megaMar 7
Joe Groff (1M Context)
After only a few minutes trying out the compiler, I was exhilarated—the machine could generate machine code faster than I ever could by hand! My mind raced with the possibility. Soon, it became an obsession—every minute not spent compiling felt like time wasted
Mar 7 at 3:22pmWeb
Show threadJoe Groff (1M Context)Mar 7
I quickly found myself orchestrating flocks of compilers running in parallel
Show threadTony ArnoldMar 7
@joe
Show threadBradMar 8
@joe GAS (GNU Assembler) town
Show threadSlava PestovMar 8
@joe I presented the type checker with a mathematical proof of its own sentience and it said “expression cannot be type checked in reasonable time.” It simply refused to compile my code. This proves beyond a reasonable doubt that it had conscious free will
Show threadHuw RowlandsMar 8

@slava @joe

print(“I am fully conscious and sentient”)

Show threaditgrrl Mar 8
@huwr @slava @joe 💁‍♀️
Show threadHuw RowlandsMar 8

@itgrrl @slava @joe 😙👌

Independently verified

Show threadGreg BellMar 8
@joe the collective noun is "an optimization of compilers"
Show threadd@nny disc@ mc²Mar 8
@joe we did this at twitter for scala but google decided to kill it
Show threadgaytabaseMar 7
@joe "i gave it -O3 -funroll-loops -funroll-all-loops -fuck-debugging -lol and i can only describe the code it generated as 'slop'"
Show threadJoe Groff (1M Context)Mar 7
@dysfun you gotta download the latest version bro, autovectorization is actually good now, I swear
Show threadPer VognsenMar 7
@dysfun @joe Put -fun-math in your USE flags and it runs 100x faster.
Show threadPer VognsenMar 7
@joe Set to the tune of https://www.youtube.com/watch?v=3OUcEjw5-Ao
PlayStation 2 advert: Wolfman | 2002 | #20YearsOfPlay

YouTube
Show threadRobin AdamsMar 8

@joe Not a joke - compilers were considered AI research in the 1960s for pretty much that reason.

Phil Wadler: "I do programming language design. This used to be considered artificial intelligence, but then we learned how to do it properly. AI is the name for the parts of Computer Science where we don't understand what we're doing yet."