logs archiveBotHelp.net / Freenode / #ada / 2015 / August / 16 / 3
charlie5
instead of doing a straigth-out GA, i've tried to model an environment in which evolution can take place 'naturally'
Lucretia
is he pissed again?
wtf are you on about?
and you've never heard of bazza white?
charlie5
ie terrain + rain + sun + organism ... organism is a collection of organs, at least one of which is a reproductive organ
Visaoni
oh geez man. you /started/ with that?
and I thought I was crazy
charlie5
organs can be leaves, branches, gonads, guts, buds (and whatever i can think of next :) )
organs are hieracical trees, so can evolve into 'complex' organisms
Lucretias thesis => 'is he pissed again?'
... and howdy :)
i've heard the name but naught else
Visaoni: part of it is hobby/interst ... it alos acts as a decent testbed for underlying components/;ibs
Lucretia
(Action) is lost re thesis
Visaoni
I think the "Lucretia thesis" is that charlie is pissed
(again)
charlie5
it exercises sprite creation/deletion and sprite expansion (amonsgt others things)
Visaoni
got it
sounds fun
if /really/ want to go crazy
charlie5
ell, not that pissed ... more bored
Visaoni
is add sensor organs that go through a neural net to become signals to other organs
which would presumably be able to do things like move
charlie5
heh, funnny you should mention that
nnets are another interest
and comining the 2 seems a natural 'next step'
combining
Visaoni
yep
charlie5
it's sort of been on my mind for years, in fct
fact
Visaoni
have also written a few toy nnets (and some of the toy genetic algs were because "fu*k back propogation")... never did get around to hooking them up to anything
but I wrote them, which was the part I found more fun :P
charlie5
:)
i think 'hooking them up' might be the really fun part :)
but, yeah, playing with them was fun alright
Visaoni
ha
charlie5
ada has some support for nnets
Visaoni
I have this habit of turning everything into a library and then never actually using the "library" I end up with
oh? what do you mean?
charlie5
not natively
(Action) tries to recall pkg
mm, it's part of a set of components
pragmarc
http://www.adapower.com/adapower1/pragmada/
Lucretia
which part?
charlie5
well, pragmada
unless it's been updated, the nnet was via a generic which allowed you to set the net params
i thin ki added a layer on top to allow saving of the nets
been ages now
Lucretia: package PragmARC.REM_NN_Wrapper
Lucretia
k
Visaoni
huh, looks like an interesting collection of stuff
Lucretia
(Action) has just been watching vulkan video...all checked are in a debug layer which is supposed to be removed in the final application, it doesn't check anything and why should it if the data is correct?
charlie5
Lucretia: i haven't got to sdlada intergration into mmi yet ... had some other probs
... like parrots
:)
Lucretia
k
Visaoni
what if you put out like... a parrot nesting... box thing?
charlie5
"Visaoni: which would presumably be able to do things like move" -> adding muscles to organs joints was next step, then next was to add brain (nnet) organ to control the nervous system (ie the muscle contractions)
Visaoni: i'm actaully working on an anti-parrot-personell auto feather target chirp triggered tri-triangulated twerper terminator
(well, i did say i was bored)
Visaoni
sounds complicated
(both things!)
charlie5
the 2nd was BS, of course
Visaoni
indeed :)
charlie5
i'd like to try the 1st sometime tho
need to get the base libs done better 1st
Damaki
!last
oliver117
!last
sparre
Shark8: Have you considered the cost of being able to generate an intermediate representation of compilation units? It is for all practical purposes equivalent to the cost of writing a complete compiler.
(code generation isn't free, but AFAICS it is much easier than writing a front-end)
nerdboy
(Action) switching from ope source graphics to headless armv5...
*open source even
Shark8
sparre: Why would it be the [whole] cost of a compiler? Aren't some of the biggest effort-sinks in the compiler the optimizations? with an IR we don't really want 'optimizations' [per se; there might be reasons to have special constructs, though] so much as a "canonical form", right?
Lucretia: How's the SDL binding coming along? Last I'd heard you thought you were getting *really* close to [essentially] finished.
Lucretia
not finished, taking a break
Shark8
Lucretia: Understandable. // I was disappointed that you weren't able to submit that generic graphics-engine thing for the contest.
Lucretia
surfaces need to be implemented (similar to textures), more texture access methods (based on pixel format) need to be done, I've only done RGBA8888 so far and YUV stuff needs doing
it wasn't a graphics engine, I was going to implement marco monster's car physics
or at least, attempt to
I was more pissed off that more people didn't enter
the linkedin Ada group were boasting of having 2000+ people in it, so I posted "feels more like 10" because no fu*ker ever posts there and if they have that many, then surely they should've done something, but no.
sparre
Shark8: You *can* of course spend as much time as you like on optimising a code generator, but I was considering a "minimal effort" compiler.
Shark8: Take a look at the budget for the first versions of GNAT. That is (practically) only Ada-to-IR effort.
Ada 2012 is somewhat more complicated, but an Ada 2012-to-AIR compiler would probably not be an order of magnitude more costly.
Lucretia
what was the budget?
Shark8: I wasn't getting at anyone in particular btw
Shark8
sparre: What was their budget?
Lucretia: I agree on the 2k subscribers / lack of involvement.
Lucretia: Sorry about confusing the graphics and physics stuff. (Probably linked it to SDL mentally.)
sparre: Isn't it some of that difficulty (GNAT IR) due to GCC's already extant IR? And what purposes does GCC's IR strive for? I don't think that it was intended for anything other than just plain HW abstraction (for GCC's multiple backends).
« prev 1 2 3 next »