Irrlicht Engine

Or why do we use batching but not instancing, is there anything that batching can do but instancing cannot?

Thank you all for the quick feedback.

does it means that if there's a cube and a sphere in the scene(the material on them are the same), they could be batched but not instanced?

another question is: could we really gain the performance benefit due to the overhead - copying vertices from meshes ...

http://irrlicht.sourceforge.net/forum// ... hp?t=39154

this link introdued a way to batch those meshes whose materials are the same, from what I know that there're similar concept in DX10 named 'instancing', are they the same thing?

Any feedback on this?

anyone successfully ogl-es branch(below link) on Android platform?

http://irrlicht.svn.sourceforge.net/viewvc/irrlicht/branches/ogl-es/

I met below error when trying to build the code:

Android NDK: There is no Android.mk under /cygdrive/d/Android/irrlicht/source/Irrlicht/Android/jni
Android NDK ...

edit: Uhm, no - horrible idea. Will not work. As there are lots of places where we pass on f32 to system-functions expecting floats (for example opengl...).

this is also my concern, we need to convert the fixed point back on the right places, such as when passing f32 to opengl as you mentioned ...

You could try replacing f32 and f64 by corresponding fixed-point classes. Can't help you much with that - it's been decades since I last worked with fixed-point math as optimization and my classes from back then probably sucked (I was learning c++ at the time). But in theory it sounds like that ...

Intel VTune, isn't that only on Intel-based Windows? Most ARM vendors have their own profiling toolchains with which you can profile on the actual device, those would give better stats, as with a PC profile you're just guessing what goes down on the device.

Also, many of the ARM profilers are able ...

Which driver do you use? Best thing is to move most calculations to the GPU, which has FP support. Doing calculations in fixed point would mean that you have to find enough consecutive calculations that would make the initial and final conversions from and to floating point neglectable. So meaning ...

I am trying to port irrlicht to android platform, the port is done, but the fps on android is very low. I heard that replacing float point with fixed point could improve the performance on mobile platform. Anyone has experience on this? Thanks very much

Hi


it seems 08.SpecialFX_vc9 is the only sample show how shadow works in irrlicht, but I noticed that that shadow is only for antimated nodes (because addShadowVolumeSceneNode() is a memeber function of CAnimatedMeshSceneNode class), my question is how to enable add shadow for other types of nodes ...

A little confused, where to call the createForsythOptimizedMesh()?