the proceedings contain 39 papers. the special focus in this conference is on Java, Low-Level Transformations, Data Distribution and High-Level Transformations. the topics include: High performance numerical computing...
ISBN:
(纸本)9783540678588
the proceedings contain 39 papers. the special focus in this conference is on Java, Low-Level Transformations, Data Distribution and High-Level Transformations. the topics include: High performance numerical computing in java;comprehensive redundant load elimination for the ia-64 architecture;minimum register instruction scheduling;unroll-based copy elimination for enhanced pipeline scheduling;a linear algebra formulation for optimising replication in data parallel programs;accurate data and context management in message-passing programs;a compiler framework for tiling imperfectly-nested loops;parallel programming with interacting processes;application of the polytope model to functional programs;multilingual debugging support for data-driven and thread-based parallellanguages;an analytical comparison of the I-test and omega test;a precise fixpoint reaching definition analysis for arrays;demand-driven interprocedural array property analysis;language support for pipelining wavefront computations;a machine-independent abstraction for managing customized data motion;optimization of memory usage requirement for a class of loops implementing multi-dimensional integrals;compile-time based performance prediction;designing the Agassiz compiler for concurrent multithreaded architectures;speculative predication across arbitrary interprocedural control flow;a geometric semantics for program representation in the polytope model;compiler and run-time support for improving locality in scientific codes;code restructuring for improving real time response through code speed, size trade-offs on limited memory embedded DSPS;symbolic analysis in the PROMIS compiler;run-time parallelization optimization techniques;an empirical study of function pointers using spec benchmarks and a parallel program model for scheduling.
A new geometric framework for parallel program representation is proposed to address the difficulties of parallel programming. the focus of this work is the expression of collections of computations and the inter-, in...
详细信息
this paper describes AntSM, a system that uses the inherent parallelism of multi-threaded programs to reduce the overhead of statistical and invariant violations detection-based debugging tools. the run-time monitorin...
详细信息
ISBN:
(纸本)9783319099675;9783319099668
this paper describes AntSM, a system that uses the inherent parallelism of multi-threaded programs to reduce the overhead of statistical and invariant violations detection-based debugging tools. the run-time monitoring of these tools leads to high overheads. the key insight of the AntSM system is that this overhead can be reduced in parallel programs by performing sampled monitoring across parallel regions of the program that are performing similar actions. AntSM implements this sampling using a combination of static and dynamic analyses to determine similar parts of the program executing in parallel and the number of threads executing those parts of the program. Experimental results, performed using the C-DIDUCE (a variant of DIDUCE for C) debugging tool on eleven Pthreads benchmarks from the PARSEC suite, show monitoring overhead is reduced by up to 18.14 times (and on average 8.73 times) on an eight-core machine relative to a naive port that performs no sampling.
Deterministic parallel programming languages are attractive as they do not allow data races, deadlocks, or similar kinds of concurrency bugs that are caused by unintended (or poorly understood) parallel execution. We ...
详细信息
ISBN:
(纸本)9783030352257;9783030352240
Deterministic parallel programming languages are attractive as they do not allow data races, deadlocks, or similar kinds of concurrency bugs that are caused by unintended (or poorly understood) parallel execution. We present here a simple programming model for deterministic parallel programming that is based on roles. the programmer specifies the role that an object plays for a task (e.g., the readonly role), and compiler and runtime system together ensure that only those object accesses are performed that are allowed by this role. An object may play different roles in the course of a program's execution, giving the programmer considerable flexibility in expressing a parallel program. the model has been implemented in a Java-like language with references and object sharing. Preliminary results indicate that the runtime overhead is moderate (compared to standard Java programs), and that the compiled programs achieve substantial parallel speedups.
In this paper, we argue that interacting processes (IP) with multiparty interactions are an ideal model for parallel programming. the IP model with multiparty interactions was originally proposed by N. Francez and I. ...
详细信息
Last year, we discussed the issues surrounding the development of languages and compilers for a general, portable, high-level SIMD With in A Register (SWAR) execution model. In a first effort to provide such a languag...
详细信息
the next generation of microprocessors, particularly IA64, will incorporate hardware mechanisms for instruction-level predication in support of speculative parallel execution. However, the compiler technology proposed...
详细信息
this paper presents an integrated compiler framework for ti- ling a class of nontrivial imperfectly-nested loops such that cache locality is improved. We develop a new memory cost model to analyze data reuse in terms ...
详细信息
暂无评论