pepijndevos has quit [Ping timeout: 240 seconds]
pepijndevos has joined #prjmistral
_whitelogger has joined #prjmistral
peepsalot has quit [*.net *.split]
ChanServ has quit [*.net *.split]
gruetzkopf has quit [*.net *.split]
Sarayan has quit [*.net *.split]
whitequark has quit [*.net *.split]
vup has quit [*.net *.split]
pepijndevos has quit [*.net *.split]
phire has quit [*.net *.split]
duck2 has quit [*.net *.split]
chipb has quit [*.net *.split]
anuejn has quit [*.net *.split]
bwidawsk has quit [*.net *.split]
Lofty has quit [*.net *.split]
_whitelogger has joined #prjmistral
emily has joined #prjmistral
peepsalot has quit [Quit: Connection reset by peep]
jevinskie[m] has joined #prjmistral
peepsalot has joined #prjmistral
_whitelogger has joined #prjmistral
fdalleau` has joined #prjmistral
peepsalot has quit [Quit: Connection reset by peep]
peepsalot has joined #prjmistral
fdalleau`` has joined #prjmistral
fdalleau` has quit [Ping timeout: 272 seconds]
fdalleau has joined #prjmistral
peepsalot has quit [*.net *.split]
peepsalot has joined #prjmistral
<
Sarayan>
Et Lofty, are you bored?
<
Lofty>
Very much so
<
Lofty>
What can I do for you, Sarayan
<
Sarayan>
Starting line 104 you have all the muxes that control the dsp behaviour
<
Sarayan>
format is mux_name:bits [list of bit positions]
<
Sarayan>
value name-of-the-selection
<
Lofty>
Holy shit, wow
<
Lofty>
That's...quite a bit
<
Sarayan>
the amusing question is understanding what the muxes actually do
<
Sarayan>
there's a bunch on labs and m10ks too :-)
<
Sarayan>
I just finished transcribing that one is all
<
Sarayan>
I remember you were wondering about of the dsp worked, well, it's more information to think about it :-)
<
Sarayan>
we'll have to document it at some point, at least enough to make synthesizing for it viable
<
Lofty>
There are...quite a few inverters here, actually
<
Sarayan>
yeah, they have inverters on the clocks usually, neg_edge here I come or something
<
Lofty>
Well, if a rumour holds true, the PIPs in Intel chips are inverting
<
Lofty>
So that might be a hardware hack to deal with reaching somewhere with an inverted signal
<
Sarayan>
Probably going to write a probe which takes a rbf and a bel coordinate and give you the state of the muxes
<
Sarayan>
I don't really want to put the intel names in public code though
<
Sarayan>
I know the LUTs are stored inverted
<
Lofty>
That...seems to be quite common
<
Sarayan>
also, lots of connections are two bits, so double inverter?
<
Sarayan>
an interesting detail, you may have noticed that in the chip planner there are no bels starting at (45, 37) all the way to the right
<
Sarayan>
ever cuts in the middle of columns 45 to 50
<
Sarayan>
well, I don't know if the bels are there or not, but the associated routing is
<
Sarayan>
they're probably not working though, used the space for the arm border or whatev'
fdalleau`` has quit [Ping timeout: 246 seconds]
fdalleau has quit [Ping timeout: 272 seconds]
mithro has quit [Ping timeout: 256 seconds]
mithro has joined #prjmistral
goldfinger has joined #prjmistral
goldfinger has quit [Ping timeout: 245 seconds]
<
Sarayan>
Lofty, still around?
<
Sarayan>
Needs posix/linux/*bsd, hopes it compiles for you
<
Sarayan>
compiles with make into a fmaker program
<
Sarayan>
fmaker without parameter gives you the help
<
Sarayan>
model gives you the list of supported models (only one for now)
<
Sarayan>
bel gives you the list of bels and they tile positions for a given model
<
Sarayan>
tile model x y gives you the mux status for the bel in the tile (except for m10k, but lab, mlab and dsp is in there)
<
Sarayan>
for lab.mlab is only the global state
<
Sarayan>
tile model x y i gives you the state for labcell number i
<
whitequark>
don't use -O9 folks
<
Sarayan>
(bad) habit
<
Lofty>
If you really want performance, please use -flto, Sarayan
<
Sarayan>
I know, I should use -O3
<
Sarayan>
Took this habit when gcc added -O2 then -O3, hard to get rid of it
<
whitequark>
they should just make it a hard error imo
<
whitequark>
and
*especially* things like -O10
<
Sarayan>
also, there's a little too much stuff coming directly from quartus in there (mux names, etc) so keep it low profile for now, heh?
<
Sarayan>
anyway, you can play with that and the chip planner to try and understand all the muxes
<
Sarayan>
I'll add m10k next week
<
Lofty>
It always amuses me that the first chip we picked was the one in the MiSTer, even though it's got an ARM SoC on there
<
Sarayan>
I pixked it
*because* it's in the mister
<
Lofty>
I think whitequark's chip should be next - can't remember what it is though
<
Sarayan>
wq's isn't open source already?
<
Lofty>
It's a different SKU
<
Sarayan>
it's a Cyclone V?
<
Lofty>
5CEBA4F23C7
<
Lofty>
She needed a dev board for nMigen's Intel backend, I think?
<
Sarayan>
Ah, A4, different die
<
Lofty>
5CEBA4, not 5CSEBA6
<
Sarayan>
EB and A4, very different die :-)
<
Sarayan>
Well, I'm building it to make it usable across the cv models
<
Sarayan>
prolly won't try outside of cv though
<
Lofty>
I dunno, it's worth experimenting a bit
<
Sarayan>
well, have fun with that toy, I'll be gone for the next 3 days
<
Sarayan>
oh, thre's a fuckup or two in the unframing of rbf files, the righmost tiles are broken somehow
<
Sarayan>
I'll fix that whenever
peepsalot has quit [Remote host closed the connection]
<
Sarayan>
did it compile for you btw?
peepsalot has joined #prjmistral
<
Lofty>
It did, yes
<
Sarayan>
have fun then :-)
_whitelogger has joined #prjmistral
<
Sarayan>
ok, wq's die is the e50f
<
Lofty>
Sounds about right
<
Sarayan>
and the model e50b
<
Lofty>
At least we know how the die naming system works
<
Sarayan>
I just run a synthesis and see with files are dumped :-)
<
Sarayan>
ok, I've added the m10k after all, so I updated fmaker.zip
<
Sarayan>
and it's bedtime, and biking for three days, so see you much later
peepsalot has quit [Quit: Connection reset by peep]
peepsalot has joined #prjmistral