WimⓂ️

λ🐫, -Ofun
probably a Luddite
we need frugal computing

blog: https://quickandtastycooking.org.uk/articles/
blog (sci/tech): https://limited.systems
pronouns: he/they/彼
verification code: MRV586644

Public Key

npub1008ngy8zyxwnr92hjzxuaxkul986wd4l29ta42lrmslj9uv9345suqz8zp

NIP-05 Address

wim_v12e_at_merveilles.town@momostr.pink

Profile Code

nprofile1qqs8hne5zr3zr8f3j4teprwwntw0jna8x6l4z476403ac0ez7xzc66gpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsetpnd9

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1008ngy8zyxwnr92hjzxuaxkul986wd4l29ta42lrmslj9uv9345suqz8zp

Show more details

Last Notes

2026-03-24 12:39:21 UTC

- reply

Maybe this is where permacomputing differs from frugal computing.
Considering we need weather forecasts, we had better run them efficiently, and that means centralised as performance is limited by communication overhead. But we can and shoould run them on "old" hardware, and use renewables to power them.
Also, when we oppose "centralisation", we should be clear what it is that we are opposing. I would say it is the centralisation of control and power that should be opposed. Physically putting compute resources in a central place does not imply this. It also depends on the scale of the compute resource, and the degree of centralisation.

2026-03-22 15:29:21 UTC

- reply

🎉🎉🎉🎉

2026-03-21 18:41:24 UTC

- reply

There's no rush. I am cooking and will be off line soon anyway.
The Donsol artwork looks amazing!

2026-03-21 18:28:34 UTC

- reply

even if they are immutable? I don't see a problem with that.

2026-03-21 18:21:09 UTC

- reply

I agree, it's about the logic. I don't think naming things itself is the issue. If the things we name are immutable, they can't be destroyed. It's the ability to modify and destroy the named things that is the issue.

2026-03-21 18:16:53 UTC

- reply

The first law of thermodynamics does not stop us from creating and destroying objects. It just says it takes energy to do work, and that the amounts must balance.
At the lowest level, we can only transform, not create or destroy, but I think that is almost a matter of semantics: if I transform an electron-positron pair into a photon, I might just as well say I destroyed an electron-positron pair and created a photon.
In computing we are concerned with information, and we can certainly create and destroy information, but it will always come at a cost, and I guess that is your point.

2026-03-10 19:20:33 UTC

- reply

Hopefully it won't be too long before I get it working, I'm keen to do a write-up.

2026-03-10 15:46:30 UTC

- reply

this one is not peep-hole, but I have a series of those as well, happy to share when I am confident it all works.

2026-03-10 11:22:02 UTC

I'm working on a little #Uxntal optimisation algorithm that is fundamentally of little use, it's one of those "I can't help myself" things, but I think it is quite nice:
What it does is, where possible, eliminate stores and loads in an <addr> <store with keep> ... <load> sequence, regardless of the intervening sequence. So it disentangles the stack juggling needed to put the address at the top of the stack for the load from the other operations.
After that is done, I have another algorithm that can eliminate the store/load pairs altogether.
I'll explain more when I actually get it to work.

2026-03-09 17:10:53 UTC

- reply

Great use of Perl source filters.
It makes me want to revisit my Haku language, because in Japanese you can also change the order to some extent using the particles as markers.

2026-03-06 19:48:50 UTC

- reply

I see, it's not about the storage but about the binding. Of course I agree, that is language-specific.

2026-03-06 19:45:20 UTC

- reply

ah, I'd call that argument binding. A stack is one option, registers another.
But I don't understand the issue. Even if you don't name your arguments, you have to store them somewhere, else how can you pass them? If you inline them it is not a function call anymore.

2026-03-06 19:20:35 UTC

- reply

what are function bindings? Is this binding a function to a name?

2026-03-01 19:34:04 UTC

- reply

Fortran to FPGA, I've done that already ^_^ It's easier than you might think though: there is a well developed approach to turn C-style code into circuit descriptions. All you have to do is go from Fortran to a restricted subset of C and you can use such a toolchain. The hard part is to get exploit the parallelism.

2026-03-01 17:00:52 UTC

- reply

I feel like I am really going against the grain here, because Fortran is at the opposite end of the language spectrum from catlangs, lisps etc. But I love the idea of running this very old language on Uxn. It was developed when hardware had similar specs. I'm using a more modern version (essentially Fortran-90), but that is only because I like it better than Fortran-77. In fact, despite its shortcomings, I rather like it.

2026-03-01 16:56:24 UTC

- reply

Not quite, as the whole thing is written in Perl, and I don't plan on creating a Perl-to-Uxntal compiler.
But I am working slowly towards my goal of having Funktal native. I am now going on a detour into optimisation, just for the hell of it, but when I'm back I will tackle paging. That will allow me to handle more than 64K, which is essential because the Funktal compiler bytecode is too large.

2026-03-01 10:57:31 UTC

I integrated my `yaku` #Uxntal assembler into my #Fortran to Uxnal compiler, so now I have a "real" compiler, which is not only quite satisfying but also an essential step for what I want to do next.

2026-02-24 16:30:15 UTC

- reply

The same tag is also used to criticise LLM so it is not easy filter out. Would it be possible to have an additional, more unique tag? 🙏

2026-01-27 19:46:34 UTC

- reply

Never even heard of uxnmin :-D
Meanwhile I found the examples so I guess I can piece it all together again ^_^

2026-01-27 19:43:51 UTC

- reply

Is this the VM and all the devices in a single file?

2026-01-27 19:36:13 UTC

I just pulled the latest #uxn from the repo, and the sources and projects are gone. What happened?

2026-01-23 17:09:27 UTC

- reply

That is a good idea. Currently I simply ignore the square brackets, adding a little state to suppress warnings shouldn't be too hard. I just need to create a token for them.

2026-01-23 17:05:47 UTC

- reply

Thanks, I had looked at tests.sh but not at acid.tal. Will check it out.
The warnings are overly strict, but it is based on all the mistakes I observed my students making. Will be nice to see if it helps this year's cohort.

2026-01-23 15:37:12 UTC

- reply

There is also a command line version, you can get it at https://codeberg.org/wimvanderbauwhede/yaku/src/branch/devel
(devel because I have not yet released all the changes to main)
There is a Perl version, which is what I use for development, and a JS version which I update every now and then.

2026-01-23 15:09:22 UTC

While the other #uxn folks are doing interesting and exiting things, I have been plodding on with the more mundane: an interpreter that aims to give warnings and error messages that are suitable for learners of Uxntal. Thanks to one of my students I got a JavaScript version with a web frontend. You can try it out at https://limited.systems/yaku/ but really, https://wiki.xxiivv.com/etc/uxnrepl/index.html is probably a better choice.

2026-01-18 08:15:50 UTC

- reply

wonderful job! you must have a pretty heavy-duty sewing machine, I don't think mine could handle such canvas.

2026-01-14 09:39:08 UTC

- reply

looks great!

2026-01-11 20:02:48 UTC

- reply

I'm sure they do. Compared to last year, I think it is now a lot easier to learn Uxntal.

2026-01-11 19:55:37 UTC

- reply

OK, good, I thought for a moment I would 's have to study it in more detail :-D
I have several students working on uxn for their final year projects, they all like it a lot and it's great fun. One is doing a POSIX style process API, another pthreads, and two others are trying to get the Uxn to work on GPU with a decent frame rate.
And in March I'm teaching uxntal again. Look what you've done to me ^_^

2026-01-11 17:30:43 UTC

- reply

I see, I never realised that. I thought the SDL version was the "official" one.

2026-01-11 17:17:42 UTC

- reply

Why do you maintain the X11 version?

2026-01-11 17:12:12 UTC

- reply

Ah, I thought it was to do with the differences between X11 and SDL.

2026-01-11 16:56:09 UTC

- reply

What makes it faster?

2026-01-11 10:02:52 UTC

- reply

Is uxn11 noticeably faster for this kind of games?

2026-01-10 18:03:13 UTC

- reply

What are those optimisations? Are they things SDL does not support?

2026-01-09 13:31:37 UTC

- reply

I had a very similar hat and did what you did, and now I have two copies.
I also took a pair of trousers apart. Creating the copy took me a very long time but I am wearing the all the time now.

2026-01-04 18:56:38 UTC

- reply

In other words, logo counterculture. Thanks for explaining the rationales for both versions, I am always interested in the why of things.

2025-12-25 19:32:02 UTC

- reply

It seems a very rewarding thing to do.

2025-12-25 16:54:43 UTC

- reply

That is very beautiful! How long does it take to make one of these?

2025-12-24 19:34:58 UTC

- reply

Thanks a lot!

2025-12-24 19:30:15 UTC

- reply

This is beautiful! Can we play with this too?

2025-12-22 08:05:17 UTC

- reply

Very cool!

2025-12-14 18:46:13 UTC

- reply

Wow, that looks great

2025-12-09 19:04:30 UTC

- reply

It's a total pain. I of course don't get paid for talks, but when I examine a student, they insist on paying and there's so much red tape.

2025-12-06 20:21:38 UTC

- reply

beautiful light though. The focus doesn't bother me at all.

2025-11-30 08:19:08 UTC

- reply

Very cool! That venue looks amazing.

2025-11-26 09:02:31 UTC

- reply

Me too, though I couldn't stay till the end. Always a pleasure to hear you talk.

2025-11-21 19:19:31 UTC

- reply

And of course, you know how it goes, I needed a displacement activity this morning so all jumps are now immediate ^_^

2025-11-20 21:06:37 UTC

- reply

Happy birthday, 100r! I am so glad I found you on the fediverse, I feel blessed to know you!

2025-11-20 15:57:07 UTC

- reply

I wrote most of that before I knew about those immediate jumps, I just didn't get around to updating it. Maybe one day ^_^

2025-11-20 09:46:39 UTC

Hi everyone, I made a repo for my various #Uxntal libraries: argument handling, string handling, formatted printing, signed arithmetic and comparison, call stack support,...
It is not very well documented but I hope it can be of some use. Please ask me anything about it!
https://codeberg.org/wimvanderbauwhede/uxntal-libs/src/branch/main/

2025-11-15 17:29:33 UTC

- reply

Ah, yes indeed, I don't know how I came to mess that up.

2025-11-15 17:19:04 UTC

- reply

🎶 Marguerite Yourcenar pushed the realms of truth so far🎶
https://www.youtube.com/watch?v=ko5hfl_9np4

2025-11-11 16:48:58 UTC

- reply

What I mean is that, when you strip away the concrete syntax, they reduce to the same underlying abstract syntax and therefore are equivalent. This is the case for the example I gave. What I wanted to say is that using a particular syntax to express a particular behaviour is merely a convention. Specifying the types would give stronger hints about the behaviour, but even then you can't be sure. It is possible to write a fast and safe version and a slow and unsafe one that look exactly the same and even have the same type.
@npub10lg…asm4

2025-11-11 07:11:51 UTC

- reply

You can have the semantics of C's strlen() with an object notation or use a function notation for a method call. It's purely a syntactical convention that the dot suggests an object, all it indicates is that this is a prefix argument.
For example, in Raku you can write arr.elems or elems arr or elems(arr)
The semantics are the same regardless of the syntax.

2025-11-10 09:22:07 UTC

- reply

To my mind, the difference between writing append(str, char) and str.append(char) is purely syntactic. I guess the difference is that str in the second case is a kind of record rather than a bare string?

2025-10-28 11:59:23 UTC

- reply

Thanks for sharing those pictures, they are very atmospheric. Love the light and the fog.

2025-10-24 17:59:00 UTC

- reply

that track is really addictive

2025-10-21 06:41:58 UTC

- reply

I don't think it is possible because the "fractran machine" uses an infinite number of named registers, rather than memory addresses. But the relationship is still very interesting.

2025-10-20 19:25:59 UTC

- reply

I think the ordering might be the problem, as fractran's accumulator is not ordered. Does e.g.
((lambda (x y) (y x)) x y)
work?

2025-10-20 19:07:28 UTC

- reply

That is very cool! So you use fractran as an evaluator for the lisp expressions. I've never considered that before.

2025-10-20 17:11:39 UTC

- reply

I tire quickly of having to roll my own numbers though

2025-10-17 10:30:35 UTC

- reply

What us a grant application juror?

2025-10-16 15:11:16 UTC

- reply

tuples don't count?

2025-10-15 19:41:59 UTC

- reply

I see, you tried to remove the lambda. That does not work because of the DUP2. Effectively, this applies the function to itself.
If you wrap it as
@fact-fix
{
,&f STR2 DUP ,&n STR
#01 NEQ ?{ #01 JMP2r }
LIT &n $1 STHk
#01 SUB LIT2 &f $2 DUP2 JSR2
STHr DUP ,&n STR MUL
JMP2r
} STH2r DUP2 JSR2
JMP2r
that should work;
else you need ;fact-fix fact-fix

2025-10-15 19:36:00 UTC

- reply

I used uxnasm+uxncli and it returns 0x78

2025-10-15 19:32:49 UTC

- reply

You know, right now I really wonder how I ever came to write something like that.

2025-10-15 19:32:17 UTC

- reply

Sorry, forgot to paste 🤦
Here are both versions:
( factorial with fixed-point )
|0100
#05
{
,&f STR2 DUP ,&n STR
#01 NEQ #04 MUL JMP #01 JMP2r
,&n LDR STHk
#01 SUB ,&f LDR2 DUP2 JSR2
STHr DUP ,&n STR MUL
JMP2r
&f $2 &n $1
} STH2r DUP2 JSR2
#18 DEO ( 120 )
BRK
( factorial with fixed-point, args stored inside the lambda, self-modifying code )
|0100
#05
{
,&f STR2 DUP ,&n STR
#01 NEQ ?{ #01 JMP2r }
LIT &n $1 STHk
#01 SUB LIT2 &f $2 DUP2 JSR2
STHr DUP ,&n STR MUL
JMP2r
} STH2r DUP2 JSR2
#18 DEO ( prints 'x', ASCII 120 )
BRK

2025-10-15 19:29:04 UTC

- reply

Here you are, bytes instead of shorts

2025-10-15 19:26:11 UTC

- reply

Let's see.

2025-10-15 19:21:20 UTC

- reply

( factorial with fixed-point )
|0100
#0005
{
,&f STR2 DUP2 ,&n STR2
#0001 NEQ2 #04 MUL JMP #0001 JMP2r
,&n LDR2 STH2k
#0001 SUB2 ,&f LDR2 DUP2 JSR2
STH2r DUP2 ,&n STR2 MUL2
JMP2r
&f $2 &n $2
} STH2r DUP2 JSR2
#18 DEO ( 120 )
BRK

2025-09-19 17:56:30 UTC

- reply

OK, what I'll do first then is share the current Uxntal API and device conventions via the mailing list, and then we can discuss any changes on IRC, and take it from there.

2025-09-19 17:36:08 UTC

- reply

Are you all on UTC-7?

2025-09-19 17:30:40 UTC

- reply

I'll do better, and implement the whole thing myself. It's not that much work (said he foolishly).

2025-09-19 17:27:32 UTC

- reply

It is a deep irony to me that I just spend a summer supervising a project to create an Uxntal network API, but then the student did not want the release the code as open source.
Is this a complete POSIX API?

2025-09-01 07:05:38 UTC

- reply

What a fabulous tribute!

2025-08-25 18:55:22 UTC

- reply

Wonderful! The eyes have it ^_^

2025-08-25 08:14:26 UTC

- reply

You always do the coolest things!

2025-08-20 15:58:47 UTC

- reply

I haven't read it. I guess the idea of being able to control a minor god or spirit must be very old.

2025-08-20 15:39:18 UTC

"You can't put the genie back in the bottle" makes no sense to me. In the old stories, humans could control the djinn through specific objects (rings, lamps). Without the explicit wish of the controlling human, the djinn would by default be in the object.
Could it be that this is a conflation with Pandora's Box, because once opened, the evils were released and could never be put back into the box. Interestingly, that"box" was actually a large jar (πίθος pithos).
At first the tribes of men had lived upon the earth
apart and free of evils and of tiresome toil
and hard diseases, which have brought to men their dooms, because by hardship mortal men are quickly aged.
But with her hands the woman raised the jar’s great lid,
released all these, devising [1] grievous cares for men.
Alone there, Elpis [2], in her indestructible home,
remained within, beneath the lip, nor by the door
escaped, because the vessel’s lid had stopped her first,
by will of aegis-bearing, cloud-compelling Zeus.
[1]. emesato= devised, intended, resolved, planned, contrived. Gantz (Early Greek Myth, p. 157) observes that this word suggests that “Pandora understood in advance the consequences of her action.”
[2]. Elpis = Hope. It seems paradoxical that all the ills are free to roam, but Hope is trapped in the jar. Gantz (Early Greek Myth, p. 157) suggests that Elpis may mean expectation (“a realistic awareness of just how bad things are and are likely to get”). By trapping Elpis Pandora saved us from true awareness of our predicament.

2025-07-26 15:51:09 UTC

- reply

I just published my article about Li Kotomi's novel "The Island where Red Spider Lilies Bloom". You might like it even if you haven't read the novel. It's really just geeking out about Japanese ^_^
https://quickandtastycooking.org.uk/articles/higanbana/

2025-07-26 12:32:49 UTC

- reply

I think not giving much headroom to deviate from the script is kind of the point of Hedy? The idea is to start very small and then extend gradually in a controlled way.
I don't actually know what is best. From a purely teaching perspective, ignoring the real world, a small untyped functional language seems best to me. But the real world is there with all its imperative languages, and it doesn't do to ignore them. And there is always some pressure to make the curriculum relevant to industry (mostly from the students, as they want a job). Still, if I had it my way, we wouldn't start with Python.

2025-07-25 19:09:29 UTC

- reply

I think it has no relation. But a smaller language with fewer concepts provides fewer distractions. I noticed that with Uxntal: it may seem very small, but you can say ,&loop JMP or ;&loop JMP2 or !&loop or ;loop JMP2 etc, so already there is a lot to take in.
From that perspective something like Hedy is probably a good idea.
One key problem for many students is that they focus too much on syntax, rather than on what the code actually does. From that perspective, Scheme is good, as there is no syntax.

2025-07-25 19:00:23 UTC

- reply

It's kind of funny. I teach 1st year students in 2nd Sem. In 1st Sem they learn Python (because my dept moved to Python long ago; before that they taught Ada as first language), then they come into my class and get Uxntal.
But the problem I perceive is not about loops and assignments. It's about mental models for programming. So I don't think Python itself is to blame (although I personally never cared for that language), it is the way it is used to teach, and what is taught using it.

2025-07-25 18:02:05 UTC

- reply

Found it, It's $64,310 but "the median annual price paid by an undergraduate who received an MIT Scholarship was $12,938⁠ for the 2023–2024 academic year."

2025-07-25 18:00:10 UTC

- reply

It's clear that the author cared about teaching programming.

2025-07-25 17:53:32 UTC

- reply

Is it very expensive to study at MIT?

2025-07-24 09:15:00 UTC

- reply

"Talk away, Mr. Holmes. I'm just loving it. It's fine!"

2025-07-23 16:56:16 UTC

- reply

Thank you very much for sharing this! It was a fun read, and following the links really sent me down the rabbit hole.
It is especially relevant to me at the moment because I am working on a blog post about a novel I finished, "The island where the red spider lilies bloom" by Li Kotomi, which is all about language.
Interestingly, in that novel they replace nihongo 日本語 by hi-no-moto kotoba ひのもとことば日の本言葉 on the one hand and nihongo written as ニホン語 on the other.
My first reaction to the mention of kotoba was, all right, but an at least equally valid word for language is gengo 言語. If the Japanese interlocutor had picked that, the while construction would fall apart; which was pointed out further down in he article.
Also, I'm with those who retorted that the connection between the different meanings of koto are maybe a but far-fetched. There are almost 50 different kanji with "koto" as a reading. But I'm no scholar, so reserve judgement.
I was a bit puzzled by the Sei Shonagon angle: OK, she wrote kotoba as 詞 which also means poetry, but she must have been far from alone in this as it is still an accepted way to write it. Anyway, I'm biased as I'm on Murasaki Shikibu's side ^_^

2025-06-11 11:27:30 UTC

- reply

Was about to remark on it too. I didn't know anything about it but the kanji caught my eye and I looked it up.

2025-06-05 11:38:55 UTC

- reply

My first impression was Rek holding an :anarchy: sign

2025-06-03 11:06:18 UTC

- reply

Verachtungserschöpfung
(I admit, I made it up from
Verachtung and Erschöpfung)

2025-05-31 08:20:27 UTC

- reply

Hence the 56 of course. In a way, it was the I Ching that made me move to Scotland.

2025-05-30 20:40:10 UTC

- reply

It sounds like something from the I Ching

2025-05-29 17:47:40 UTC

- reply

"B" as in the predecessor of the C language?

2025-05-20 08:08:26 UTC

- reply

This is how I used it to formalise my Gannet machine, which is a parallel reduction machine programmed in a very Scheme-like language:
https://eprints.gla.ac.uk/6546/1/6546.pdf

2025-05-11 17:46:34 UTC

- reply

It's a formal description of how the language works. It's mostly popular amongst people doing type theory, esp. the notation used in Pierce's Types and Programming Languages. But Iike the approach Matthias Felleisen has taken for Scheme.

2025-05-11 17:44:37 UTC

- reply

No, don't even know what it is.

2025-05-11 10:07:48 UTC

- reply

Soon, Uxntal will be one of the best specified languages out there :-D
Next up, operational semantics?

2025-05-06 07:05:37 UTC

- reply

That sounds like a great idea. I would love to see a book like that. Happy to help/contribute though I assume it's something you'd rather do on your own.