Impact of Reverse Computing on Information Locality in Register Allocation for High Performance Computing

Authors:
Mouad Bahi;Christine Eisenbeis
Affiliations:
INRIA Saclay Île-de-France, Orsay, France and LRI, Université de Paris-Sud 11, Orsay, France and LERMA, Observatoire de Paris, Paris, France;INRIA Saclay Île-de-France, Orsay, France and LRI, Université de Paris-Sud 11, Orsay, France
Venue:
International Journal of Parallel Programming
Year:
2014

Citing 14
Cited 0

Time/space trade-offs for reversible computation

SIAM Journal on Computing
Rematerialization

PLDI '92 Proceedings of the ACM SIGPLAN 1992 conference on Programming language design and implementation
Register allocation via graph coloring

Register allocation via graph coloring
Efficient optimistic parallel simulations using reverse computation

PADS '99 Proceedings of the thirteenth workshop on Parallel and distributed simulation
A systematic generation of initial register-reuse chains for dependence minimization

ACM SIGPLAN Notices
Minimum Register Instruction Sequencing to Reduce Register Spills in Out-of-Order Issue Superscalar Architectures

IEEE Transactions on Computers
NREVERSAL of Fortune - The Thermodynamics of Garbage Collection

IWMM '92 Proceedings of the International Workshop on Memory Management
URSA: A Unified ReSource Allocator for Registers and Functional Units in VLIW Architectures

PACT '93 Proceedings of the IFIP WG10.3. Working Conference on Architectures and Compilation Techniques for Fine and Medium Grain Parallelism
Register allocation & spilling via graph coloring

SIGPLAN '82 Proceedings of the 1982 SIGPLAN symposium on Compiler construction
Using Reversible Computing To Achieve Fail-Safety

ISSRE '97 Proceedings of the Eighth International Symposium on Software Reliability Engineering
Value-driven redundancy elimination

Value-driven redundancy elimination
Compiler Optimizations to Reduce Security Overhead

Proceedings of the International Symposium on Code Generation and Optimization
Logical reversibility of computation

IBM Journal of Research and Development
Rematerialization-based register allocation through reverse computing

Proceedings of the 8th ACM International Conference on Computing Frontiers

Quantified Score

Hi-index	0.00

Visualization

Abstract

Reversible computing aims at keeping all information on input and intermediate values available at any step of the computation, making information virtually present everywhere. Rematerialization in register allocation amounts to recomputing values instead of spilling them in memory when registers run out. In this paper we detail a heuristic algorithm for exploiting reverse computing for register materialization. This improves information locality as it provides more opportunities for retrieving data. Rematerialization adds instructions and we show on one specifically designed example that reverse computing may alleviate the impact of these additional instructions on performance. We also show how thread parallelism may be optimized on GPUs by performing register allocation with reverse recomputing that increases the number of threads per Streaming Multiprocessor. This is done on the main kernel of Lattice Quantum Chromo Dynamics simulation program where we gain a 11 % speedup.