<joshua_>
[16:50:01] <whitequark[cis]> you know things are going well when your SCK looks like this <-- if you keep this up I am going to mail Esden some littleprobes to send you next time he sends you a care package with hyperrams or something. please, my people are dying
<whitequark[cis]>
I might in fact keep this up
<whitequark[cis]>
I think at this point the Glasgow IO performance is in fact limited by the ICE40 clock-to-out (about 3 ns (!!), though it depends how you count) and the level shifter in-to-out (about 3ns either way, iirc) latency
<whitequark[cis]>
by the time you reach 100 MHz you essentially exhaust your timing budget, and this matches what I'm seeing
<whitequark[cis]>
things sort of function at even 90 MHz, but I'm not really capturing any amount of correct data. the absolute highest I went with 0/4M BER is about 75 MHz
<whitequark[cis]>
and that required careful manual tuning of DDR buffer input clock through a delay line made out of LUTs
<whitequark[cis]>
with a something like 500 ps resolution and a window of only a bit bigger
<joshua_>
can you have iCE40's PLL give you a 90 degree phase shifted version?
<whitequark[cis]>
that you can
<joshua_>
re: Glasgow to pmod: it seems fine to do except the one problem with doing it well is that I cannot find obvious keyed 2x10 2.54mm THT connectors on LCSC
<joshua_>
hm I guess you could enforce correct insertion by having the pmod side stick out off the bottom of the board rather than the top, so that it mechanically interferes if you insert the adapter the wrong way around
<whitequark[cis]>
yea I mean I can live with using the silk to align
<whitequark[cis]>
I want something to play with a flash that wouldn't have the inductance of a transcontinental hvdc line
<whitequark[cis]>
* something to use to play with, * with a QSPI flash that
<joshua_>
this includes LCSC part number annotations for jlcpcb assy, but realistically you do not want them to do assy because this counts as a double sided assy and so they would charge you lots for it
<joshua_>
pop whatever resistors you want, 0 ohm, 50 ohm, whatever
notgull has joined #glasgow
notgull has quit [Ping timeout: 268 seconds]
<whitequark[cis]>
hm, i upgraded libc
<Darius>
can you run it in a container?
<whitequark[cis]>
i don't run containers
<whitequark[cis]>
oh, aptitude failed and now i'll have to spend 40 more minutes changing versions
<whitequark[cis]>
i give up lol
<whitequark[cis]>
i have decided thatnobody needs kicad 8
<whitequark[cis]>
* i have decided that nobody needs kicad 8
<Darius>
let the suffering commence
<whitequark[cis]>
bleh
<Darius>
just run it under wine </s>
<whitequark[cis]>
the stable linux abi is win32, yeah
<joshua_>
containers are the static linking of the future!
<Darius>
I used to think stuff like static linking was kinda dumb but for "mega apps" like KiCAD or browsers it's not so stupid
<joshua_>
gosh, I remember the brief craze that swept the world of 'bare metal' hypervisor guests
<joshua_>
at some point someone was like "wow, you know, it would be nice to provide at least a little bit of services to these. like maybe we could give the HV guests a little bit of persistent storage and manage it for them. and maybe they should be able to interact with other HV guests. we would need some kind of way to call these mechanisms from the system, though, some sort of 'system call'"
<whitequark[cis]>
i still think those are a pretty good idea
<whitequark[cis]>
i mean, even microsoft not only got on the "library OS" train but have implemented it as well
<whitequark[cis]>
fifty syscalls are a much more maintainable and defendable boundary than five hundred syscalls or five thousand library entry points
<whitequark[cis]>
you do have to make sure that they're like, actually good
<joshua_>
Plan 9 Developers would like to access your location.
<whitequark[cis]>
i don't like plan 9 but it does have some good ideas
<whitequark[cis]>
although, i should clarify: it's perfectly fine research
<whitequark[cis]>
i think in this particular case the problem isn't so much static or dynamic linking but like
<whitequark[cis]>
apt is just bad at dependency resolution
<joshua_>
and having particularly old systems with mixed package sources is how things go to hell, yes
<joshua_>
anyway, above are gerbers, I assume you can just hand them to a fab if you don't want anybody to place parts for you, because they are really not pushing the limits of tehcnology on anything
<whitequark[cis]>
yea i just shoved them into pcbway
<whitequark[cis]>
hm, what was the size of the PCB?
<whitequark[cis]>
i was tired so i just typed in 50x50 mm
<joshua_>
46 x 35 apparently
<whitequark[cis]>
close enough. thanks!
<joshua_>
差不多
<miek>
fwiw, i think the numbers on the pmod side are reversed? (at least how i'd read them)
<joshua_>
yeah, I was thinking in terms of mapping them downwards on the connector, but also yes they are reversed in reasonable readings