I suppose multi-threading wasn't a huge problem for most applications. But I'm not sure how a mutex could ever be properly implemented in a multithreaded environment without an atomic compare and swap, but they've certainly been a thing for a very long time. In terms of atomics as a assembler intrinsic, I think that only became popular when intrinsics started being supported in C compilers, but I doubt the patent was holding that back - unless you know different?