- definition of arguments an configuration parameters is centralized
in rwgame/RWConfig.inc
- argument parsing is tested
- the try/catch in main is less weird now (imho)
It lets complier to make extra optimization.
See:
https://gcc.godbolt.org/#compilers:!((compiler:g6,options:'-O2+-Wall++-
xc%2B%2B+-
std%3Dc%2B%2B11',sourcez:MYGwhgzhAECC0G8BQAHArgIxAS2ALiWiOgDdsAnAFzTBFIHtsATaAMwAoBKaAXmgAYA3EgC%2BwpKEgwAQtDzR0WXHESpMOfIWIlGLDt1bYAdrWj0SAU3LlmF4WKQTwUaAGE5C9cvjJFGgsQMzGxcZpbWtvbiOsGs9PTs0gBk0BjcyIEYAHT69o4xLBhg5OyuKcDpWkTAOVz2QAA%3D%3D)),filterAsm:
(commentOnly:!t,directives:!t,intel:!t,labels:!t),version:3
How?
Calling stepsimulation each frame,
bullet interpolate vehicles and characters for us.
Removed unneeded code:
"float alpha = fmod(dt, GAME_TIMESTEP) / GAME_TIMESTEP;"
It looks like alpha shift is
I also removed unneeded variable clock
and refactored names to match bullet's example.
We should alse think about problem of
crossing the range of float.
This was just a useless container for a set of modules. We only have
one module now so don't bother with it. This means we can remove some
more raw new & deletes from RWGame too.
This involves a few changes. The first changes involve
allocating GameWindow and WorkContext on the heap, so that
RWGame still owns them but chooses when they're freed.
The work queue is given a method to stop the worker thread
without destroying the work context, so that subsystems
relying on the work context may still function to shut down.
Then RWGame is rearranged to cleanup separate subsystems
in an order that does not conflict (i.e., stop the work queue,
shut down other subsystems, then the renderer, *then* the window.)
The window needs to be cleaned up *after* the renderer because it
owns the OpenGL context.
* Fix VM Global Addressing
* Modify VM structures to simplify storage
* Add explicit GameWorld::createPlayer() method
* Move gameTime to GameState for storage
* Add SaveGame class for reading + writing
* New Dependancy: cereal
