Netflix Optimized FreeBSD's Network Stack More Than Doubled AMD EPYC Performance

[Space Heater]

"AMD's tools are lacking (even on Linux)."
AMD's tools are lacking, even more so on Linux. FTFY.

Could you explain this a bit more? What AMD performance tools are available for FreeBSD but not Linux?

[wizard69]

I'm not trying to start a software licence war, but I assume all their code is closed source and these improvements wont make it back to FreeBSD?

Why these absolute paranoid posts? First off you could have investigated the status of the patches. Second if they did keep this internal the software license mean nothing.

it looks like much of this has in fact made it back to FreeBSD and I would imagine it will go through the same review process seen for other patches just like stuff for Linux gets reviewed. You do realize that stuff developed by external companies gets rejected all the time for these projects. If not completely rejected it could take years before everybody is satisfied for kernel inclusion. The use of an open source license does not assure inclusion of ones code. It is very much a two way street!

[Space Heater]

Author here.. Every optimization is either (a) in head now, or (b) has a review pending. The slides reference the FreeBSD SVN revision numbers where patches landed, and the differential review numbers where the open reviews are.

Drew, is there a mailing list or review site to watch development of this sort of stuff? if it's on https://reviews.freebsd.org/ - then I'm not able to find it.

[rhavenn]

I'm not trying to start a software licence war, but I assume all their code is closed source and these improvements wont make it back to FreeBSD?

Nagh, they upstream a lot of their code. It's easier to support that way and like someone else said. They're not making money selling appliances, so it doesn't hurt their bottom line any. I'm sure they have application code to do the actual distribution, etc...that isn't open source, but it wouldn't be if they were using Linux either.

[rhavenn]

Could you explain this a bit more? What AMD performance tools are available for FreeBSD but not Linux?

I don't think that's what they said / meant. I think they meant that they don't have the tools and that even under Linux they don't have them. ie: It's not hurting them to be using FreeBSD for this because even on Linux it wouldn't be any better.

[ssokolow]

...on mobile?

When I see a YouTube video I always open the article in an incognito tab, but if by accident I open it in the normal ones, I must clear all my cookies and cache.

If you're on Android, uMatrix is compatible with Firefox Mobile.

No clue whether it's compatible with any of the WebKit skins on iOS though I wouldn't rule it out. It's basically a frontend for synthesizing and injecting Content-Security-Policy headers to leverage the browser's built-in blocking functionality and iOS WebKit does support CSP 2.0.

[drewg123]

Could you explain this a bit more? What AMD performance tools are available for FreeBSD but not Linux?

None.. FreeBSD has a much worse version of AMDuProf which just outputs a CSV with CPU utilization per core, and a few other things. It looks like the Linux tool is closer to something like Vtune.

But in both cases, neither tool (AFAIK) can access any information regarding infinity fabric utilization. This contrasts with Intel, where both Vtune and the Intel PCM Tools can access QPI utilization information.

[Space Heater]

I don't think that's what they said / meant. I think they meant that they don't have the tools and that even under Linux they don't have them. ie: It's not hurting them to be using FreeBSD for this because even on Linux it wouldn't be any better.

The author of the presentation is saying that neither platform is supported well by AMD, but the person I'm quoting isn't the author of the presentation.

Instead, they are claiming that "AMD's tools are lacking, even more so on Linux." and given that this is about AMD on FreeBSD the natural implication is that they are saying "AMD's tools are lacking, even more so on Linux [compared to FreeBSD]."

[drewg123]

Drew, is there a mailing list or review site to watch development of this sort of stuff? if it's on https://reviews.freebsd.org/ - then I'm not able to find it.

KTLS affinity (trivial): https://reviews.freebsd.org/D21648
TCP_REUSEPORT_LB_NUMA: https://reviews.freebsd.org/D21636

[mrugiero]

The kernel is GPLv2. They only have to share the source with people who legally received the corresponding binaries. That's the whole reason the AGPL was created.

What you're thinking of does exist though... it's the reason that Open Watcom C/C++ can't get into Debian or Fedora package repositories. (The Sybase Open Watcom Public License requires you to publish the source to your changes to the compiler, even if you're only distributing the modified binaries within your company. Debian and Red Had have decided that doesn't meet their definition of a Free Software License... I assume because it is a de facto violation of the "No discrimination against fields of endeavor, like commercial use." clause in the Debian Free Software Guidelines.)

You're right, I forgot about internal use not counting as redistribution.