1. I'm still unhappy about the loop that fills the count histogram, as I noted earlier today. It at least needs a decent comment and some overflow protection, and I'm not entirely convinced that it doesn't have more bugs than the overflow issue.
Attached patch is focused on fixing this. The "frac" variable overflow is evaded by making it int64. I hope comments is clarifying something. In general this loop copies behaviour of histogram constructing loop of compute_scalar_stats function. But instead of values array we've array of unique DEC and it's frequency.
