pub const INT_PRECISION: u32 = 136;
Maximum number of significant bits to use for integer values.
Values exceeding this are hard errors. This is at least be high enough to compute full products of u64 and i64.