Parallelization across multiple cores is also supported by ispc, making it possible to write programs that achieve performance improvement that scales by both number of cores and vector unit size.