5 ms·

I'm hoping that the Rust compiler is good enough to compile this back to a bit-shift instruction?

by chrisseaton 5y ago

jjgreen 5y ago
That would be impressive, could someone check?
- SekstiNi 5y ago
  Unfortunately, but expectedly, it cannot reason about it [1]. Even if it could I'm not sure it would be allowed to elide the allocation(s). [1] https://godbolt.org/z/63xEcfaEE https://godbolt.org/z/63xEcfaEE
  - jjgreen 5y ago
    Pity, if it ever does then I'll put aside my distaste for the whole "curl <url> | sh" thing and learn the language.
  - h2odragon 5y ago
    The metaphysics of that are going to be interesting. "Do what I mean" programming meets all the stories about why one shouldn't accept bargains with fairies or djinn. "I can't let you use `strcpy()` there, Dave" is the least of it.
- coldtea 5y ago
  No need to check, that would be almost impossible, unless they could read the programmer's mind for the intent...
  - AnIdiotOnTheNet 5y ago
    Which I think is the argument behind declarative languages. If we can encode the intent precisely, then at least in theory the compiler can find the optimal solution for the given architecture and other constraints. Of course in practice that probably requires strong AI to accomplish.
yholio 5y ago
You are surely joking. It would have to make a static inference that a numeric variable is converted to a string variable, then processed as a string to base 2 representation, then truncated as a string that can only contain the '1' and '0' characters, then converted back into binary.
- chrisseaton 5y ago
  > It would have to make a static inference Not necessarily a static inference... could be speculative. Code folding golf is a bit of a hobby of mine. For example you can optimise through the string-integer conversion by storing a special string that remembers it was originally an integer and returning that value again instead of re-parsing. I work on a compiler that does that one for real, for example. I don't seriously think Rust does this right now, but would be a challenge to try.
reitzensteinm 5y ago
Could Graal? That would be pretty impressive.
spaetzleesser 5y ago
That would go way further than existing optimizers can go.
- dev_tty01 5y ago
  And a lot further than we should want them to attempt to go.

I'm hoping that the Rust compiler is good enough to compile this back to a bit-shift instruction?

jjgreen 5y ago

That would be impressive, could someone check?

SekstiNi 5y ago
Unfortunately, but expectedly, it cannot reason about it [1]. Even if it could I'm not sure it would be allowed to elide the allocation(s). [1] https://godbolt.org/z/63xEcfaEE https://godbolt.org/z/63xEcfaEE
- jjgreen 5y ago
  Pity, if it ever does then I'll put aside my distaste for the whole "curl <url> | sh" thing and learn the language.
- h2odragon 5y ago
  The metaphysics of that are going to be interesting. "Do what I mean" programming meets all the stories about why one shouldn't accept bargains with fairies or djinn. "I can't let you use `strcpy()` there, Dave" is the least of it.
coldtea 5y ago
No need to check, that would be almost impossible, unless they could read the programmer's mind for the intent...
- AnIdiotOnTheNet 5y ago
  Which I think is the argument behind declarative languages. If we can encode the intent precisely, then at least in theory the compiler can find the optimal solution for the given architecture and other constraints. Of course in practice that probably requires strong AI to accomplish.

yholio 5y ago

You are surely joking. It would have to make a static inference that a numeric variable is converted to a string variable, then processed as a string to base 2 representation, then truncated as a string that can only contain the '1' and '0' characters, then converted back into binary.

chrisseaton 5y ago
> It would have to make a static inference Not necessarily a static inference... could be speculative. Code folding golf is a bit of a hobby of mine. For example you can optimise through the string-integer conversion by storing a special string that remembers it was originally an integer and returning that value again instead of re-parsing. I work on a compiler that does that one for real, for example. I don't seriously think Rust does this right now, but would be a challenge to try.

reitzensteinm 5y ago

Could Graal? That would be pretty impressive.

spaetzleesser 5y ago

That would go way further than existing optimizers can go.

dev_tty01 5y ago
And a lot further than we should want them to attempt to go.