Quantifying the interpretation overhead of Python

While Python has become increasingly popular for its convenience, it is also criticized suboptimal performance. To figure out what burdens the interpreter of and provide insights into possible optimizations, we conduct this empirical study on CPython's performance via sampling-based profiling. This approach incurs a low runtime overhead does not require any modification application code, thus providing convincing experimental results. Specifically, use 48 benchmarks from pyperformance project to analyze interpreter. We compare usage different opcodes decompose at various granularities (e.g., files, functions, statements). It turns that most parts contribute small portion overhead, promising improvements lie in minority, such as name access reference counting functions. Furthermore, pay attention four specific performance-affecting issues: access, dynamic typing, garbage collection, opcode dispatch. The issue reveals several optimization techniques, register-based virtual machine architecture tracing-based well few fruitless points, operator overloading