I cross-compiled for ARM in 2012.... using gcc. I suspect that hasn't changed much, and Gerbil is somewhat not in the picture as it does not compile to machine code, it transpiles to Gambit Scheme Code. Gambit "can" compile to C, and it uses gcc a lot of the time, or whatever C compiler you have around.
3 replies
To repeat: Gerbil does not compile to x86 or ARM, it outputs scheme code. Have a look at ~/.gerbil/lib/static/ to see that compilation for ARM is way way way outside of what Gerbil is capable of. It's a Meta-Scheme :)
We already have a working
cond-expand, yeah?
that's very nice
we'll have to do some trickery to espand to gambit cond-expand or learn about
compilation-target
because we handle cond-expand ourselves
26 replies
drewc@nyx:~]$ cd `mktemp -d`
[drewc@nyx:/tmp/tmp.pwv69MS6Sl]$ cat > testme.ss <<'EOF'
(##inline-host-declaration "console.warn('This is JS!');")
EOF
[drewc@nyx:/tmp/tmp.pwv69MS6Sl]$ gxc -v -d "`pwd`" -s -S testme.ss
compile testme.ss
compile testme
compile /tmp/tmp.pwv69MS6Sl/testme__0.scm
compile /tmp/tmp.pwv69MS6Sl/testme__rt.scm
compile /tmp/tmp.pwv69MS6Sl/testme.ssi
[drewc@nyx:/tmp/tmp.pwv69MS6Sl]$ cp testme__0.scm testme.scm
[drewc@nyx:/tmp/tmp.pwv69MS6Sl]$ gsc -verbose -exe -label-namespace 'a' -target js testme.scm
Parsing:
"declare"
"expr"
Compiling:
Compilation finished.
cat testme.js > "testme.o"
cat testme_.js > "testme_.o"
echo "#! /usr/bin/env node" > "testme"
cat testme_.o testme.o "/nix/store/wxfj3jryhqc4r3c19757wj75qvripsa8-gambit-unstable-2021-04-05/gambit/lib/_gambit.js" >> "testme"
chmod +x "testme"
[drewc@nyx:/tmp/tmp.pwv69MS6Sl]$ ./testme
This is JS!du testme -h => 4.8M, that includes pretty much all of gsi and libgambit, only smaller.
Shortly, 5.5M to 6.66m or so will have most of Gerbil's run time and interactive.
gzipped will be less than 1 meg most likely.
JS rules, can create smaller 'execs' then C :P
Ok, ok, but how big is the linux kernel and glibc?
drewc @drewc has really gotten back into JS these past couple seasons :P
I just want to say thanks for all the cool stuff you guys are working on. Sounds like writing full SPAs in gerbil might be an option in the future!
Have a look here for the
npm packages: https://github.com/drewc/gxjs
Don't try to install them, they likely won't work for you at all and they are old, but lol:
That's what does the loading I did at the command line above, and then transpiles the JS code into something smaller more readable and lexically declared for all variable ... no globals.
But, lol, though it works great and no spider web of globals, the issue with having the Gambit runtime use "object Environment Records" rather than "declarative Environment Records" is performance. For my tests,
with(RTS){ ...}, where RTS is an object I specifically crafted to have the correct properties defined upon first set and always "has" an _ prefixed symbol ,(fib 20) took about 1 second.
When they are all declared, 0.2 seconds.
500% difference is well worth the
_pollution, which I am thinking of replacing with ᵍpollution, even though it adds 18k to the compressed runtime size 'cause of the extra bits and bytes.
Anyone checked out Goblin? Is Racket trying to compete with Gerbil with actors? https://docs.racket-lang.org/goblins/
I don’t think racket has a reason to compete with anything. actor models aren’t exactly novel among schemes either (check out guile’s 8sync). most likely because the actor model predates scheme [citation needed]
I think I read on c2 wiki about the actor model inspiring scheme but I can’t find it right now. generally, as long as you have first class continuations and tco you can implement an efficient actor system, which is probably why most schemes have one. i’m not an expert though
Goblins implements an Object Capabilities (OCaps) system (like E) and is more oriented towards decentralized trustless communication https://docs.racket-lang.org/goblins/captp.html. The original ActivityPub author is making it. I think there are some good ideas and designs there that would likely be implemented better on top of Gerbil/Gambit than Racket. I think they are using Racket just because it is the "Python" of schemes and is becoming popular.
The larger project is called "spritely" https://gitlab.com/spritely
I think I read on c2 wiki about the actor model inspiring scheme but I can’t find it right now. generally, as long as you have first class continuations and tco you can implement an efficient actor system, which is probably why most schemes have one. i’m not an expert though
@drewc educated me on this a while back. Basically Steele and Sussman tried to implement both the actor model and lambda calc and it ended up being the same. Its on the regular wikipedia page now I think. Individual actor systems seem to be wildly different though. I gave up trying to get a universal definition of actors...
Having said that, SCHEMERS might be more universal if we were allowed 8 letters in a file name :P
In short: If you want to enter the Turning Tarpit that is your choice, and if you are a student that's a tarpit you are likely in and yet do not yet know it.
And the sarcastic closing remark: What match/game are Racket and Gerbil in competition over? What leg are we on, and do away goals count?
drewc @drewc is watching the pre-season of Vancouver's WCFC right now and we're up by 1 after 54 minutes ... so is pretending he knows anything about computing while watching us press
3-1!
drewc @drewc ... and we're Vancouver, Vancouver Whitecaps FC! We're the finest football team, Canada has ever seen.
@belmarca Our coach is still Marc Dos Santos ... who not only has a good first name (right?) but was born on a certain royal mountain and coached L'impact when they one the USL cup in '09. To make it on topic, that was the second last time I was in Montréal. The last time, in '19, was for Gambit's 30 :)
To make it even more personal, on my certificat baptismal mes first nom is "André" ... Marc-(Dos Santos)André(Michel Groleau-Crampsie)! :D
@belmarca Our coach is still Marc Dos Santos ... who not only has a good first name (right?) but was born on a certain royal mountain and coached L'impact when they won the USL cup in '09. To make it on topic, that was the second last time I was in Montréal. The last time, in '19, was for Gambit's 30 :)
lol ... tried to s/one/won there
So many chat apps... should go back to the matrix.
@drewc Haha. If your name is Groleau then check this out, if you haven't seen it already :laughing: https://www.youtube.com/watch?v=uX7T3F9En1s