The best way is to validate using datasets from highly curacted genomic records, such as O. sativa, E. coli, M. tuberculosis and H. sapiens. Remember to try with genomes with different characteristics (high repetitive content, high or low GC content, many chromosomes, large genomes, SINE/LINE regions ...), as they may tricky your (and almost all) algorithm for genome alignment. Also try to validate using genomes with structural variations already validated (ex: Article Genome reduction in Leptospira borgpetersenii reflects limit...
).
Are you going to compare your algorithm to MUMmer and LAST?