M(10^24) = 7,189,337,839 and M(10^25) = -258,560,632,948 are computed using an optimized O(x^{2/3+epsilon}) algorithm, extending the record by two orders of magnitude.
Computing the Mertens function on a GPU
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
A GPU implementation of an algorithm to compute the Mertens function in O(x2/3+{\ko}) time is discussed. Results for x up to $10^{22}$, and a new extreme value for $M(x)/x^{1/2}$, -0.585768 ($M(x) \approx -1.996 \ast 10^9$ at $x \approx 1.161 \ast 10^{19}$), are reported.An approximate algorithm is used to examine values of M(x) for x up to $\exp{(10^{15})}$.
fields
math.NT 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Practical Computations of the Mertens Function: $M(10^{24})$ and $M(10^{25})$
M(10^24) = 7,189,337,839 and M(10^25) = -258,560,632,948 are computed using an optimized O(x^{2/3+epsilon}) algorithm, extending the record by two orders of magnitude.