00:45 <_whitenotifier-f> [YoWASP/yosys] whitequark pushed 1 commit to develop [+0/-0/±1] https://github.com/YoWASP/yosys/compare/001bd2adf85e...3c3f21badc4b

00:45 <_whitenotifier-f> [YoWASP/yosys] whitequark 3c3f21b - Update dependencies.

03:49 <cr1901> Is there an algorithm for the following problem: "Given a multiply by a rational number, spit out a minimal sequence of L/R shifts and adds to multiply by the numerator and divide by the denominator"?

03:54 <cr1901> I guess I could derive the naive algorithm by doing some in my head- e.g. "n" times 9/5 is "shift n left by 3, add n, shift right by 2, sub n"

03:55 <whitequark[cis]> what's this for?

03:56 <cr1901> Convert Celsius to Fahrenheit from a temp sensor (1.8*t + 32). I think I have code for that somewhere already.

03:57 <whitequark[cis]> do it in fixed point?

03:59 <cr1901> Hmmm lemme try that

04:44 <cr1901> http://gopher.wdj-consulting.com:70/store/fixed.py

04:44 <cr1901> SB_CARRY 15

04:44 <cr1901> SB_LUT4 93

04:44 <cr1901> I feel like this is a lot, but I did have to extend the precision to get close to 1.8 (not exactly representable in base 2)

04:48 <whitequark[cis]> that is really not a lot imo

04:48 <whitequark[cis]> for 1-cycle

04:52 <cr1901> Well, at least I have decimal places after the ones now :P. And yea I was just curious about 1-cycle LUT usage. God help me if I start seriously playing with fixed point types in Amaranth.

05:39 <cr1901> Updated to final version. With an intermediate precision of Q8.3 input, Q9.18 intermedia (27 bits), and Q10.3 output. 23 SB_CARRY and SB_LUT4 170.

05:39 <cr1901> But the model is adjustable, so I can get a feel for how fixed-point precision affects resource usage

05:40 <cr1901> s/intermediate// (first one)

07:04 <galibert[m]> how do you count the usage?

13:53 <sporniket> Hello, I defined a clock domain with e.g. myDomain = ClockDomain("foo", local=True) and setup the clk signal, and then I use it. Then amaranth complains, saying "amaranth.hdl.cd.DomainError: Domain 'foo' is used but not defined". Actual code :

13:53 <sporniket> https://github.com/sporniket/the-cradle-colorlight-i9-ecp5-amaranth-hdl/blob/3-test-screen-through-hdmi-port/src/the_cradle/the_cradle.py#L109

13:53 <sporniket> did I miss something to do ?

13:54 <sporniket> (using the domain means that I use DomainRenamer("foo")(MyOtherElaboratable(...)) )

13:54 <sporniket> e.g. https://github.com/sporniket/the-cradle-colorlight-i9-ecp5-amaranth-hdl/blob/3-test-screen-through-hdmi-port/src/the_cradle/the_cradle.py#L120

13:56 <galibert[m]> Where's Sequencer?

14:01 <galibert[m]> or, in fact, what do I need to try to run your code?

14:01 <galibert[m]> (I don't see anything obvious so I'd like to try)

14:05 <sporniket> oh there is a "redeploy" script that will try to build and program the ECP5 (that is what I am doing with my colorlight board plugged on my computer). The "amaranth-stuff" is there if you want to see the code of those other modules like sequencer, dvitmdsencoder, etc https://github.com/sporniket/amaranth-stuff/tree/main/src/amaranth_stuff/modules

14:06 <sporniket> (the redeploy script just launch the redeployer.py in the project)

14:08 <galibert[m]> getting fun errors like "no module named black"

14:09 <galibert[m]> missing a bunch of dependencies anyway, like Colorlight_I9_V7_2_Platform

14:09 <galibert[m]> anyway, can I have the exact error message?

14:10 <galibert[m]> (fixing some stuff in the deps, but whatever)

14:11 <sporniket> this what I get at the end of running "redeploy" :

14:11 <sporniket> ========================[ START OF Deployment ]============================

14:11 <sporniket> Traceback (most recent call last):

14:11 <sporniket>   File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main

14:11 <sporniket>     return _run_code(code, main_globals, None,

14:11 <sporniket>   File "/usr/lib/python3.10/runpy.py", line 86, in _run_code

14:11 <sporniket>     exec(code, run_globals)

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/the_cradle/deployer.py", line 48, in <module>

14:11 <sporniket>     Deployer(Colorlight_I9_V7_2_Platform(), TheCradle()).deploy()

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/the_cradle/deployer.py", line 43, in deploy

14:11 <sporniket>     self.endpoint.build(self.payload, do_program=True)

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/amaranth/build/plat.py", line 105, in build

14:11 <sporniket>     plan = self.prepare(elaboratable, name, **kwargs)

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/amaranth/build/plat.py", line 147, in prepare

14:11 <sporniket>     fragment._propagate_domains(self.create_missing_domain, platform=self)

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/amaranth/hdl/ir.py", line 362, in _propagate_domains

14:11 <sporniket>     new_domains = self._create_missing_domains(missing_domain, platform=platform)

14:11 <sporniket>   File "/home/dsporn/.local/lib/python3.10/site-packages/amaranth/hdl/ir.py", line 340, in _create_missing_domains

14:12 <charlottia> (You might consider http://paste.debian.net/ for long text like this.)

14:12 <sporniket> (real life is calling, I have to go out, if you have other questions I will answer them as soon as I am back)

14:13 <sporniket> charlottia ok, got it

14:16 <sporniket> galibert[m] I believe that to run the code, only those 2 lines (the first is to self install) http://paste.debian.net/1292480/

14:16 <galibert[m]> yeah, I'm missing some stuff like nextpnr-ecp5

14:18 <galibert[m]> Hmm, I've done impolite things to the code to try to make it compile without a platform, but then I'm getting no errors

14:18 <galibert[m]> guess I need to fake ecp5 instead

14:20 <charlottia> do you need nextpnr-ecp5 to get to that point? It's still elaborating; it hasn't gotten to calling the toolchain yet.

14:21 <galibert[m]> I don't actually need it, I just had to remove it from required_tools()

14:21 <charlottia> ahhh, nice.

14:22 <galibert[m]> Yes, got the error now

14:22 <galibert[m]> amaranth.hdl.cd.DomainError: Domain 'pixel' is used but not defined

14:23 <galibert[m]> I bet it has just been missed in one place but a better error message is hell to generate without the new ir

14:24 <galibert[m]> or maybe it's an order issue

14:29 <charlottia> galibert: if it's convenient for you, try adding e.g. `print(self)` around line 332 of your `amaranth/hdl/ir.py` temporarily.

14:29 <charlottia> Then you'll see which Fragment didn't have the domain defined on it.

14:30 <galibert[m]> oh good idea

14:30 <galibert[m]> <amaranth.hdl.ir.Fragment object at 0x7f8b22bcc810>

14:30 <galibert[m]> friggin' python

14:30 <charlottia> Hah :3

14:31 <charlottia> You could try printing self.__dict__. It'll be noisy!

14:32 <charlottia> Going by that traceback, though, it looks like the toplevel Elaboratable maybe?

14:32 <charlottia> Hey, um, where's the m.domains += ...?

14:32 <charlottia> I think that'll be it :3

14:33 <galibert[m]> DUH

14:33 <galibert[m]> of course

14:33 <galibert[m]> I had forgotten about that

14:34 <galibert[m]> sporniket: what she says, you need a m.domains += [the domains you created]

14:36 <galibert[m]> * sporniket: what she says, you need a `m.domains += \[the domains you created\]`

14:40 <galibert[m]> * need a `m.domains +=, * you created]`

14:57 <sporniket> I'm back, on it now

14:59 <sporniket> ok no more errors :)

14:59 <charlottia> Yay!

15:02 <sporniket> thanks

15:03 <charlottia> Thanks galibert!

15:05 <sporniket> when I will get something on screen, I'll will tell you (now I have nothing, so I will have to check various things), see you later :)

15:06 <galibert[m]> Have fun

15:58 <cr1901> galibert[m]: To count usage, I do (requires an amaranth PR not merged yet): python fixed.py > fixed.v; yosys -p 'synth_ice40' fixed.v. I have python code to invoke yosys and run a script, but wish to rework it

16:03 <galibert[m]> interesting, thanks