4 ms·
Scheme is highly dynamic and can be compiled to surprisngly small and fast binaries. SBCL can produce blazing fast (though not small) binaries for highly dynami
by divs1210 6y ago
Scheme is highly dynamic and can be compiled to surprisngly small and fast binaries. SBCL can produce blazing fast (though not small) binaries for highly dynamic Common Lisp programs.
The claim that Python is slow because of its nature rather than its implementations is wrong and flawed.
PyPy is pretty fast. Python compiled to Chicken Scheme / SBCL should be in the same ballpark, if not faster.
- rzzzwilson 6y ago> Scheme is highly dynamic and can be compiled to surprisngly small and fast binaries. Yes, after almost 60 years of compiler development. PyPy may get there in another 40 years. > The claim that Python is slow because of its nature rather than its implementations is wrong and flawed. The python core developers don't lack in brain power. If it's just a matter of poor implementation then cpython would be much faster than it is. Or do you claim the core developers are incompetent? > PyPy is pretty fast. The PyPy people only claim 4.4 times faster than cpython, which is itself about 20 times slower than C on comparable code and much slower in many cases, so "pretty fast" is questionable. Still, in 40 year's time maybe cpython will match the fast languages, but I doubt it.
- divs1210 6y agoTurns out there IS a third party optimizing static compiler for Python: https://github.com/Nuitka/Nuitka https://github.com/Nuitka/Nuitka Also, a 4.4 times boost in case of pypy is enormous. It translates to 4x less energy, 4x less resources to manage, 4x less operational costs, etc The goal is not to have the fastest implementation, but a fast enough implementation that can be distributed easily.