Show HN: Less Slow C++
a day ago
- #Performance
- #C++
- #Benchmarking
- Benchmarks focus on performance-oriented software design in C++.
- Uses C++20 features and is optimized for GCC and Clang on Linux.
- Covers topics from micro-kernels to parallel algorithms and coroutines.
- Highlights include faster trigonometry, compiler optimizations, and matrix multiplication.
- Discusses GPU programming, CUDA, and differences between CPU and GPU code.
- Includes practical examples like JSON handling without memory allocations.
- Recommends Linux environment setup for best compatibility.
- Build process involves cloning the repo, installing dependencies, and compiling with CMake.
- Supports various third-party libraries for enhanced functionality.
- Provides flags to control benchmark output and filter specific tests.
- Encourages disabling SMT for stable benchmark results.
- Includes hand-written Assembly and PTX IR for hardware-specific optimizations.
- Condensed list of Google Benchmark features for writing benchmarks.