Integrated Register Allocation and Instruction Scheduling with Constraint Programming
2014 (English)Licentiate thesis, comprehensive summary (Other academic)
This dissertation proposes a combinatorial model, program representations, and constraint solving techniques for integrated register allocation and instruction scheduling in compiler back-ends. In contrast to traditional compilers based on heuristics, the proposed approach generates potentially optimal code by considering all trade-offs between interdependent decisions as a single optimization problem.
The combinatorial model is the first to handle a wide array of global register allocation subtasks, including spill code optimization, ultimate coalescing, register packing, and register bank assignment, as well as instruction scheduling for Very Long Instruction Word (VLIW) processors. The model is based on three novel, complementary program representations: Linear Static Single Assignment for global register allocation; copy extension for spilling, basic coalescing, and register bank assignment; and alternative temporaries for spill code optimization and ultimate coalescing. Solving techniques are proposed that exploit the program representation properties for scalability.
The model, program representations, and solving techniques are implemented in Unison, a code generator that delivers potentially optimal code while scaling to medium-size functions. Thorough experiments show that Unison: generates faster code (up to 41% with a mean improvement of 7%) than LLVM (a state-of-the-art compiler) for Hexagon (a challenging VLIW processor), generates code that is competitive with LLVM for MIPS32 (a simpler RISC processor), is robust across different benchmarks such as MediaBench and SPECint 2006, scales up to medium-size functions of up to 1000 instructions, and adapts easily to different optimization criteria.
The contributions of this dissertation are significant. They lead to a combinatorial approach for integrated register allocation and instruction scheduling that is, for the first time, practical (it robustly scales to medium-size functions) and effective (it yields better code than traditional heuristic approaches).
Place, publisher, year, edition, pages
Stockholm, Sweden: KTH Royal Institute of Technology, 2014. , 48 p.
TRITA-ICT-ECS AVH, ISSN 1653-6363 ; 14:13
Research subject Computer Science
IdentifiersURN: urn:nbn:se:kth:diva-154599ISBN: 978-91-7595-311-3OAI: oai:DiVA.org:kth-154599DiVA: diva2:758121
2014-11-27, Sal A, Electrum, Kistagången 16, Kista, Stockholm, 14:00 (English)
van Beek, Peter, Professor
Schulte, Christian, ProfessorCarlsson, MatsSander, Ingo, Associate Professor
QC 201411172014-11-172014-10-242014-11-17Bibliographically approved
List of papers