Beyond LLVM: Evaluating Fast Code Generation Alternatives for Query Compilation in PostgreSQL

The evolution of query compilation in database management systems traces back to System R, which pioneered a code generation scheme where small machine code fragments were stitched together to form a specialized routine to process a given SQL statement. Subsequent approaches shifted to generating C code, compiling it with system compilers like GCC into dynamic libraries, and loading them at runtime. The current state-of-the-art standard for dynamic query compilation is the LLVM framework, which bypasses frontend compiler overhead by directly generating intermediate representation, enabling machine-independent optimizations and efficient machine code generation. LLVM's resource-intensive nature, primarily designed as an optimizing compiler, however, can lead to compilation times that are orders of magnitude longer than query execution times, particularly problematic for queries with millisecond-level interpretation costs. This paper evaluates two lightweight code generation frameworks for x86-64 architecture as alternatives to LLVM in PostgreSQL, assessing their code generation speed and the quality of emitted machine code. We present a qualitative comparison with LLVM, analyzing trade-offs between compilation latency and runtime performance across databases of varying sizes. Experimental results demonstrate that lightweight code generation can not only outperform LLVM on small-scale datasets but also maintain competitive performance on larger ones.

1. PostgreSQL – open Source DBMS. [Online], Available at: https://www.postgresql.org, accessed 05.05.2025.

2. G. Graefe, “Volcano - An Extensible and Parallel Query Evaluation System,” IEEE Trans. Knowl. Data Eng., vol. 6, no. 1, pp. 120-135, 1994. DOI: 10.1109/69.273032. [Online], Available at:

3. https://doi.org/10.1109/69.273032.

4. T. Neumann, “Efficiently Compiling Efficient Query Plans for Modem Hardware,” PVLDB, vol. 4, no. 9, pp. 539-550, 2011. DOI: 10.14778/2002938.2002940. [Online], Available at: http://www.vldb.org/pvldb/vol4/p539-neumann.pdf.

5. PostgreSQL mailing lists, “[GSoC] Push-based query executor discussion.” [Online], Available at: https://www.postgresql.org/message-id/87shm1zfnz.fsf%40ispras.ru, accessed 05.05.2025.

6. A. Shaikhha, M. Dashti and C. Koch, «Push vs. pull-based loop fusion in query engines», CoRR, abs/1610.09166, 2016 arXiv: 1610.09166. Available at: https://arxiv.org/abs/1610.09166.

7. PostgreSQL 10.0 Release Notes, [Online] Available at: https://www.postgresql.org/docs/release/10.0, accessed 05.05.2025.

8. GitHub, Mirror of the official PostgreSQL GIT repository, “Faster expression evaluation and targetlist projection.” Commit SHA: b8d7f053c5c2bf2a7e8734fe3327f6a8bc711755 [Online], Available at: https://github.com/postgres/postgres/commit/b8d7f053c5c2bf2a7e8734fe3327f6a8bc711755, accessed 05.05.2025.

9. PGCon 2017. A. Freund Integrating Just In Time Compilation, [Online] Available at: https://www.pgcon.org/2017/schedule/attachments/462_jit-pgcon-2017-05-25.pdf, accessed 05.05.2025.

10. The LLVM Foundation, The LLVM Compiler Infrastructure. [Online] Available at: https://llvm.org, accessed 05.05.2025.

11. PGCon 2018. A. Freund A.The State of Postgres JIT – 2018 Edition, [Online] Available at: https://anarazel.de/talks/2018-06-01-pgcon-state-of-jit/state-of-jit.pdf, accessed 05.05.2025.

12. A. Engelke and T Schwarz. 2024. Compile-Time Analysis of Compiler Frameworks for Query Compilation. In Proceedings of the 2024 IEEE/ACM International Symposium on Code Generation and Optimization (CGO '24). IEEE Press, 233–244. DOI: https://doi.org/10.1109/CGO57630.2024.10444856.

13. A. Kohn, V. Leis, T. Neumann “Adaptive Execution of Compiled Queries”, 2018 IEEE 34th International Conference on Data Engineering (ICDE), Paris, France, 2018, pp. 197-208, DOI: 10.1109/ICDE.2018.00027.

14. T. Kersten, V. Leis, T. Neumann “Tidy Tuples and Flying Start: fast compilation and fast execution of relational queries in Umbra” The VLDB Journal 30, 5 (Sep 2021), 883–905. [Online], Available at: https://doi.org/10.1007/s00778-020-00643-4.

15. Transaction Processing Performance Council, "TPC-H Benchmark." [Online]. Available at: http://www.tpc.org/tpch, accessed 05.05.2025.

16. GCC, the GNU Compiler Collection, [Online], Available at: https://gcc.gnu.org/, accessed 05.05.2025.

17. V8 JavaScript Engine, [Online], Available at: https://v8.dev/, accessed 05.05.2025.

18. OpenJDK, The HotSpot Group, [Online], Available at: https://openjdk.org/groups/hotspot/, accessed 05.05.2025.

19. DotNet, RyuJIT compiler overview, [Online], Available at: https://github.com/dotnet/runtime/blob/main/docs/design/coreclr/jit/ryujit-overview.md, accessed 05.05.2025.

20. Parrot Virtual Machine, [Online], Available at: http://www.parrot.org/, accessed 05.05.2025.

21. MoarVM, a VM for NQP and Rakudo, [Online], Available at: https://www.moarvm.org/, accessed 05.05.2025.

22. GNU LibJIT – backend for DotGNU Portable.NET JIT engine, [Online], Available at: https://www.gnu.org/software/libjit, accessed 05.05.2025.

23. DynASM – dynamic assembler for code generation engines. [Online], Available at: https://luajit.org/dynasm.html, accessed 05.05.2025.

24. AsmJIT – lightweight library for low-latency machine code generation. [Online], Available at: https://asmjit.com, accessed 05.05.2025.

25. GNU lightning – library that generates assembly language code at run-time. [Online], Available at: https://www.gnu.org/software/lightning/, accessed 05.05.2025.

26. sljit – a low-level, platform-independent JIT compiler. [Online], Available at: https://zherczeg.github.io/sljit/, accessed 05.05.2025.

27. Xbyak – a JIT assembler for x86/x64 architectures. [Online], Available at: https://github.com/herumi/xbyak, accessed 05.05.2025.

28. M. Pall, “LuaJIT 2.0 intellectual property disclosure and research opportunities”, [Online], Available at: http://lua-users.org/lists/lua-l/2009-11/msg00089.html, accessed 05.05.2025.

29. Erlang/OTP documentation “BeamAsm, the Erlang JIT”, [Online] Available at: https://www.erlang.org/doc/apps/erts/beamasm.html, accessed 05.05.2025.