crabbedhaloablut has quit [Ping timeout: 276 seconds]
winterflaw has quit [Ping timeout: 276 seconds]
crabbedhaloablut has joined #riscv
winterflaw has joined #riscv
hendursaga has joined #riscv
nvmd has quit [Quit: Later, nerds.]
<jrtc27>
dibi58: define knows
<dibi58>
status, where to buy if available
<jrtc27>
it exists, upstream kernel support isn't happening any time soon due to its lack of coherent DMA, don't know now the indiegogo campaign is closed
<palmer1>
we're trying to work through how to deal with the D1 upstream, they've recently published some docs which was the big blocker but I haven't had time to look at them
<jrtc27>
would be interesting to see what you end up doing..
<jrtc27>
all I've seen so far is an incoherent flip-flopping stream of patches that don't learn from feedback
<palmer1>
the implementations have all been a mess, but the rough plan is to implement the standard non-coherent stuff and then describe whatever the D1 is doing as errata against that
<palmer1>
then there'll at least be some sanity to the process
<jrtc27>
ok, that kinda makes sense if it can be made to fit
<jrtc27>
do their cmo's at least look like normal-ish ones?
* jrtc27
won't be racing to implement this in FreeBSD though...
BOKALDO has quit [Quit: Leaving]
<palmer1>
doesn't really matter that much, the flush interface is really simple
<palmer1>
it's more about sorting out what semantics the arch/riscv primitives have, and then using a big hammer to make the D1 stuff fit that
<jrtc27>
ack
<palmer1>
if we do it the other way around we'll start to have a bunch of code that depends on the D1 way of doing things, and it'll be a headache to untangle that for the future stuff
<jrtc27>
yeah no that's not a good idea and also sets a bad precedent
<jrtc27>
similarly with the vector mess...
<palmer1>
that's a whole different problem -- at least for the V stuff they were sort of trying to play by the rules
<jrtc27>
the similarly was referring to precedent
<jrtc27>
and ehhh not really
<jrtc27>
it was a draft standard, but it was clearly something that you shouldn't implement
<jrtc27>
anyway
<palmer1>
there was a lot put out by the RISC-V folks saying that it should be implemented
<jrtc27>
hm, guess there's some backstory I wasn't aware of
<palmer1>
likely ;)
<jrtc27>
probably for the best
<jrtc27>
I only get involved with internal politics stuff when I feel it's impeding technical progress..
Andre_H has joined #riscv
smartin has quit [Quit: smartin]
<jimwilson>
nezha is available from aliexpress.com
Amanieu has quit [Ping timeout: 272 seconds]
<jimwilson>
drafts were supposed to be implemented to get experience to improve the next iteration, but they weren't supposed to be implemented in SoCs, they were only supposed to be experimental designs to simulate or put on FPGAs
<dh`>
if implementors don't look at drafts, they don't progress beyond drafts :-)