https://hgpu.org/?p=10869
Architectural improvements and 28 nm FPGA implementation of the APEnet+ 3D Torus network for hybrid HPC systems