pacak has quit [Read error: Connection reset by peer]
pacak has joined #yosys
kraiskil has joined #yosys
unkraut has quit [Remote host closed the connection]
unkraut has joined #yosys
<thardin>
is it possible to prevent yosys from replacing memory with registers?
<tnt>
(* nomem2reg *) IIRC
strubi has joined #yosys
<thardin>
read_verilog -nomem2reg works
<thardin>
I think replacing the state machine with a ROM based one is better use for the BRAM for now
<daveshah>
Be careful with nomem2reg, it can cause missynthesis
<thardin>
yep, saw that
<thardin>
sweet, using a state mechine LUT brings LC use down by 6%
<thardin>
perhaps I can do something similar with division
<strubi>
hi there, just tuning in...is anyone aware of the current situation on the pyosys API, is it used much at all? I think I recall someone mentioning integration with nmigen, but not sure if that was using pyosys
<mwk>
that's not pyosys
<strubi>
Allright, thought they were connected (from whitequark's comments on github WRT python API)
m4ssi has joined #yosys
<lambda>
can someone explain to me why this two-stage evaluation of DFF inputs is necessary in the btor backend? I don't see why it wouldn't be possible to just recurse on the D port's signal immediately when exporting the cell https://github.com/YosysHQ/yosys/blob/master/backends/btor/btor.cc#L70
FFY00 has quit [Quit: dd if=/dev/urandom of=/dev/sda]
FFY00 has joined #yosys
<lambda>
huh, `test_cell` has been broken for $[s]sh(l|r) since december 2019 (it generates B_SIGNED=1 with 50% probability) and apparently nobody noticed yet
X-Scale has quit [Ping timeout: 256 seconds]
kraiskil has quit [Ping timeout: 264 seconds]
m4ssi has quit [Remote host closed the connection]
emeb has joined #yosys
<thardin>
can yosys do simulation? or is iverilog the way to go?
<whitequark>
yosys has a sim pass
<whitequark>
though it's pretty basic
<whitequark>
it also has the cxxrtl backend, which is similar to verilator
<lambda>
$shift and $shiftx ignore \A_SIGNED; is that intentional? it seems to be accounted for in satgen.h so `check_cells` doesn't catch it, but it's highly unintuitive
<thardin>
it's hard to impress on people just how godawful everything is in terms of correctness or even just not crashing or having horrible 0days
emeb_mac has joined #yosys
<ZipCPU>
"Language security"? The concept looks interesting ... untrusted inputs into a formal model to turn them into something trusted, hmm ... okay. I came across something else similar recently under the title "Promise theory".
<thardin>
the E language is based around promise pipelining if I remember correctly