sorear changed the topic of #riscv to: RISC-V instruction set architecture | https://riscv.org | Logs: https://libera.irclog.whitequark.org/riscv | Matrix: #riscv:catircservices.org
aredridel47 has quit [Quit: The Lounge - https://thelounge.chat]
aredridel47 has joined #riscv
KREYREN has joined #riscv
BootLayer has joined #riscv
vagrantc has quit [Quit: leaving]
naoki has quit [Quit: naoki]
heat has quit [Ping timeout: 240 seconds]
mlw has joined #riscv
naoki has joined #riscv
freakazoid332 has quit [Ping timeout: 260 seconds]
frkazoid333 has joined #riscv
riktw has quit [Quit: ZNC 1.8.2+deb2+b1 - https://znc.in]
riktw has joined #riscv
mlw has quit [Ping timeout: 245 seconds]
BootLayer has quit [Quit: Leaving]
JanC has quit [Ping timeout: 252 seconds]
zBeeble24 has joined #riscv
zBeeble42 has quit [Remote host closed the connection]
mlw has joined #riscv
indy_ is now known as indy
davidlt has joined #riscv
handsome_feng has joined #riscv
jobol has joined #riscv
alperak has joined #riscv
davidlt has quit [Ping timeout: 252 seconds]
clever has quit [Ping timeout: 240 seconds]
clever has joined #riscv
KREYREN has quit [Remote host closed the connection]
handsome_feng has quit [Quit: Connection closed for inactivity]
fuwei has joined #riscv
fuwei has quit [Ping timeout: 264 seconds]
fuwei has joined #riscv
naoki has quit [Quit: naoki]
handsome_feng has joined #riscv
hightower2 has quit [Ping timeout: 268 seconds]
fuwei has quit [Remote host closed the connection]
fuwei has joined #riscv
esv has joined #riscv
Tenkawa has joined #riscv
davidlt has joined #riscv
billchenchina has joined #riscv
billchenchina has quit [Remote host closed the connection]
Noisytoot has quit [Ping timeout: 264 seconds]
hightower2 has joined #riscv
psydroid has joined #riscv
luca_ has joined #riscv
luca_ is now known as OwlWizard
OwlWizard has quit [Quit: OwlWizard]
JanC has joined #riscv
Trifton has quit [Read error: Connection reset by peer]
Trifton has joined #riscv
hightower2 has quit [Ping timeout: 256 seconds]
davidlt has quit [Quit: Leaving]
davidlt has joined #riscv
Stat_headcrabed has joined #riscv
heat has joined #riscv
JanC has quit [Remote host closed the connection]
JanC has joined #riscv
mfny has joined #riscv
<mfny> Hi all
<mfny> wondering if anyone here has eperciance with the Sifive HiFive Unmatched board at all ?
Stat_headcrabed has quit [Quit: Stat_headcrabed]
<drmpeg> mfny: I have one.
<palmer> mfny: should be a bunch of people who do, are you looking for anything specific?
stolen has joined #riscv
<mfny> I am potentially looking to buy one. And have some questions like what would I require in addition to the board to get it working exactly ? What GPUs are supported ? and Aside from Linux support is there any BSD support of any kind ?
<palmer> there's an old list on the kickstarter, but it's essenially just a case/PSU: https://www.crowdsupply.com/sifive/hifive-unmatched . We had good luck with AMD cards, the system is so slow even fairly old stuff was fast enough that the GPU wasn't a bottleneck. IDK about BSD ;)
<q66> radeon/amdgpu largely works
<q66> the board is slow as hell though
<palmer> ya, anything that has decent upstream support should be fine. It's so slow it doesn't even really matter what GPU you have, though ;)
<Tenkawa> palmer: even if we can get this Star64 mainline working I'm worried the x1 lane is still going to be its weakness too in that regard....
BootLayer has joined #riscv
ln5 has quit [Remote host closed the connection]
<mfny> which AMD cards in particular would be recommended ?
handsome_feng has quit [Quit: Connection closed for inactivity]
<drewfustini> conchuod: i'll post the current th1520 clk series today. It does work for me with a "real" clock property for mmc now. there are some things that still need to be improved but i will post today so its not just hidden away in my personal repo.
<smaeul> mfny: depends on your requirements. Radeon RX 550 or Radeon RX 6400 are relatively inexpensive and don't require external power
<smaeul> anything RDNA or newer requires this kernel patch series: https://lore.kernel.org/linux-riscv/20240329072441.591471-1-samuel.holland@sifive.com/
<mfny> ok
<davidlt> mfny, RX500 works out of the box today
<gurki> is there any point in plugging a gpu in there? like ... how bad is it, really?
<davidlt> You can watch 4K movies with it
<gurki> thats .. the gpu doing all the work with hw decoding
<Tenkawa> does the Unmatched come with a Imagination on it?
<mfny> does HD 6000 work out of box also ?
<davidlt> Yeah, GPU decoding it, but user experience is fine
<gurki> ah. the latter was where im getting at
<davidlt> browser experience is kinda slow, that's annoying
<davidlt> I would advice that for JIRA :)
<Tenkawa> davidlt: nice headless server eh?
<Tenkawa> git box
<davidlt> I would prefer more horse power on my servers
<davidlt> large git repo checkout/processing is kinda slow especially if there is no hw rng.
<Tenkawa> Yeah.. most of my "copilers" all have 8 cores or more
<Tenkawa> er compilers
<Tenkawa> some 12-16
<drmpeg> mnfy: Have you looked at the StarFive VisionFive 2? That's what most folks are using these days.
Stat_headcrabed has joined #riscv
<drmpeg> Much cheaper for similar performance.
<smaeul> mfny: probably. I think most testing is done with the `amdgpu` kernel driver, but `radeon` probably works too
<smaeul> looks like the Milk-V Pioneer ships with R5 230, which is also Terascale, so that suggests it will be fine
EchelonX has joined #riscv
rbmarliere has quit [Quit: rbmarliere]
raym has quit [Ping timeout: 268 seconds]
Noisytoot has joined #riscv
psydroid has quit [*.net *.split]
Stat_headcrabed has quit [*.net *.split]
mfny has quit [*.net *.split]
frkazoid333 has quit [*.net *.split]
Finde has quit [*.net *.split]
haritz has quit [*.net *.split]
llu has quit [*.net *.split]
alexghiti has quit [*.net *.split]
drmpeg has quit [*.net *.split]
bitoff_ has quit [*.net *.split]
peepsalot has quit [*.net *.split]
somlo has quit [*.net *.split]
smaeul has quit [*.net *.split]
josuah has quit [*.net *.split]
The_Decryptor has quit [*.net *.split]
sjs has quit [*.net *.split]
[exa] has quit [*.net *.split]
stefanct has quit [*.net *.split]
dobson has quit [*.net *.split]
jljusten has quit [*.net *.split]
indyZ has quit [*.net *.split]
arcade_droid has quit [*.net *.split]
rsalveti has quit [*.net *.split]
dlan has quit [*.net *.split]
agraf has quit [*.net *.split]
knielsen has quit [*.net *.split]
alperak has quit [*.net *.split]
mlw has quit [*.net *.split]
DesRoin has quit [*.net *.split]
bjdooks has quit [*.net *.split]
paddymahoney has quit [*.net *.split]
JTL has quit [*.net *.split]
GenTooMan has quit [*.net *.split]
cronos has quit [*.net *.split]
clandmeter has quit [*.net *.split]
clemens3 has quit [*.net *.split]
Esmil has quit [*.net *.split]
duckworld has quit [*.net *.split]
crabbedhaloablut has quit [*.net *.split]
test924 has quit [*.net *.split]
foxbat has quit [*.net *.split]
lu_zero has quit [*.net *.split]
troglodito has quit [*.net *.split]
shachaf has quit [*.net *.split]
nelnire has quit [*.net *.split]
nathanchance has quit [*.net *.split]
paulbarker has quit [*.net *.split]
NishanthMenon has quit [*.net *.split]
geist has quit [*.net *.split]
conchuod has quit [*.net *.split]
mturquette has quit [*.net *.split]
mlw has joined #riscv
mturquette has joined #riscv
conchuod has joined #riscv
geist has joined #riscv
paulbarker has joined #riscv
NishanthMenon has joined #riscv
nathanchance has joined #riscv
nelnire has joined #riscv
shachaf has joined #riscv
troglodito has joined #riscv
foxbat has joined #riscv
lu_zero has joined #riscv
test924 has joined #riscv
crabbedhaloablut has joined #riscv
rbmarliere has joined #riscv
ganboing has joined #riscv
Stat_headcrabed has joined #riscv
mfny has joined #riscv
psydroid has joined #riscv
alperak has joined #riscv
DesRoin has joined #riscv
frkazoid333 has joined #riscv
bjdooks has joined #riscv
Finde has joined #riscv
paddymahoney has joined #riscv
haritz has joined #riscv
GenTooMan has joined #riscv
llu has joined #riscv
JTL has joined #riscv
cronos has joined #riscv
clandmeter has joined #riscv
clemens3 has joined #riscv
Esmil has joined #riscv
JTL has quit [Ping timeout: 268 seconds]
peepsalot has joined #riscv
somlo has joined #riscv
The_Decryptor has joined #riscv
sjs has joined #riscv
alexghiti has joined #riscv
drmpeg has joined #riscv
indyZ has joined #riscv
bitoff_ has joined #riscv
smaeul has joined #riscv
josuah has joined #riscv
[exa] has joined #riscv
stefanct has joined #riscv
arcade_droid has joined #riscv
jljusten has joined #riscv
dobson has joined #riscv
rsalveti has joined #riscv
knielsen has joined #riscv
agraf has joined #riscv
dlan has joined #riscv
duckworld has joined #riscv
duckworld has joined #riscv
duckworld has quit [Changing host]
duckworld has quit [Max SendQ exceeded]
bitoff_ has quit [Remote host closed the connection]
duckworld has joined #riscv
bitoff_ has joined #riscv
JTL has joined #riscv
nmeum has quit [Remote host closed the connection]
nmeum has joined #riscv
another| has quit [Remote host closed the connection]
hightower2 has joined #riscv
Trifton has quit [Quit: Client exited]
<mfny> drmpeg: i looked into the StarFive VisionFive 2, is it really similar performance to the Unmatched ? i note that GPU drivers are not there yet tho acording to what ive found ..
<q66> it's supposedly a little faster
<q66> but not by much since it's effectively the same cores
<q66> still raspberry pi 3 tier
another has joined #riscv
<drmpeg> Yeah, I guess you have to use some vendor kernel for the GPU.
<drmpeg> I have a VF2, but I never really did much with it.
<drmpeg> The Unmatched will run a completely vanilla kernel.
<q66> vendor kernel and proprietary userland, and a distro that can support said proprietary userland
<mfny> so if i want the most OS support and GPU support the unmatched would be the choice ?
stolen has quit [Quit: Connection closed for inactivity]
<drmpeg> My use case is Linux kernel testing. I use it headless and do testing for linux-stable and the latest kernel.
<drmpeg> I can build and run 5.15.x, 6.1.x, 6.6.x LTS kernels no problem. 5.15.x requires a few patches, but 6.1.x and later require none.
hightower3 has joined #riscv
<drmpeg> And I just run Ubuntu.
<drmpeg> I'll probably load 24.04 pretty soon.
hightower2 has quit [Ping timeout: 260 seconds]
<palmer> IIUC the main reason VF2 needs that vendor kernel is for the workaround for their PCIe bugs, their "just slow down a bit" stuff is unlikely to land upstream so someone would need to figure it out
<drmpeg> IF you can wait, SiFive has a new board on the way. https://www.sifive.com/boards/hifive-premier-p550
<drmpeg> Supposedly available in July.
<mfny> yeah but will it be cheaper or more expensive then the Unmatched tho
<mfny> Unmatched is about as expensive as i can go
<gurki> mfny: what do you want to do with that system?
<drmpeg> Dunno. I'll guess about the same.
<gurki> i feel like boards are still mostly targeted at developers trying to evaluate $thing
<mfny> gurki: i want to "play" with RISCV in a desktop PC like setting
eightthree has quit [Ping timeout: 264 seconds]
<palmer> mfny: unless you want to hack on upstream kernel stuff, you might be better off getting one of the cheaper SBCs then?
<palmer> Unmatched is super expensive for what you get, and it's not really usable as a desktop because it's extremely slow
eightthree has joined #riscv
<mfny> i see
<drmpeg> There isn't even a Firefox package for Ubuntu.
<Tenkawa> drmpeg: The VF2 is probably the best utilized and mainlined RISC-V sbc out there currently with imo performance/per cost
<drmpeg> Mainlined?
<Tenkawa> drmpeg: yes
<drmpeg> I thought PCI was still pending.
<Tenkawa> so? pending != not-exists
<drmpeg> Then it's not mainline.
<Tenkawa> most of RV is in flux continually
<Tenkawa> drmpeg: it will be in transition for yeats
<Tenkawa> er years
<Tenkawa> so was ARM
<drmpeg> I'm not sure what you're saying.
<Tenkawa> I unfortinately had a front row seat for that one..
<Tenkawa> what I'm saying is "mainline" is a marketing term
<Tenkawa> people use it too much
<drmpeg> I'm saying I can git clone linux, build it without any patches and run it.
<drmpeg> Which is no problem on Unmatched.
<drmpeg> That's "mainline" to me.
<Tenkawa> Yeah and with about a hour of work I have about 20x the performance and x*more features on the VF2
<Tenkawa> I'm still likely to buy another one to replace mine that got a power hit
<drmpeg> 20x?
<Tenkawa> yes
<Tenkawa> the unmatched is terribly slow
<Stat_headcrabed> Are eswin guys working on mainlining that p550 chip?
<Tenkawa> Stat_headcrabed: hey you had some Milk-V equip didn't you?
<Stat_headcrabed> No, VF2 only
<Tenkawa> darn.. we found something really odd
<Stat_headcrabed> milkv has tooo much problems(C906)
<Stat_headcrabed> What happens?
<Tenkawa> if you take a Mars CM4 and put it on a RPI IO Board it acts completely different (memory/eeprom) than using a Waveshare one...
<Tenkawa> It is the oddest thing
<Stat_headcrabed> vendor uboot?
<Tenkawa> seems to be u-boot related
<Tenkawa> yeah
<Stat_headcrabed> might be something wrong on vendor u-boot
<mfny> i did look at the Milk_V Pioneer but its way to expensive
<mfny> 2.5k for a system
<Stat_headcrabed> mfny: better wait for mikv oasis
<Tenkawa> the Waveshare reports wrong ram but right info about board... RPI... right ram.. .wrong info about board
<Stat_headcrabed> That's using SG2380 and much cheaper
<Stat_headcrabed> also has RVV1.0 support
<Tenkawa> its crazy that the carrier board can make such a big impact
<Tenkawa> (kinda scary)
<mfny> when is the Oasis out ?
<Stat_headcrabed> Maybe a carrier board auto detection mechanism, but behaves wrong
<Stat_headcrabed> mfny: according to official, about end of this year
<Stat_headcrabed> But I think we may wait longer for it
<Tenkawa> Well I'll be nice and engrained into RPI5 land by then... *shrug*
<mfny> oh i cant wait that long lol
<Stat_headcrabed> So what do you expect from buying pioneer?
<Stat_headcrabed> C910 is also "QUITE BUGGY"
<Stat_headcrabed> (also C920v1, but not v2
* Tenkawa will be waiting for someone to need him to buy something to debug
<Tenkawa> that's essentially how I end up with equip like that nowadays lol
<Stat_headcrabed> The FPU problem on them already can be seen when using OpenBLAS
<Stat_headcrabed> And seems can't be solved in software
ximp has joined #riscv
<Tenkawa> Stat_headcrabed: sounds like that old fpu x86 problem from 20 yrs ago or so
<Tenkawa> 20+
<Stat_headcrabed> Ahhhh
<Stat_headcrabed> Also I heard that someone is making a chip with 8xC908
<Stat_headcrabed> I want to get one when released
<Tenkawa> ouch... 30 years
<Tenkawa> I'm old
<Stat_headcrabed> LOL
BootLayer has quit [Quit: Leaving]
<Tenkawa> I remember that when it happened
<Stat_headcrabed> Maybe we should just disable FPU for those IPs
<Stat_headcrabed> Windows was doing the same
<Tenkawa> heheheh yep
<Tenkawa> they had to
<Tenkawa> lol
<Stat_headcrabed> Thankfully it is fixed on C920v2 and C908
<conchuod> Stat_headcrabed: IIRC I did mention disabling the FPU on those systems to Palmer in November - but the impact would be pretty bad, seeing as you'd need a userspace that doesn't use the FPU.
heat_ has joined #riscv
heat has quit [Read error: Connection reset by peer]
<Stat_headcrabed> Maybe a fallback to software FP using glibc?
<sorear> Stat_headcrabed: how do you observe the issue with openblas? does openblas use the exception flags?
<Tenkawa> conchuod: yeah is there even a build of Debian/Ubuntu anymore without FPU?
<Stat_headcrabed> is it possible?
<sorear> glibc no, bbl maybe
<Tenkawa> musl even needs it too right?
<conchuod> bbl being berkley boot loader sorear ?
<Tenkawa> although I doubt musl is used in riscv yet
<sorear> the "userspace" softfloat code lives in libgcc/compiler-rt and it's already supported in both for glibc, musl, and uclibc targets
<Tenkawa> oh...
<conchuod> "Are eswin guys working on mainlining that p550 chip?" I've yet to see a patch for it, but it is also not yet released. I'd be surprised tbh if someone like Samuel got one that he'd not end up adding support.
<Stat_headcrabed> sorear: not me, but T-head has a contest to college students to fix that openblas problem
<Stat_headcrabed> emmmm
<sorear> glibc has a copy of the softfloat code but I don't really understand when it's used
<sorear> i can personally confirm that uclibc and musl targets work fine with just the libgcc support
<sorear> it's also possible to emulate floating point using the illegal instruction handler; this is a standard feature of bbl (the "berkeley boot loader" even though it does no actual boot loading) and was widely used in the 2010s
<Stat_headcrabed> Maybe porting that to opensbi?
<conchuod> I don't think those systems use bbl though, right? So it'd be the SBI implementations job I guess.
<Stat_headcrabed> Just like unaligned access on U74?
<sorear> the opensbi maintainers explicitly vetoed that years ago
<conchuod> sorear: not as if t-head use upstream opensbi anyway :)
<Stat_headcrabed> Oh
<conchuod> But ye, clearly not a solution that'd work for anything other than their vector stack if it was kept in their opensbi tree given that tree's other problems
<Stat_headcrabed> Emmm
<sorear> Stat_headcrabed: do you have a link for the openblas thing? all I can find is https://github.com/OpenMathLib/OpenBLAS/issues/3506 and that's a "user attempted to configure openblas to use RVV 1.0 on a chip that doesn't support RVV 1.0" issue not a float errata issue
<sorear> if you need to emulate floating point on a c906 (this is not confirmed to be a real issue for typical use, see above) it's a little bit more complicated than just "use bbl" or "copy bbl logic into opensbi" because you need floating point instructions to generate exceptions.
<sorear> the c906 doesn't implement writable misa.F, unlike sifive cores, so the only way to do that is to set mstatus.FS = Off. but now whatever kernel you're using needs to have a "virtual sstatus.FS"
<Stat_headcrabed> sorear: Sorry I can't find that, but I'm 100% sure it exists
<sorear> if you find anything, let us know
<Stat_headcrabed> ok
<Stat_headcrabed> Btw
<Stat_headcrabed> Seems C906 doesn't completely support double precision on it's rvv implementation
<Stat_headcrabed> According to OpenBLAS readme
<palmer> is it the same rounding bugs as the scalar stuff, or something new?
<Stat_headcrabed> Possibly something new
zBeeble24 has quit [Ping timeout: 268 seconds]
ganboing has quit [Quit: Connection closed for inactivity]
Stat_headcrabed has quit [Quit: Stat_headcrabed]
<gurki> so is it "not supported" or "broken"?
<sorear> by "rounding bugs" do you mean "FE_UNDERFLOW flag setting bugs" or something more generally applicable?
<palmer> I forget the exact bug, just those that showed up on the glibc lists
<palmer> I'm not really an FP guy, so I just sort of put all these "it does't quite compute what IEEE says it should" bugs in the same bin ;)
Andre_Z has joined #riscv
heat_ has quit [Read error: Connection reset by peer]
heat_ has joined #riscv
unnick has quit []
davidlt has quit [Ping timeout: 264 seconds]
unnick has joined #riscv
EchelonX has quit [Quit: Leaving]
zBeeble has joined #riscv
unnick has quit []
jfsimon1981 has joined #riscv
raym has joined #riscv
zBeeble42 has joined #riscv
zBeeble has quit [Remote host closed the connection]
zBeeble42 has quit [Read error: Connection reset by peer]
zBeeble has joined #riscv
eightthree has quit [Ping timeout: 260 seconds]
eightthree has joined #riscv
mlw has quit [Ping timeout: 256 seconds]
ximp has quit [Quit: Leaving]
alexghiti has quit [Ping timeout: 260 seconds]
alexghiti has joined #riscv
jobol has quit [Quit: Leaving]
Noisytoot has quit [Excess Flood]
Noisytoot has joined #riscv
motherfsck has quit [Ping timeout: 255 seconds]
unnick has joined #riscv
psydroid has quit [Quit: KVIrc 5.0.0 Aria http://www.kvirc.net/]
esv has quit [Remote host closed the connection]
heat_ has quit [Remote host closed the connection]
heat_ has joined #riscv
jfsimon1981 has quit [Remote host closed the connection]
cousteau has joined #riscv
another is now known as another|
heat has joined #riscv
cousteau_ has joined #riscv
heat_ has quit [Read error: Connection reset by peer]
cousteau has quit [Ping timeout: 255 seconds]
motherfsck has joined #riscv
cousteau_ has quit [Remote host closed the connection]
cousteau_ has joined #riscv
sakman has quit [Remote host closed the connection]
heat has quit [Remote host closed the connection]
naoki has joined #riscv
Andre_Z has quit [Quit: Leaving.]
cousteau` has joined #riscv
cousteau_ has quit [Ping timeout: 255 seconds]
cousteau` is now known as cousteau
KREYREN has joined #riscv
alperak has quit [Quit: Connection closed for inactivity]
Trifton has joined #riscv
DesRoin has quit [Ping timeout: 268 seconds]
DesRoin has joined #riscv