Skip to main content
. 2010 Jan 11;11:20. doi: 10.1186/1471-2105-11-20

Figure 11.

Figure 11

Mean computation time of one single alignment (including standard deviation) and speedup (sp) after each optimising step: (1) partitioning, (2) vectorisation, (3) reduce branches, (4) avoid Int32 multiplies, (5) explicit unroll. Speedup (sp) compares the optimised solution to a simple partitioning (1) on the CBE and shows the effect of each optimisation step.