@PhiNotPi Please do.

That would be so funny.

I mean, we could submit just the assembly...?

@quartata Yeah, but tell the story.

I bet you'll get one of the top voted questions, and no answers for ever.

Also, I popped in to say that this is probably PhD-level work and I'd be surprised if it wasn't publishable in some academic journal.

And that you guys inspired me to make my first tweet since ever.

@Hosch250 I might

@Hosch250 I have suggested the idea to Phi before.

@El'endiaStarman I would be interested if I knew what it would involve / what we would actually put in the paper.

You put everything in the paper.

Basically, you put the problem, how you solved it, and what was interesting about it.

^^ That's what I was thinking too. We'd probably mostly just stick the challenge answers end-to-end with a little glue in between.

That + change of writing style

It would probably involve a lot of writing and editing, and perhaps a few going back and forths with papers. If you have a hardware/comp sci professor at your college, talk to them.

Yeah, there'd definitely have to be some harmonization between parts.

Writing a paper is a lot of work, but you guys already did probably 3/4th of it (the research).

but :| shouldn't you try to golf it first :P

No :P

The guys over at conwaylife suggested this, if it doesn't need to be Spartan or a constructor it can be made much smaller

> Conditional jumps are relative, rather than absolute

Honestly, practically the only place to golf it now - and make significant gains - is the OTCA metapixel. It'd be fairly easy to rewrite that particular part of the paper if needed.

Well, if you still want to keep using the current design

Yeah, most of it seems to be quite sound. The OTCA metapixel is basically the only part that people looked at and said "y'know, that could be better".

But :( why don't you want to try a more compact computer

I'm extremely doubtful we could get anywhere close to 60%, maybe 75% of the current size. On the other hand, Dave Greene has said a few times that they think a specialized metapixel, or one better suited for this work, could fit into a 512x512 bounding box, maybe a 256x256 bounding box with a lot of blood, sweat, and tears. That'd reduce the whole computer to 25% of its current size.

25% * 25%

but isn't this already way smaller

Whoops, yes, you're right.

@ASCII-only How long would the tapes have to be to construct the computer?

if QFTASM can be adapted to work in it (doesn't seem easy, especially with the relative jumps) then it should work fine :P

@El'endiaStarman construct the computer?

Oh, I was thinking of using that constructor to build the computer we have now.

Oh haha the tapes would be massive (actually wait it wouldn't be able to construct it)

Well, y'know, you can always take a stab at it yourself.

Yeah, I'm going to try and do that (once the people over at the Discord reply haha)

What discord is this?

The ConwayLife one

You guys have enough content for several papers.

The gates alone is probably enough for a paper.

Then you can discuss VarLife, and the wires, and building the "computer" components.

I don't either, but I'm sure this is publishable.

I'd go for IEEE :)

Who wouldn't?

And they say how many it is on their online journal...

And they can count page numbers for the physical one, if they are one of the ones who still publishes real ones.

Oh, lol.

Pages are integers, though.

They must have so many useless blank pages come out of their printer.

@Mego Integer addition results in a float? That's new...

@Mego Please don't remind me. I use it all day at work.

We are planning on going to TypeScript, thank goodness.

I was thinking C# at first (my primary language).

Yeah.

Thanks.

We use Typescript as one of the languages at my work too. I can definitely see how it'd be invaluable for a company product.

And no such thing as a date that is for sure a date, and not a string or an "int" or something.

LOL.

@Mego Did you get a chance to peek at the backend yet

@El'endiaStarman I tried TypeScript once and was rather frustrated by the compiler itself

@Mego If I'm not mistaken that is @DaveGreene. It was linked earlier

sorry about your eyeballs, it's a little sparse on comments

@Mego Oh hey, most recent post is today!

Yeah, Dave Greene did link it in here a couple days ago or something.

I can kind of explain the overall situation if that would help, unless you're scared by a particular part

Ah so you were looking at the machine description

I'll put comments in the patterns since how that works isn't immediately obvious, but the match_* takes whatever argument goes there and gives it a number. It's like a capture group in a funky LISP regex

@Mego go ahead if you want

The worst part is probably qftasm.c. The way that works is it's basically initializing a poor man's vtable with functions about the architecture

All of this was written before the compiler was ported to C++...

@Mego (or @PhiNotPi potentially) The easiest thing for you to try your hand at is finishing qft_cbranch, which takes in a code representing the kind of comparison (see the first switch) and another number representing the order and type of the arguments (address,address;address,immediate;immediate,address) and returns a string of QFTASM that does a conditional jump. I have greater than, less than and not equals but am missing equals, greater than or equal and less than or equal

In the string %1 and %2 will be replaced with the arguments and %3 the jump position

use address 5 for scratch (I have it as a constant CC_ADDRESS), take a peek at the greater than example

don't sweat it if you don't want to try it though

that's fair which is why I also pinged Phi aka QFTASM swag lord

> QFTASM swag lord

you golfed down my multiplication I am in your debt

if you can't C ask Mego

this part basically looks like Java though other than the macro concat

@quartata It's great that this is also a pun.

yes

btw it's at the top of qftasm.c

I have the switch with the condition types laid out so you don't need to worry about the scary GET_CODE RTX crap

However if you want to check the value of an immediate you'll need to use INTVAL which is another magic macro. I use it check for special cases like x < 0

I don't think these ones have any possible optimizations though

ooh I just noticed a typo

@quartata I'll take a look

So I actually did find a GCC bug

A machine description with no expander patterns won't compile

I had to add a dummy: (define_expand "*dummy" [] "" "")