This Linux Magazine article http://www.linux-mag.com/id/792 explains the difference in the way threads are implemented in Linux as compared to commercial Unixs such as Solaris. In summary, Linux uses a 1-to-1 mapping of user threads to kernel threads, while Solaris uses a many to many mapping. The article implies that this might give Solaris a performance edge, especially when switching between threads. Before I spend the time to test this, has anyone already done it?
This Linux Magazine article http://www.linux-mag.com/id/792 explains the difference in the way threads are implemented
Share
People used to think that M:N threading was a good idea, but everyone has been moving to 1:1 threading; it has lower overheads and works better in an SMP environment. Solaris moved from M:N threading to 1:1 threading in Solaris 9, I believe. That article you linked to seems to be from 2001, so it’s a bit outdated (it’s subscriber only, so I couldn’t read it).