Figure 7

GPU and CPU execution times. In all the subplots the x-axis indicates the number of sink particles and the y-axis the number of source particles used. For subplots (a), (b) and (c) the raw execution times are presented and indicated with the colors. Plot (d) does not present the execution time but rather which of the configuration gives the best performance. If the GPU is faster than the 8 cores of the CPU we indicate by how much faster the GPU performs. To use more than a single CPU core we use OpenMP. Note that the colors are scaled per plot and are not comparable between the different subplots. All the GPU times include the time required to copy data between the host and device.