64bit and multi-threading
Posted: 28.03.2012, 22:03
Hi guys,
This subject may have been covered already, so my apologies if it's duplicated.
Firstly, I was wondering if it was now possible to build a 64bit version of Celestia to take advantage of the latest 64bit hardware and OS's (Windows 7 in particular).
More importantly, is there any motivation amongst the DEVs to make Celestia a multi-threading application?
IMO, at this point in time, this should be a priority over adding new features as the performance gains on multi-core processors could be quite spectacular.
I'm assuming that you could target certain parts of the code for multi-threading (computations that rely on the CPU rather than the GPU)... my guess would be script execution, reference frame / positional calculations, etc, might be areas that would benefit greatly from this.
For example, in my models. multiple objects are positioned using reference frames (many objects are often positioned relative to the same object), and the trajectories and orientations of many moving objects are computed by lua/celx scripts. There can end up being a huge number of computations to be done between render cycles, and being able to use all processor cores would I think improve frame-rates significantly in this type of scenario, as I suspect that the more objects you control with celx/lua scripting, the more likely you are to be CPU-bound, as all the computations for objects in the FOV must be completed between each rendered frame.
Regards
CC
This subject may have been covered already, so my apologies if it's duplicated.
Firstly, I was wondering if it was now possible to build a 64bit version of Celestia to take advantage of the latest 64bit hardware and OS's (Windows 7 in particular).
More importantly, is there any motivation amongst the DEVs to make Celestia a multi-threading application?
IMO, at this point in time, this should be a priority over adding new features as the performance gains on multi-core processors could be quite spectacular.
I'm assuming that you could target certain parts of the code for multi-threading (computations that rely on the CPU rather than the GPU)... my guess would be script execution, reference frame / positional calculations, etc, might be areas that would benefit greatly from this.
For example, in my models. multiple objects are positioned using reference frames (many objects are often positioned relative to the same object), and the trajectories and orientations of many moving objects are computed by lua/celx scripts. There can end up being a huge number of computations to be done between render cycles, and being able to use all processor cores would I think improve frame-rates significantly in this type of scenario, as I suspect that the more objects you control with celx/lua scripting, the more likely you are to be CPU-bound, as all the computations for objects in the FOV must be completed between each rendered frame.
Regards
CC
