Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the intricate relationship between software engineering and underlying hardware platforms in this comprehensive conference talk that challenges common assumptions about performance optimization. Examine how contemporary programming languages, paradigms, and tools often overshadow critical hardware considerations, potentially negating the benefits of modern hardware advancements. Delve into the mapping of high-level language constructs like references, lambdas, inheritance, and object representation to actual platform-level code, understanding their performance costs. Investigate out-of-order execution beyond surface-level claims, discovering what specifically operates out-of-order in modern x86 platforms and its performance implications. Navigate the complexities of memory hierarchy operations, including logical, physical, and geometrical address spaces, their translations, and strategies for hiding memory latency. Analyze why universal vectorization remains elusive and examine data dependencies alongside compiler optimization techniques. Cover branching control, ABI aspects, parameter passing, and compilation unit scope impacts on performance. Address claims about x86 obsolescence and the potential of ARM and RISC-V architectures. Gain insights from performance optimization experiences in network engineering, electronic trading, and high performance computing, while critically evaluating promises made by modern programming languages like Rust, Go, Zig, and D regarding performance, security, and development efficiency.
