2022-08-10 13:06:17 <KipIngram   > You know, every time I visit StackExchange while just nosing around for info on something (might be programming, might be physics, whatever) I wind up annoyed.
2022-08-10 13:06:38 <KipIngram   > Seems like a big part of the SE culture is to nitpick one another, accuse people of not answering questions, and so on.
2022-08-10 13:06:52 <KipIngram   > It's annoying as hell - how did those folks get so damn proud of themselves?
2022-08-10 13:07:15 <KipIngram   > It's just people sharing thoughts - get over yourselves.
2022-08-10 13:07:32 <KipIngram   > I used to try to participate, but I just gave up on it years ago.
2022-08-10 13:07:37 <KipIngram   > I just lurk these days.
2022-08-10 13:08:47 <thrig       > I have to keep my skills halfway dull somehow
2022-08-10 13:08:47 <KipIngram   > And inevitably there will be a response along the lines of "Why would you even want to know that?  You're an idiot to even be thinking that way."
2022-08-10 13:09:20 <KipIngram   > The main point of the response being to showcase how the responder is NOT an idiot and is SO SO MUCH better than the person asking the question.
2022-08-10 13:13:55 <KipIngram   > Anyway, it just doesn't come off to me as a group of people genuinely interested in helping one another - it's more like a bunch of folks that want to strut and think of themselves as some sort of elite.
2022-08-10 15:01:36 <KipIngram   > I'm working through a series of machine learning lectures as part of my work continuing education requirement.
2022-08-10 15:01:55 <KipIngram   > Neural networks are pretty cool, and touch on parallel computing enough to be fun in that respect too.
2022-08-10 15:02:16 <thrig       > neural networks: how to mostly get XOR right after burning a lot of CPU
2022-08-10 15:03:04 <KipIngram   > :-)
2022-08-10 15:03:22 <KipIngram   > Zarutian_HTC: Bitgrid would be a worthwhile way to think about implement NN's.
2022-08-10 15:04:23 <KipIngram   > Though it's probably going more "primitive" than absolutely necessary - there's an example where increasing the sophistication of the nodes instead of staying entirely generic would probably increase final performance vs. hardware bang for the buck.
2022-08-10 15:05:09 <KipIngram   > You know in NN you're going to need multiply/accumulate operations, so no need to use architecture flexibility to make those - you'd be better off hardwiring them.
2022-08-10 16:51:33 <vms14       > I'm not sure if I want a return stack
2022-08-10 16:51:58 <vms14       > never wanted it, but asked some guy cool features to add and he said 'tco'
2022-08-10 16:52:11 <vms14       > the return stack provides it
2022-08-10 16:52:33 <vms14       > and also allows for continuations
2022-08-10 16:53:51 <vms14       > also it triggers me to have one stack when this channel says "If you have two (or more) stacks you're welcome
2022-08-10 16:54:04 <vms14       > :/
2022-08-10 17:01:32 <eris[m]123  > why dont you want a return stack?
2022-08-10 17:01:39 <eris[m]123  > you will have one anyway
2022-08-10 17:03:58 <vms14       > eris[m]123: I'm  using perl to write a concatenative lang, I don't really see the need of a return stack, I mean, when I see a colon word (they are perl arrays) I just make a foreach and interpret all the elements.
2022-08-10 17:05:05 <vms14       > But I'm rewriting it cause it was very different and it had no "parser", it was just a bunch of regex converting the code into an array. Now I want to do it a bit more like forth
2022-08-10 17:05:27 <vms14       > that is, words are able to read from the source code and they're really part of the parser
2022-08-10 17:05:46 <vms14       > but the return stack can provide me tail cail optimization
2022-08-10 17:12:59 <MrMobius    > Can the words call other user defined words?
2022-08-10 17:15:06 <vms14       > they're  actually a list of strings xD
2022-08-10 17:16:25 <vms14       > I mean it's exactly the same as if I were reading code, a colon word is a list and code in this lang is a list
2022-08-10 17:16:51 <vms14       > when a colon word is found it will call interpret_list
2022-08-10 17:17:14 <vms14       > which is just a foreach that interprets all the atoms, if it finds a word there will call interpret_list again
2022-08-10 17:17:59 <vms14       > but for example looping forever will blow the perl stack
2022-08-10 17:18:11 <vms14       > a return stack avoids that
2022-08-10 17:18:46 <MrMobius    > so you do have a return stack then if the parser is recursive. you just don't see it
2022-08-10 17:19:09 <MrMobius    > parser/interpret_list
2022-08-10 17:19:13 <vms14       > it's the perl call stack, I have crossed recursion between interpret atom interpret list and execute word
2022-08-10 17:19:32 <vms14       > execute word calls interpret list if it's a colon word
2022-08-10 17:19:40 <vms14       > interpret list calls execute word
2022-08-10 17:19:50 <vms14       > (if it finds a word)
2022-08-10 17:21:35 <vms14       > and actually I'm rewriting everything from scratch so it's a good time to try with a return stack
2022-08-10 17:22:00 <vms14       > atm I only want make the words able to read code, like forth does
2022-08-10 17:23:13 <vms14       > but even if it's likely I don't need a return stack, nor even tco, now I could start having both
2022-08-10 17:24:56 <vms14       > what happens in the return stack when the INTERPRET colon word is executed for the first time?
2022-08-10 17:26:00 <vms14       > I mean, the return stack once forth starts has already the interpret word there
2022-08-10 17:26:52 <vms14       > so I just put it there at first, in my case they'll be arrays, I don't have memory allocation at all, so I would put the colon word + the index
2022-08-10 17:28:21 <vms14       > but maybe it's just conterproductive
2022-08-10 17:28:27 <vms14       > counter*
2022-08-10 17:28:53 <MrMobius    > You can always jump to the first word instead of a subroutine call if you don't want the address in the stack
2022-08-10 17:31:20 <vms14       > how I'm supposed to jump there
2022-08-10 17:32:04 <vms14       > I'd like to avoid the crossed recursion I have, so it's likely the return stack is the best option
2022-08-10 17:32:44 <vms14       > I'll try, but I'm not sure about what I should do at the start
2022-08-10 17:33:05 <vms14       > I just put the word INTERPRETER and a index of 0?
2022-08-10 17:33:55 <vms14       > I need to put the name + index cause there's no memory address
2022-08-10 17:34:31 <vms14       > I shouldn't put the whole array I suppose, so just the name
2022-08-10 17:35:04 <vms14       > so every item in the return stack would be an array of two items, the name and the index