<mwk>
I'll look into it in a month or two (once I'm done with my current tasks)
<Cerpin>
Awesome!
<mwk>
this is a hard issue unfortunetely, as it ties into a long-overdue full redesign of yosys blockram inference
<mwk>
but, this is finally on the roadmap
<Cerpin>
Ahh, right
<Sarayan>
what is an integrated logic analyzer actually? I mean at a concrete level, not a high-level idea?
<mwk>
Sarayan: in its simplest variant, a big RAM where you write values of all your monitored signals every cycle starting from your trigger cycle, until you run out of it
<mwk>
with a secondary port connected to some interface that allows you to read it out at your leisure (eg. JTAG)
<Sarayan>
so that's something hardware only, or it has a software sim side too?
<mwk>
well there's not much reason to do such a thing in sim
<mwk>
given that you already see all your signals in simulation anyway
<Sarayan>
I see
<mwk>
it's something that you use when your design works in simulation, but craps itself on real hardware, and you have no idea why
<Sarayan>
having that in a generated fpga firmware is very heisenberg though, right? I probably massively changes the behaviour in the first place?
<Sarayan>
It
<mwk>
it could
<Sarayan>
especially timings
<mwk>
but FPGAs are mostly deterministic
<Sarayan>
which, from what I understand, is the #1 case of "it's not running in hardware"
<mwk>
unless CDCs are involved
<Sarayan>
ah yeah
<mwk>
or unless you don't pass timing, but that is an obvious issue you should fix anyway
<mwk>
but... there's a different problem
<mwk>
inserting an ILA could very well cause you to no longer pass timing
<mwk>
so yeah, tricky
<Cerpin>
This is a fun discussion considering I work with someone who uses Xilinx's ILA as a debugging tool of first resort :\
<mwk>
also — another case when you need an ILA is when you're interfacing to some other hardware which you cannot easily simulate
<mwk>
like, no amount of simulation will help you if it turns out that this external chip you talk to behaves differently than you thought
<Sarayan>
yeah indeed
<mwk>
... yes, ILAs can be used as a first-line debugging tool
<mwk>
which is a supremely bad idea
<mwk>
but that's never stopped anyone, now has it
<Sarayan>
so I guess the question I should ask myself is how to make (nmigen+)yosys a nicer environment for debugging before hitting hardware
<Sarayan>
because I suspect it could be much better
<Sarayan>
but that requires smart UI work
<Sarayan>
dunno if I'm smart enough :-)
<Cerpin>
Supremely bad indeed
fsasm has joined #yosys
emeb has joined #yosys
<ZirconiumX>
Why does nextpnr output solely to stderr?
<emily>
all it outputs is diagnostics, right?
<ZirconiumX>
If you consider "the program output" to be diagnostics, sure
<emily>
stdout would be for if it was, like, dumping a bitstream on stdout