<davidlt[m]>
Oh, I need to reconfigure Kojira to avoid dist repo being garbage collected
<davidlt[m]>
Asahi Lina live stream in the background is always nice.
davidlt has quit [Ping timeout: 260 seconds]
zsun has joined #fedora-riscv
davidlt has joined #fedora-riscv
<neil>
davidlt[m]: that IPC Classes design looks really interesting
Kevinsadminaccou has joined #fedora-riscv
<davidlt[m]>
Oh, there is also something interesting mentioned on the web. JH7120.
<neil>
🤔
* neil
is excited
<davidlt[m]>
No details were shared on the patch :)
<neil>
i just want my JH7110 to come in time for xmas
<davidlt[m]>
I would prefer to see majority of patches on the mailing list before the board lands in the hands :)
<neil>
okay.. fair :)
<neil>
I _suppose_ a usable board is better than a brick ;)
<davidlt[m]>
I am truly surprised how upstream is going on.
<davidlt[m]>
A lot have been posted for a review already (minus PCIe of course).
<davidlt[m]>
If they get their PCIe driver in U-Boot that will be gold.
<davidlt[m]>
Another interesting thing, some cores are Sv39, some Sv48.
<neil>
Yes that would be very ideal
<davidlt[m]>
There is some discussions on OpenSBI mailing-list how to handle this. They wanted to change boot hart, as that defined which MMU mode is picked.
* neil
finds the riscv isa manual to refresh on these terms :)
<davidlt[m]>
It's how many bits are used in the address.
<davidlt[m]>
39-bit, 48-bit, 57-bit.
<neil>
yeah, interesting. i've not looked at this low-level stuff since university heh
<davidlt[m]>
57-bit VA has it's own problem as I mentioned few days ago. That will break things like OpenJDK, Golang, etc.
<davidlt[m]>
and QEMU will default to the max supported (which just happened to be Sv57).
<neil>
right... so basically fun* upstream work to try and get them compatible with 57-bit VA?
<davidlt[m]>
RISC-V doesn't have internal bits wired like arm64 or x86_64 where mmap + hints allow userspace to request memory from new higher space.
<palmer>
we should probably add them, though
<davidlt[m]>
I also a bit excited about Zbb optimized string functions in the kernel :)
<davidlt[m]>
I wonder if that could be applied to JH7110.
<davidlt[m]>
palmer: from what I understand this is how arm64 and x86_64 works (didn't look at the code, just documentation and LWN articles).
<palmer>
ya, that's all I did too
zsun has quit [Quit: Leaving.]
<palmer>
do you have a jh7110 yet? I've gotten some Kickstarter messages, but no HW
<davidlt[m]>
I also quickly looked at Golang and something else. It's all random guesses (let's call it magic) to calculate the hint properly.
<neil>
palmer: same position here. messages but no HW as of yet
<davidlt[m]>
palmer: yes, I got dev sample from PINE64, but didn't play too much (it comes with no software).
<davidlt[m]>
So it's a bit expensive time wise to do a lot of guess work.
<davidlt[m]>
Icenowy has a tree, but didn't want (couldn't?) share a disk image to play around.
<davidlt[m]>
It's being produced for some time now. I think there should be an update on KS with some pictures.