Adrian SampsonWe read the 1989 Self paper for #cs6120 today, and everybody wanted to talk about this sentence:
> Researchers seeking to improve performance should improve their compilers instead of compromising their languages.
It made me think about meeting Dave Ungar at the first SNAPL. He made it *real* clear that this was still an animating philosophy for him. To him, it was a moral issue: screw all these nitpicky, overbearing languages that squash creativity and subjugate humans to what the machine wants! Performance is our problem (PL/compiler people’s); it’s unseemly to foist it onto everyday programmers.
I also tried to think about, aside from anyone’s personal opinion, what the industry “thinks” about this. I actually think it’s pretty stark:
* To what degree has the industry invented nice, clean, PL-nerd-preferred languages with performance as a first priority and gotten people to use them?
* To what degree has it poured billions into heroically salvaging performance from languages people already want to use, regardless of their messy, dynamic semantics?
Obviously some of both, but it seems clear the Ungars have won if you measure by sheer lines of code.
Bill Harris@adrian as someone with lots of early years in PL research and a few recent years in Google eng, I've been most amazed at how they've poured money and time into making C++ as (internally) safe and readable as they practically can
@overfull_hitbox I agree; this is truly impressive. (now when is Rust-in-google3 coming? :)