Vectorization past dependent branches through speculation

Authors:
Majedul Haque Sujon;R. Clint Whaley;Qing Yi
Affiliations:
University of Texas at San Antonio, San Antonio, TX, USA;Louisiana State University, Baton Rouge, LA, USA;University of Colorado Colorado Springs, Colorado Springs, CO, USA
Venue:
PACT '13 Proceedings of the 22nd international conference on Parallel architectures and compilation techniques
Year:
2013

Citing 20
Cited 0

Compiler transformations for high-performance computing

ACM Computing Surveys (CSUR)
The LRPD test: speculative run-time parallelization of loops with privatization and reduction parallelization

PLDI '95 Proceedings of the ACM SIGPLAN 1995 conference on Programming language design and implementation
Exploiting superword level parallelism with multimedia instruction sets

PLDI '00 Proceedings of the ACM SIGPLAN 2000 conference on Programming language design and implementation
Speculative dynamic vectorization

ISCA '02 Proceedings of the 29th annual international symposium on Computer architecture
Automatic intra-register vectorization for the Intel architecture

International Journal of Parallel Programming
A Vectorizing Compiler for Multimedia Extensions

International Journal of Parallel Programming
Subword Parallelism with MAX-2

IEEE Micro
A compiler framework for speculative analysis and optimizations

PLDI '03 Proceedings of the ACM SIGPLAN 2003 conference on Programming language design and implementation
A Unified Compiler Framework for Control and Data Speculation

PACT '00 Proceedings of the 2000 International Conference on Parallel Architectures and Compilation Techniques
Vectorization for SIMD architectures with alignment constraints

Proceedings of the ACM SIGPLAN 2004 conference on Programming language design and implementation
Compiler Estimation of Load Imbalance Overhead in Speculative Parallelization

Proceedings of the 13th International Conference on Parallel Architectures and Compilation Techniques
Superword-Level Parallelism in the Presence of Control Flow

Proceedings of the international symposium on Code generation and optimization
Tuning High Performance Kernels through Empirical Compilation

ICPP '05 Proceedings of the 2005 International Conference on Parallel Processing
Automated empirical optimization of high performance floating point kernels

Automated empirical optimization of high performance floating point kernels
Auto-vectorization of interleaved data for SIMD

Proceedings of the 2006 ACM SIGPLAN conference on Programming language design and implementation
Software behavior oriented parallelization

Proceedings of the 2007 ACM SIGPLAN conference on Programming language design and implementation
Introducing Control Flow into Vectorized Code

PACT '07 Proceedings of the 16th International Conference on Parallel Architecture and Compilation Techniques
Trace Scheduling: A Technique for Global Microcode Compaction

IEEE Transactions on Computers
Outer-loop vectorization: revisited for short SIMD architectures

Proceedings of the 17th international conference on Parallel architectures and compilation techniques
Whole-function vectorization

CGO '11 Proceedings of the 9th Annual IEEE/ACM International Symposium on Code Generation and Optimization

Quantified Score

Hi-index	0.00

Visualization

Abstract

Modern architectures increasingly rely on SIMD vectorization to improve performance for floating point intensive scientific applications. However, existing compiler optimization techniques for automatic vectorization are inhibited by the presence of unknown control flow surrounding partially vectorizable computations. In this paper, we present a new approach, speculative vectorization, which speculates past dependent branches to aggressively vectorize computational paths that are expected to be taken frequently at runtime, while simply restarting the calculation using scalar instructions when the speculation fails. We have integrated our technique in an iterative optimizing compiler and have employed empirical tuning to select the profitable paths for speculation. When applied to optimize 9 floating-point benchmarks, our optimizing compiler has achieved up to 6.8X speedup for single precision and 3.4X for double precision kernels using AVX, while vectorizing some operations considered not vectorizable by prior techniques.