Why does the sampling result need to be multiplied by 256？

[Joll123]

Hi, in project_and_sample.h of Lyra-0.0.2-release, why does the sampling result be multipled by 256?
float f_result = mean + scale * log((1.0f - prob) / prob);
int result = std::min(static_cast(std::numeric_limits<int16_t>::max()),
std::max(static_cast(std::numeric_limits<int16_t>::min()),
static_cast(f_result * 256)));