GPU Radix Sort written in Apple Metal

I was hoping one would be announced at WWDC but no such luck.

I need a GPU Radix Search written in Metal for integer key value pairs. Currently using thrust::stable_sort_by_key on Nvida GPU's but want to get working in Metal.

One option would be to look at CUDPP: [login to view URL]

But I know there are faster options:

[login to view URL]

Merrill, D. and Grimshaw, A. High Performance and Scalable Radix Sort: A case study of implementing dynamic parallelism for GPU computing, Parallel Processing Letters 21 (2011)

Ideally would like to have as a single compute kernel function.

// thrust::stable_sort_by_key(thrust:: device_ptr<uint>(dGridParticleHash),

// thrust:: device_ptr<uint>(dGridParticleHash + numParticles),

// thrust:: device_ptr<int>(dGridParticleIndex));

C++ Programming CUDA GPGPU Swift

Project ID: #17139435

About the project