00:11 <bl0x_> Dankeschön =)

00:58 <_whitenotifier-9> [YoWASP/yosys] whitequark pushed 1 commit to develop [+0/-0/±1] https://github.com/YoWASP/yosys/compare/fe8e729ab034...d119fc35d75d

00:58 <_whitenotifier-9> [YoWASP/yosys] whitequark d119fc3 - Update dependencies.

07:10 <d1b2> <Nate> amaranth is the way

12:55 <whitequark> it's definitely one of the possible ways

13:09 <d1b2> <Olivier Galibert> it's one of the nice ones at least

13:20 <cr1901> whitequark: (Sorry, been busy) Would you be willing to correct my code while it's still somewhat fresh in both our heads? https://libera.irclog.whitequark.org/amaranth-lang/2023-03-13#1678743443-1678743527;

16:02 <crzwdjk> Has anyone written parsers in Amaranth? Mostly curious if people have figured out some best practices and such.

16:03 <whitequark> I tried writing parsers in Migen and that's a part of why Amaranth came into existence

16:03 <whitequark> don't think there is anything recent though

16:03 <whitequark> usually people manually unroll the FSM

16:04 <crzwdjk> What to you mean by unroll?

16:05 <whitequark> well the reason you want a parser generator in a HDL is that you typically process n-wide streams, right?

16:05 <whitequark> like for USB you have a 32-bit datapath but a bunch of things are specified as per-byte

16:05 <whitequark> PCIe has a similar thing though I think anything between 16 and 64 can be used

16:07 <whitequark> a parser is an FSM (and sometimes a stack, though the languages we parse in hardware are almost universally regular)

16:09 <crzwdjk> Yeah, in my case it's a regular language.

16:09 <crzwdjk> And yeah, it's just a big old FSM. Though I'm not doing something like USB or PCIe but something rather more ridiculous and frivolous.

16:11 <whitequark> what is it, if I may ask?

16:11 <crzwdjk> Terminal escape codes, because I am making a terminal.

16:14 <d1b2> <Olivier Galibert> and you don't plonk in some cpu core?

16:14 <d1b2> <Olivier Galibert> afaict all terminals were cpu-based

16:15 <whitequark> oh I also went through that

16:15 <whitequark> I mean, people put TCP/IP in hardware

16:15 <whitequark> terminal escape codes are considerably easier

16:16 <whitequark> I'd probably just implement that manually

16:17 <d1b2> <Olivier Galibert> heh, I did udp in hardware (including arp) but never tcp

16:17 <d1b2> <Olivier Galibert> In hardware meant there was a phy but no mac (fast ethernet era)

16:17 <bl0x> http://fy.chalmers.se/~f96hajo/fakernet/

16:17 <crzwdjk> It's more fun this way

16:18 <bl0x> A colleague of mine came up with that

16:19 <whitequark> azonenberg did a full TCP/IP stack

16:20 <d1b2> <Olivier Galibert> your colleague sounds fun 🙂

16:20 <d1b2> <Olivier Galibert> azo is insane (and fun)

16:20 <whitequark> I'm going to do it again in Amaranth some day, for Glasgow with Ethernet

16:20 <bl0x> I've written a severely limited encoder decoder for register rw access over UART.

16:20 <crzwdjk> I don't think I will be doing TCP in hardware though, I am not sure I would have the space for all that anyway

16:22 <bl0x> Håkan is definitely fun. Right now he's doing some ps-level time measurements on Artix boards.

20:31 <adamgreig[m]> i've done gigabit ethernet with udp and arp in amaranth and it was fun and easy. i shudder to imagine how complicated tcp could end up being instead :P

20:31 <whitequark> I'll let you know :p