Computational Physics: The Future of Network Code Development?

[cscott]

Would you ever be writing network code as a computational physicist?

 

[PerennialII]

... network code, are you referring to 'common network operations' (in a network of computers) or am I just missing something?

 

[cscott]

I just mean writing the code for communication over a network. (ie. sending data back and forth)

Yes, I realize this is kind of a random question

 

[cscott]

Speaking of computational physics... I have an indepedent study (HS student) on differential equations and I was wondering if anyone knows anything simple enough that I'd be able do. I haven't haven't taken any classes on numerical analysis or anything but I can get by programming in C or Python.

 

[PerennialII]

I just mean writing the code for communication over a network. (ie. sending data back and forth)

Yes, I realize this is kind of a random question

... would say not the primary forte, but comp phys involves lots of computational work , and for example anything bit larger is nowadays done parallel using multiple CPUs/clusters etc. and as such knowledge of what "goes on in there" is important for example when writing software which would process your work efficiently. And then there is of course the weird thing that comp phys does involve quite a bit of 'practical' work in the hardware/software areas and even not so usual tasks seem to find their way to peoples desks .

 

[PerennialII]

Speaking of computational physics... I have an indepedent study (HS student) on differential equations and I was wondering if anyone knows anything simple enough that I'd be able do. I haven't haven't taken any classes on numerical analysis or anything but I can get by programming in C or Python.

Implementing something 'complex/simple enough' to numerically solve some differential equation problem might be interesting. Lots of numerical methods available for different equations, which don't really need that much numerical analysis education to grasp. Identifying an interesting application (or anything which wouldn't blow the workload beyond "humanly" doable) and building a discretization around it & solving using some numerical methodology (the discretization referring to selecting an appropriately complex/simple method and problem formulation to work on) which fits the problem might be it. 'What' is a tougher question, wouldn't hurt if you'd had some previous info and experience about the problem area would start working on.

 

[cscott]

Thanks for your insights Perennial.

As for the numerical stuff for my project, I will take your advice; I'll search for some problems and see what you think.

 

[neurocomp2003]

Numerical Recipes in C C++ or Fortran(i think they published in fortran)
pick up one of those books...go to a university library and pay the ~$50-100 fee...so you can borrow the books =]

start off with some simple stuff...write vector code 2D,3D,4D, write Matrix Code 2x2,3x3,4x4...code all the operations...write some functional code for integration(quadtratures) and Differentation, write some rooting finding stuf based on Newtons law ...man my memory is shot i can't remember any of the other algorithmic names. Then do some higher ODE/PDE stuff like RK4

These are standard stuff from an intro to num. anal. course

 

[cscott]

Sorry for the delay, but thanks neurocomp.

I wonder about the employability of computational scientists compared to their theoretical and experimental couterparts? What about in the industry?

I can only see large scale computation being used more and more in the furture, no?