Three programs to compute a set of DNA sequences where each sequence in the set differs from every other sequence by a minimum of a given number of nucleotides. This is an interesting problem as it becomes more difficult the larger the set becomes.

Three implementations are available:

• ser_barcodes.cc is the serial implementation in C
• cu_barcodes.cu and cu_barcodes_curand.cu are CUDA C implementations differing by the way the random sequences are generated.

The CUDA C code is between 2x and more than 30x faster than the serial C code depending on the difficulty of the problem. For example a set of 15,000 sequences of 100 nucleotides differing by 60 nucleotides takes about 220 seconds to complete using the CUDA C code (slightly faster using the cuRAND variation) and about 1900 seconds on an AMD Phenom(tm) 9750 Quad-Core Processor machine equiped with an NVIDIA GTX 580 graphics card, a performance gain of 8.6x.