WEBVTT 00:00.000 --> 00:14.840 Hello everyone, my name is Kashyab-Chamersy and I work for Red Hat. Unfortunately, I 00:14.840 --> 00:18.000 really like to speak. Cospeaker David cannot make it, so he left me to pick up the 00:18.000 --> 00:27.040 pieces. I hope he can hear me in the back. So I began working with five 11 months ago, 00:27.040 --> 00:32.480 so last first step, just about then. But before that, my background is in virtualization 00:32.480 --> 00:37.360 and virtualization based management tools. So yeah, today we'll talk a little bit about 00:37.360 --> 00:45.680 Federal State of Affairs and what's going on. Yeah, this is a collective effort, so I'm 00:45.680 --> 00:51.520 presenting many people to work here. Some people might be in the room, so without them, all 00:51.520 --> 00:57.440 of this won't be possible. A quick, a timeline slide. It's almost 10 years of 00:57.440 --> 01:03.200 a risk-five in the Federal lands. 2016, Richard Jones is not here at the moment, but I started 01:03.200 --> 01:08.640 the bootstrap and David, my close-picker, started right after that, and then there were 01:08.640 --> 01:14.160 cummi-based builders and slowly there's a packaging, porting grind of, you know, porting 01:14.240 --> 01:22.880 all the 24,000-plus packages to risk five, over the years, and slowly hardware came into existence, 01:22.880 --> 01:30.240 so builders that are based on hardware. And most recently last year, we've had several highlights, 01:30.960 --> 01:36.400 talk a bit more in the next slide, but this is also one of the last year, we had, I think the first 01:36.400 --> 01:41.920 time, both the main primary architecture, Federal images, and the risk-five architecture. 01:44.400 --> 01:54.400 I went out on the same day. Sorry. I went out on the same day, so there was also a rail 10 01:54.400 --> 02:04.960 developer preview on this five. Yeah, in Federal 43 rebuild, and so on. 26, some more improvements, 02:04.960 --> 02:11.680 we'll talk a bit. So that's sort of the rough happenings in the recent past year. One of the main 02:11.760 --> 02:17.200 things is we moved to a new cozy server, the build infrastructure, so David used to maintain 02:18.000 --> 02:26.240 another overlay, which tracks the non upstream changes in Fedora, the packaging related changes, 02:26.240 --> 02:32.560 patches that are not upstream or in progress, so they build. And that's now moved to the primary 02:32.560 --> 02:38.640 Fedora infrastructure, so that's a, that's starting on Fedora infrastructure, this new server, 02:38.720 --> 02:43.760 and that's a good step towards risk-five becoming a primary architecture, because we want the 02:43.760 --> 02:49.040 builders also to be running on the Fedora infrastructure, so this is a good step towards that. 02:50.240 --> 02:58.720 And during the Federal 43, we had a nasty bug, most of the 43 was ready for a long time, and we 02:58.720 --> 03:05.520 hit a sort of nasty bug in debugger, that the program that provides scripts and other tools to 03:06.480 --> 03:11.840 properties to produce tracing and debugging data, but that's sort of the work with upstream 03:11.840 --> 03:19.040 containers, bin noodles, folks, debugger it, maintainers, and that's sort of the unblocked some 15 03:19.040 --> 03:24.880 or packages that were not being, that were blocked due to this issue, but that's not solved. 03:26.320 --> 03:32.240 And of course, there's the chimneysweep task of streaming changes that are not yet 03:33.200 --> 03:40.880 streamed from the separate code you server, we'll maintain to Fedora proper, so that's the classic 03:40.880 --> 03:47.840 upstream work, and yeah, some more improvements like in the past, Fedora images were built locally, 03:47.840 --> 03:51.440 now they're built in the standard infrastructure, you're classic tooling that the excited 03:51.440 --> 04:00.320 excel to use is, and most recently we've got QMU emulated risk-five sea troupes in the 04:00.320 --> 04:08.240 infrastructure, so package your scan, now test their software and produce arpins, the copper is 04:09.280 --> 04:16.160 a place where people can upload custom arpins for people to test, for example if you have, 04:16.160 --> 04:23.280 I don't know some bug fix, an OVMF firmware, a virtual firmware software, and you want to provide 04:23.280 --> 04:28.640 bills for testing, that's a good place to do that before they go into Fedora main, so that's, 04:28.960 --> 04:35.360 where you know you have these are QMU emulated sea troupes, but still that's something nice to have, 04:36.000 --> 04:43.280 then nothing. Okay, we have a situation, and that's the hardware, what's going on? 04:44.000 --> 04:50.480 Now before we go, last year, Emil, a season in the room, and did an excellent talk on all the 04:50.640 --> 04:57.840 different kinds of risk-five boards for the past six years, and I highly recommend you to watch it 04:57.920 --> 05:01.840 if you're new to this like a system, so I'm not going to cover all that things, so Emil gives 05:01.840 --> 05:08.160 a really good overview of all these different boards, and so yeah, I recommend you to watch that, 05:08.160 --> 05:16.160 and this we will see what's new relatively. This is my, the first board, actually I have, 05:16.160 --> 05:20.800 this is the only board, actually I have, I'm testing, now some of my desk, it has a it's standard 05:20.880 --> 05:30.720 form factor, the P550 was released 2024, and it's a standard ITX mini ITX I guess, yes, and the fan 05:30.720 --> 05:37.440 is a bit loud, but this is the main machine where I'm doing all the build work, and most of Fedora 05:37.440 --> 05:43.360 builders are also, the heavy duty stuff happened on this, and it's based on the easwing system 05:43.360 --> 05:48.800 on ship, and that's being upstream, some of them are upstream parts of the drivers are upstream, others 05:48.800 --> 05:55.280 like clock and so on, are still being upstream, this version 9, I guess, that one, clock is it, 05:55.280 --> 06:00.720 I'm not a kernel guide, so I don't pretend to be that, but those who are interested in that, so you can 06:00.720 --> 06:11.440 check the details. Okay, this is a bit of a novelty board, so this is from 10 starrants, this comes as a 06:11.440 --> 06:16.480 PCIe card that you can slot in, and it can run in so on Linux, and it's a black hole thing, 06:16.560 --> 06:25.680 and it's the use case mostly for AI inferencing, this is my AI moment of the talk, not my there, 06:25.680 --> 06:32.800 I guess, so yeah, it's for AI inferencing and training, and I haven't had a chance to try this 06:32.800 --> 06:40.800 out yet, maybe soon, and it can boot with mainline kernel, and there's some limitations, I think it's 06:40.880 --> 06:47.040 only like UART serial console kind of a thing, and Jufa Stini, both the patches, and that's being 06:47.040 --> 07:01.200 worked on. This is somewhat relatively new, this from milk 5, that's based on a chip call URDP 07:02.160 --> 07:10.640 1000, UR is the ultra risk, and that's, that has four more cores than the board we saw before, 07:10.640 --> 07:19.600 this side 5, high 5, premier p5, 5, 0, and I have access to an evil board of this CPU to not 07:19.600 --> 07:27.280 the type in itself, but that's, so let's add the risk by some it from Fuwei, one of the community 07:27.280 --> 07:35.120 members who did have an access to type in itself, and this is definitely, in my limited testing, 07:35.120 --> 07:40.320 I only got access like two weeks ago, in my limited testing, it's been reasonably good in the 07:40.320 --> 07:46.560 bills I'm doing, and mostly all the work I'm doing is for bills, so it's pretty decent so far, 07:46.560 --> 07:51.600 but remains to be seen, I have some benchmark, like very basic simple benchmarks, poor 07:52.160 --> 07:57.920 benchmarks, I'll show a couple of them, to just give a sense of it, again the upstreaming some 07:57.920 --> 08:08.000 parts are upstreamed, others are still in progress. Okay, this was announced like two days ago, 08:08.560 --> 08:14.640 I'm going to find Jupiter 2, a couple of URLs move find this, another one called side p, 08:14.720 --> 08:24.240 I guess they're also selling this thing, and this is based on SpaceMet Core K3, and it has a 08:24.240 --> 08:28.640 couple of wallplanes, I don't have access to it, so I can't say much meaningful things about it, 08:29.920 --> 08:37.760 but it is supposed to be the compatible with the latest ISA specification, and also has 08:37.760 --> 08:42.960 hypervisor support it says, so I'm curious about all the hardware virtualization stuff, so I think 08:43.040 --> 08:49.440 it remains to be seen, this is, it's possible to order maybe some of already placed orders, I guess, 08:49.440 --> 08:58.400 in this room. Okay, this one is again from 10 star, and I saw this at 08:58.960 --> 09:04.560 Risfy Summit, not America last year, a couple of months ago, that thing itself, what you see on the 09:04.560 --> 09:10.800 screen is not the board, obviously, it's the FPGA that apparently costs as much as a sportscar, 09:10.880 --> 09:15.360 which is not that much when you're working in this semiconductor industry, I'm told, 09:16.400 --> 09:24.240 and that's running their Askelon CPU for the Atlantis platform, that's the board that 09:24.240 --> 09:29.840 is supposed to come out. Next quarter, there was a bit of delay I heard, but I don't know much about 09:29.840 --> 09:36.000 how that's progressing, but the current estimated ETA is later this year, and you can actually 09:36.080 --> 09:41.840 try this out in emulated hardware, in QMU, the quick emulator, the rest-style program, and 09:42.800 --> 09:47.360 for a simple command line, you can try, of course, you can construct very complex command lines 09:47.360 --> 09:53.280 that QMU too. Again, for this one also, upstreaming a started in the kernel too, there's some 09:53.280 --> 09:57.840 links for those who are interested, I haven't uploaded slides yet, but I will do it later after this. 09:58.800 --> 10:04.160 So yeah, the actual board will be obviously much smaller than this one, and this is, yeah. 10:06.960 --> 10:09.760 I think this was running doom or something, that's a classic game, right? 10:11.440 --> 10:16.560 Okay, this is the benchmarks I was talking about, this is again, simple compression of a 10:16.560 --> 10:23.520 federized file with exact one for an eight course, and for single core, you almost have like 50 10:23.520 --> 10:31.840 person performance overhead for P550, but I mean, benefit or lead for DP 1000. 10:32.400 --> 10:41.920 When I say P550, I'm talking about the sci-fi high-fi premier P550, just for short, 10:43.440 --> 10:51.040 and that has four cores, I say, mentioning, and DP 1000 is eight cores, obviously, it scales 10:51.040 --> 10:58.720 a bit better, few on more details I have linked there, yeah, this is the other bin noodles, 10:58.800 --> 11:04.160 because we were debugging that debugging it saga, so I think that's one of the reasons I did 11:04.160 --> 11:08.560 bin noodles, or maybe David asked me, my course speaker, to do, to try the bin noodles build, 11:08.560 --> 11:13.760 and that's sort of the rough, without testing the benchmark itself, I kind of drops the 11:14.480 --> 11:20.400 time in the kernel space by half, and quite, quite decent bill speed, if you're doing a lot of 11:20.400 --> 11:26.640 bills, this is what you're looking for, this kind of speeds. I haven't done the kernel benchmark, 11:26.960 --> 11:34.400 there's a KC bench tool and other stuff like that, so LVM, which will take two days, maybe more. 11:35.840 --> 11:42.480 So yeah, that's surely driven by its higher combination, it's cache combination, L3 and the 11:42.480 --> 11:52.480 last level cache, and also higher IPC, yeah, this is the basic thing, didn't do anything more beyond this. 11:52.720 --> 11:59.840 Okay, I don't want to get more into details, this is just for completeness, all of 12:00.960 --> 12:07.360 Bernstein has a really nice web page with excellent benchmarks for developers to do, RVV, okay, 12:07.360 --> 12:20.080 there he is, with the back of the room, for developers to try performance portable core, so definitely 12:20.160 --> 12:26.240 check out that link once slides are up, and there are also some more benchmarks a couple of years 12:26.240 --> 12:33.680 ago by Richard and Andrea from the risk of community, which shows again, bill times for different 12:33.680 --> 12:38.960 packages, because that's what we're doing in federal rights, a lot of bills, a lot of other things. 12:41.200 --> 12:47.760 Okay, so we're here, what do you want to pay, if you're just getting started? I think there are many 12:47.760 --> 12:53.920 risk-fight developers here, I see many, many of some human faces, but for those who are new to this, 12:55.280 --> 13:00.960 as I mentioned at the beginning, Emil's talk, last year's talk still has, all the advice he talks 13:00.960 --> 13:05.760 about there is much of it is still valid, so if it depends on what you're trying to do, which 13:05.760 --> 13:13.120 what to choose, if you want something quick and cheap, try the star 5 version 5 2, it also has 13:13.120 --> 13:19.280 upstream support, but you are something slightly, if you have slightly more fans to burn, the 13:20.080 --> 13:27.360 sci-fi P50 has pretty decent performance, and a lot of federal rights bills like GCC, LVM, 13:27.360 --> 13:34.000 are all like QTE, all these are targeted at this board, and this is also what I've been using, 13:34.000 --> 13:39.280 I've been pretty happy with it, mostly plug and play, more due to some small configurations at the 13:39.600 --> 13:45.920 beginning, and if you're interested in the vector stuff, there's two boards, I'm going to apply 13:45.920 --> 13:52.240 Jupiter and the banana pie is also vector capable, I'm going to add a vector challenge to myself, 13:52.240 --> 13:58.320 because it's still a new to the ecosystem, but I haven't tried those boards myself, but you can 13:59.040 --> 14:03.840 try them out here and probably check all last benchmarks if you're interested in the needy 14:03.920 --> 14:12.400 security. Okay, last, once the RE823 capable boards are here, should be around the corners, 14:12.400 --> 14:19.360 so we are told, try one of these Atlantis, probably that's, I'm looking forward to that one, 14:19.360 --> 14:27.520 for sure, and yeah Titan is also around the corner should be out available, March, I was told, 14:27.600 --> 14:35.840 early March, and yeah, the most recently announced Jupiter 2, so that's, that's about it there. 14:39.200 --> 14:45.120 Okay, let's talk a little bit about what we have in store for upcoming releases and beyond. 14:46.320 --> 14:52.800 Don't be scared, so this, I saw this slide myself last week, so how the kernels are built today, 14:52.880 --> 14:58.720 the federal 43 kernels, the rest of the kernels, board specific kernels, so when they're trees, 14:58.720 --> 15:05.120 Jason Montelian from the rest of the community does this, I'm on collaborating with others, 15:06.400 --> 15:12.160 so different when they're a kernel trees, re-based them or forward both the patches onto the mainline, 15:12.160 --> 15:18.800 stable kernels, and then you know apply some federal RP and specific sauce and stuff like that, 15:19.120 --> 15:25.040 and then you know he uploads some stuff to his personal space, some RPM to maybe test 15:25.040 --> 15:31.520 more experimental patches or something, and the rest of them he runs it through the RIS5 build system 15:31.520 --> 15:39.440 and they end up in the RIS5 image, so that's, it's a lot, and there's a lot of federal specific kernels, 15:39.440 --> 15:45.680 so that's a problem, right? The solution here is to come up with a unified kernel that will reduce 15:45.760 --> 15:52.240 this complexity much more to the single kernel that will work across boards, but that requires a lot of testing, 15:52.800 --> 15:59.360 so all the forward-porting and re-baseding will be done on a single always ready kernel, 15:59.360 --> 16:05.920 the arc kernel repo, and that's the plan, and he also already has Jason already has 16:07.600 --> 16:13.920 some testing with the latest 6.19 RC1 and RC2 kernels, some features work with several boards, 16:13.920 --> 16:19.200 some don't, he has a spreadsheet that is tracking somewhere and maybe we'll be published in a 16:19.200 --> 16:27.280 more in a different format over the weeks, as the dust settles down, and Jason's kernel arc repo 16:27.280 --> 16:35.440 has more details, and he'll get up repos also around with the vendor trees, so the earlier 16:36.320 --> 16:42.880 mess of a diagram will simplify like the top one, let's ignore the bottom one for a sec, 16:42.880 --> 16:52.080 so it's form was simplified, the vendor trees and the patches that are in flight on LKML or elsewhere 16:52.080 --> 16:57.440 are applied to the kernel arc repository and then run it through the build system, produce the build, 16:58.080 --> 17:04.800 and these are apparently because these are for the framework laptop at the bottom, the LTS alternatives, 17:06.800 --> 17:12.000 trees, these have some bugs with the graphics and so on, they don't quite work with the mainline 17:12.080 --> 17:20.480 kernels, so therefore Jason chose the long-term kernels that's still work, so that will 17:20.480 --> 17:26.800 result over time, I guess, so she's more of the mention for the completeness sake there, so that's 17:26.800 --> 17:33.120 one of the main things and also looking forward, so that's sort of this of boards that Jason and others 17:33.200 --> 17:44.160 are testing, it's a group effort and about 1718 boards, that's the bottom is the evaluation board 17:44.160 --> 17:52.000 of the DP1000, that look fine, it's going to come up with, it's a lot of work, but it's going to be exciting 17:54.560 --> 18:01.920 what else, yeah, it's again chimneys weeping, there's still a bunch of factors that need some work, 18:03.280 --> 18:11.440 usual suspects the kernel and the shim, open duty K and so on, and mostly failing tests that 18:11.440 --> 18:18.400 you need to work with upstream and follow them up, give their good reproducer and trial the patches 18:18.400 --> 18:24.560 and get them solved, so working with them bugging some LLVM containers and they're reasonably responsive, 18:25.120 --> 18:30.000 so that's good, it's just a box standard upstream work, there's nothing special in that sense there, 18:30.960 --> 18:36.640 there's a respite tracker, it's about 100 packages or so, so, not super important, but 18:37.840 --> 18:43.840 yeah, you know, to get them done, eventually, the details are in the respite tracker there, 18:46.160 --> 18:51.600 what else, yeah right now most of the hardware is in other people's deaths, which is not exactly 18:51.680 --> 19:01.120 resilience, so we won't move the builders, Koji builders, the systems that do all they have 19:01.120 --> 19:05.600 lifting in terms of builds in classic federal infrastructure hardware, that's also a part of 19:06.400 --> 19:15.840 a requirement for making a primary architecture on, yeah, in Inferdora, so we're working with the 19:15.920 --> 19:21.200 federal engineering steering committee and others to, you know, make that happen, it will take some time, 19:21.920 --> 19:29.760 just need to get the right things in place, like this, and once the Avi 23 boards are available, 19:29.760 --> 19:36.400 preferably those because they will stand, you know, you don't garbage them two quick, so 19:36.800 --> 19:44.000 because if you use the older boards likely to go, it's still soon, soon, it's maybe, 19:44.880 --> 19:51.520 so, but anyway, hope it will be our hope or expect, rather, our Avi 23 systems to be racked, 19:53.440 --> 20:01.600 and yeah, working with the federal team and others to sort out the details there, just process. 20:01.680 --> 20:11.680 Okay, this is a big topic, what about the baseline, right? Right now, like, 20:11.680 --> 20:20.480 other prominent distributions, federal also uses the RV64 GC as the baseline, and still, 20:20.480 --> 20:28.720 the community is still discussing, went to move and sort out the details, it's an open discussion, 20:28.880 --> 20:33.040 David has some thoughts on it, so if you're on metrics, I think, should be around, maybe 20:33.040 --> 20:39.760 he's even watching the stream, so it's an open question and we're still figuring out, once, 20:39.760 --> 20:45.440 what big thing, it's open because we're in prayer reference, I'm going to back from, right? So once 20:45.440 --> 20:55.600 this Atlantis or Space Mid-K3 boards are available, we can then use that to, as I say, target that, 20:55.680 --> 21:02.880 so that's evolving situation, but what about the software side? Right, so for that, 21:04.640 --> 21:09.520 Drew Jones, current developer in the nearby space, did a really nice talk at the 21:09.520 --> 21:16.560 RISFICE summits NA last year, where he covers a lot of ground in terms of the, how 21:17.840 --> 21:24.960 software ecosystem stays, you know, tries to catch up with the extensions that are being produced, 21:24.960 --> 21:29.040 you often hear that hardware is ready, software is not ready, but he makes a good case that 21:29.040 --> 21:35.840 it's not quite the situation, software is also catching up, and he points out one of the things, 21:35.840 --> 21:41.360 I just put there from that talk, is the system called the hardware prop, which will let you, it's a 21:41.360 --> 21:49.520 key value pair that will let you detect the extensions reliably for user space programs, so 21:50.480 --> 21:57.520 the details are in the kernel documentation, so we can check out that talk, and of course, 21:57.520 --> 22:05.120 the RVA-23 kernels, the latest spec kernels, should support or will support, because 22:06.960 --> 22:14.240 kernel doesn't break the user space, right? So should support the older baseline user space, 22:15.200 --> 22:20.160 and, again, well, that's an evolving situation, we don't have a concrete answer yet, 22:20.160 --> 22:24.880 it's the figuring out the details of it, just want to rush into something that turns out into a mess, 22:25.440 --> 22:32.880 so just want to be more thoughtful and careful about it. What else, I think that's that's that's that one 22:32.880 --> 22:42.880 of this. Yeah, actually, I don't have a lot more here, so obviously, you don't always get a chance 22:42.880 --> 22:50.640 to enable new architectures, so it can be a good opportunity to try out how all the ecosystem works, 22:50.640 --> 22:57.200 there's a lot of variables, I myself, I'm still learning, and so yeah, come join if you're interested 22:57.200 --> 23:03.760 in this sort of thing, test the packages that you, you're really care about, report the bugs, 23:03.760 --> 23:09.120 work without the strain, it's a classic upstream work, and yeah, through each out, there's a 23:09.200 --> 23:14.480 metric channel, that's where most of the active communication happens, on the Federer 23:14.480 --> 23:19.200 metric server, and there's also discussion forum and Federer devil list obviously as well, 23:19.920 --> 23:29.680 so that's the main areas of communication. Yeah, a bunch of references, so that's pretty much it, 23:29.760 --> 23:40.800 I've gone actually. Questions, comments? What is Federer VForce? Okay, so the question is what 23:40.800 --> 23:46.320 is the Federer VForce? That's a good question, Federer VForce is another community by the Chinese 23:46.320 --> 23:52.880 Academy of Sciences, and China, they're producing vendor specific borne, so that's not official Federer. 23:53.200 --> 23:57.760 So official Federer is all the stuff that happens in the Federer metric channel, and the Federer 23:57.760 --> 24:04.720 list, and so on, that is when there are specific borne, so that's why the branding is different, 24:04.720 --> 24:14.320 I'm not a branding specialist, but yeah, that's not primary for Dora, I hope that answers, yeah, 24:14.320 --> 24:37.760 thank you. Go for it. Please speak a bit. Flat packs. It's really hard to listen here. 24:37.760 --> 24:44.400 Yeah, I don't know much about flat packs, but there is some discussion, so the question is about 24:44.400 --> 24:55.120 flat packs, and you're building on P550, did you say? The Wild5 Jupiter, so I don't know much about 24:55.120 --> 25:00.320 flat packs myself, there is some discussion in the community I heard from the Federer project leader, 25:00.320 --> 25:04.800 as well, there's some kerfuffle going on in the community, I was, I'm not up to speed on that, 25:04.800 --> 25:11.120 but we can discuss on the metric channel, so I don't have meaningful answer to that one of our 25:11.200 --> 25:20.160 fries. Yes, the food loader has triggered the food loader's life in there, for the Lord, 25:20.160 --> 25:25.360 I'm a checker, we have a package, you maintain the food, we are able to move for different 25:25.360 --> 25:30.240 hardware, let's say let's go to the vendor, because that's one might have a different system, 25:30.240 --> 25:34.400 how are you able to work with that with the different hardware that you have now for a spy? 25:34.400 --> 25:38.960 Are you going to maintain the safety of the package, are you going to have been there? 25:38.960 --> 25:50.000 Okay, the question is about U-boot, and what we're going to do, I actually have absolutely 25:50.000 --> 25:56.000 no idea how we're going to deal with that, so we have to figure that out on the, on the, 25:56.000 --> 26:00.880 there's so many things to track, so it's impossible to keep up speed on everything, 26:00.880 --> 26:06.800 but you can join the metric channel and ask us there, sure David has a response, 26:06.800 --> 26:13.280 because he has all the federal cash in his memory, so he's, so he's been building it for 26:13.280 --> 26:18.480 like 10 years or so long, sorry David, I'm putting it in a spot in the ether, 26:18.480 --> 26:27.680 what, I just question please, what about the deviant route? 26:27.680 --> 26:42.240 Okay, the question is, Ubuntu decided to be RVH23, the change the baseline, or at least I read, 26:42.240 --> 26:47.280 I don't know, they did or not, but I read some noise, what about Fedora? Is that the, is that the 26:47.280 --> 26:52.480 question? Yes, okay, well as I was saying a bit ago, we don't have a concrete decision yet, 26:52.480 --> 26:58.000 we don't rush into something to market something, Fedora really cares about staying up close 26:58.000 --> 27:02.880 enough stream and you know doing things the right way, even if it takes a bit longer, so it's 27:02.880 --> 27:10.560 being discussed and I think that's, that's my estimate, it's not a, it's not a set in stone answer 27:10.560 --> 27:17.440 anything, likely that we'll switch to that sooner once the baseline is available, yes, so that's 27:17.440 --> 27:23.840 the, and so it's still an evolving situation, somebody doesn't have to be figured out, yes, 27:23.840 --> 27:30.480 I just want to clear up that the, yes, the lane is going to ask which to RVH23, okay? 27:30.480 --> 27:42.160 The node is then another test that's still using RVH22, yeah, so, okay, okay, okay, so, 27:42.160 --> 27:50.160 I'm going to clarify that, okay, I'm going to clarify that the latest, the one to, 27:50.160 --> 28:04.160 kernels are RVH23 based, and some of the older boards are RVH64GC, any other questions? 28:04.160 --> 28:12.000 I think we good? That's kind of totally implemented for it, is there anything that you're 28:12.320 --> 28:17.920 majorly missing in terms of toolkits outside of just hardware that runs and stuff, 28:17.920 --> 28:23.600 if what else can hardware people give you now? How do you mean how do I, maybe, how do I 28:23.600 --> 28:32.320 order to begin with? Sorry, the question was, what else can the hardware people give besides, 28:32.320 --> 28:37.600 are you missing anything, so more seriously, are you missing anything? It's a good question, 28:37.600 --> 28:46.320 I'm not sure if I can give a deep answer to this, first of all, yeah, of course, 28:46.320 --> 28:52.320 yeah, better performance, my personal interest in virtualization, so I would like to see better 28:52.320 --> 28:56.880 hypervisor supports, that says the personal answer, David would have a different answer, 28:56.880 --> 29:04.880 maybe he's watching and on the stream and all half-and-tots, so yeah, there is one thing I would 29:04.880 --> 29:14.000 like to say, you don't need to have risk-fight hardware to help. Yeah, okay, Marseille is a contributor 29:14.000 --> 29:20.000 also in the federally system, so he is clarifying that you don't really need a hardware to contribute 29:20.000 --> 29:28.080 the package, because he himself uses heavy, cumumier-based emulators, it is a simulator to do all 29:28.080 --> 29:31.920 his bill work, so he's just making a point that if you have a strong enough arm system or 29:32.000 --> 29:41.840 x86, that can run good cumumier binaries, emulated VMs, that is slightly faster than some of the hardware 29:41.840 --> 29:47.920 that you have, that's fair, yes, but not everyone has access to such heavy beefy missions, 29:47.920 --> 29:53.440 so that's the other side, so how are we doing at that time, by anything else, yeah, five minutes 29:53.440 --> 30:05.760 or can take some time to breathe? Okay, I think that's about it, thank you very much.