## Which programming language should I use?

I was wondering which programming language I should learn for general purpose calculations and physics simulations.

I already know a bit java.

 PhysOrg.com science news on PhysOrg.com >> 'Whodunnit' of Irish potato famine solved>> The mammoth's lament: Study shows how cosmic impact sparked devastating climate change>> Curiosity Mars rover drills second rock target
 I suppose it would depend upon what compilers were available on the computer you intend to use. Fortran is the old standby, but there are others out there that are probably just as effective.
 Recognitions: Gold Member That seems to be a bit of a theological question around here (and probably everywhere). Different folks will SWEAR to you that only is the right one. Java will do you fine, as will others. FORTRAN, as old and creaky as it is, has TONS of subroutines that have accumulate over the years specifically for scientific applications so that's in its favor. If you plan on doing much in modern computing, especially outside of science programming, I recommend that whatever you pick, it be fully OO like Java or C++ or C# or VB.NET

## Which programming language should I use?

 Quote by yup790 I was wondering which programming language I should learn for general purpose calculations and physics simulations. I already know a bit java.
If you need speed, then anything where you can compile to an optimum representation like the machine code for your desired platform.

You might even have to learn vector based CPU instructions, if speed is super super critical. An example of SIMD instructions are the SSE and SSE2 instruction sets for common x86-type platforms.

 Fortran 77. Is there another language yet that directly handles complex variables? For 'general' use (implying graphs, diagrams and 'free') I was planning to take a good look at the current state of VPython sometime. Has anyone any experience with VPython?

 Quote by phinds FORTRAN, as old and creaky as it is, has TONS of subroutines that have accumulate over the years specifically for scientific applications so that's in its favor. If you plan on doing much in modern computing, especially outside of science programming, I recommend that whatever you pick, it be fully OO like Java or C++ or C# or VB.NET
I just want to point out that while Olde FORTRAN (77 and earlier) is indeed old and creaky (though still very useful), modern Fortran (90 and later) is comparable to the other languages mentioned, and is fully OO now. And as of the latest 2008 standard, it's the only language with built-in parallel processing support in the form of coarrays, if you're planning on doing expensive calculations. I'm just saying, it's not your grandpa's FORTRAN anymore, though it's still more physicist-friendly than others.

But if I were you, I'd just stick with Java if you already know it well. I know that the big particle-in-cell codes are mostly written in C++ or Fortran, but Java is starting to show up more and more, and they all have comparable speeds. These days it doesn't matter as much. It really comes down to using what you're most comfortable with.

Recognitions:
Gold Member
 Quote by PICsmith I just want to point out that while Olde FORTRAN (77 and earlier) is indeed old and creaky (though still very useful), modern Fortran (90 and later) is comparable to the other languages mentioned, and is fully OO now.
Thanks for that update. I'm pretty clear now that *I* am whats old and creaky in my post ... I have not kept up w/ FORTRAN so was not aware it had made such great strides (FORTRAN 77 is very NEW by my standards --- I stopped programming in FORTRAN prior to that).

I would hazard a guess however that if it is indeed fully OO it really doesn't much resemble the original FORTRAN at all so I pretty clearly need to just stop thinking of it as FORTRAN at all

Mentor
 Quote by cmb Fortran 77. Is there another language yet that directly handles complex variables?
C++ does.

Code:
#include <iostream>
#include <complex>

using namespace std;

int main ()
{
complex<double> z1, z2, z3;
z1 = complex<double> (1.0, 2.0);
z2 = complex<double> (3.0, 4.0);
cout << "z1 = " << z1 << endl;
cout << "z2 = " << z2 << endl;
z3 = z1 + z2;
cout << "Sum = " << z3 << endl;
z3 = z1 * z2;
cout << "Product = " << z3 << endl;
return 0;
}
 Quote by PICsmith But if I were you, I'd just stick with Java if you already know it well.
Even if you don't already know it well, but have made a fair start on it, I'd stick with it and get lots of practice programming in it before moving on to another language. Good program design, coding and debugging techniques are applicable in most languages. If you know how to "program well" in general, you can easily pick up a new language when necessary, and apply those general techniques.

 OK, so what does C++ return when you enter cos-¹(2)? F77 returns a complex value.
 Mentor No go... the C++ standard library doesn't seem to include a version of acos() for complex numbers. It does include automatically-overloaded complex versions of cos(), cosh(), exp(), log(), log10(), pow(), sin(), sinh(), sqrt(), tan(), tanh(). (I tried cos().) However, you can write your own complex inverse trig functions using the complex logarithmic forms: http://en.wikipedia.org/wiki/Inverse...arithmic_forms Code: #include #include #include using namespace std; // my attempt at a complex-valued acos() function complex acos (const complex& x) { const complex eye = complex(0,1); const complex one = 1; return -eye*log(x + eye*sqrt(one - x*x)); } int main () { complex z = 2; cout << acos(z) << endl; // this outputs "(0,1.31696)" return 0; } Of course a built-in library function might use a different method which is more accurate or efficient.

Recognitions:
 Quote by jtbell C++ does.
Yep, and with it's operator and function overloading it's surprising how well this added class feels like a language inbuilt feature. Also I've found it's precision is actually better than the Fortran built in type (at least for the only two freeware F95 implementations that I've found). I'm not sure why, but both the gnu "g95" fortran95 compiler and the Silverfrost ftn95 have serious accuracy problems with double precision complex numbers. No issues to report with the gcc c++ compilers implementation of double precision complex however.

I'd hope that a commercial compiler like ifort can handle double precision complex properly, but I've never had the opportunity to test it.

I've got to admit Fortran 90 (and onward) has come a long way, and it's probable still a contender for number crunching programs. I'm not really fan of Fortran as a general purpose language though. It still carries some legacy stuff that I hate. I don't like implicit typing (or having to "implicit none" for every function) and I really hate Fortran I/O.

 Quote by uart I really hate Fortran I/O.
Once you learn to pipe data outputs into text streams and such [much less chance of crashing mid-analysis], then post-process the data, it all makes much more sense to treat F77 as a computer-based 'engineer's calculator'.

If you want (/need) flash and bling, buy MatLab.

Recognitions:
 Quote by cmb If you want (/need) flash and bling, buy MatLab.
Or get one of the freeware clones like Octave.

Actually that is a bit of a plus for Fortran, it's close syntax similarity to matlab is very handy.

Recognitions:
 Quote by cmb OK, so what does C++ return when you enter cos-¹(2)? F77 returns a complex value.
That's well spotted. I'd never noticed that the C++ complex implementation didn't include any inverse trig functions. Yeah I agree that having as complete as possible function coverage is a very important feature for a complex type. It makes it so much easier, especially when you're translating an algorithm, originally written for reals, to handle complex. With a good inbuilt complex type and complete function coverage it can often make the job almost trivial.

I'm surprised that they didn't include inverse trig's in the c++ complex implementation, as jtbell pointed out above it's not very difficult. I just wrote the following inverse trig library in about 15 minutes if anyone wants it. I even threw in the hyp-trig's at no extra cost.

Code:
/* File complex_itrig.h */
/* Uart's complex inverse trig library */

#include <complex>
typedef std::complex<double> cmplx;
const cmplx _I = cmplx(0.0,1.0);
const cmplx _1 = cmplx(1.0,0.0);

cmplx atan(cmplx x) {
return 0.5*_I*( log(_1 - _I*x) - log(_1 + _I*x) );
}

cmplx asin(cmplx x) {
return atan( x / sqrt(_1-x*x) );
}

cmplx acos(cmplx x) {
return atan( sqrt(_1-x*x) / x );
}

cmplx acosh(cmplx x) {
return -_I * acos(x);
}

cmplx asinh(cmplx x) {
return -_I * asin(_I*x);
}

cmplx atanh(cmplx x) {
return -_I * atan(_I*x);
}

Blog Entries: 47
Recognitions:
Gold Member
Homework Help