As I always do, I ran the new version against a suite of primes sieve code done with older versions.
A December 18, 2024 1.15 development versions is 10-12 seconds faster against the final release for one large input value, and generally faster all around.
I ran the development version against the latest source code, but I hadn’t compiled that code with the final release. When I noticed that, I ran that code with the final release and it was the same time.