I don’t think you can generally control the cache with such granularity since modern processors do all sorts of instruction level parallelism and cache coherency voodoo