of their cons west coast + east coast + austin ?

[coder]

>> Otherwise, you could end up pouring lots and lots of cash into a relatively small number of projects with little to show for it,

Right right, like skype wasnt a waste of money ?

These big corps throw tons of cash around with little to show for it

I thought we were talking about The Linux Foundation.

notice the top bar has big corp icons to their stuff but no codeberg?

No agendas. No politics. Just stuff developers need to build and maintain projects.

I'm not familiar with that organization, but it seems as if they're aiming to provide similar services as Linux Foundation. If so, why would they funnel their resources though LF? The fact that they exist in parallel to it suggests some difference in approach or philosophy, but I can't exactly say.

Linux foundation has big rich cats as platinums but can only afford to have conference in austin,tx?

I can't comment on that, but there might be travel concerns still lingering from the pandemic and its aftershocks.

If you are trying to help grow why isn't even one of their cons west coast + east coast + austin ?

In this day and age, probably most of their attendees will be virtual anyhow.

I thought your question of "why Austin" was more aimed at "why not also a Euro conference". IMO, it'd make more sense to have any additional conferences on different continents (e.g. Europe, Asia, India, Africa, South America).