I wrote a small benchmark in C (compiled O3), Java, Python2, Python3, PyPy, and Perl. I kept it small so that I could make the remarkably similar between the languages. The benchmark basically looked for Prime numbers to an arbitrary target. What really shocked me is that the PyPy version was the fastest, besting even the compiled C version. Obviously, a prime finder is a couple of tightly-nested loops, and tight loops is where PyPy excels (from what I understand). Still, though, for an interpreted language to beat compiled machine code is incredible.
i think this project is excellent, most will continue to use cPython sure, but to see that we can take our experience to a high performance environment is just brilliant, i am not sure why people seem to feel this is a bad thing, I am really impressed by the image processing.
like they say: "it's not a real solution, it's just a toy"... applies very well to pypy. Simply because whenever you need performance you write a C library. The numpy guys even reimplement numpy so that it's only 10 times slower.... wow great. Then just allow to use C libraries... but no that seems against their belief : the belief that everything should be written in python and be fast. Where belief lives open-mind dies. In the end, just use the standard Python interpreter with (optimised in C whenever required) libraries and consider pypy as their authors do: a toy... at least until they grow up and stop re-implementing everything for the sake of running python code only.
That tie almost reaches his knees!
I wrote a small benchmark in C (compiled O3), Java, Python2, Python3, PyPy, and Perl. I kept it small so that I could make the remarkably similar between the languages. The benchmark basically looked for Prime numbers to an arbitrary target. What really shocked me is that the PyPy version was the fastest, besting even the compiled C version. Obviously, a prime finder is a couple of tightly-nested loops, and tight loops is where PyPy excels (from what I understand). Still, though, for an interpreted language to beat compiled machine code is incredible.
Armin did not panic, he's using his terminal as a presentation tool! That's just so COOL B)
i think this project is excellent, most will continue to use cPython sure, but to see that we can take our experience to a high performance environment is just brilliant, i am not sure why people seem to feel this is a bad thing, I am really impressed by the image processing.
I LOVE THIS GUY!!!
he panicked i guess!
best at 11:44 and 12:44
of course of course! because because!! XD ..
hilarious..
all in all great talk!
brilliant talk!! :D .. and the guy at 20:30 is awesome too btw!
like they say: "it's not a real solution, it's just a toy"... applies very well to pypy. Simply because whenever you need performance you write a C library.
The numpy guys even reimplement numpy so that it's only 10 times slower.... wow great. Then just allow to use C libraries... but no that seems against their belief : the belief that everything should be written in python and be fast.
Where belief lives open-mind dies. In the end, just use the standard Python interpreter with (optimised in C whenever required) libraries and consider pypy as their authors do: a toy... at least until they grow up and stop re-implementing everything for the sake of running python code only.
Armin Rigo reminds me of Wikus van de Merwe (Sharlto Copley) from the movie district9 !
The 3rd guy was so damn nervous...
+Andi Rady Kurniawan That's just his mannerism, he holds talks for huge audiences all the time
boy is he compensating for something with that tie or what.
That tie
This is a great way to turn someone off to PyPy
hahahaha seriusly.
The guy with the speech impediment that looks like a woman?
是不是印度人说英语比东欧人好听一万倍?
cpython 0.0x，好囧。。。
Just when Java people slowed down with the "faster than C" claims stupidity, here comes the amazing python community.