benelot
commented
9 years ago According to Erwin Coumans, this is very far from ready and also not cross-platform at all. Postpone this or discard completely.
Open benelot opened 9 years ago
benelot
commented
9 years ago According to Erwin Coumans, this is very far from ready and also not cross-platform at all. Postpone this or discard completely.
(This might have changed in the future:
Usage
The demo executables will be located in the bin folder. The Bullet 2 demo starts with App_AllBullet2Demos* The Bullet 3 demo starts with App_Bullet3_OpenCLDemos*
You can just run it though a terminal/command prompt, or by clicking it.
There are some command-line options, you can see using the --help option. For example, this will perform a benchmark writing to some files:
./App_Bullet3_OpenCL_Demos_clew_gmake --benchmark [--selected_demo=] Start with a selected demo] Run the benchmark for frames] Choose a certain OpenCL device] Choose a certain OpenCL platform] Change default demo settings (x,y,z)] Load AABB's from disk for the PairBench] Change the resolution of the shadowmap] Change the worldspace size of the shadowmap
[--benchmark] Run benchmark and export results to file
[--maxFrameCount=
[--dump_timings] Print the profile timings to console
[--cl_device=
[--cl_platform=
[--disable_cached_cl_kernels] Disable loading cached binary OpenCL kernels
[--x_dim=
[--pair_benchmark_file=
[--no_instanced_collision_shapes] Disable collision shape instancing (for tests)
[--no_shadow_map] Disable shadowmap rendering
[--shadowmap_resolution=
[--shadowmap_size=
[--use_uniform_grid] Use uniform grid broadphase (no all scenes work)
[--use_jacobi] Use GPU parallel Jacobi solver (instead of PGS)
[--use_large_batches] Use a different strategy for the constrains solver
[--debug_kernel_launch] Show debug info at start/end of each kernel launch
[--use_dbvt] Use the CPU dynamic BVH tree broadphase
[--allow_opencl_cpu] Allow to use an OpenCL CPU device
)