For a large dataset (295M reads) the GPU version took 50 seconds. Just running zcat on the files takes 10 minutes! Decompression and parsing FASTQ is a major bottleneck . Instead of using kseq we moved this work parsing to the GPU which delivers amazing throughput.