Offline library adaptation using automatically generated heuristics

Mesmay, Frédéric de; Voronenko, Yevgen; Püschel, Markus

doi:10.1109/ipdps.2010.5470479

Cited by 17 publications

(11 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Examples of such systems include ATLAS [37], PhiPAC [38], and OSKI [11] for linear algebra, [39], FFTW [40] and SPIRAL [41] for signal processing, [42], [43], [44] for stencil computations, and [45] for sorting.…”

Section: Related Workmentioning

confidence: 99%

Nitro: A Framework for Adaptive Code Variant Tuning

Muralidharan

Shantharam

Hall

et al. 2014

2014 IEEE 28th International Parallel and Distributed Processing Symposium

View full text Add to dashboard Cite

Abstract-Autotuning systems intelligently navigate a search space of possible implementations of a computation to find the implementation(s) that best meets a specific optimization criteria, usually performance. This paper describes Nitro, a programmer-directed autotuning framework that facilitates tuning of code variants, or alternative implementations of the same computation. Nitro provides a library interface that permits programmers to express code variants along with metainformation that aids the system in selecting among the set of variants at run time. Machine learning is employed to build a model through training on this meta-information, so that when a new input is presented, Nitro can consult the model to select the appropriate variant. In experiments with five real-world irregular GPU benchmarks from sparse numerical methods, graph computations and sorting, Nitro-tuned variants achieve over 93% of the performance of variants selected through exhaustive search. Further, we describe optimizations and heuristics in Nitro that substantially reduce training time and other overheads.

show abstract

Section: Related Workmentioning

confidence: 99%

Nitro: A Framework for Adaptive Code Variant Tuning

Muralidharan

Shantharam

Hall

et al. 2014

2014 IEEE 28th International Parallel and Distributed Processing Symposium

View full text Add to dashboard Cite

show abstract

“…Compiler researchers have proposed to use machine learning models to focus search on beneficial areas of the optimization search space [6,10,11,14,20,25]. An important step in using these models is to characterize (or construct features for) the programs being optimized.…”

Section: Characterizing the Programmentioning

confidence: 99%

Using graph-based program characterization for predictive modeling

Park

Cavazos

Álvarez

2012

Proceedings of the Tenth International Symposium on Code Generation and Optimization

View full text Add to dashboard Cite

Using machine learning has proven effective at choosing the right set of optimizations for a particular program. For machine learning techniques to be most effective, compiler writers have to develop expressive means of characterizing the program being optimized. The current state-of-the-art techniques for characterizing programs include using a fixed-length feature vector of either source code features extracted during compile time or performance counters collected when running the program. For the problem of identifying optimizations to apply, models constructed using performance counter characterizations of a program have been shown to outperform models constructed using source code features. However, collecting performance counters requires running the program multiple times, and this "dynamic" method of characterizing programs can be specific to inputs of the program. It would be preferable to have a method of characterizing programs that is as expressive as performance counter features, but that is "static" like source code features and therefore does not require running the program.In this paper, we introduce a novel way of characterizing programs using a graph-based characterization, which uses the program's intermediate representation and an adapted learning algorithm to predict good optimization sequences. To evaluate different characterization techniques, we focus on loop-intensive programs and construct prediction models that drive polyhedral optimizations, such as auto-parallelism and various loop transformation.We show that our graph-based characterization technique outperforms three current state-of-the-art characterization techniques found in the literature. By using the sequences predicted to be the best by our graph-based model, we achieved up to 73% of the speedup achievable in our search space for a particular platform, whereas we could only achieve up to 59% by other state-of-the-art techniques we evaluated.

show abstract

“…However, many of the proposed techniques for exploring optimizations (e.g., genetic algorithms [5], random search [21], statistical techniques [18], or exhaustive search [20]) are expensive, which limits their practical use. This has led compiler researchers to propose using "intelligent" prediction models that focus exploration to beneficial areas of the optimization search space [4,7,11,13,22]. Prediction models can reduce the cost of finding good optimizations, but increase complexity in the design of the search function because models require characterizing the program being optimized (e.g., with source code features or performance counters), generation of training data, and a training phase.…”

Section: Learning To Optimizementioning

confidence: 99%

An evaluation of different modeling techniques for iterative compilation

Park

Kulkarni

Cavazos

2011

Proceedings of the 14th International Conference on Compilers, Architectures and Synthesis for Embedded Systems

View full text Add to dashboard Cite

Iterative compilation techniques, which involve iterating over different sets of optimizations, have proven useful in helping compilers choose the right set of optimizations for a given program. However, compilers typically have a large number of optimizations to choose from, making it impossible to iterate over a significant fraction of the entire optimization search space. Recent research has proposed to "intelligently" iterate over the optimization search space using predictive methods. In particular, state-the-art methods in iterative compilation use characteristics of the code being optimized to predict good optimization sequences to evaluate. Thus, an important step in developing predictive methods for compilation is deciding how to model the problem of choosing the right optimizations.In this paper, we evaluate three different ways of modeling the problem of choosing the right optimization sequences using machine learning techniques. We evaluate a novel prediction modeling technique, namely a tournament predictor, that is able to effectively predict good optimization sequences. We show that our tournament predictor can outperform current state-of-the-art predictors and the most aggressive setting of the Open64 compiler (-Ofast) on an average by 75% in just 10 iterations over a set of embedded and scientific kernels. Moreover, using our tournament predictor, we achieved on average 10% improvement for a set of MiBench applications.

show abstract

Offline library adaptation using automatically generated heuristics

Cited by 17 publications

References 16 publications

Nitro: A Framework for Adaptive Code Variant Tuning

Nitro: A Framework for Adaptive Code Variant Tuning

Using graph-based program characterization for predictive modeling

An evaluation of different modeling techniques for iterative compilation

Contact Info

Product

Resources

About