Getting Customized Runtime Hardware Details at Compile-Time for Thread Mapping on Any Machine
Explore a 27-minute conference talk from code::dive 2020 featuring Iulia Ştirb, who discusses obtaining customized runtime hardware details at compile-time for thread mapping on any machine. Delve into the challenges of finding information such as the number of logical cores and cores per CPU at compile-time, which is crucial for static mapping algorithms like NUMA-BTDM. Learn about the recent approach of using Intermediate Representation (IR) in compiler code to gather hardware architecture information, and examine the complexities of adapting this method to various hardware configurations. Gain insights from Iulia Ştirb, a Ph.D. candidate in Computer Science with expertise in compilers, energy consumption optimization, and image processing. Discover how this research contributes to improving runtime performance and energy efficiency in computational systems.