gopybench contains implementations in both Go and Python, designed to assess memory usage and performance across the entire lifecycle of object creation, initialization, and utilization.

All the tests are performed on sample object Person with:

• members:
  • strings: First Name and Last Name,
  • integer: Age
• methods:
  • Get Full Name, returning string where First and Last Names merged,
  • Increase Age, where Age is increased by 1

sample.go implements a Go program that iteratively creates instances of a Person struct, including associated methods, while monitoring memory allocation and timing with following approaches:

• single thread
• multiple threads, where the number of threads equals the number of cores

Python program sample.py with lib.py implements a logic that achieves the same task as the Go implementation using the following approaches:

• object creation via:
  • default manner: Person in bench.sh output below
  • __slots__ enabled: PersonSlots
  • dataobject from recordclass: PersonDataObject
• execution with:
  • single thread/process
  • multiple threads
  • multiple processes
• with different interpreters:
  • Python, CPython 3.12
  • Cython, 3.0.10 with libc.pyx converted from lib.py
  • Pypy, 3.10.14

=== GOLANG
golang: single thread 530.055568ms
golang: multiple threads 396.051268ms
=== PYTHON
Person object creation: single thread 5.23 seconds
Person object creation: 2 threads 5.22 seconds
Person object creation: 2 processes 2.53 seconds
PersonSlots object creation: single thread 4.61 seconds
PersonSlots object creation: 2 threads 4.60 seconds
PersonSlots object creation: 2 processes 2.20 seconds
PersonDataObject object creation: single thread 2.43 seconds
PersonDataObject object creation: 2 threads 2.41 seconds
PersonDataObject object creation: 2 processes 1.27 seconds
=== CYTHON
Person object creation: single thread 3.22 seconds
Person object creation: 2 threads 3.19 seconds
Person object creation: 2 processes 1.74 seconds
PersonSlots object creation: single thread 3.16 seconds
PersonSlots object creation: 2 threads 3.26 seconds
PersonSlots object creation: 2 processes 1.73 seconds
PersonDataObject object creation: single thread 3.20 seconds
PersonDataObject object creation: 2 threads 3.26 seconds
PersonDataObject object creation: 2 processes 1.65 seconds
=== PYPY
Person object creation: single thread 1.61 seconds
Person object creation: 2 threads 1.47 seconds
Person object creation: 2 processes 1.67 seconds
PersonSlots object creation: single thread 1.60 seconds
PersonSlots object creation: 2 threads 1.45 seconds
PersonSlots object creation: 2 processes 1.67 seconds
PersonDataObject object creation: single thread 1.59 seconds
PersonDataObject object creation: 2 threads 1.43 seconds
PersonDataObject object creation: 2 processes 1.66 seconds

• see bench.log for full output

Based on the benchmark results comparing memory usage and performance for object creation between Go and Python implementations, several key findings emerged:

• Go:

  • multiple threads implementation showcased the best timing, completing in 0.396 seconds.

• Python:

  • dataobject from the recordclass library with multiple processes achieved the best timing for default Python interpreter, completing in 1.27 seconds,
  • __slots__ and dataobjects implementations didn't significantly impact timing for both Cython and Pypy,
  • Cython with multi-process implementations were found to be more efficient than its other approaches,
  • Pypy with multithreaded implementation showed better performance than its other approaches.

These results underscore the efficiency of Go's concurrency model, particularly in multi-threaded scenarios. However, Python demonstrates competitive performance and can be further optimized through strategic techniques such as leveraging multiple processes with dataobject from recordclass for default interpreter.