Exserohilum turcicum is a devastating fungal pathogen that infects both maize and sorghum, leading to severe leaf diseases of the two crops. According to host specificity, pathogenic isolates of E. turcicum are divided into two formae speciales, namely E. turcicum f. sp. zeae and E. turcicum f. sp. sorghi. To date, the molecular mechanism underlying the host specificity of E. turcicum is marginally known. In this study, the whole genomes of 60 E. turcicum isolates collected from both maize and sorghum were resequenced, which enabled identification of 147,847 high-quality SNPs in total. Based on the SNPs, all isolates were clustered into four genetic groups that had a close relationship with host source. This observation was validated by the result of principal component analysis. The analysis of population structure revealed that there was obvious genetic differentiation between maize and sorghum host populations. Further analysis showed that 5,431 SNPs, including 612 nonsynonymous SNPs, were completely co-segregated with host source. These nonsynonymous SNPs were located in 539 genes in which 18 genes were predicted to encode secretory proteins, including six putative effector genes. The sequence polymorphism analysis of the six effector genes in 60 isolates indicated that these genes were perfectly co-segregated with host source. All SNVs in the coding regions of these genes were non-synonymous substitutions, suggesting that these genes were subject to strong positive selection pressure. These findings provide new insights into the molecular basis of host specificity in E. turcicum.