Scale one 1e9-scaled SVI magnitude down by the fraction of anchored time
remaining, returning it at 1e18.
The roll-down result is kept at 1e18 because the fraction is applied to values
that are themselves tiny on short-dated surfaces: a 1e9 floor here costs up to a
whole raw unit of a, and a short-dated a is only about ten raw units, so the
truncation alone moves the digital by percent-scale amounts. At 1e18 the same
floor is a billionth of that.
The u256 intermediate keeps the product exact for any expiry_ms rather than
relying on a bound on the anchored horizon. The result is at most
value * 1e9 < 2^64 * 1e9, so narrowing to u128 never truncates.
Scale one 1e9-scaled SVI magnitude down by the fraction of anchored time remaining, returning it at 1e18.
The roll-down result is kept at 1e18 because the fraction is applied to values that are themselves tiny on short-dated surfaces: a 1e9 floor here costs up to a whole raw unit of
a, and a short-datedais only about ten raw units, so the truncation alone moves the digital by percent-scale amounts. At 1e18 the same floor is a billionth of that.The
u256intermediate keeps the product exact for anyexpiry_msrather than relying on a bound on the anchored horizon. The result is at mostvalue * 1e9 < 2^64 * 1e9, so narrowing tou128never truncates.