Skip to main content
. 2011 May 4;6(5):e18539. doi: 10.1371/journal.pone.0018539

Figure 11. Performance Profile.

Figure 11

Figure shows the performance profile of our GPU implementation with respect to where GPU time is spent during the simulation (shown as a percentage for each GPU kernel corresponding with figure 8). The figure in brackets next to the vertical axis label indicates the total number of times the kernel function is called over single “learn step” with a total time simulation period of T = 128 and Inline graphic. An additional amount of CPU time is also required however this is negligible in the scale of the overall simulation. A: Represents the case for our model without lateral connections. B: Represents a case with lateral connections which is the same however it includes a kernel function “actionCellLateralSpikePropagation” which performs the lateral spike propagation simulation.