I still haven't understood the [INLINE] square brackets thing (yet) but I've begun to understand the templates concept amongst all the syntax MUCH better when I found this Article here on Code Project.
An Idiots Guide to C++ Templates-Part-1
So I guess I'm an idiot since that page helped me more than any I've read previous. A happy idiot now.
(c) It appears to be a scope resolution :: to the member function c_str() but what is the [inline] bracket syntax?
Where did you get this template example from? The only place I've found the [inline] syntax is here (gcc 4.6.3 api docs). That syntax does not show up in the current API docs, nor does it show up in the libstdc++ sources for gcc 4.6.3 or gcc 8.2. It is also not described in the the MS c++ template docs or the template docs at cpprefereence.com. I'm thinking that's a documentation note that c_str() may be declared as an inline function, much the same as the man page for a utility might show optional parameters e.g. find [-H] [-L] [-P] [-D debugopts] [-Olevel] [starting-point...] [expression]
Thanks k5054, that must be it. Probably only novice learning dummies like me fail to ascertain that. You have cleared up my confusion and I can stop searching for this new mysterious syntax that I've never seen before.
We have a C++ MFC MDI application with a very heavy computational part.
And we want to run it on HPC cluster. Do we need to rewrite the application some specific way to do it? If yes, then what needs to be changed. What the simplest Windows cluster could look like and configured? How to deploy the app and distribute the job?
The overriding concept of makefiles is the ability to specify the dependencies in a simple tree. But since you are the one who knows which module(s) depend on which, you are the person who will need to create the tree.
If you get the original function back after the reverse, you have probably coded the transform correctly.
The problem is likely to be that your sampling rate is not a multiple of the basic frequency of sin(0.1*x). This means that you get a large FFT component close to the frequency, but you also get "noise" throughout the range, due to the fact that the residues do not cancel out.
I would try taking an FFT of sin(x), sin(2*x), etc., and see if these functions give the expected values.
Freedom is the freedom to say that two plus two make four. If that is granted, all else follows.
-- 6079 Smith W.
Thank You for the hint. I have tried with sin(nx), sin(2pi*x/n) and sin(2pi*x*n) too, but results were similar.
The most promising result was for the option with sin(2pi*x*n): here is a result with n=10. Values for x>270 are 0.
I used decimation in frequency FFT instead of decimation in time version, but it still doesn't show good values.