sorear[m] changed the topic of #riscv to: Matrix users: #riscv:libera.chat will be ending operation NET Jul 25; please test #riscv:catircservices.org as a replacement | RISC-V instruction set architecture | https://riscv.org | Logs: https://libera.irclog.whitequark.org/riscv
paulk-bis has quit [*.net *.split]
jmdaemon has quit [*.net *.split]
xypron has quit [*.net *.split]
pabs3 has quit [*.net *.split]
zjason has quit [*.net *.split]
joev1 has quit [*.net *.split]
s1b1 has quit [*.net *.split]
riktw has quit [*.net *.split]
sami has quit [*.net *.split]
olofj has quit [*.net *.split]
edef_ has quit [*.net *.split]
aurel32 has quit [*.net *.split]
zapb_ has quit [*.net *.split]
[Reinhilde] has quit [*.net *.split]
justache has quit [*.net *.split]
balrog has quit [*.net *.split]
gatecat has quit [*.net *.split]
kito-cheng has quit [*.net *.split]
jrjsmrtn has quit [*.net *.split]
sami has joined #riscv
olofj has joined #riscv
paulk-bis has joined #riscv
gatecat has joined #riscv
joev1 has joined #riscv
zapb_ has joined #riscv
kito-cheng has joined #riscv
jrjsmrtn has joined #riscv
justache has joined #riscv
justache has quit [Max SendQ exceeded]
riktw has joined #riscv
clonorizer has quit [Remote host closed the connection]
jmd_ has joined #riscv
balrog has joined #riscv
xypron has joined #riscv
edef has joined #riscv
xypron has quit [Changing host]
xypron has joined #riscv
pabs3 has joined #riscv
pierce has quit [Ping timeout: 245 seconds]
davidlt[m] has quit [Ping timeout: 245 seconds]
aurel32 has joined #riscv
justache has joined #riscv
heat has joined #riscv
clonorizer has joined #riscv
s1b1 has joined #riscv
frkzoid has quit [Ping timeout: 260 seconds]
frkazoid333 has joined #riscv
kaaliakahn has joined #riscv
Reinhilde has joined #riscv
clonorizer has quit [Remote host closed the connection]
frkazoid333 has quit [Ping timeout: 245 seconds]
EchelonX has quit [Quit: Leaving]
EchelonX has joined #riscv
jmd_ has quit [Ping timeout: 250 seconds]
jmd_ has joined #riscv
jacklsw has joined #riscv
MaxGanzII__ has joined #riscv
MaxGanzII_ has quit [Remote host closed the connection]
heat has quit [Ping timeout: 246 seconds]
crabbedhaloablut has quit []
BootLayer has joined #riscv
EchelonX has quit [Quit: Leaving]
vagrantc has quit [Quit: leaving]
stolen has joined #riscv
GenTooMan has quit [Ping timeout: 260 seconds]
handsome_feng has joined #riscv
GenTooMan has joined #riscv
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
JanC has quit [Ping timeout: 240 seconds]
JanC has joined #riscv
BootLayer has quit [Quit: Leaving]
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
clonorizer has joined #riscv
clonorizer has quit [Remote host closed the connection]
<unlord>
mps: is --accel tcg,thread=multi necessary? The documentation says it will be enabled by default if the emulated platform supports it
<mps>
unlord: I didn't noticed big difference with or without it, but I didn't tested riscv in qemu for more than a year
<unlord>
mps: I will know in a few minutes if it helps my workloads
<unlord>
hmm, hard to say
<unlord>
maybe it is faster?
<mps>
unlord: I experimented with thread=multi but at the end removed it and now can't remember why
<courmisch>
I know I sound like a broken record but don't use system QEMU if you want to optimise for perf
<mps>
and just noticed that my guide is now outdated and wrong, have to find some time to test all these again and update
<courmisch>
soft MMU is never going to be fast
<mps>
courmisch: yes, I use qemu-user for building riscv things
<mps>
but it is also not so fast even on apple silicon m1pro
<courmisch>
I think an adhoc RV to A64 recompiler could be plenty fast, much faster than TCG
clonorizer has joined #riscv
<courmisch>
but nobody has written it
clonorizer has quit [Excess Flood]
<courmisch>
I'm stuck with x86 so no motivation :)
hightower2 has joined #riscv
<mps>
not sure if riscv binaries could run in FEX emulator
<mps>
or it is only for x86
<courmisch>
I thought that only ate x86
<courmisch>
I was more interested in implementing the opposite direction but A64 is so much more complicated than RV
<courmisch>
and RV has no place to store the NCVZ and SP bits
<Tenkawa>
mps: do you have a test scenario string with qemu I can run on this M1 MBP? I want to compare it vs my Paralles VM (which to be honest flies)
billchenchina has joined #riscv
<Tenkawa>
er Parallels
<mps>
Tenkawa: only this url I posted above, but it is outdated (as I wrote)
<mps>
I think I will try to update it on weekend and post mail to starfive ML
<mps>
uhm, no starfive ML but to riscv foundation front man (Jeff)
<Tenkawa>
It should be enough to give me a test template
<courmisch>
not sure aboit gentoo. Debian binfmt just worked for me IIRC
<mps>
adapt it to your working dirs
<Tenkawa>
you can chroot without all the binds usually just not do next to anything...(In every scenario I've used)
<Tenkawa>
you can open a shell... just don't try to run apps...
<courmisch>
or you ask your boss to buy you a MilkV
<mps>
Tenkawa: yes, binfmt must be 'activated'
<mps>
courmisch: is MilkV available on market
<courmisch>
I dunno, I didn't follow up since it's clearly not my hobbyist self
<Tenkawa>
mps: I don't think so yet... sadly
<Tenkawa>
The RPI4 needs a competitor heheh
<courmisch>
LPi4a ? but no NVMe unlike VF2
* mps
never bought any RPi and using arm for 20 years
<unlord>
Tenkawa: right, I would have expected the chroot to work even if I hadn't mounted everything through
<Tenkawa>
unlord: without the binfmt enabled kernel side on the host it won't
<courmisch>
RPi's success is to a significant extent about software support. it's so much easier to work with than most comparable embedded SBCs
<unlord>
Tenkawa: binfmt is enabled kernel side on the host
<unlord>
I am able to run a riscv64 binary on the host
<unlord>
ahh, I had to copy the static built qemu-riscv64 into /mnt/gentoo/usr/bin
<unlord>
I'm in
clonorizer has joined #riscv
<unlord>
wow, this is way faster for building
<Tenkawa>
heheh
<Tenkawa>
this is what's funny on my box.. I have both 64 and 32 bit env configured... not that I'd ever use 32 bit
<courmisch>
ofc,hardware MMU and host kernel...
<mps>
I have riscv64 lxc for alpine development
<Tenkawa>
nice
<mps>
so I can ssh to it over net
<Tenkawa>
mps: I was quite happy to see how well this M1 works with ARM64/RV both since that's all I work on... I don't do anything x86 related anymore.
<mps>
for short 'living' tasks I spin chroot but for long needed ENVs I use lxc
<mps>
Tenkawa: I stopped to use x86 about 12-13 years ago for anything personal
<courmisch>
I wonder what horrible hack user QEMU uses on fork()
<Tenkawa>
Same... about 2010
<courmisch>
maybe I don't want to know
<unlord>
courmisch: this is great
<mps>
Tenkawa: hah, also at this time
<courmisch>
unlord: now you can take over from me to port all OSS multimedia to RV
<unlord>
courmisch: perfect
<unlord>
as long as I get to do it in gentoo
handsome_feng has quit [Quit: Connection closed for inactivity]
<courmisch>
there goes my job safet^W^Wahem business plan
<unlord>
I am kind of confused how this chroot works
<unlord>
/proc/cpuinfo shows the host CPU
<courmisch>
well yeah that's one or the limitation
<unlord>
so programs that parse this to try and figure out what extensions RISCV has will clearly not work
<courmisch>
it's really more of a build env
<conchuod>
They should stop parsing that garbage anyway :)
<courmisch>
dbus and display server probably don't work
<unlord>
yeah, so building llvm did not work
<courmisch>
parsing cpuinfo is so x86
<unlord>
weird, running it again worked
<mps>
`setarch` maybe could help
vagrantc has joined #riscv
heat has joined #riscv
<courmisch>
sudo - somebody implement riscv_hwprobe in QEMU user
awita has quit [Ping timeout: 258 seconds]
<courmisch>
well looks like it's already been done
<courmisch>
but it doesn't have Zb and V yet :/
MarvelousWololo has joined #riscv
meta-coder has quit [Quit: leaving]
ldevulder has quit [Quit: Leaving]
jacklsw has joined #riscv
jacklsw has quit [Client Quit]
<palmer>
courmisch: ya, we just got the first round. IIRC it was before the other stuff had been released, but I forget what's shown up. Robin is on vacation for like a month, so it'd be a bit before it gets added
<courmisch>
those people, not only paid to work but taking vacations too
<unlord>
palmer: what about getauxval() for riscv?
<courmisch>
it's a one line patch, as I see
<courmisch>
but it's not been done
<courmisch>
need to unmask 'V' in linux-user/elfload.c
stolen has joined #riscv
<palmer>
ya, I'm not surprised there's some uABI missing from QEMU -- that's kind of always the case
<palmer>
I don't know of anyone working on it, should be a simple patch ;)
<unlord>
is it simple enough that someone who has never contributed to QEMU can write it?
<palmer>
sure, just go for it
<palmer>
you can pretty much send patches for anything, it might just take a bit longer if it's your first time
<unlord>
OK, I'm very interested to see getauxval() fixed
<courmisch>
last time I submitted code to QEMU somebody dared to find a bug in my patchset