Lofty changed the topic of #prjmistral to: Project Mistral: Yosys (and hopefully nextpnr) on Cyclone FPGAs - https://github.com/ZirconiumX/mistral - logs: https://freenode.irclog.whitequark.org/prjmistral
<Sarayan> chipb: you're no spokesman but you make sense
<Lofty> Ping Sarayan
<Sarayan> pong?
<Lofty> Sarayan: you probably saw the bunch of issues filed by gatecat; seems we've got some DB work to do :P
<Sarayan> I actually didn't :-)
<Lofty> Pushed some updates to the schematic, so you might have to kick off another doc rebuild
<Lofty> But anyway, Sarayan
<Lofty> We got attosoc to compile and run on real hardware
<Lofty> So things like the routing graph seem okay
<Sarayan> excellent
<Sarayan> do you have linux?
<Sarayan> and python-sphinx installed (it tends to be)
<Sarayan> if yes, go in the docs subdirectory and run builddoc.sh
<Lofty> Sarayan: I do, but I don't have latexmk (and can't find it in the Arch repos)
<JJJollyjim> 70% sure it’s included in texlive
<JJJollyjim> yeah it is
<JJJollyjim> (texlive-core specifically, but you probably want at least texlive-most if you have disk space)
<Lofty> Okay, so now I need fncychap.sty
<JJJollyjim> yeah that’s in texlive-latexextra
<JJJollyjim> (which is included in -most)
<Sarayan> lemme check, I'm on arch too
<Sarayan> and yeah, I confirm it's in texlive-latexextra
<Sarayan> (It's better if you can generate the docs too :-)
<Lofty> > Total Installed Size: 2027.16 MiB
<JJJollyjim> it might be possible to use https://tectonic-typesetting.github.io/en-US/ which downloads things on-demand
<JJJollyjim> not sure how it interacts with latexmk though
<Lofty> Okay, built and pushed
<Sarayan> cool
<Sarayan> so.... that means you can FF the same lut output on two different clocks/enables
<Sarayan> that's interesting
<Sarayan> makes having two ffs per lut possibly more interesting
<Lofty> Sarayan: I mean, I suspect that the only time it gets used in practice is if the same LUT is used with two different enables
<Sarayan> Lofty: Yeah, that's pretty much the only way that makes sense in practice, even if I suspect there potentially other funky cases, like clocks with the same frequency but different phases
<Sarayan> or the edges of the clock
<Sarayan> niiiiice
<Sarayan> still lot of work, but really nice
<gatecat> Sarayan: in the CMUXHG diagram (for now I'm only considering the general routing inputs, a tiny bit of general routing in the clock path isn't generally a big deal to start with), do you know why CLKIN feeds both 27 and 33?
<Sarayan> I'm not entirely convinced it makes any real sense no
<Sarayan> you could use the direct clock pin -> cmux path
<gatecat> yeah, they need to be added but I started with clocks from fabric because that's most general
<Sarayan> gatecat: I kinda suspect that between the clock pin and pll counters direct links they're not that used in practice
<gatecat> they would be useful if we want to divide a clock in fabric, which we might need for testing before plls are working
<gatecat> but yeah, a good design and a correct PCB layout* shouldn't need them at all
<Sarayan> Sure, but I mean eventually in real designs :-)
<Sarayan> yeah
* gatecat has never messed up a pcb layout and not used a correct clock pin, definitely never :p
<Sarayan> ne-ver
<Sarayan> I suspect wherever the clock is coming from, you want to plonk it in a pll anyway
<gatecat> mmm
<gatecat> I presume there is some dedicated IO to PLL routing, too ?
<Sarayan> clock pin yes, otherwise it's the usual DCMUX
<gatecat> right, that makes sense
<Sarayan> there's pll-pll direct links too, to cascade them
<gatecat> nice
<gatecat> are these in mistral already?
<Sarayan> Not yet
<Sarayan> I've started studying the plls, they're interesting
<gatecat> fun
<Sarayan> 9 counter ouputs + m, bunch of direct links all over the place
<Sarayan> they really worked their clocks
<Sarayan> of course, given how important they are, they better
<gatecat> indeed
<gatecat> are there a bunch of weird analog parameters for the PLL, too?
<Sarayan> I don't think so
<gatecat> good, that makes things easier
<Sarayan> for sure
<gatecat> lattice have them, but ecppll kinda fudges them and probably isn't giving optimum loop bandwidth in all cases