Released this week was FluidX3D 3.7, the latest feature update to this computational fluid dynamics (CFD) software that is CPU/GPU accelerated by way of OpenCL.
Making FluidX3D 3.7 exciting is a significant local memory optimization to its Q-criterion isosurface rendering kernel. Lead developer Dr. Moritz Lehmann is reporting that this new optimization can yield up to a 2x speed-up over prior versions of the software. The performance gains are most pronounced with FluidX3D 3.7 on older GPUs. The work is also enough that it shifts the kernel from being memory-bound to compute-bound.
Beyond this up to 2x faster Q-criterion isosurface rendering, FluidX3D 3.7 also has other micro-optimizations and cleaning up its code.
Downloads and more details on FluidX3D CFD software at GitHub. It will be fun to run some new FluidX3D benchmarks for looking at the performance impact of this optimization work on a broader range of hardware.