2026-08-26 00:00:08 <lisbeths    > when you optimize your compiler for subtraction less than or equal to there can be some peformance advantages to subtraction
2026-08-26 00:00:12 <xentrac     > bjorkint0sh: it might make sense, it might not; it might depend on your design constraints
2026-08-26 00:00:39 <bjorkint0sh > talk of overloading an operation.
2026-08-26 00:00:41 <lisbeths    > also remember on an x86_64 machine the optimizer is going to jump ahead and predict future subtractions, and that is much easier to do on an incredibly small state machine like one that runs subleq
2026-08-26 00:00:54 <xentrac     > in Calculus Vaporis https://github.com/kragen/calculusvaporis I have subtraction and NAND, but no addition
2026-08-26 00:00:56 <lisbeths    > on microcontrollers without opptimizers there would be an extreme performance hit because no branch prediction
2026-08-26 00:00:58 <bjorkint0sh > lisbeths: do you have an fpga board to play with?
2026-08-26 00:01:03 <bjorkint0sh > it might be highly instructive.
2026-08-26 00:01:20 <lisbeths    > no but I hope there is some software where I can simulate an fpga over a period of time
2026-08-26 00:01:27 <lisbeths    > I can get my boss to buy me an fpga soon here
2026-08-26 00:01:37 <xentrac     > lisbeths: probably not an *extreme* performance hit, probably only about 30%
2026-08-26 00:01:39 <hello-operator> https://github.com/howerj/subleq/blob/master/subleq.fth shows that implementation of all those operators
2026-08-26 00:01:40 <bjorkint0sh > they're not really expensive.
2026-08-26 00:02:06 <hello-operator> And here it is running on an FPGA https://github.com/howerj/subleq-vhdl
2026-08-26 00:02:13 <xentrac     > I mean interpreting a SUBLEQ instruction is probably about 15-20 cycles, and only about three to five of those are going to be spent on not taking the jump
2026-08-26 00:02:29 <bjorkint0sh > there was a book I found in my college library years ago...
2026-08-26 00:02:40 <lisbeths    > so this type of abstract machine, for our purposes, is not ideal for when performance matters
2026-08-26 00:02:52 <lisbeths    > its for when you want true portability and isolation of your code
2026-08-26 00:03:18 <lisbeths    > its the difference between a good sandbox and a hypothetically perfect sandbox
2026-08-26 00:03:22 <xentrac     > Calculus Vaporis was designed to use as few gates as possible, but the Intersil 6100 is just as few gates and considerably more performant
2026-08-26 00:03:34 <bjorkint0sh > https://www.amazon.com/dp/1402074166?lv Computer Architecture: A Minimalist Perspective
2026-08-26 00:03:52 <xentrac     > also, the 6502
2026-08-26 00:04:05 <lisbeths    > I mean we could do this all day: come up with more and more efficient abstract machines that have been implemented before
2026-08-26 00:04:16 <bjorkint0sh > yes.
2026-08-26 00:04:20 <bjorkint0sh > let's do that, lisbeths.
2026-08-26 00:04:22 <xentrac     > I *did* do this all day on several occasions
2026-08-26 00:04:26 <xentrac     > they were days well spent
2026-08-26 00:04:26 <bjorkint0sh > there has to be the ULTIMATE one!
2026-08-26 00:04:29 <bjorkint0sh > to rule them all.
2026-08-26 00:04:32 <bjorkint0sh > the ONE instruction.
2026-08-26 00:04:37 <xentrac     > no, there's a Pareto optimality frontier
2026-08-26 00:04:45 <lisbeths    > if you insist but I argue that is not the intellectually fruitful part is trying to find the most efficient abstract machine
2026-08-26 00:04:52 <bjorkint0sh > pfft. talk to the Moore, xentrac.
2026-08-26 00:05:00 <lisbeths    > second of all I am trying to explain to you that ive nearly implemented an abstract machine that is compiled not interpreted
2026-08-26 00:05:15 <lisbeths    > so seeing my compiled abstract machine might speed up your search
2026-08-26 00:05:18 <bjorkint0sh > what did you write it in, lisbeths?
2026-08-26 00:05:24 <bjorkint0sh > asm?
2026-08-26 00:05:42 <lisbeths    > I wrote it in awk and it compiles to a subset of c that gets injected in the middle of a c file
2026-08-26 00:05:47 <lisbeths    > it is gpl2 but the documentation is written by ai
2026-08-26 00:05:58 <xentrac     > right, fastspungus
2026-08-26 00:06:05 <lisbeths    > yes and it has been updated recently
2026-08-26 00:06:17 <lisbeths    > https://github.com/memesmith67/fastspungus
2026-08-26 00:06:38 <lisbeths    > if you happened to learn a technique or two from fastspungus that would be sooooooo unfortunate
2026-08-26 00:06:49 <lisbeths    > I am sure the designer would care so much if you figured out how it worked
2026-08-26 00:07:04 <bjorkint0sh > the designer, namely lisbeths.
2026-08-26 00:07:22 <lisbeths    > it would be such a shame if someone learned its design secrets
2026-08-26 00:08:02 <xentrac     > IIRC crc pointed out that some of the examples in the README don't work
2026-08-26 00:08:04 <bjorkint0sh > lisbeths: eh. no formatting?
2026-08-26 00:08:17 <lisbeths    > yeah cause the readme is formed by ai that doesnt truly understand how fastspsungus works
2026-08-26 00:08:28 <lisbeths    > I am awful at writing documentation
2026-08-26 00:08:31 <xentrac     > hmm, well, maybe you should delete it
2026-08-26 00:08:36 <xentrac     > the README, I mean
2026-08-26 00:08:37 <bjorkint0sh > it's written like Arthur WHitney wrote his code for k.
2026-08-26 00:08:38 <xentrac     > not the project
2026-08-26 00:08:51 <lisbeths    > the readme is the only non gpl document I can share with people about fastspungus
2026-08-26 00:08:59 <xentrac     > well, then you should make sure it's correct
2026-08-26 00:09:14 <lisbeths    > as ive said one of my skills is not documentation
2026-08-26 00:09:34 <lisbeths    > each year i will just ask ai to rewrite better documentation and it will get better each year
2026-08-26 00:09:38 <bjorkint0sh > if that's all you lack, it'll be the cherry on top once you master that too.
2026-08-26 00:09:57 <xentrac     > I guess that's possible?
2026-08-26 00:10:26 <bjorkint0sh > there's also the option of literate programming where you intertwine the two.
2026-08-26 00:10:44 <crc         > disclosure: I skimmed through the readme and ran the code against a few of the examples; didn't look at the actual fastspungus code in any detail since it's GPL, which I mostly avoid
2026-08-26 00:11:02 <lisbeths    > I mean bro
2026-08-26 00:11:03 <xentrac     > (a valid choice, though not one I share)
2026-08-26 00:11:13 <lisbeths    > it would be sUCH a shame if you did look a that gpl code and didnt honor the gpl
2026-08-26 00:11:23 <lisbeths    > I am sure the author would be sooooo upset
2026-08-26 00:11:50 <xentrac     > the author's successors in interest might be delighted because it gave them a license to sue for damages
2026-08-26 00:12:01 <lisbeths    > yeah that is the bind I am in it seems
2026-08-26 00:12:20 <lisbeths    > as far as I know I am able to license it to you if I choose
2026-08-26 00:12:20 <xentrac     > it hardly seems like a bind.  if you wrote the code you can license it however you want
2026-08-26 00:12:35 <xentrac     > unless it's a work for hire or something
2026-08-26 00:12:40 <lisbeths    > if you are able to work with a laywer with me then we could figure out how i can license it to you safely
2026-08-26 00:12:44 <lisbeths    > but I cant afford a lawyer
2026-08-26 00:12:57 <bjorkint0sh > lisbeths: what if you removed the license altogether?
2026-08-26 00:12:59 <crc         > Given that I do commercial work with stack based languages, I am cautious about what I read code-wise
2026-08-26 00:13:10 <xentrac     > bjorkint0sh: that would make the situation worse, not better
2026-08-26 00:13:20 <bjorkint0sh > 
ACTION is clearly not getting it.

2026-08-26 00:13:24 <lisbeths    > there are legally technical reasons why I dont use public domain. including that in many nations public domain isnt recognized
2026-08-26 00:13:29 <crc         > bjorkint0sh: no license to me implies no rights to the code at all
2026-08-26 00:13:42 <lisbeths    > in the end it will likely either be gpl 2.0 or 0 clause bsd
2026-08-26 00:13:43 <bjorkint0sh > think of it as a pastebin.
2026-08-26 00:13:49 <xentrac     > you could use a permissive license like the Apache-2.0 or 2-clause BSD
2026-08-26 00:13:57 <lisbeths    > 0 clause bsd is superior
2026-08-26 00:14:01 <crc         > I use a fairly short, liberal license for what I release publicly
2026-08-26 00:14:04 <lisbeths    > written by mark landley
2026-08-26 00:14:13 <xentrac     > bjorkint0sh: sadly, the Berne Convention thinks of pastebins as radioactive waste
2026-08-26 00:14:31 <lisbeths    > so I *can* relisence it as long as I can guarantee that I can still use it how I want like I would under gpl
2026-08-26 00:14:37 <lisbeths    > I dont know enough about the law to safely relicense it
2026-08-26 00:14:51 <lisbeths    > I desire to give you access believe me, its just not currently within my power
2026-08-26 00:15:59 <xentrac     > oh, nobody can stop you from using software you own because you licensed it to them
2026-08-26 00:16:09 <crc         > As the author, you retain the copyright; you can offer the code under different licenses to different people
2026-08-26 00:16:17 <lisbeths    > I understand that but you arent attourneys
2026-08-26 00:16:25 <xentrac     > (that you own the copyright for)
2026-08-26 00:16:32 <crc         > What they can do with the code depends on the license you provide them
2026-08-26 00:16:53 <lisbeths    > I have already in the past been in tricky situations over bsd 0 clause
2026-08-26 00:16:57 <lisbeths    > many people threatened to sue me
2026-08-26 00:17:02 <xentrac     > there is no "BSD 0 clause"
2026-08-26 00:17:11 <lisbeths    > yes it is just named that it doesnt actually ahve zero clauses
2026-08-26 00:17:17 <xentrac     > a 0-clause license would be the empty string ""
2026-08-26 00:17:19 <lisbeths    > it was written by robert landley I believe his name is
2026-08-26 00:17:34 <lisbeths    > so the thing to understand is that legally a bsd 3 clause has more than 3 clauses
2026-08-26 00:17:51 <lisbeths    > so when you remove 3 of the clauses from bsd 3 clause the namesake is bsd 0 clause despite it having clauses
2026-08-26 00:18:05 <hello-operator> What trouble did you get into?
2026-08-26 00:18:25 <lisbeths    > people just kept threatening to sue me very frequently
2026-08-26 00:18:37 <lisbeths    > they seemed to believe that they could just fork my code into proprietary or gpl code and that this would allow them to sue me
2026-08-26 00:19:00 <lisbeths    > even if they couldnt sue me the constant threat of lawsuits was draining for me
2026-08-26 00:19:04 <lisbeths    > under gpl nobody threatens to sue me
2026-08-26 00:19:14 <hello-operator> No, it wouldn't, that's not how things work
2026-08-26 00:19:25 <lisbeths    > I am somewhat aware that you are correct yes
2026-08-26 00:19:37 <lisbeths    > but most coders are actually incompetent when it comes to licensing law
2026-08-26 00:19:43 <lisbeths    > and they believe htey can sue me
2026-08-26 00:20:53 <bjorkint0sh > that's odd, lisbeths.
2026-08-26 00:21:05 <hello-operator> It seems like they are trying to scare you into using a specific license, or for other ulterior goals.
2026-08-26 00:21:15 <lisbeths    > indeed. and it worked! XD
2026-08-26 00:21:30 <xentrac     > lisbeths: you should probably spend less time interacting with delusional people
2026-08-26 00:21:35 <lisbeths    > I suffer no consequence from using gpl2 only people who want to implement fast spungus in another license have the consequence
2026-08-26 00:21:50 <lisbeths    > unfortunately I frequently interact with very strange people
2026-08-26 00:21:57 <xentrac     > tht's often very draining
2026-08-26 00:22:02 <hello-operator> I would ignore them. You should use the GPL because you believe in it, or not use because you do not believe in it, not because of nonsense
2026-08-26 00:22:26 <lisbeths    > I would prefer for reform to software law so that globally that public domain was redfined as gpl 0 clause
2026-08-26 00:22:46 <xentrac     > there has never, to my knowledge, been a lawsuit because someone forked code into proprietary or gpl code and then sued the author
2026-08-26 00:22:50 <crc         > As a meaningless data point, I have been using liberal licenses for the better part of three decades without any legal headaches
2026-08-26 00:22:58 <lisbeths    > I am very chinese in my design philosophy for coders tools. i want the intellectual property to spread when it comes to coders tools
2026-08-26 00:23:22 <xentrac     > perhaps you could be very Chinese like Dr. C.H. Ting and write good documentation
2026-08-26 00:23:24 <lisbeths    > there are only two ways for me to transfer fastspungus onto someone who does not want to read the gpl code
2026-08-26 00:23:28 <lisbeths    > the first is to relicense it
2026-08-26 00:23:35 <lisbeths    > the second is to find some other way to transfer the knowledge of how it works
2026-08-26 00:23:55 <bjorkint0sh > lisbeths: Re-write it in Rust.
2026-08-26 00:24:23 <lisbeths    > I think that would still open me up to lawsuit threats of *that* implementation
2026-08-26 00:24:42 <bjorkint0sh > really? I didn't know it was so fraught.
2026-08-26 00:24:43 <lisbeths    > its lik eone of those games where everyone pretends to close their eyes but someone always peaks. eventually the design of fastspungus will leak
2026-08-26 00:25:22 <lisbeths    > it not necessarliy that I can be successfully sued, its that they can try to sue me
2026-08-26 00:25:32 <lisbeths    > and it immediately creates social friction with me when a coder gets it in their heads they want to sue me
2026-08-26 00:25:52 <lisbeths    > the moment a coder feels they want ot sue me thats all they ever talk aobut ever again. they become a clockwork parrot
2026-08-26 00:26:08 <lisbeths    > perhaps cause they are a lesser mind...
2026-08-26 00:26:15 <lisbeths    > when I am working under gpl nobody ever threatens to sue me
2026-08-26 00:26:51 <hello-operator> If you want, we could just say we will sue if you don't relicense it XD
2026-08-26 00:27:02 <xentrac     > it isn't so fraught
2026-08-26 00:27:16 <xentrac     > there are just some crazy people out there
2026-08-26 00:28:19 <lisbeths    > yeah but youd be suing the free software foundation most likely
2026-08-26 00:28:34 <lisbeths    > so it makes me not htat worried
2026-08-26 00:28:43 <crc         > If you're that worried about lawsuits, register the code with your countries copyright office so you have a clear record of ownership and move on.
2026-08-26 00:28:55 <lisbeths    > there is a record of my authorship on github
2026-08-26 00:29:09 <crc         > GPL doesn't mean the FSF has anything to do with the code in question
2026-08-26 00:29:27 <lisbeths    > I understand but what I am saying is that I have little fear of someone suing me over gpl code
2026-08-26 00:29:41 <hello-operator> Yeah, I think you're really confused about copyright and how things work
2026-08-26 00:29:52 <lisbeths    > no *I'm* not other engineers are
2026-08-26 00:30:06 <lisbeths    > and many other engineers out there believe they can sue you over their fork of your bsd code
2026-08-26 00:30:11 <xentrac     > no, it sounds like you are too
2026-08-26 00:30:18 <hello-operator> It really does
2026-08-26 00:30:29 <lisbeths    > I dont have the tiem or the resources to deal with a lawsuit even a frivalrous one
2026-08-26 00:30:46 <lisbeths    > furthermore having programmers frequently think they should sue me causes social friction immensely
2026-08-26 00:31:05 <lisbeths    > it has wasted countless man hours of my life during the period when I switched to bsd license
2026-08-26 00:31:51 <hello-operator> Most people do not either have the resources either, but I think you were being trolled
2026-08-26 00:32:04 <jmiven      > threatening is free, I'd wait at least until they pay an attorney to worry about that
2026-08-26 00:32:12 <lisbeths    > yes but that still causes problems in a way that is difficult to ignore
2026-08-26 00:32:17 <hello-operator> Even a letter before action
2026-08-26 00:32:18 <jmiven      > yeah, exactly what hello-operator said
2026-08-26 00:32:37 <lisbeths    > an engineer who previously would have been cooperative with me now only sees me as a target for duscissions about lawsuits
2026-08-26 00:32:41 <lisbeths    > do you see how that could be problematic
2026-08-26 00:33:03 <hello-operator> I can see how that "engineer" might be problematic!
2026-08-26 00:33:34 <jmiven      > yeah, it's good news if that means you do not use more of that engineer's code
2026-08-26 00:35:13 <jmiven      > I do understand wanting to avoid the *stress* of the situation though
2026-08-26 00:36:01 <lisbeths    > its not the stress
2026-08-26 00:36:25 <lisbeths    > there is a common misconception amongst some coders that you can just fork a bsd project than sue the author
2026-08-26 00:36:41 <lisbeths    > and these people cant be convinced the inverse is true they are typically very beligerant on this point
2026-08-26 00:36:59 <lisbeths    > and it has ruined several productive friend groups I was collaborating with
2026-08-26 00:38:30 <KipIngram   > xentrac: Apparently it thinks it uses AI to analyze the page layouts.
2026-08-26 00:38:51 <crc         > 
ACTION wonders what kind of projects you've encountered these issues with

2026-08-26 00:38:53 <hello-operator> I've never come across this before, but these people are mistaken, and don't seem like good people to be hanging around.
2026-08-26 00:39:00 <KipIngram   > My ebook management software will do a pdf to epub conversion, but the results are CRAP, so apparently it's not a slam dunk problem.
2026-08-26 00:41:33 <crc         > PDFs can be very irritating to work with
2026-08-26 00:43:01 <crc         > E.g., one of the programs we use at work generates pdfs of orders, but the internal structure is a complete mess.
2026-08-26 00:43:20 <crc         > Parts of words get split up and assembled separately on the page, in a weird, inconsistent grid
2026-08-26 00:44:05 <crc         > (I've been slowly working on a parser for these for onenof my projects, but it's a very frustrating endeavor so far)
2026-08-26 00:44:12 <lisbeths    > well lets have a go at describing fastspungus to you
2026-08-26 00:44:32 <lisbeths    > I assently have a few op codes that compile to standard c
2026-08-26 00:44:41 <lisbeths    > for example one of the op code merely emits a switch statement
2026-08-26 00:44:54 <lisbeths    > one of the opcodes deals with doing some arithmetic on two variables then storing the result in another variable
2026-08-26 00:45:02 <lisbeths    > oneof hte op codes just emits a while loop
2026-08-26 00:45:14 <lisbeths    > one of the op codes just emits the end to a code block }
2026-08-26 00:45:25 <lisbeths    > one of the op codes emits the function prototype to a function
2026-08-26 00:45:40 <lisbeths    > one of the op codes calls a function and stores its output
2026-08-26 00:45:59 <lisbeths    > one of the op codes emits a return statment to return a value
2026-08-26 00:46:33 <lisbeths    > one of the op codes does nothing but typecasts a variable into another variable of a different type
2026-08-26 00:46:48 <lisbeths    > one op code just emits a dype declaration of a new name in namespace
2026-08-26 00:46:52 <lisbeths    > and assigns it to zero
2026-08-26 00:47:07 <lisbeths    > one op code emits a literl case statement in a switch
2026-08-26 00:47:15 <lisbeths    > one op code emits a literal break statement in a switch
2026-08-26 00:47:40 <lisbeths    > as far as I can see the very last op code emits a default statement for a switch
2026-08-26 00:47:55 <lisbeths    > so everything I have listed here should be possible in raw asm or in any standard c program
2026-08-26 00:48:13 <lisbeths    > and it is relatively abstract except for function calls
2026-08-26 00:48:35 <lisbeths    > it is what c coders refer to as a c-- type language
2026-08-26 00:49:06 <lisbeths    > it treats the basic c structures and operations as machine instructions in a high level portable asm-like lanuage
2026-08-26 00:54:09 <lisbeths    > oh ive forgotten
2026-08-26 00:54:34 <lisbeths    > one op code dereferences a piece of data at a pointer, the other one takes a piece of data and stores it inside the location at a pointer
2026-08-26 00:54:43 <crc         > What's the actual goal for this?
2026-08-26 00:54:50 <lisbeths    > 2 parts
2026-08-26 00:55:04 <lisbeths    > the first exercise is just to explore abstract machines that are very small for embedded systems
2026-08-26 00:55:17 <lisbeths    > but fastspunugs is an attempt to make that abstract machine compiled and not interpreted
2026-08-26 00:55:25 <lisbeths    > and it is a very good attempt if I do say so myself
2026-08-26 00:55:48 <xentrac     > KipIngram: are there different options for the AI?
2026-08-26 00:57:28 <xentrac     > 22:32 < lisbeths> an engineer who previously would have been cooperative with me now only sees me as a target for duscissions about lawsuits
2026-08-26 00:57:42 <xentrac     > sounds like you dodged a bullet; imagine if you'd gotten more deeply entangled with such a person?
2026-08-26 00:57:58 <lisbeths    > its just about that spefic person
2026-08-26 00:58:14 <lisbeths    > it has caused massive flame wars and hours of arguments in the fastlisp lawsuit proceedings
2026-08-26 00:58:39 <xentrac     > crc: some PDFs are better than others
2026-08-26 00:59:00 <bjorkint0sh > because don't you know it, PDFs are partially programs.
2026-08-26 00:59:31 <xentrac     > there is no fastlisp lawsuit
2026-08-26 00:59:41 <xentrac     > bjorkint0sh: no, PDFs are not partially programs
2026-08-26 00:59:53 <xentrac     > I mean I guess in the sense that they can embed JS and TrueType
2026-08-26 01:00:08 <xentrac     > but that's like saying that your filesystem is partially a program because you can store C programs in it
2026-08-26 01:00:46 <bjorkint0sh > yes. in the sense they're scriptable in JS.
2026-08-26 01:00:51 <xentrac     > PDFs are passive data structures, unlike PostScript.  That's the biggest difference
2026-08-26 01:01:06 <xentrac     > none of the PDF readers I normally use support the JS scripting
2026-08-26 01:01:23 <crc         > xentrac: true. I'm just stuck dealing with terrible ones most of the time
2026-08-26 01:01:45 <xentrac     > yeah.  PDFs that are terrible for parsing are pretty common because they still print fine
2026-08-26 01:02:09 <KipIngram   > xentrac: Don't know yet.  I just ran the basic command and it croaked on the RAM.  Maybe it has some options I can tune.
2026-08-26 01:08:16 <xentrac     > Did it download a giant file of weights from HuggingFace?
2026-08-26 06:14:17 <lisbeths    > okay so I have a compiled binary that is supposed to contain a pure lambda expression inside of it
2026-08-26 06:14:39 <lisbeths    > so I am hoping someone here who knows x86_64 can read how the lambda is implemented
2026-08-26 06:15:03 <lisbeths    > then we can settle the debate on the performance of haskells ghc compiler
2026-08-26 06:17:15 <lisbeths    > this appears to be the encoding of hte lambda in haskell https://www.irccloud.com/pastebin/SEkoRXO4/
2026-08-26 06:17:42 <lisbeths    > this is the lambda we encoded fun = (\x -> (\y -> ((\x -> (\y -> (\z -> ((z x y )))) x))))
2026-08-26 06:27:24 <lisbeths    > i believe it was xentrac who requested this https://www.irccloud.com/pastebin/NPs9xAr2/
2026-08-26 06:55:29 <xentrac     > it seems to be four instructions?  and a bunch of data
2026-08-26 06:59:14 <xentrac     > the data seems to be three objects whose first word is some kind of a type tag or code pointer (a cfa, you might say) followed by more words
2026-08-26 06:59:34 <xentrac     > one of the objects points to the other two, which seem to be strings, specifically "Foo" and "main"
2026-08-26 06:59:48 <xentrac     > I think the compiler has optimized away your lambda because you didn't use it
2026-08-26 07:03:40 <xentrac     > oh, and there's a fourth object, which is a closure for the four instructions and which doesn't seem to be referenced by anything else, although it is .globl, so visible outside this file
2026-08-26 07:53:13 <lisbeths    > xentrac they are asking you to com einto haskell rather htan me pasting back and forth responses
2026-08-26 07:53:16 <lisbeths    > into #haskell
2026-08-26 17:02:00 <X-Scale     > Peloa, aAh,/whepl dtman34