Efficient Personalized and Non-Personalized Alltoall Communication for Modern Multi-HCA GPU-Based Clusters | IEEE Conference Publication | IEEE Xplore