libera/#sicl - IRC Chatlog
Search
4:48:55
Bike
i could write up the bytecode stuff for ELS, but while it's working it's not really... done, to the level i'd like to be. doesn't seem interesting enough. i haven't even started on interconverting between bytecode and cleavir IR.
4:49:18
Bike
i have way too much stuff that's like three quarters done. the abstract interpreter has been sitting there unused for like a year now
5:01:03
beach
Maybe drmeister can be persuaded to hire another Common Lisp developer. Provided the company is doing well, of course.
5:01:49
beach
I find I have a hard time finishing a thing if I am the only one working on that thing. I need the help and inspiration of a second person.
5:04:16
Bike
that might have to wait a bit, unfortunately. we're still in the hemorrhaging capital investment stage. plus i suspect that the more valuable new employee would be a secretary
5:04:41
beach
If I had a multi-million € research grant, I would hire scymtym. I think his talent is largely wasted on that dayjob. :)
7:45:11
jcowan
beach: I have that exact problem too. I invented pair programming along with a co-worker back in the 1970s (of course we didn't publish so it doesn't count). The reason for it was that we only had one computer
7:46:13
jcowan
a few decades later we actually worked together again (he recommended me for the job) and we slipped right into the old work pattern, smooth as could be
12:37:02
karlosz
Bike: if you are serious about writing up the bytecode compiler to submit for ELS, i could collaborate with you on that. there's probably still some design rationales i have in my head i didn't actually commit to writing yet for the design and implementation
12:37:47
karlosz
Bike: in particular i feel like the fixup algorithm i cam up with to do 1-pass cell optimization and assembly comprsesion (i.e. "no holes") is actually somewhat novel
12:38:39
karlosz
i actually find the bytecode stuff we came up with to be a particularly sweetspot in compile-speed/execution speed
12:40:02
karlosz
also apropos hiring mroe developers: i would've liked to stay on a little longer to finish out the bytecode stuff and do the fasl stuff but yeah money issues
12:46:54
karlosz
beach: i was contracted on to help them get started with the bytecode stuff by designing the vm and general compiler architecture but as mentioned before funds are not so limited so i only stayed on until the bytecode stuff started working well
13:38:20
Bike
karlosz: alright, i'm up for that. i can start writing something up. also by the way, yesterday i finally implemented the pass so it uses save-sp restore-sp instead of entry exit when possible.
13:41:44
Bike
karlosz: https://github.com/clasp-developers/clasp/pull/1424/commits/ae40bdd32c0e9ff2d8d866017692ab450b00e818 is the lisp version - basically just does what you already wrote, but with the extra dynenv information so local unwinding is done correctly