A Deep Dive into Python Execution: CPython Bytecode, PVM, and Online Platforms

Abstract

Python’s popularity in artificial intelligence, education, and software development is driven by its intuitive syntax and flexible execution model. Unlike traditional compiled languages, Python code is first transformed into bytecode and then interpreted bythe Python Virtual Machine (PVM), making it highly portable and adaptable. This paper explores the internal workings of CPython the reference implementation of Python, by examining how source code is compiled into bytecode,how the PVM executes these instructions, and how recent changes in Python’s bytecode enhance performance. Additionally, the paper discusses the growth of online Python interpreters, which make coding accessible from any device without complex setup. By analyzing these technical processes, the paper aims to help learners and practitioners better understand how Python operates internally and how modern tools are shaping its use in education and development.

Introduction

I. Python's Role in AI and Software Development

• Growth of AI has increased demand for programming languages that are high-level yet flexible.

• Python has emerged as a leading language due to:

  • Simple syntax

  • Vast standard library

  • Integration with AI frameworks (e.g., TensorFlow, PyTorch)

• Suitable for both research and production, supporting rapid prototyping and large-scale deployment.

• Python consistently ranks among the top 3 programming languages globally due to its:

  • Developer community

  • Academic and industrial relevance

  • Platform independence and active ecosystem

II. Traditional Compilation vs Python Execution

• Traditional compiled languages (like C/C++):

  • Source code is compiled into machine code (platform-specific).

  • Execution is fast but inflexible across platforms.

• Python (via CPython):

  • Compiles source code into bytecode, a platform-independent intermediate form.

  • Bytecode is executed by the Python Virtual Machine (PVM), not directly by hardware.

III. CPython Execution Process

• CPython is the default and most used implementation of Python.

• Involves two key steps:

  1. Compilation to Bytecode:

     • Stored in .pyc or .pyo files.

     • Made of opcodes (operation codes) and arguments.

  2. Execution by PVM:

     • PVM reads bytecode instructions and performs operations via an evaluation stack.

     • Supports dynamic typing, memory management, and exception handling.

IV. Bytecode Analysis Tools

1. dis.dis(): Disassembles code into readable bytecode.

2. code.co_code: Reveals raw bytecode stream.

3. dis.get_instructions(): Provides detailed, structured inspection of bytecode.

4. Key Bytecode Instructions:

   • LOAD_GLOBAL, LOAD_CONST, PRECALL, CALL, RETURN_VALUE, and new RESUME (Python 3.11+).

5. PVM Functioning:

   • Operates in a stack-based manner.

   • Interprets opcodes in a loop.

   • Written in C (in CPython).

   • Does not generate native binaries; outputs are runtime behaviors, not compiled artifacts.

V. Online Python Interpreters

• Tools like Replit, Google Colab, Jupyter Notebooks, Trinket allow Python code execution in browsers.

• Useful for:

  • Education (bootcamps, MOOCs)

  • Prototyping

  • Coding interviews

  • Lightweight collaboration

How They Work

• Use client-server model:

  • Code runs on remote servers (or locally on mobile apps).

  • Some use WebAssembly or JavaScript transpilers (e.g., Pyodide, Skulpt).

• Cloud platforms may offer access to GPUs or TPUs for high-performance tasks.

Usage Trends

• Increasingly used in education.

• Estimated to account for 10–20% of overall Python use.

• Popular in quick tasks, testing, and collaborative coding.

Pros and Cons

• Advantages:

  • No installation needed

  • Accessible anywhere

  • Ideal for beginners and education

• Limitations:

  • Restricted access to system features

  • Dependency on internet

  • Limited third-party library support

Future Outlook

• Growth expected due to:

  • Advances in cloud computing and WebAssembly

  • Integration of AI-assisted tools

  • Enhanced support for debugging and full-stack development

• Despite growth, local environments will remain critical for production and complex projects.

Conclusion

The execution model of Python, particularly through its CPython implementation,highlights the language’s focus on portability, dynamism, and runtime flexibility. By compiling high-level source code into platform-independent bytecode and executing it via the PVM, Python enables rapid development and cross-platform compatibility. The disassembly and bytecode inspection tools discussed in this study such as dis, co_code, and get_instructions() offer valuable insight into Python\'s internal workings, bridging the gap between abstraction and execution. In parallel, the growth of online Python interpreters has expanded the accessibility of programming environments, removing traditional barriers like installation and setup. Theseplatformsareespeciallybeneficialineducationalcontexts,smallscalesoftwaredevelopment,and collaborative workflows. While they face limitations related to system access, performance, and extensibility, their role in democratizing programming remains significant. Looking ahead, advances in web technologies, cloud infrastructure, and AI integration are likely to enhance the functionality of online interpreters, making them an increasingly practical option for broader development scenarios. This paper thus connects Python’s internal execution mechanisms with the evolving landscape of its runtime platforms, offering both technical insight and practical relevance for learners, educators, and developers alike.

References

[1] T.E.Oliphant,“PythonforScientificComputing,”ComputinginScience&Engineering,vol. 9, no. 3, pp. 10–20, May/Jun. 2007. DOI: 10.1109/MCSE.2007.58. [2] M.Lutz, LearningPython,5th ed.,Sebastopol,CA,USA:O’ReillyMedia,2013. [3] L.Hetland,PythonAlgorithms:MasteringBasicAlgorithmsin thePythonLanguage,2nded., Berkeley, CA, USA: Apress, 2014. [4] B.Slatkin,EffectivePython:90SpecificWaystoWriteBetterPython,2nded.,Boston,MA, USA: Addison-Wesley, 2019. [5] G.vanRossumandJ.deBoer,“InteractivelyTestingRemoteServersUsingthePython Programming Language,” CWI Quarterly, vol. 4, no. 4, pp. 283–303, 1991. [6] J.VanderPlas,PythonDataScienceHandbook:EssentialToolsforWorkingwithData, Sebastopol, CA, USA: O’Reilly Media, 2016. [7] I.Goodfellow,Y.Bengio,andA.Courville,DeepLearning,Cambridge,MA,USA:MIT Press, 2016. [8] S.vanderWalt,S.C.Colbert,andG.Varoquaux,“TheNumPyArray:AStructurefor [9] EfficientNumericalComputation,”ComputinginScience&Engineering,vol.13,no.2,pp.22–30, Mar./Apr. 2011. DOI: 10.1109/MCSE.2011.37. [10] W.McKinney,PythonforDataAnalysis:DataWranglingwithPandas,NumPy,andIPython, 2nd ed., Sebastopol, CA, USA: O’Reilly Media, 2017. [11] D.BeazleyandB.K.Jones,PythonCookbook:RecipesforMasteringPython3,3rded., Sebastopol, CA, USA: O’Reilly Media, 2013. [12] A.Godbole,DemystifyingComputers,NewDelhi,India:McGrawHillEducation,2013,ch. 9.1. [13] A.V.Aho,M.S.Lam,R.Sethi,andJ.D.Ullman,Compilers:Principles,Techniques,and Tools, 2nd ed., Boston, MA, USA: Addison-Wesley, 2006. [14] C.GhezziandM.Jazayeri,ProgrammingLanguageConcepts,3rded.,Wiley,2002. [15] J.Wiedemeieretal.,“PYLINGUAL:TowardPerfectDecompilationofEvolvingHigh- Level Languages,” unpublished. [16] C.Cooketal.,“BenefitsandPitfallsofJupyterNotebooksintheClassroom,”ACM Transactions on Computing Education, 2020. [17] M.Arayaetal.,“JOVIAL:Notebook-basedAstronomicalDataAnalysisintheCloud,” arXiv, 2018. arXiv:1805.00000. [18] S.Ghoshetal.,“EnhancingPython-basedCloudComputingEducationUsingJupyterHub and BinderHub,” in Proc. 2020 ACM Conf. Innovation and Technology in Computer Science Education, 2020. DOI: 10.1145/3341525.3387397. [19] A.Rule,A.Tabard,andJ.D.Hollan,“ExplorationandExplanationinComputational Notebooks,” in Proc. 2018 CHI Conf. Human Factors in Computing Systems, 2018. DOI: 10.1145/3173574.3173606. [20] B.E.GrangerandF.Pérez,“Jupyter:ThinkingandStorytellingWithCodeandData,” Journal of Open Source Software, vol. 6, no. 62, 2021. DOI: 10.21105/joss.03014. [21] J.M.Perkel,“WhyJupyterisDataScientists’ComputationalNotebookofChoice,”Nature, vol. 563, pp. 145–147, 2018. DOI: 10.1038/d41586-018-07196-1. [22] A.Headetal.,“ManagingMessesinComputationalNotebooks,”inProc.2019CHIConf. Human Factors in Computing Systems, 2019. DOI: 10.1145/3290605.3300500. [23] F. Pérez, B. E. Granger, et al., “The Jupyter Notebook as a Tool for Open Science,” in Positioningand Powerin AcademicPublishing:Players,AgentsandAgendas,Springer,2016. [24] M.Meieretal.,“pyodide:BringingthePythonRuntimetotheBrowserviaWebAssembly,” arXiv, 2021. arXiv:2107.11408. [25] M.Abadietal.,“WebAssemblyforCloud-NativePython:ExperienceswithPyodideand Beyond,” arXiv, 2023. arXiv:2304.06789. [26] B.Karreretal.,“AcceleratedPythonintheCloud:UseofGPUsandTPUsinJupyter Environments,” in Proc. 2022 IEEE Int. Conf. Cloud Engineering, 2022. DOI: 10.1109/IC2E53769.2022.00019. [27] A.Paszkeetal.,“PyTorch:AnImperativeStyle,High-PerformanceDeepLearningLibrary,” in Advances in Neural Information Processing Systems, vol. 32, pp. 8024–8035, 2019. [28] N.P.Jouppietal.,“In-DatacenterPerformanceAnalysisofaTensorProcessingUnit,”in Proc. 44th Annual Int. Symp. Computer Architecture, pp. 1–12, 2017. DOI: 10.1145/3079856.3080246. [29] A.Bisong,BuildingMachineLearningandDeepLearningModelsonGoogleCloud Platform: A Comprehensive Guide for Beginners, Berkeley, CA, USA: Apress, 2019. [30] E.Frankfordetal.,“AnOnlineIntegratedDevelopmentEnvironmentforAutomated Programming Assessment Systems,” arXiv, 2025. arXiv:2503.13127.

Copyright

Copyright © 2025 Gajanan Santosh Koleshwar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.