https://hgpu.org/?p=10268
A Memory Efficient Algorithm for Adaptive Multidimensional Integration with Multiple GPUs