No Sane Compiler Would Optimize Atomics

Explore the intricacies of compiler optimization for atomics in this comprehensive CppNow conference talk. Delve into how compilers actually optimize atomics, memory accesses around atomics, and leverage architecture-specific knowledge. Learn about the speaker's efforts to encourage more compiler optimizations, promote programmer reliance on these optimizations, and advocate for new atomic tools from hardware vendors. Examine common misconceptions about volatile keywords, benign races, and non-temporal accesses. Gain insights into the complexities of memory orderings and their importance in concurrent and parallel programming. Discover close-to-the-metal concurrency and parallelism tools being standardized. Note that the last 15 minutes of audio are missing due to a microphone failure, but slides are available for reference.

JF Bastien: No Sane Compiler Would Optimize Atomics