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

[email protected]

Profile Code

nprofile1qqs8hne5zr3zr8f3j4teprwwntw0jna8x6l4z476403ac0ez7xzc66gprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmnta53wmu

Publishing to

relay.momostr.pink

Author Public Key

npub1008ngy8zyxwnr92hjzxuaxkul986wd4l29ta42lrmslj9uv9345suqz8zp

Show more details

Last Notes

2026-08-06 09:17:00 UTC

- reply

I had not come across POP-2 before. As everything you do, it sounds really cool.

2026-07-22 16:26:38 UTC

- reply

Are you running dos in an emulator?

2026-06-20 15:53:22 UTC

- reply

First of all, thank you for thinking of me!
I am listening to the album right now, about 10 minutes in, and liking it a lot so far. To my ears it's like Mogwai with beautiful Japanese voices -- and screaming ^_^.

2026-06-07 16:25:30 UTC

- reply

OK, I understand now. So the assembler does not need to know at all, it's part of the source code. Thanks for your patience.

2026-06-07 16:22:30 UTC

- reply

Ah, now I see, that ;meta #06 DEO2 call must be the first thing in the program.

2026-06-07 16:07:37 UTC

- reply

Sorry, I don't understand. If the metadata is not identified by name or by position, then how is it identified?

2026-06-07 16:03:06 UTC

- reply

thanks! does the assembler look fir @meta (so `meta` is a reserved word), or does it have to be at the end of the source code?

2026-05-27 08:51:42 UTC

- reply

I'm sure a tiny Perl script can handle this ^_^

2026-05-26 18:56:28 UTC

- reply

let the fun begin! though it will have to wait until I'm done with the boring work stuff

2026-05-26 18:53:23 UTC

- reply

Ah, how come that is different from https://paulbourke.net/dataformats/hershey/japanese.gz ?

2026-05-26 18:15:23 UTC

- reply

amazing! is this the font format you posted about earlier?

2026-05-26 18:07:07 UTC

- reply

It looks amazing! That font is very beautiful. And doing this in Uxn is awesome. Move over, LaTeX :-D

2026-05-17 15:58:49 UTC

I submitted a paper to the LIMITS2026 workshop ("Computing within limits") on running #uxn on GPU, mostly based on work my students did.
It was rejected, not because of of the technical contribution but because it was considered not a good fit for the workshop.
I am kind of surprised, because the reason to do this work is to see if we can run resource constrained VMs like uxn not only on the CPU but also on the GPU, which is there anyway so we might as well use it.

2026-04-29 18:49:42 UTC

- reply

rightly so, it's beautiful

2026-04-20 15:35:39 UTC

- reply

where did abba go?

2026-04-20 15:26:41 UTC

- reply

abracadabr ^_^

2026-04-20 15:20:09 UTC

- reply

and a "d"

2026-04-20 15:04:44 UTC

- reply

You need an "r" in there somewhere 😄

2026-04-09 10:36:44 UTC

bad.horse is back!
traceroute -m 40 bad.horse

2026-04-05 16:23:20 UTC

- reply

I didn't know her, thanks for sharing!

2026-04-05 16:13:25 UTC

"Take a little hand" by Yoko Kanno
https://youtu.be/UjBm0aiD-CE?si=tuMHWeUlJBfZl0vf
#GitS

2026-04-05 10:23:09 UTC

- reply

Looks very cool. I have a question about the first example: you say the sets are unordered. So is the outcome "[this is a programming language]" an accident of the evaluator implementation? If it is unordered, elements could occur in any order in the set, isn't it?

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!