Sunday Bla Bla

[mayorgreener 37]

OK since today it is Sunday I will do some bla bla bla and since it is Sunday all of you have to read it and digest it

OK, I was thinking if some team can re-build the SC4 in new Unity engine with all fixes and few optional excitements ???

Recently I joined a FB group of indie developers and I am stunned by the effects and products they have made. Simulation in city builder games is the main thing though.

Thank you all

[A Nonny Moose 8,262]

Good luck to you.  The main limitation of any stochastic simulation of this type is running out of processor cycles.  It doesn't matter how you do the multiprocessing, sooner or later at least one of the processors will start blocking on time out when there are not enough cycles to complete some part of the simulation loop.  These things tend to be hyperbolic and go towards infinity as the number of objects in the universe of discourse increases.

Notwithstanding any myth that games programming is something special and needs super deluxe techniques, it is no worse than building an operating system.  So, in effect, it is just adroit programming.  Whatever programming language you choose, make sure that there is an assembler interface wrapper so that when the chips are down and the optimizer fails, some parts can be written in machine code.  If it wasn't for the staggering overhead, I'd suggest C++, but you would be better off, I think, using the pre-Soustrup object model in pure C.  C has a well-defined assembler interface.  Stick to the x86 architecture and stay away from lock in products like Visual Studio and .NET.  Your best bet is the GNU C compiler (gc).  It is always up to the latest version of the standard.

Pick a public 3D package such as Ogre  This may require knowledge of several other packages, notably the GNU suite of development packages including GTK.  Stay with the public domain stuff, and you won't be locked in by some snap back in a EULA.

Recent experience in C# has shown it no better than C++ with respect to overheads.  So some testing is probably needed. 

I suspect a project of this magnitude with a team of three system programmers will take five to seven years.  Crowd coding with co-ordination will take longer (see the wine development team) but in the end the product will be fine tuned.  Try to avoid shooting at a moving target and avoid trying to be everything to everyone's hardware and software.  Pick a target configuration and go with that.  Upgrades to newer stuff can follow as updates.

Make a specification and stick to it.  Quality is conformance to requirements.  The initial documents must be good enough to work from.

  Edited August 9, 2015 by A Nonny Moose