<davidlt[m]>
So technically we have the 1st core IP compatible with RISC-V Profiles (which is not ratified), but typically this still means 1-3 years away from some silicon (if someone makes a public one).
<davidlt[m]>
RVA23 is the next major one to replace RVA20 (aka RV64GC stuff we use today).
<davidlt[m]>
RVA22 is just a stopgap solution (?).
<davidlt[m]>
RVA23 will require vector, vector crypto, and support for higher page sizes than 4K IIRC.
<davidlt[m]>
Kinda brings the last few bits.
<davidlt[m]>
Thus I would expect a minor psABI update soonish too. RVA23 should get defined early 2023 as there is an interest to rush this one out.
<davidlt[m]>
If you look at SiFive slides P670 looks like a minor update for P650, not a major one.
<davidlt[m]>
By the time we get Horse Creek with P550 it's an ancient thing :)
esv_ has quit [Quit: Leaving]
jcajka has quit [Quit: Leaving]
esv_ has joined #fedora-riscv
esv has quit [Ping timeout: 252 seconds]
somlo has quit [Remote host closed the connection]
<davidlt[m]>
rwmjones: check nujive
<davidlt[m]>
ping seems to be off by two hours, shows load, but no jobs
<nirik>
davidlt[m]: asked some questions about the aws instance in that discuss thread if you have some time to look sometime. :)
<davidlt[m]>
nirik: checking...
<davidlt[m]>
It's probably too big too, but who know as it's a VM with CPU over-commiting.
<davidlt[m]>
If you do a full setup and management, you pick what needs :) If it's me and we basically just moving existing configuration/setup in that case it should be good enough (something smaller probably is fine).
<davidlt[m]>
Data wise, today I am moving 200-300G to the board hosted at my place, and eat probably 30G a day of /mnt/koji.
<davidlt[m]>
This will probably increase as I still don't run boards 24/7 and still have another 3 to connect :)
<nirik>
I'll try and spin something up in the next few days. Oh, can you point me to your ssh pub key?
<davidlt[m]>
Not to mention new boards showing up on the market ;)
<nirik>
yeah
<davidlt[m]>
I can generate a new one just for this, and send you over the email.
<davidlt[m]>
Somehow I am in the mood to generate new SSH keys for different things in the last couple of years :)
<nirik>
sure, sounds fine
<davidlt[m]>
nirik: so if you want to manage it and do things in a proper Fedora way, you pick all the things and we just import repositories into a new thing :)
<davidlt[m]>
Otherwise we just re-create the existing minimal setup for now. Koji over DB, /mnt/koji, etc. and simply restart.
<davidlt[m]>
So far I mostly focused (and almost completely ignored everything else) on catching up to F37, which is now almost done.
<davidlt[m]>
Over 80+% of all packages seem to be in. Almost approaching 5000 successful builds for 30 days, and delta with ABI/SOVERSION bumps between F38 is not that big so far.
<nirik>
yeah, I was thinking just minimal/ad hoc setup... putting everyting in CM and such can wait until minlining I think.
<nirik>
so, on to f38? ;)
<davidlt[m]>
Not yet, but probably this month.
<nirik>
oh, and BTW, should builders update? or not until we land f37 ?
<davidlt[m]>
There are still a few K of F37 packages in pipeline. Some are still important.
<davidlt[m]>
So the ideal would be to have everything on F37.
<davidlt[m]>
I will start toying and testing F37 disk images most likely next week.
<davidlt[m]>
I still need to get some packages, and rebase a few thing (e.g. kernel-ark to 6.0.X)
<nirik>
yeah, my builder says it's f33. ;)
<davidlt[m]>
Well technically it's not exactly F33 ;)
<davidlt[m]>
It's mostly F33 with some parts being from Rawhide at a time.
* nirik
nods.
<nirik>
anyhow, I'll try and get something spun up. You may want to write up a 'here's the current state' on discussion sometime soon...
<davidlt[m]>
Ah yes, we didn't have a nice update post.
somlo has joined #fedora-riscv
<davidlt[m]>
nirik: should we continue to use my fedora.riscv.rocks, or should that move under some fedora domain?
<nirik>
so, is the riscv summit likely to be worth attending? I've not looked at the other talks, but I am interested in the new board. ;)
<davidlt[m]>
All of the things will be available on YouTube :) So not sure :)
<nirik>
yeah, I was pondering on that... probibly best to just keep using the current name. If we use fedoraproject.org people might start expecting auth to work with kerberos, etc...
<davidlt[m]>
I have Star64, but still didn't attempt to build a custom disk image for it.
<davidlt[m]>
Well at some point that would be a nice thing to have. I would like everything to be able to login without me creating new accounts based on certificates.
<davidlt[m]>
But that's maybe sometime later as I am only running boards like 16-18 hours a day, which is not useful for all the time zones.
<nirik>
in order to use kerberos we need all the machines on the same network for ipa... so yeah, I think thats more for when we roll into primary koji or perhaps just before that if we want to have a step with a internal secondary koji.
<davidlt[m]>
I still think we are bit far away from primary koji. I mean we had these discussion for years, and I don't think the right hardware will be available for RH to host at their colo.
<davidlt[m]>
Not to mention that other interesting topics like standards, and standards based hardware :)
<nirik>
yeah, it's down the road for sure. But we will get there I am sure of it. ;)
<davidlt[m]>
The most important is the binaries, that follow RISC-V Profiles and psABI and few other things.
<davidlt[m]>
The rest is just how you assemble various components to meet the standard requirements.
<davidlt[m]>
So I am not really worried about it as long as those RPMs and binaries are cooked properly.
<davidlt[m]>
The goal would be to do more and more with each Fedora release to get closer to the way official stuff is done.
<davidlt[m]>
Small thing, but I am considering avoiding importing noarch packages and attempting to build those a few times before finally importing.
<davidlt[m]>
Maybe playing with pungi. Or neil might do it ;)