Fig. 3
From: GPU accelerated adaptive banded event alignment for rapid comparative nanopore signal analysis
Performance comparison of ABEA on CPU vs GPU over a wide range of systems. Runtime for C-opti (left bars) and the G-opti (right bars). Runtime for the GPU has been broken down into: compute kernel time; different overheads (memory copying to/from the GPU, data serialisation time), and the extra CPU time due to CPU processing of the reads. The compute kernel time includes the sum of time for all GPU kernels. The extra CPU time is the additional time spent by the CPU to process very long reads and ultra long reads (see Supplementary materials) assigned to the CPU (excluding the processing time that overlaps with the GPU execution, i.e. only the extra time which the GPU has to wait after the execution is included)