rt，可以把

	u32 lim = 1;
	for (; lim < N; lim <<= 1);
	inv[0] = inv[1] = 1;
	for (u32 i = 2; i ^ lim; ++ i) inv[i] = inv[mod % i] * (mod - mod / i);
	fac[0] = 1;
	for (u32 i = 1; i ^ lim; ++ i) fac[i] = fac[i - 1] * i;
	ifac[0] = 1;
	for (u32 i = 1; i ^ lim; ++ i) ifac[i] = ifac[i - 1] * inv[i];

放进 constexpr 函数里提前编译加快运行时间么？

或者说还有什么优化这一段代码的方法？