[itchyjunk] has quit [Remote host closed the connection]
ttmrichter has joined #riscv
rlittl01 has quit [Quit: -a- Connection Timed Out]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
hendursaga has quit [Ping timeout: 276 seconds]
hendursaga has joined #riscv
winterflaw has quit [Ping timeout: 276 seconds]
solrize has quit [Changing host]
solrize has joined #riscv
solrize has left #riscv [Leaving]
riff-IRC has quit [Remote host closed the connection]
riff-IRC has joined #riscv
BOKALDO has joined #riscv
cmuellner_ has joined #riscv
aredridel has quit [Read error: Connection reset by peer]
peeps[zen] has joined #riscv
aredridel has joined #riscv
sirn- has joined #riscv
wolfshappen_ has joined #riscv
zapb__ has joined #riscv
Xark has joined #riscv
abelvesa_ has joined #riscv
smartin has joined #riscv
oaken-so1rce has joined #riscv
apalos- has joined #riscv
cmuellner__ has quit [*.net *.split]
wolfshappen has quit [*.net *.split]
sirn has quit [*.net *.split]
peepsalot has quit [*.net *.split]
reinhardt has quit [*.net *.split]
dh` has quit [*.net *.split]
wigyori has quit [*.net *.split]
Xark_ has quit [*.net *.split]
abelvesa has quit [*.net *.split]
zapb_ has quit [*.net *.split]
apalos has quit [*.net *.split]
oaken-source has quit [*.net *.split]
cs_0x6373 has quit [*.net *.split]
Bigcheese has quit [*.net *.split]
ahs3 has quit [*.net *.split]
sirn- is now known as sirn
cs_0x6373 has joined #riscv
cronos has joined #riscv
ahs3 has joined #riscv
wigyori has joined #riscv
mobius has quit [Ping timeout: 250 seconds]
mobius has joined #riscv
Jmabsd has joined #riscv
<Jmabsd>
guys just curious, does anyone have a Sifive Unmatched board - how many watts does it consume really???? SiFive's spec says 150W, that sounds absolutely excessive to me
<sorear>
don't forget the power supply also has to supply power to whatever graphics card you choose and other peripherals
<pierce>
Jmabsd: i am not sure, tbh, i don't have anything to measure it with
<Jmabsd>
pierce: does it get hot, is it noisy???
<Jmabsd>
sorear: sure! but that is "extra-spec" relative to CPU+motherboard
<drmpeg>
The stock CPU fan is very noisy.
<pierce>
<Jmabsd> "Pierce: does it get hot, is it..." <- as a whole? nah
<Jmabsd>
drmpeg: oh can it be replaced with anything more silent? i like really silent
<pierce>
mine's currently overclocked to 1.5GHz and it seems to be fine
<Jmabsd>
overall are you happy with it?
<Jmabsd>
oh, default is 1.2 is it
<Jmabsd>
in say 18 months from now, do you believe you will still have use of it?
<Jmabsd>
or it' smore of a curious historical artefact waiting to be binned
<pierce>
i ripped off the stock fan, as drmpeg said, it's crazy noise
<pierce>
it's practically silent
<pierce>
and i managed to shoehorn on an old AIO water cooler
<pierce>
it's just the GPU that makes noise now
<pierce>
PSU doesn't turn its fans on
<pierce>
which would be a good indicator of it not consuming much power
<Jmabsd>
mm
<Jmabsd>
nanoPSU also works, those come in 150W
<pierce>
Jmabsd: yeah, it's early adopter stuff though
<pierce>
there are a couple rough edges
<pierce>
like USB dropping out
<pierce>
Jmabsd: me personally, yeah, i'll continue to use it as my porting box
<pierce>
Jmabsd: i see it as both
<pierce>
i like the fact that this is practically the first desktop risc-v device
<Jmabsd>
yeah it is a real icebreaker
<pierce>
but it is very usable for what i want it for
<pierce>
software still needs to catch up though
<Jmabsd>
utility-wise, i guess it's maybe 3x or 4x the utility of a Raspberry Pi 4
<Jmabsd>
the CPU is a bit slower, but the IO is absolutely great
<Jmabsd>
and the RAM is enough for build etc
<Jmabsd>
and it's ECC. so it's kind of, quality home server, just a bit slow cpu
<pierce>
whatever browser you use, you won't be running it quickly
<Jmabsd>
pierce: chrome gets jammed really? i know there's a RISCV backend for V8 now
<Jmabsd>
is it slow even then?
<Jmabsd>
i thiiiiiink a Raspberry Pi 4 will do web browsing O K doesn't it? ...
<pierce>
yeah if you are building for RISC-V, you'd want to look at cross compilation if you're impatient
<Jmabsd>
the SiFive has a real AMD GPU too so should be superior by far for browsing?
<Jmabsd>
i guess web browsers offload as much of graphics as they can to the GPU
<pierce>
Jmabsd: i haven't tried the V8 stuff as reported recently, i think someone is currently getting it working outside of opensuse
<Jmabsd>
pierce: for real world web browsing i'd presume that patch will make a world of difference.
<pierce>
Jmabsd: the problem here is the missing JIT for epiphany/webkit2
<Jmabsd>
i would presume a decent AMD GPU together with that ported V8, should make it a totally decent web browsing computer
<Jmabsd>
pierce: if you use Chromium that's not an issue or what?
<Jmabsd>
you are talking Firefox now right
<pierce>
and i can't get firefox or chromium to build from master
<pierce>
so the changes to get it to build are sitting in forks somewhere
<Jmabsd>
so there is RISCV64 OpenSuse which does get chromium to build??
<pierce>
from what i read, yeah
<pierce>
i'm currently running ubuntu server with the gnome shell on the umatched
<pierce>
and yeah, the only browser i got on there is gnome-web
<pierce>
i think with the release of ubuntu 20.10 there is expected to be a bump in performance as per the phoronix review
<pierce>
but i haven't tried the pre-release
<pierce>
something for me to look forward to later this month
<drmpeg>
I think that performance bump mentioned in the Phoronix review is just clock rate.
<drmpeg>
1.0 to 1.2 GHz
<pierce>
ah okay, i didn't read it completely, and just skimmed over the graphs
<pierce>
i guess the question for you Jmabsd is, what do you expect to do with it?
<rjek>
IO on the Unmatched is excellent, but the CPU is about one quarter of a RPi4
<Jmabsd>
pierce: Can you share with me any reference that mentions that OpenSuse now incorporates the V8 patch for RISCV64?
<rjek>
I don't recall how many PCIe channels come out of the SoC
<pierce>
sure, i'll see if i can find it
<Jmabsd>
pierce: Ubuntu 20.10 will become faster for RISCV64 due to some kernel tweak?
<Jmabsd>
pierce: web browsing such!
<Jmabsd>
rjek: I think 8 PCIe v3 channels come out of the SoC.
* rjek
nods, in-order and puny cache don't help the CPU
<Jmabsd>
rjek: so as a web browsing computer then, the GPU will be vastly faster than the CPU ha ha
<Jmabsd>
what's a "puny cache"?
<drmpeg>
No, Ubuntu 21.10 is just clocked at 1.2 GHz instead of 1.0 GHz for 21.04.
<Jmabsd>
ahaa, okok
<Jmabsd>
rjek: when you say quarter of the speed to a Pi, do you mean the Pi with the BCM2711?
<Jmabsd>
= Pi 4
<rjek>
Jmabsd: Webkit and Firefox are too slow on it to be worth using :)
<drmpeg>
But most of us have modifed U-boot for 1.4 or 1.5 GHz.
<rjek>
Jmabsd: Yees, it's probably a bit slower than that too
<ddevault>
I wonder (entirely speculatively) if it's a clock issue
<ddevault>
I know the dts have a lot of shit with respect to clocks and pcie, which has gone through some churn
BOKALDO has quit [Quit: Leaving]
pabs3 has joined #riscv
cwebber has quit [Remote host closed the connection]
Ivii has quit [Quit: Leaving]
ivan_ has joined #riscv
jamtorus has joined #riscv
jellydonut has quit [Ping timeout: 252 seconds]
rlittl01 has quit [Quit: -a- Connection Timed Out]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
jjido has joined #riscv
jjido has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
hendursa1 has quit [Quit: hendursa1]
hendursaga has joined #riscv
BOKALDO has joined #riscv
mahmutov has joined #riscv
guerby has quit [Read error: Connection reset by peer]
guerby has joined #riscv
<palmer1>
ddevault: gfx_v8_0_ring_test_ring() certainly seems like the sort of thing that could blow up because we don't enforce PCIe memory ordering
<ddevault>
mind leaving a comment to that effect on the gitlab thread?
freakazoid343 has joined #riscv
<jrtc27>
there's an mb in amdgpu_ring_commit though, that's fence iorw, iorw
<palmer1>
that only orders to the PCIe controller, not through it
<palmer1>
IIUC there are some bits of PCIe that expose orderings we can't encode within the ISA
<jrtc27>
the preceding writes were to main memory
<jrtc27>
I believe
freakazoid12345 has quit [Ping timeout: 268 seconds]
<palmer1>
my guess would be the timeout is the issue: if the timeout starts ticking before the GPU actually sees the command, we're not really ticking the right timeout
<jrtc27>
so just a lone write to a PCIe address, dunno if memory or config, to update the GPU's wptr
<jrtc27>
timeout seems likely
<palmer1>
yep
<palmer1>
IIUC it doesn't really matter if it's memory or config, they both can be configured in this manner
<palmer1>
I'm not really a PCIe person, though, so I'm never entirely sure
<jrtc27>
though it looks like the timeout should be AMDGPU_MAX_USEC_TIMEOUT which is 100ms
<jrtc27>
exceeding that seems very bad...
<palmer1>
ya, that seems excessive
<palmer1>
maybe we've just got a bug in usleep (or whatever it ends up calling)?
<ddevault>
fwiw the observed behavior involves a tens-of-seconds delay between modprobe amdgpu and oops
elastic_1 has quit [Quit: elastic_1]
<palmer1>
ya, but that could just be it kicking off the tests because something in userspace actually turned on the driver
elastic_dog has joined #riscv
<ddevault>
yeah, maybe
<ddevault>
and it happens in fini, which could happen far from this
<palmer1>
a lot of driver probing is deferred in these big drivers that support a bunch of devices, so you might not even get it to talk deeply to the chip before trying to spin up graphics
<palmer1>
isn't the stack trace from fini, not the test?
<ddevault>
no, and it's from v6 also
<ddevault>
but similar arguments apply
<ddevault>
it's test that has the bug, apparently
<ddevault>
but fini where the bug manifests in an oops
<palmer1>
ya, that would be my guess
<palmer1>
it probably triggers something to drop the driver, do all the cleanup, and then fire the issue along
awordnot has joined #riscv
[itchyjunk] has quit [Remote host closed the connection]
<palmer1>
jimwilson: Andrew's yelling at me already so I'm not going to say anything on this thread, but v-0.7 was _definately_ called "stable" and ret-conning that just adds to the confusion
<palmer1>
but I'm suppposed to be staying out of it ;)
jamtorus has quit [Quit: jamtorus]
Narrat has joined #riscv
Jmabsd has joined #riscv
<jrtc27>
"This is a draft of a stable proposal for the vector specification to be used for implementation and evaluation. Once the draft label is removed, version 0.7 is intended to be stable enough to begin developing toolchains, functional simulators, and initial implementations, though will continue to evolve with minor changes and updates."
<jrtc27>
that appears in the v0.7.1 tag
<jrtc27>
then "This is a draft of a stable proposal for the vector specification to be used for implementation and evaluation. Once the draft label is removed, version 0.8 is intended to be stable enough to begin developing toolchains, functional simulators, and initial implementations, though will continue to evolve with minor changes and updates."
<xentrac>
wouldn't you need toolchains, functional simulators, and initial implementations to evaluate the proposal?
Jmabsd has quit [Ping timeout: 240 seconds]
devcpu has quit [Quit: leaving]
devcpu has joined #riscv
<jimwilson>
palmer1, stable doesn't mean what you think it means, there were thousands of different versions of the draft, the "stable" ones were ones that people were supposed to use for experimenting to test the draft, and come up with improvements for the next draft, stable is not OK for a SoC
<palmer1>
OK, well, I'll stay out of it
<palmer1>
sorry
<jimwilson>
but I agree that the terminology was confusing to some
<palmer1>
yep -- I don't actually think I was confused (I was in the room when we decided to call it stable, pointed out that calling it stable would make a mess, and got overridden). IMO the real issue here is the vendor that built an SOC being confused
<palmer1>
but I don't really care that much about those guys
<jimwilson>
if you looked at the history of the V spec, the stable draft was clearly not safe for implementing in a SoC, but if you only looked at the v0.7.1 draft and not the history, it wasn't clearly unsafe
<xentrac>
you'd think reading a few versions of the spec would take less effort than writing and verifying the necessary Verilog
<xentrac>
like, a lot less
<jrtc27>
you'd only do that if you thought there was reason to
<jrtc27>
if you believe it's ready to implement, you're not going to think to do that
<jrtc27>
if you don't believe it's ready to implement then you go have a look
<palmer1>
IMO the bigger issue is that this rule applies to every spec: they're all in flux until they're marked as not in flux, and this was clearly described as being ready to implement
<jrtc27>
so naive vendors aren't going to notice
<palmer1>
that's my real worry here, as from where I'm standing we're in the same spot with all these current specs
<xentrac>
maybe the vendor just didn't care because the alternative was to write their own vector or SIMD spec or to not ship a product
<palmer1>
you guys should go look at the announcments, it was pretty strongly messaged that SiFive was implementing this as well
jellydonut has joined #riscv
<xentrac>
thanks :)
<palmer1>
I have a post back from a year or so ago where I enumerated three or four announcments at foundation events around the 0.7.1 release time, if you're just looking at that public messaging it's not at all clear that this spec was going to be abandoned right after it was released
<palmer1>
I specifically remember using the term "stable but useless" internally to describe what was going on here, to point out how much confusion this would cause
<palmer1>
but anyway, i'm supposed to be staying out of this ;)
<palmer1>
so I guess I'm goilg to clos IRC for abit...
<xentrac>
palmer1: I very much appreciate the insight
ivan_ has quit [Remote host closed the connection]
jellydonut has quit [Quit: jellydonut]
jellydonut has joined #riscv
jimbzy has quit [Ping timeout: 252 seconds]
Sofia has quit [Ping timeout: 276 seconds]
jjido has joined #riscv
Sofia has joined #riscv
jimbzy has joined #riscv
jimbzy has quit [Changing host]
jimbzy has joined #riscv
X-Scale has quit [Ping timeout: 265 seconds]
X-Scale` has joined #riscv
X-Scale` is now known as X-Scale
peepsalot has quit [Read error: Connection reset by peer]
peepsalot has joined #riscv
jjido has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
vagrantc has joined #riscv
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
rlittl01 has quit [Excess Flood]
rlittl01 has joined #riscv
crabbedhaloablut has quit [Ping timeout: 276 seconds]
loki_val has joined #riscv
peepsalot has quit [Read error: Connection reset by peer]