Abstract
Lattice rules are implemented in CUDA for many-core computations on GPUs. A high-speed evaluation of Feynman loop integrals is presented, based on lattice rules and suitable transformations. The accuracy and efficiency of the method are compared for higher order sin(m)-transformations. Extensive results are reported for classes of diagrams including 2-loop box and 3-loop self-energy diagrams with massive internal lines. The method is further combined with an extrapolation with respect to the dimensional regularization parameter.