Loading…
Back To Schedule
Friday, November 4 • 3:00pm - 3:45pm
Performance improvements in libcxx

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

We want to discuss current state of libcxx performance w.r.t. libstdc++. What are the regressions that we know and plan to fix them. A few of the regressions are: 

PR21192 - Reading from stdin is 1-2 orders of magnitude slower than using libstdc++ 
PR19708 - std::find is significantly slower than libstdc++. 
PR20837 - libc++'s std::sort is O(N^2) in the worst case (instead of O(N*ln(N))). 
PR26886 - libc++'s std::stable_sort also has a worst-case complexity issue. 
PR15456 - A faster implementation of std::function is possible 
PR16747 and PR21275 - Our unordered_multimap insert is much slower than libstdc++'s 
- Removing Heap usage 
- Missing features (and performance issues) in <regex> 
- Non-conforming allocators in the test suite 
- Identifying missing attributes in functions 
- Debug mode for the library

Speakers
avatar for Aditya Kumar

Aditya Kumar

Senior Compiler Engineer, Facebook
I've been working on LLVM since 2012. I've contributed to GVNHoist, Hot Cold Splitting, Hexagon specific optimizations, clang static analyzer, libcxx, libstdc++, and graphite framework of gcc.
avatar for Sebastian Pop

Sebastian Pop

Samsung Austin R&D Center
Loop optimizations, testing, benchmarks, performance tracking.


Friday November 4, 2016 3:00pm - 3:45pm PDT
3 - BoF (Rm LL21CD)