Skip to main content

Über dieses Buch

This book constitutes the proceedings of the 24th International Conference on Compiler Construction, CC 2015, held as part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2015, in London, UK, in April 2015.

The 11 papers presented in this volume were carefully reviewed and selected from 34 submissions. They deal with compiler engineering and compiling techniques; compiler analysis and optimisation and formal techniques in compilers. The book also contains one invited talk in full-paper length.



Invited Paper


A Graphical Model for Context-Free Grammar Parsing

In the compiler literature, parsing algorithms for context-free grammars are presented using string rewriting systems or abstract machines such as pushdown automata. Unfortunately, the resulting descriptions can be baroque, and even a basic understanding of some parsing algorithms, such as Earley’s algorithm for general context-free grammars, can be elusive. In this paper, we present a graphical representation of context-free grammars called the Grammar Flow Graph (GFG) that permits parsing problems to be phrased as path problems in graphs; intuitively, the GFG plays the same role for context-free grammars that nondeterministic finite-state automata play for regular grammars. We show that the GFG permits an elementary treatment of Earley’s algorithm that is much easier to understand than previous descriptions of this algorithm. In addition, look-ahead computation can be expressed as a simple inter-procedural dataflow analysis problem, providing an unexpected link between front-end and back-end technologies in compilers. These results suggest that the GFG can be a new foundation for the study of context-free grammars.
Keshav Pingali, Gianfranco Bilardi

Compiler Engineering and Compiling Techniques


A Refactoring Library for Scala Compiler Extensions

Compiler plugins enable languages to be extended with new functionality by adding compiler passes that perform additional static checking, code generation, or code transformations. However, compiler plugins are often difficult to build. A plugin can perform arbitrary code transformations, easily allowing a developer to generate incorrect code. Moreover, the base compiler assumes many complex, sometimes undocumented invariants, requiring plugin developers to acquire intimate knowledge of the design and implementation of the compiler. To address these issues in the context of the Scala compiler plugin framework, we introduce Piuma. Piuma is a library that provides, first, an API to perform many common refactoring tasks needed by plugin writers, and, second, a DSL to eliminate much of the boilerplate code required for plugin development. We demonstrate the usefulness of our library by implementing five diverse compiler plugins. We show that, using Piuma, plugins require less code and are easier to understand than plugins developed using the base Scala compiler plugin API.
Amanj Sherwany, Nosheen Zaza, Nathaniel Nystrom

Feature-Specific Profiling

High-level languages come with significant readability and maintainability benefits. Their performance costs, however, are usually not predictable, at least not easily. Programmers may accidentally use high-level features in ways that compiler writers could not anticipate, and they may thus produce underperforming programs as a result.
This paper introduces feature-specific profiling, a profiling technique that reports performance costs in terms of linguistic constructs. With a feature-specific profiler, a programmer can identify specific instances of language features that are responsible for performance problems. After explaining the architecture of our feature-specific profiler, the paper presents the evidence in support of adding feature-specific profiling to the programmer’s toolset.
Vincent St-Amour, Leif Andersen, Matthias Felleisen

A Synchronous-Based Code Generator for Explicit Hybrid Systems Languages

Modeling languages for hybrid systems are cornerstones of embedded systems development in which software interacts with a physical environment. Sequential code generation from such languages is important for simulation efficiency and for producing code for embedded targets. Despite being routinely used in industrial compilers, code generation is rarely, if ever, described in full detail, much less formalized. Yet formalization is an essential step in building trustable compilers for critical embedded software development.
This paper presents a novel approach for generating code from a hybrid systems modeling language. By building on top of an existing synchronous language and compiler, it reuses almost all the existing infrastructure with only a few modifications. Starting from an existing synchronous data-flow language conservatively extended with Ordinary Differential Equations (ODEs), this paper details the sequence of source-to-source transformations that ultimately yield sequential code. A generic intermediate language is introduced to represent transition functions. The versatility of this approach is exhibited by treating two classical simulation targets: code that complies with the FMI standard and code directly linked with an off-the-shelf numerical solver (Sundials CVODE).
The presented material has been implemented in the Zélus compiler and the industrial Scade Suite KCG code generator of Scade 6.
Timothy Bourke, Jean-Louis Colaço, Bruno Pagano, Cédric Pasteur, Marc Pouzet

Faster, Practical GLL Parsing

Generalized LL (GLL) parsing is an extension of recursivedescent (RD) parsing that supports all context-free grammars in cubic time and space. GLL parsers have the direct relationship with the grammar that RD parsers have, and therefore, compared to GLR, are easier to understand, debug, and extend. This makes GLL parsing attractive for parsing programming languages.
In this paper we propose a more efficient Graph-Structured Stack (GSS) for GLL parsing that leads to significant performance improvement. We also discuss a number of optimizations that further improve the performance of GLL. Finally, for practical scannerless parsing of programming languages, we show how common lexical disambiguation filters can be integrated in GLL parsing.
Our new formulation of GLL parsing is implemented as part of the Iguana parsing framework. We evaluate the effectiveness of our approach using a highly-ambiguous grammar and grammars of real programming languages. Our results, compared to the original GLL, show a speedup factor of 10 on the highly-ambiguous grammar, and a speedup factor of 1.5, 1.7, and 5.2 on the grammars of Java, C#, and OCaml, respectively.
Ali Afroozeh, Anastasia Izmaylova

Analysis and Optimisation


A Backend Extension Mechanism for PQL/Java with Free Run-Time Optimisation

In many data processing tasks, declarative query programming offers substantial benefit over manual data analysis: the query processors found in declarative systems can use powerful algorithms such as query planning to choose high-level execution strategies during compilation. However, the principal downside of such languages is that their primitives must be carefully curated, to allow the query planner to correctly estimate their overhead. In this paper, we examine this challenge in one such system, PQL/Java. PQL/Java adds a powerful declarative query language to Java to enable and automatically parallelise queries over the Java heap. In the past, the language has not provided any support for custom user-designed datatypes, as such support requires complex interactions with its query planner and backend.
We examine PQL/Java and its intermediate language in detail and describe a new system that simplifies PQL/Java extensions. This system provides a language that permits users to add new primitives with arbitrary Java computations, and new rewriting rules for optimisation. Our system automatically stages compilation and exploits constant information for dead code elimination and type specialisation. We have re-written our PQL/Java backend in our extension language, enabling dynamic and staged compilation.
We demonstrate the effectiveness of our extension language in several case studies, including the efficient integration of SQL queries, and by analysing the run-time performance of our rewritten prototype backend.
Hilmar Ackermann, Christoph Reichenbach, Christian Müller, Yannis Smaragdakis

Staged Points-to Analysis for Large Code Bases

Bug checker tools for Java require fine-grained heap abstractions including object-sensitive call graphs, field information for objects, and points-to sets for program variables to find bugs in source codes. However, heap abstractions coined commonly as points-to analysis, have high runtime-complexity especially when the points-to analysis is context- sensitive, and, hence, state-of-the-art points-to analyses do not scale for large code bases.
In this paper, we introduce a new points-to framework that facilitates the computation of context-sensitive points-to analysis for large code bases. The framework is demand-driven, i.e., a client queries the points-to information for some program variables. The novelty of our approach is a pre-analysis technique that is a combination of staged points-to analyses with program slicing and program compaction. We implemented the proposed points-to framework in Datalog for a proprietary bug checker that could identify security vulnerabilities in the OpenJDKTM library which has approximately 1.3 million variables and 500,000 allocation-sites. For the clients that we have chosen, our technique is able to eliminate about 73% of all variables and about 95% of allocation-sites. Thus our points-to framework scales for code bases with millions of program variables and hundreds of thousands of methods.
Nicholas Allen, Bernhard Scholz, Padmanabhan Krishnan

Exact and Approximated Data-Reuse Optimizations for Tiling with Parametric Sizes

Loop tiling is a loop transformation widely used to improve spatial and temporal data locality, to increase computation granularity, and to enable blocking algorithms, which are particularly useful when offloading kernels on computing units with smaller memories. When caches are not available or used, data transfers and local storage must be software-managed, and some useless remote communications can be avoided by exploiting data reuse between tiles. An important parameter of tiling is the sizes of the tiles, which impact the size of the required local memory. However, for most analyzes involving several tiles, which is the case for inter-tile data reuse, the tile sizes induce non-linear constraints, unless they are numerical constants. This complicates or prevents a parametric analysis with polyhedral optimization techniques.
This paper shows that, when tiles are executed in sequence along tile axes, the parametric (with respect to tile sizes) analysis for inter-tile data reuse is nevertheless possible, i.e., one can determine, at compiletime and in a parametric fashion, the copy-in and copy-out data sets for all tiles, with inter-tile reuse, as well as sizes for the induced local memories. When approximations of transfers are performed, the situation is much more complex, and involves a careful analysis to guarantee correctness when data are both read and written. We provide the mathematical foundations to make such approximations possible. Combined with hierarchical tiling, this result opens perspectives for the automatic generation of blocking algorithms, guided by parametric cost models, where blocks can be pipelined and/or can contain parallelism. Previous work on FPGAs and GPUs already showed the interest and feasibility of such automation with tiling, but in a non-parametric fashion.
Alain Darte, Alexandre Isoard

Optgen: A Generator for Local Optimizations

Every compiler comes with a set of local optimization rules, such as x + 0 → x and x & x → x, that do not require any global analysis. These rules reflect the wisdom of the compiler developers about mathematical identities that hold for the operations of their intermediate representation. Unfortunately, these sets of hand-crafted rules guarantee neither correctness nor completeness. Optgen solves this problem by generating all local optimizations up to a given cost limit. Since Optgen verifies each rule using an SMT solver, it guarantees correctness and completeness of the generated rule set. Using Optgen, we tested the latest versions of gcc, icc and llvm and identified more than 50 missing local optimizations that involve only two operations.
Sebastian Buchwald

Formal Techniques


Towards a Scalable Framework for Context-Free Language Reachability

Context-Free Language Reachability (CFL-R) is a search problem to identify paths in an input labelled graph that form sentences in a given context-free language. CFL-R provides a fundamental formulation for many applications, including shape analysis, data and control flow analysis, program slicing, specification-inferencing and points-to analysis. Unfortunately, generic algorithms for CFL-R scale poorly with large instances, leading research to focus on ad-hoc optimisations for specific applications. Hence, there is the need for scalable algorithms which solve arbitrary CFL-R instances.
In this work, we present a generic algorithm for CFL-R with improved scalability, performance and/or generality over the state-of-the-art solvers. The algorithm adapts Datalog’s semi-naïve evaluation strategy for eliminating redundant computations. Our solver uses the quadtree data-structure, which reduces memory overheads, speeds up runtime, and eliminates the restriction to normalised input grammars. The resulting solver has up to 3.5x speed-up and 60% memory reduction over a state-of-the-art CFL-R solver based on dynamic programming.
Nicholas Hollingum, Bernhard Scholz

Protocols by Default

Safe MPI Code Generation Based on Session Types
This paper presents a code generation framework for type-safe and deadlock-free Message Passing Interface (MPI) programs. The code generation process starts with the definition of the global topology using a protocol specification language based on parameterised multiparty session types (MPST). An MPI parallel program backbone is automatically generated from the global specification. The backbone code can then be merged with the sequential code describing the application behaviour, resulting in a complete MPI program. This merging process is fully automated through the use of an aspect-oriented compilation approach. In this way, programmers only need to supply the intended communication protocol and provide sequential code to automatically obtain parallelised programs that are guaranteed free from communication mismatch, type errors or deadlocks. The code generation framework also integrates an optimisation method that overlaps communication and computation, and can derive not only representative parallel programs with common parallel patterns (such as ring and stencil), but also distributed applications from any MPST protocols. We show that our tool generates efficient and scalable MPI applications, and improves productivity of programmers. For instance, our benchmarks involving representative parallel and application-specific patterns speed up sequential execution by up to 31 times and reduce programming effort by an average of 39%.
Nicholas Ng, Jose Gabriel de Figueiredo Coutinho, Nobuko Yoshida

Verifying Fast and Sparse SSA-Based Optimizations in Coq

The Static Single Assignment (SSA) form is a predominant technology in modern compilers, enabling powerful and fast program optimizations. Despite its great success in the implementation of production compilers, it is only very recently that this technique has been introduced in verified compilers. As of today, few evidence exist on that, in this context, it also allows faster and simpler optimizations. This work builds on the CompCertSSA verified compiler (an SSA branch of the verified CompCert C compiler). We implement and verify two prevailing SSA optimizations: Sparse Conditional Constant Propagation and Global Value Numbering. For both transformations, we mechanically prove their soundness in the Coq proof assistant. Both optimization proofs are embedded in a single sparse optimization framework, factoring out many of the dominance-based reasoning steps required in proofs of SSA-based optimizations. Our experimental evaluations indicate both a better precision, and a significant compilation time speedup.
Delphine Demange, David Pichardie, Léo Stefanesco


Weitere Informationen

Premium Partner