Document Open Access Logo

Eliminate Branches by Melding IR Instructions

Authors Yuze Li , Srinivasan Ramachandra Sharma , Charitha Saumya , Ali R. Butt , Kirshanthan Sundararajah

PDF

File

Thumbnail PDF
PDF
LIPIcs.ECOOP.2026.18.pdf
  • Filesize: 1.34 MB
  • 30 pages

Document Identifiers

Related Versions

Subject Classification

ACM Subject Classification
  • Software and its engineering → Compilers
  • Software and its engineering → Software performance
Keywords
  • Control-flow Transformation
  • Branch Elimination

Metrics

Abstract

Branch mispredictions cause catastrophic performance penalties in modern processors, leading to performance loss. While hardware predictors and profile-guided techniques exist, data-dependent branches with irregular access patterns remain challenging. Traditional if-conversion eliminates branches via software predication but faces limitations on architectures like x86. It often fails on paths containing memory instructions or incurs excessive instruction overhead by fully speculating large branch bodies. This paper presents MERIT (MElding IR InsTructions), a compiler transformation that eliminates branches by aligning and melding similar operations from divergent paths at the IR instruction level. By observing that divergent paths often perform structurally similar operations with different operands, MERIT adapts sequence alignment to discover merging opportunities and employs safe operand-level guarding to ensure semantic correctness without hardware predication. Implemented as an LLVM pass and evaluated on 102 programs from four benchmark suites, MERIT achieves a 1.52× geometric mean speedup on 24 branch-heavy microbenchmarks (peak 32× on toUpper) and reduces branch mispredictions by 48.9% on average. On more realistic workloads - SPECrate 2017 (16 benchmarks), SQLite TPC-H (22 queries), and CPython pyperformance (40 benchmarks) - MERIT with profile-guided function selection achieves a positive geomean (∼1.01×) across all three suites, demonstrating consistent improvement without regressions.

Cite As Get BibTex

Yuze Li, Srinivasan Ramachandra Sharma, Charitha Saumya, Ali R. Butt, and Kirshanthan Sundararajah. Eliminate Branches by Melding IR Instructions. In 40th European Conference on Object-Oriented Programming (ECOOP 2026). Leibniz International Proceedings in Informatics (LIPIcs), Volume 372, pp. 18:1-18:30, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026) https://doi.org/10.4230/LIPIcs.ECOOP.2026.18

Author Details

Yuze Li
  • Virginia Tech, Blacksburg, VA, USA
Srinivasan Ramachandra Sharma
  • Virginia Tech, Blacksburg, VA, USA
Charitha Saumya
  • Intel Corporation, Santa Clara, CA, USA
Ali R. Butt
  • Virginia Tech, Blacksburg, VA, USA
Kirshanthan Sundararajah
  • Virginia Tech, Blacksburg, VA, USA

Funding

This work is sponsored in part by the NSF under the grants: CSR-2106634 and CSR-2312785.

Supplementary Materials

References

  1. Andreas Abel and Jan Reineke. uops.info: Characterizing latency, throughput, and port usage of instructions on intel microarchitectures. In Iris Bahar, Maurice Herlihy, Emmett Witchel, and Alvin R. Lebeck, editors, Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems, ASPLOS 2019, Providence, RI, USA, April 13-17, 2019, pages 673-686. ACM, 2019. doi:10.1145/3297858.3304062. Google Scholar
  2. KA Jordan Alexander and Kim Nikolai. Ir-level versus machine-level if-conversion for predicated architectures. In Proceedings of the 10th Workshop on Optimizations for DSP and Embedded Systems (ODES-10), pages 3-10, 2012. doi:10.1145/2443608.2443611. Google Scholar
  3. John R. Allen, Ken Kennedy, Carrie Porterfield, and Joe D. Warren. Conversion of control dependence to data dependence. In John R. Wright, Larry Landweber, Alan J. Demers, and Tim Teitelbaum, editors, Conference Record of the Tenth Annual ACM Symposium on Principles of Programming Languages, Austin, Texas, USA, January 1983, pages 177-189. ACM Press, 1983. doi:10.1145/567067.567085. Google Scholar
  4. David I. August, Wen-mei W. Hwu, and Scott A. Mahlke. A framework for balancing control flow and predication. In Mark Smotherman and Tom Conte, editors, Proceedings of the Thirtieth Annual IEEE/ACM International Symposium on Microarchitecture, MICRO 30, Research Triangle Park, North Carolina, USA, December 1-3, 1997, pages 92-103. IEEE, ACM/IEEE Computer Society, 1997. doi:10.1109/MICRO.1997.645801. Google Scholar
  5. David I. August, Wen-mei W. Hwu, and Scott A. Mahlke. The partial reverse if-conversion framework for balancing control flow and predication. Int. J. Parallel Program., 27(5):381-423, 1999. doi:10.1023/A:1018787007582. Google Scholar
  6. Grant Ayers, Nayana Prasad Nagendra, David I. August, Hyoun Kyu Cho, Svilen Kanev, Christos Kozyrakis, Trivikram Krishnamurthy, Heiner Litz, Tipp Moseley, and Parthasarathy Ranganathan. Asmdb: understanding and mitigating front-end stalls in warehouse-scale computers. In Srilatha Bobbie Manne, Hillery C. Hunter, and Erik R. Altman, editors, Proceedings of the 46th International Symposium on Computer Architecture, ISCA 2019, Phoenix, AZ, USA, June 22-26, 2019, pages 462-473. ACM, 2019. doi:10.1145/3307650.3322234. Google Scholar
  7. Rastislav Bodík, Rajiv Gupta, and Mary Lou Soffa. Interprocedural conditional branch elimination. In Marina C. Chen, Ron K. Cytron, and A. Michael Berman, editors, Proceedings of the ACM SIGPLAN '97 Conference on Programming Language Design and Implementation (PLDI), Las Vegas, Nevada, USA, June 15-18, 1997, pages 146-158. ACM, 1997. doi:10.1145/258915.258929. Google Scholar
  8. James Bucek, Klaus-Dieter Lange, and Jóakim v. Kistowski. Spec cpu2017: Next-generation compute benchmark. In Companion of the 2018 ACM/SPEC International Conference on Performance Engineering, pages 41-42, 2018. doi:10.1145/3185768.3185771. Google Scholar
  9. Dehao Chen, Xinliang David Li, and Tipp Moseley. Autofdo: automatic feedback-directed optimization for warehouse-scale applications. In Björn Franke, Youfeng Wu, and Fabrice Rastello, editors, Proceedings of the 2016 International Symposium on Code Generation and Optimization, CGO 2016, Barcelona, Spain, March 12-18, 2016, pages 12-23. ACM, 2016. doi:10.1145/2854038.2854044. Google Scholar
  10. Chris Cummins, Volker Seeker, Dejan Grubisic, Baptiste Rozière, Jonas Gehring, Gabriel Synnaeve, and Hugh Leather. LLM compiler: Foundation language models for compiler optimization. In Daniel Kluss, Sara Achour, and Jens Palsberg, editors, Proceedings of the 34th ACM SIGPLAN International Conference on Compiler Construction, CC 2025, Las Vegas, NV, USA, March 1-2, 2025, pages 141-153. ACM, 2025. doi:10.1145/3708493.3712691. Google Scholar
  11. Yangruibo Ding, Benjamin Steenhoek, Kexin Pei, Gail E. Kaiser, Wei Le, and Baishakhi Ray. TRACED: execution-aware pre-training for source code. In Proceedings of the 46th IEEE/ACM International Conference on Software Engineering, ICSE 2024, Lisbon, Portugal, April 14-20, 2024, pages 36:1-36:12. ACM, 2024. doi:10.1145/3597503.3608140. Google Scholar
  12. Agner Fog. The microarchitecture of intel, amd and via cpus: An optimization guide for assembly programmers and compiler makers. Software optimization resources, 2016. URL: https://www.agner.org/optimize/microarchitecture.pdf.
  13. Kevin P. Gaffney, Martin Prammer, Laurence C. Brasfield, D. Richard Hipp, Dan R. Kennedy, and Jignesh M. Patel. Sqlite: Past, present, and future. Proc. VLDB Endow., 15(12):3535-3547, 2022. doi:10.14778/3554821.3554842. Google Scholar
  14. Wen-mei W. Hwu, Scott A. Mahlke, William Y. Chen, Pohua P. Chang, Nancy J. Warter, Roger A. Bringmann, Roland G. Ouellette, Richard E. Hank, Tokuzo Kiyohara, Grant E. Haab, John G. Holm, and Daniel M. Lavery. The superblock: An effective technique for VLIW and superscalar compilation. J. Supercomput., 7(1-2):229-248, 1993. doi:10.1007/BF01205185. Google Scholar
  15. Yasuo Ishii, Jaekyu Lee, Krishnendra Nathella, and Dam Sunwoo. Rebasing instruction prefetching: An industry perspective. IEEE Comput. Archit. Lett., 19(2):147-150, 2020. doi:10.1109/LCA.2020.3035068. Google Scholar
  16. Daniel A Jiménez. Multiperspective perceptron predictor. In 5th JILP Workshop on Computer Architecture Competitions (JWAC-5): Championship Branch Prediction (CBP-5), 2016. URL: https://jilp.org/cbp2016/paper/DanielJimenez1.pdf.
  17. Daniel A. Jiménez. Multiperspective perceptron predictor with TAGE. In 5th JILP Workshop on Computer Architecture Competitions (JWAC-5): Championship Branch Prediction (CBP-5), 2016. URL: https://jilp.org/cbp2016/paper/DanielJimenez2.pdf.
  18. Daniel A. Jiménez and Calvin Lin. Dynamic branch prediction with perceptrons. In Proceedings of the Seventh International Symposium on High-Performance Computer Architecture (HPCA'01), Nuevo Leone, Mexico, January 20-24, 2001, pages 197-206. IEEE Computer Society, 2001. doi:10.1109/HPCA.2001.903263. Google Scholar
  19. Tanvir Ahmed Khan, Nathan Brown, Akshitha Sriraman, Niranjan K. Soundararajan, Rakesh Kumar, Joseph Devietti, Sreenivas Subramoney, Gilles A. Pokam, Heiner Litz, and Baris Kasikci. Twig: Profile-guided BTB prefetching for data center applications. In MICRO '21: 54th Annual IEEE/ACM International Symposium on Microarchitecture, Virtual Event, Greece, October 18-22, 2021, pages 816-829. ACM, 2021. doi:10.1145/3466752.3480124. Google Scholar
  20. Tanvir Ahmed Khan, Muhammed Ugur, Krishnendra Nathella, Dam Sunwoo, Heiner Litz, Daniel A. Jiménez, and Baris Kasikci. Whisper: Profile-guided branch misprediction elimination for data center applications. In 55th IEEE/ACM International Symposium on Microarchitecture, MICRO 2022, Chicago, IL, USA, October 1-5, 2022, pages 19-34. IEEE, IEEE, 2022. doi:10.1109/MICRO56248.2022.00017. Google Scholar
  21. Hyesoon Kim, Onur Mutlu, Yale N. Patt, and Jared Stark. Wish branches: Enabling adaptive and aggressive predicated execution. IEEE Micro, 26(1):48-58, 2006. doi:10.1109/MM.2006.27. Google Scholar
  22. Chit-Kwan Lin and Stephen J. Tarsa. Branch prediction is not a solved problem: Measurements, opportunities, and future directions. CoRR, abs/1906.08170, 2019. doi:10.48550/arXiv.1906.08170. Google Scholar
  23. LLVM Project. llvm::simplifycfgpass class reference, 2025. LLVM Doxygen documentation. Accessed: 2026-02-13. URL: https://llvm.org/doxygen/classllvm_1_1SimplifyCFGPass.html.
  24. Scott A. Mahlke, David C. Lin, William Y. Chen, Richard E. Hank, and Roger A. Bringmann. Effective compiler support for predicated execution using the hyperblock. In Wen-mei W. Hwu, editor, Proceedings of the 25th Annual International Symposium on Microarchitecture, Portland, Oregon, USA, November 1992, pages 45-54. ACM / IEEE Computer Society, 1992. doi:10.1109/MICRO.1992.696999. Google Scholar
  25. Pierre Michaud. An alternative tage-like conditional branch predictor. ACM Trans. Archit. Code Optim., 15(3):30:1-30:23, 2018. doi:10.1145/3226098. Google Scholar
  26. Frank Mueller and David B. Whalley. Avoiding conditional branches by code replication. In David W. Wall, editor, Proceedings of the ACM SIGPLAN'95 Conference on Programming Language Design and Implementation (PLDI), La Jolla, California, USA, June 18-21, 1995, pages 56-66. ACM, 1995. doi:10.1145/207110.207116. Google Scholar
  27. Dorit Nuzman, Ayal Zaks, and Ziv Ben-Zion. If-convert as early as you must. In Proceedings of the 33rd ACM SIGPLAN International Conference on Compiler Construction, CC 2024, pages 26-38, New York, NY, USA, 2024. Association for Computing Machinery. doi:10.1145/3640537.3641562. Google Scholar
  28. Maksim Panchenko, Rafael Auler, Bill Nell, and Guilherme Ottoni. BOLT: A practical binary optimizer for data centers and beyond. In Mahmut Taylan Kandemir, Alexandra Jimborean, and Tipp Moseley, editors, IEEE/ACM International Symposium on Code Generation and Optimization, CGO 2019, Washington, DC, USA, February 16-20, 2019, pages 2-14. IEEE, IEEE, 2019. doi:10.1109/CGO.2019.8661201. Google Scholar
  29. Maksim Panchenko, Rafael Auler, Laith Sakka, and Guilherme Ottoni. Lightning BOLT: powerful, fast, and scalable binary optimization. In Aaron Smith, Delphine Demange, and Rajiv Gupta, editors, CC '21: 30th ACM SIGPLAN International Conference on Compiler Construction, Virtual Event, Republic of Korea, March 2-3, 2021, pages 119-130. ACM, 2021. doi:10.1145/3446804.3446843. Google Scholar
  30. Joseph CH Park and Mike Schlansker. On predicated execution. Hewlett-Packard Laboratories Palo Alto, California, 1991. URL: https://shiftleft.com/mirrors/www.hpl.hp.com/techreports/91/HPL-91-58.pdf.
  31. Glenn Reinman, Brad Calder, and Todd M. Austin. Fetch directed instruction prefetching. In Ronny Ronen, Matthew K. Farrens, and Ilan Y. Spillinger, editors, Proceedings of the 32nd Annual IEEE/ACM International Symposium on Microarchitecture, MICRO 32, Haifa, Israel, November 16-18, 1999, pages 16-27. IEEE, ACM/IEEE Computer Society, 1999. doi:10.1109/MICRO.1999.809439. Google Scholar
  32. Rodrigo C. O. Rocha, Pavlos Petoumenos, Zheng Wang, Murray Cole, and Hugh Leather. Function merging by sequence alignment. In 2019 IEEE/ACM International Symposium on Code Generation and Optimization (CGO), pages 149-163. IEEE, 2019. doi:10.1109/CGO.2019.8661174. Google Scholar
  33. Rodrigo C. O. Rocha, Pavlos Petoumenos, Zheng Wang, Murray Cole, and Hugh Leather. Effective function merging in the SSA form. In Proceedings of the 41st ACM SIGPLAN Conference on Programming Language Design and Implementation, pages 854-868, 2020. doi:10.1145/3385412.3386030. Google Scholar
  34. Rodrigo C. O. Rocha, Charitha Saumya, Kirshanthan Sundararajah, Pavlos Petoumenos, Milind Kulkarni, and Michael F. P. O'Boyle. Hybf: A hybrid branch fusion strategy for code size reduction. In Clark Verbrugge, Ondrej Lhoták, and Xipeng Shen, editors, Proceedings of the 32nd ACM SIGPLAN International Conference on Compiler Construction, CC 2023, Montréal, QC, Canada, February 25-26, 2023, pages 156-167. ACM, 2023. doi:10.1145/3578360.3580267. Google Scholar
  35. Charitha Saumya, Kirshanthan Sundararajah, and Milind Kulkarni. DARM: control-flow melding for SIMT thread divergence reduction. In Jae W. Lee, Sebastian Hack, and Tatiana Shpeisman, editors, IEEE/ACM International Symposium on Code Generation and Optimization, CGO 2022, Seoul, Korea, Republic of, April 2-6, 2022, pages 1-13. IEEE, IEEE, 2022. doi:10.1109/CGO53902.2022.9741285. Google Scholar
  36. André Seznec. A 64 kbytes isl-tage branch predictor. In JWAC-2: Championship Branch Prediction, 2011. URL: https://jilp.org/jwac-2/program/cbp3_03_seznec.pdf.
  37. André Seznec. Tage-sc-l branch predictors. In JILP-Championship Branch Prediction, 2014. URL: https://jilp.org/cbp2014/paper/AndreSeznec.pdf.
  38. André Seznec. Tage-sc-l branch predictors again. In 5th JILP Workshop on Computer Architecture Competitions (JWAC-5): Championship Branch Prediction (CBP-5), 2016. URL: https://jilp.org/cbp2016/paper/AndreSeznecLimited.pdf.
  39. André Seznec and Pierre Michaud. De-aliased hybrid branch predictors. PhD thesis, INRIA, 1999. doi:10.1145/2155620.2155635. Google Scholar
  40. Temple F Smith, Michael S Waterman, et al. Identification of common molecular subsequences. Journal of molecular biology, 147(1):195-197, 1981. doi:10.1016/0022-2836(81)90087-5. Google Scholar
  41. Shixin Song, Tanvir Ahmed Khan, Sara Mahdizadeh-Shahri, Akshitha Sriraman, Niranjan K. Soundararajan, Sreenivas Subramoney, Daniel A. Jiménez, Heiner Litz, and Baris Kasikci. Thermometer: profile-guided btb replacement for data center applications. In Valentina Salapura, Mohamed Zahran, Fred Chong, and Lingjia Tang, editors, ISCA '22: The 49th Annual International Symposium on Computer Architecture, New York, New York, USA, June 18 - 22, 2022, pages 742-756. ACM, 2022. doi:10.1145/3470496.3527430. Google Scholar
  42. Victor Stinner. The python performance benchmark suite — python performance benchmark suite 1.0.6 documentation. URL: https://pyperformance.readthedocs.io/.
  43. Transaction Processing Performance Council (TPC). TPC benchmark^tm A: standard specification. In Jim Gray, editor, The Benchmark Handbook for Database and Transaction Systems (2nd Edition), Revision 3.0.1. Morgan Kaufmann, May 1993. URL: https://www.tpc.org/tpc_documents_current_versions/pdf/tpc-h_v2.17.1.pdf.
  44. Li Wang, Hong An, Yongqing Ren, and Yaobin Wang. Profile guided optimization for dataflow predication. In 13th Asia-Pacific Computer Systems Architecture Conference, ACSAC 2008, Hsinchu, China, August 4-6, 2008, pages 1-8. IEEE, IEEE Computer Society, 2008. doi:10.1109/APCSAC.2008.4625471. Google Scholar
  45. Nancy J. Warter, Scott A. Mahlke, Wen-Mei W. Hwu, and B. Ramakrishna Rau. Reverse if-conversion. In Proceedings of the ACM SIGPLAN 1993 Conference on Programming Language Design and Implementation (PLDI), pages 290-299, 1993. doi:10.1145/155090.155118. Google Scholar
  46. Siavash Zangeneh, Lizy K. John, and Andreas Gerstlauer. Branchnet: A convolutional neural network to predict hard-to-predict branches. In Proceedings of the 53rd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO), pages 118-130. IEEE, 2020. doi:10.1109/MICRO50266.2020.00020. Google Scholar
  47. Eric J Zimmerman. Profile-directed If-Conversion in Superscalar Microprocessors. PhD thesis, University of Illinois at Urbana-Champaign, 2005. URL: https://llvm.org/pubs/2005-07-ZimmermanMSThesis.pdf.
Any Issues?
X

Feedback on the Current Page

CAPTCHA

Thanks for your feedback!

Feedback submitted to Dagstuhl Publishing

Could not send message

Please try again later or send an E-mail
© 2023-2026 Schloss Dagstuhl – LZI GmbH Schloss Dagstuhl – LZI GmbH About DROPS Imprint Privacy Contact