app state. If the app state is attached then it's 'enabled'
state can be toggled with the F6 key.
It displays a continnuously updating bar graph of the application
wide frame timings for 'update' and 'render'... where 'update'
is all of the parts that aren't 'rendering', ie: running
enqueued tasks, updating states, updating controls, etc..
Implementors can provide their own AppProfiler
implementation to collect and/or visualize timing
stats however they want, even at the viewport level
if they choose.
A basic profiler implementation will be following
shortly that does simple update vs render frame
timings.
around some commit issue with RenderManager. Actually
calling of these hooks will follow immediately.
Added hooks for detailed performance profiling.
Implementors can provide their own AppProfiler
implementation to collect and/or visualize timing
stats however they want, even at the viewport level
if they choose.
A basic profiler implementation will be following
shortly that does simple update vs render frame
timings.
originally from Lemur. This provides more convenient
enabled-state management than the normal AbstractAppState
class. In addition, it exposes convenience methods for
getting to other app states, the application instance,
etc..
Fix using AssetManager.loadModel() with fbx file causing to throw
AssetLoadException.
Implemented loading of new file format .fba - .fbx file with animation
mapping.
first one to follow other conventions. Seems like
Material is the only thing using it (and I think it
could get away with a vec3 really but no harm done).
to signal the beginning of a frame. Currently this
just grabs the time and tpf values once so as to avoid
per-material Float allocation and time drift within a
frame.
RenderManager now calls this at the beginning of the
root-level render() method.
its redundant.
Modified setGuiBounding to call the new setCenter(x,y,z)
method instead of allocating a Vector3f each time. (This
method is called any time someone sneezes.)
Fixed a missing line feed in a javadoc comment.
collision. The old code had an implementation but
it missed tons of cases.
For 100,000 random points, the old way would process
that in about 12+ ms. The new way does it in about
18+ ms... but the old way missed 643 collisions
(180 versus 823).
Plus, with the new way accurate contact normal and
contact point can be provided trivially... so it is.
(so far only for lwjgl because I'm lazy)
Went ahead and added in divisor support while I was
at it because it doesn't cost anything and there might
be use-cases for it.
Mesh now knows about instances and you can directly
ask it for an instance count. I haven't added support
for proper bounding box calculation but it should
at least be possible now.
pspeed42
Nehon
shadowislord
Hannes Nevalainen
Rena4ka
tort32
Jan Ivenz
Christian Gärtner