2022-04-22 05:22:32 <eris[m]     > =/ is a lil state machine
2022-04-22 10:33:13 <KipIngram   > Please do tell - I'm still curious.
2022-04-22 10:33:48 <KipIngram   > It's pretty clear from the results I'm seeing on my tests that it has that kind of behavior.
2022-04-22 10:34:00 <KipIngram   > It's clearly not "just a calculation."
2022-04-22 10:37:47 <KipIngram   > There seems to be enough "rigorous consistency" in APL to really pique my interest - I've always liked things like that, because they sort of provide a "playground" for logical thinking.  But there also seems to be a dash of "arbitrary" (which can still be rigorous).
2022-04-22 10:38:22 <KipIngram   > Or maybe I just haven't figured out the things that seem arbitrary yet - haven't seen the "rule" yet.
2022-04-22 11:46:58 <sennomo     > APL is more consistent than standard math in some ways.
2022-04-22 11:47:11 <sennomo     > Precedence, for example, is always positional.
2022-04-22 12:12:24 <KipIngram   > Yes.  I picked that up from Iverson's paper.
2022-04-22 12:13:19 <KipIngram   > I'm not sure I always see the "relationship" between a symbol's unary meaning and it's binary meaning - I'd like it if that was extremely consistent.  Maybe it is and I'm just not seeing how yet.
2022-04-22 12:13:39 <KipIngram   > Sometimes they seem to fit together well; just not always.
2022-04-22 12:14:35 <eris[m]     > for example?
2022-04-22 12:14:40 <eris[m]     > anyway, state machine
2022-04-22 12:14:42 <KipIngram   > I knew someone would ask that.
2022-04-22 12:14:55 <eris[m]     > its more easily demonstrated with ones and zeros and scan
2022-04-22 12:15:23 <eris[m]     > basically, you can imagine the state for =\ being a single bit
2022-04-22 12:15:42 <KipIngram   > I can't give you a "best" example.  The only one that's coming to mind at the moment is one that I didn't have too much trouble swalling.  The ceil and floor operators - sometimes they truncate to an integer and sometimes they don't.
2022-04-22 12:15:56 <KipIngram   > Like I said, that one isn't too painful to me, but it IS a different behavior.
2022-04-22 12:16:15 <KipIngram   > I do see, though, that the binary behavior just isn't really applicable when there's only one argument.
2022-04-22 12:16:32 <KipIngram   > Max or min of a single number?  Could just "skip it."
2022-04-22 12:17:58 <eris[m]     > the behaviour of the =\ state machine is more accurately described with its not variant
2022-04-22 12:18:02 <KipIngram   > If I think of any others I'll let you know - I'm just recollecting "thoughts" I've had over the last few days.
2022-04-22 12:19:44 <eris[m]     > 
ACTION uploaded an image: (4KiB) < https://libera.ems.host/_matrix/media/r0/download/chat.unix.lgbt/018929da308a460f1c9e3c384542407fa6d98572/image.png >

2022-04-22 12:19:53 <eris[m]     > essentially
2022-04-22 12:20:08 <eris[m]     > if it recieves a 1, it changes state
2022-04-22 12:20:13 <eris[m]     > and starts producing 1s
2022-04-22 12:20:19 <eris[m]     > or stops producing them
2022-04-22 12:20:50 <TCZ         > 'LEAVE' should be used only with definitive loops?
2022-04-22 12:21:46 <eris[m]     > this behaviour produces the sirpenski triangle in certain scenarios
2022-04-22 12:23:10 <eris[m]     > 
ACTION uploaded an image: (29KiB) < https://libera.ems.host/_matrix/media/r0/download/chat.unix.lgbt/d5ea9d92a878a03d02d919384911e3cda346f8be/image.png >

2022-04-22 12:23:13 <eris[m]     > this is of course, K
2022-04-22 12:24:00 <eris[m]     > APL doesn't have an analogue for monad\ in K, which is something called scan-fixpoint
2022-04-22 12:28:01 <KipIngram   > Oh wow - that second image is really something.
2022-04-22 12:28:20 <eris[m]     > yea
2022-04-22 12:28:34 <KipIngram   > To bad it won't run on mine.  :-)
2022-04-22 12:28:43 <eris[m]     > scan-fixpoint is a lovely primitive
2022-04-22 12:28:48 <KipIngram   > See, that's what I mean, though, when I say "I sense huge potential."
2022-04-22 12:29:11 <KipIngram   > Being able to get at large structures like that, in such a simple way - that's cool.
2022-04-22 12:29:59 <eris[m]     > have you encountered ⍣ yet?
2022-04-22 12:31:33 <eris[m]     > its called the power operator
2022-04-22 12:31:45 <eris[m]     > give it a numerical right arg and it'll do something that many times
2022-04-22 12:31:55 <KipIngram   > No, I was tinkering with ⍨ and ¨ last night, though.
2022-04-22 12:32:00 <eris[m]     > however
2022-04-22 12:32:03 <eris[m]     > if you pass it ≡
2022-04-22 12:32:08 <eris[m]     > as a right operand
2022-04-22 12:32:12 <eris[m]     > it gets very fun
2022-04-22 12:32:19 <eris[m]     > it'll do something until it stops changing
2022-04-22 12:32:26 <KipIngram   > :-)
2022-04-22 12:32:45 <KipIngram   > Well, this definitely gives me new angles to look at my =/ tinkering from now.
2022-04-22 12:34:22 <eris[m]     > 
ACTION uploaded an image: (2KiB) < https://libera.ems.host/_matrix/media/r0/download/chat.unix.lgbt/97d6f621ae8082a971c71d1393dfa289fe45d06d/image.png >

2022-04-22 12:34:25 <eris[m]     > i really wish APL had currying
2022-04-22 12:34:33 <eris[m]     > add one 7 times
2022-04-22 12:52:05 <Zarutian_iPad> been looking into the complexity of implementing a BitGrid cell
2022-04-22 12:54:29 <Zarutian_iPad> so basically 16 input MUX whose selector lines are the NEWS von Neuman neighbours cell output
2022-04-22 12:55:11 <Zarutian_iPad> each input of the mux being a four bit bus and hence the output too
2022-04-22 12:56:19 <Zarutian_iPad> those mux inputs come from a 16x4 sram memory
2022-04-22 12:56:49 <Zarutian_iPad> so 64 sram cells
2022-04-22 13:04:10 <KipIngram   > Yeah - 4 output bits for each of 16 possible input bit patterns.  One 64-bit cell for configuration.
2022-04-22 13:04:45 <KipIngram   > I figure that's how it would look to the outside - one 64-bit word in your "configuration space," one word for each cell.
2022-04-22 14:06:32 <KipIngram   > Zarutian_iPad: I felt like the most straightforward way to make a BitGrid chip would be to start with an existing RAM chip design, where you already have a logic structure that "houses" all the config bits - then just "spreread it open" a little and drop the selectors in there and interconnect them.  It would be four 16-1 selectors for each cell.
2022-04-22 14:07:06 <KipIngram   > That makes it sound "easy" - I'm sure there would be a TON of "sub-digital" issues to cope with.  But that's about the size of it as far as I can see.
2022-04-22 14:21:15 <KipIngram   > eris: I'm still really tickled with that second image you posted earlier.  The one with the fractal pattern.
2022-04-22 14:29:18 <eris[m]     > :)
2022-04-22 14:38:43 <KipIngram   > eris: So in this expression:
2022-04-22 14:38:49 <KipIngram   > 'λ' (4.5 2.2) 'string' = 'r'
2022-04-22 14:39:10 <KipIngram   > everything to the left of the = sign forms a list and that list is the left operator of =?
2022-04-22 14:39:21 <KipIngram   > Because stranding binds more tightly than function application?
2022-04-22 14:39:53 <KipIngram   > And the the = 'r' is applied to every scalar in that left list structure, because of "scalar pervasion"?
2022-04-22 14:40:37 <KipIngram   > So you wind up with 0  (0 0)  (0 0 1 0 0 0)
2022-04-22 15:01:06 <Zarutian_iPad> KipIngram: yeah but the question is how that configuration space is accessed
2022-04-22 15:02:49 <Zarutian_iPad> is it via row and column select and word buses or hilbert curve snaked spi/jtag ?
2022-04-22 15:04:27 <Zarutian_iPad> the former allows for usage of BitGrid blocks as memory but can eat up die space both polysilicon and metal layers
2022-04-22 15:05:47 <Zarutian_iPad> hmm.. no I suspect that they are mostly equiv in that regard as spi/jtag needs extra sram cells for the spi/jtag chain
2022-04-22 15:07:19 <eris[m]     > stranding binds tighter yes
2022-04-22 15:07:41 <Zarutian_iPad> so doing the former is both cheaper and more versatile
2022-04-22 15:09:50 <Zarutian_iPad> heck one could even use the same address decoding for readonly debug access to each BitGrid cells inputs
2022-04-22 16:04:29 <KipIngram   > I assumed it would be accessed by the ram chip's existing design.
2022-04-22 16:04:39 <KipIngram   > That's why I wanted to "start with a RAM chip."
2022-04-22 16:04:59 <KipIngram   > All of the circuitry to WRITE to the config space is exactly what's already on the chip.
2022-04-22 16:05:10 <KipIngram   > the selectors would be connected directly to the data bits themselves.
2022-04-22 16:05:44 <KipIngram   > Rather, the 16 inputs of the selector would come from the existing data cells in the  RAM - the four select inputs would come from the neighbors selector OUTPUT.
2022-04-22 16:06:19 <KipIngram   > I'd leave all of the stuff that lets you read and write the RAM exactly as it is, logically.
2022-04-22 16:06:31 <KipIngram   > We just need to make room for the selectors and interconnect them.
2022-04-22 16:07:48 <KipIngram   > Heh.  "Just."  I shouldn't say it like that - LOGICALLY it's a "just," but so far as engineering goes it might be pretty hard.  I don't know about delicate RAM design issues enough to say.
2022-04-22 16:08:14 <KipIngram   > eris:  I'm reading this right now, and so far it's working really well for me:
2022-04-22 16:08:16 <KipIngram   > https://mlochbaum.github.io/OuterProduct/commentary.html
2022-04-22 16:10:34 <KipIngram   > Uh oh.  I found something I'm not immediately understanding.
2022-04-22 16:10:46 <KipIngram   >       3.5 ⌊ 4.7
2022-04-22 16:10:47 <KipIngram   > 3.5
2022-04-22 16:10:49 <KipIngram   > I get that.
2022-04-22 16:10:58 <KipIngram   > two-operator ⌊ is "min."
2022-04-22 16:11:22 <KipIngram   > Oh, wait - never mind.  I misread - it makes sense.
2022-04-22 16:28:18 <eris[m]     > mlochbaum also made an array lang :)
2022-04-22 16:32:29 <KipIngram   > So, I get why 6 +/ 1 2 3 4 5 "can't do anything," since there aren't 6 items in the array to window over.  But it returns something.  Then 7 +/ 1 2 3 4 5 throws an error.
2022-04-22 16:32:41 <KipIngram   > What's special about "one too many"?
2022-04-22 16:33:49 <KipIngram   > I guess windowing is a feature of function created by having / operate on a function?
2022-04-22 16:33:57 <KipIngram   > "any function"
2022-04-22 16:40:46 <KipIngram   > Operator actions come before right to left evaluation?
2022-04-22 16:41:33 <KipIngram   > This: ÷⍤+/ wound up behaving like (÷⍤+)/ as far as I can tell.
2022-04-22 16:42:22 <eris[m]     > operators bind tighter yea
2022-04-22 16:42:32 <eris[m]     > it goes stranding then operators then rtl
2022-04-22 16:45:58 <KipIngram   > Good - I'm starting to "see" some of these things.  :-)  Making more sense.
2022-04-22 16:47:57 <KipIngram   > Oh, HA HA.  Precedence order was the very next slide.
2022-04-22 17:33:58 <eris[m]     > hahaha