Fortran (Was: The "does Python have variables?" debate)

[Alain Ketterlin]

Mark H Harris <harrismh777 at gmail.com> writes:

> On 5/10/14 8:42 AM, Roy Smith wrote:

>> http://tinyurl.com/mr54p96

> 'Julia' is going to give everyone a not so small run for competition;
> justifiably so,  not just against FORTRAN.
>
> Julia is Matlab and  R, Python, Lisp, Scheme; all rolled together on
> steroids. Its amazing as a dynamic language, and its fast, like
> lightning fast as well as multiprocessing (parallel processing) at its
> core. Its astounding, really.

Hmmm...

> Its number concept is unified,

What exactly is unified? There is no implicit promotion between
primitive types and BigInt/Float. Conversion happens because all
arithmetic is overloaded for various types, and julia does
multi-dispatch. There is little chance number-crunching applications
will shine in julia, unless everything is statically typed.

> BigFloats are by default arbitrary precision with full scientific and
> transcendental functions built-in, everything complex just works, and
> did I mention its fast?

Yes, julia wraps GMP and MPFR, just like many compilers already do
(e.g., gcc). Very good, but nothing specific to julia here.

And for linear algebra, julia uses... BLAS and LAPACK. And fftw for FFT.
And so on.

Nothing new for a Fortran programmer, really.

> The bench-marks are within 2x of C across the boards; makes Matlab
> look like a rock, and is well ahead of python (NumPy SciPy) for
> technical computing.

2x for syntactic sugar is a lot. But yes, numpy/scipy developers should
watch out...

> Julia is still very much beta in my opinion but its maturing fast. Its
> open free (libre) and cross platform and did I mention it flatout
> screams? Not only will it replace FORTRAN completely if things keep
> progressing, but also Matlab, Mathematica, NumPy, & SciPy (and
> others). Keep your eye on her fellows.

Well, Fortran experts around me are skeptic.

-- Alain.