400 Gflops in your home PC!

Thanks to Lars Nyland (NVIDIA) I’ve got 400 Gflops on Radeon HD 4870 in the N-Body simulation (N = 4096)
It is similar to the CAL based simulations and performance may be twiced by the loop-unrolling technique. N-Body is an amazing but a little special application. Note that OES_texture_float is implemented only in the nightly Chromium yet. Unfortunately vertex_texture2D is not supported by WebGL yet and works with OpenGL drivers only.
I’m collecting “WebGL and simulations on GPU” notes at