Yinliang Zhao

Prophet: A Speculative Multi-threading Execution Model with Architectural Support Based on CMP

Speculative Multithreading (SpMT) has been proposed as a perspective method for sequential programs to benefit from the increasing computing resources provided by Chip Multiprocessors (CMP). This paper analyzes the extraction of ihread-level parallelism from general-purpose programs and presents a speculative multi-threading execution model, Prophet. The architectural support for Prophet execution model is designed based on CMP. In Prophet the inter-thread data dependences are reduced by precomputation slice (p-slice). Multi-versioning Cache system along with thread state control mechanism are designed for buffering the speculative data and also a snooping bus based cache coherence protocol is used to detect data dependence violation. The experiment results show that Prophet system could achieve significant speedup for general-purpose programs.

An Overview of Prophet

Speculative Multithreading (SpMT) has been proposed as a perspective method to exploit Chip Multiprocessors (CMP) hardware potential. This paper researches speculative thread-level parallelism(TLP) for general-purpose programs. The Prophet system consists of a SpMT compiler framework and a simulator prototype based on the SpMT execution model. In the Prophet system, procedures are represented as weighted control flow graph (WCFG), the thread generator uses structural analysis and heuristic algorithm to partition the WCFG into sub-graphs which represent the candidate threads. Inter-thread data dependences are predicted by pre-computation slice (p-slice) to reduce RAW violation. Since the partition is speculative, thread state control mechanism and multi-versioning cache system are designed to buffer the speculative data, and a snooping bus based cache coherence protocol is used to detect data dependence violation. Simulation-based evaluation shows that the Prophet system could achieve significant speedup for general-purpose programs.

A Thread Partitioning Method for Speculative Multithreading

Speculative Multithreading (SpMT) is an effective mechanism for parallelizing irregular programs which are hard by conventional approaches. SpMT technology can be applied to exploit Thread-Level Parallelism effectively through allowing multiple threads executed in the presence of ambiguous data and control dependences while the correctness of the programs maintained by hardware support. This paper focuses on the thread partitioning method for Prophet, a SpMT compiler system. The thread partitioning method is based on the weighted control flow graph (WCFG) of the procedure. A structural analysis and some heuristic rules are used to partition the WCFG into sub-graphs and each sub-graph represents a candidate thread. The inter-thread data dependences are predicted by pre-computation slice to reduce RAW violation. The thread partitioning method proposed in this paper deals with both loop region and non-loop region partitioning for the consideration of general purpose programs partitioning. Experimental results show that an average speedup of 30% can be achieved by our thread partitioning method.

A Novel Thread Partitioning Approach Based on Machine Learning for Speculative Multithreading

Speculative multithreading (SpMT) is a thread-level automatic parallelization technique to accelerate sequential programs on multi-core. The existing heuristic-based approaches are only suitable for one kind of programs and cannot guarantee to get the optimal solution of thread partitioning. In this paper, we propose a novel thread partitioning approach based on machine learning to partition irregular programs into multithreads. It mainly includes: generating sufficient training samples, building and applying the prediction model to partition the irregular programs. By using the thread partition approach, an unseen irregular program can obtain a stable, much higher speedup than the heuristic-based approaches. On the Prophet, which is a SpMT processor to evaluate the performance of multithreaded programs, the novel thread partitioning approach is evaluated and reaches an average speedup of 1.80 on 4-core processor. Experiments show that our proposed approach can obtain a significant increase in speedup and Olden benchmarks deliver a better performance improvement of 5.41% than the traditional heuristic-based approach.

Least squares littlewood-paley wavelet support vector machine

The kernel function of support vector machine (SVM) is an important factor for the learning result of SVM. Based on the wavelet decomposition and conditions of the support vector kernel function, Littlewood-Paley wavelet kernel function for SVM is proposed. This function is a kind of orthonormal function, and it can simulate almost any curve in quadratic continuous integral space, thus it enhances the generalization ability of the SVM. According to the wavelet kernel function and the regularization theory, Least squares Littlewood-Paley wavelet support vector machine (LS-LPWSVM) is proposed to simplify the process of LPWSVM. The LS-LPWSVM is then applied to the regression analysis and classifying. Experiment results show that the precision is improved by LS-LPWSVM, compared with LS-SVM whose kernel function is Gauss function.

Effective algorithm of mining frequent itemsets for association rules

The efficiency of mining association rules is an important field of KDD. The algorithm Apriori is a classical algorithm in mining association rules. It is a breadth first search on the lattice space of itemsets. Though it makes use of anti-monotone of itemsets to reduce searching breadth, the algorithmic complexity of time is still the exponential quantity. In this article, the concepts of the generation and the ordinal itemsets tree are introduced. The ordinal itemsets tree is the dynamic description of mining relation of itemsets, and the vegetal ability of the ordinal itemsets tree is described by the generation. Through the study of the association rules, the conclusion that all frequent itemsets are not all vegetal itemsets and all vegetal itemsets are all frequent itemsets is discovered. With this conclusion, the number of the candidate itemsets can be reduced further to improve the efficiency of mining association rules and reduce the searching breadth. According to the generation, the AprioriFREQ algorithm, which is the improvement algorithm of Apriori, is designed in this article. By testing, the efficiency of the AprioriFREQ algorithm is obviously higher than the Aprioris.

Granule-oriented programming

A program will become obsolete or less effective in solving domain problems due to many reasons. One of the main reasons can be the fact that the program does not fit its context. The context of a program is defined as a collection of functionalities that support the program to solve domain problems, e.g., runtime environmental supports, meta-strategies, architectural supports, etc. Unfitness phenomena exist in many software systems, which lead the systems prematurely end their life cycles, or decrease their performance and accuracy in solving problems. In existing programming systems, from the perspective of language expressivity, little attention has been paid to this unfitness problem. Granule-oriented programming is an evolutionary metaphor in which programs are ground into code granules in order to localize their unfitness parts as explicitly as possible and then the code granules are compounded into the target program, in which a code granulation space, ont to express program in a well-formed and multi-layered framework, is formed. In this paper, we propose and briefly describe the notion of granule-oriented programming.

Algorithm of mining fuzzy association rules in network management

This paper discusses the current status of the research about mining association rules in a database, which points out the shortcoming of classical a prioris algorithm, and presents some theorems of mining association rules based on reducing records in a larger database. It also applies the theory of fuzzy sets to processing fuzzy data of traps in a network management. According to those theorems and the theory of fuzzy sets, we have designed the AprioriFuzzy algorithm to mine fuzzy association rules in database. Through the performance of AprioriFuzzy algorithm is analyzed and evaluated in this paper, this algorithm can save mining time about 30% and can mine fuzzy association rules in fuzzy data effectively. It has been implemented on PC in Visual C++6.0 and has been applied to mine the traps information in the network management based on SNMP protocol.

A scalable scheme for certificate revocation

This paper proposes a scalable scheme for certificate revocation in public key infrastructure. In this scheme, the set of revoked certificates is divided into groups. In each group, proofs for certificate status are computed using one-way accumulators, while all groups are still organized in a hash tree. The main advantage of the proposed scheme is that it can adjust traffic between CA-to-directory and directory-to-user according to certificate update rate and query rate in applications, thus overall traffic for certificate revocation can be remarkably reduced and a wider range of scenarios can be accommodated. Compared with Naors dynamic hash tree scheme, results show it can reduce traffic by about 50% in typical environments.

Prophet Synchronization Thread Model and Compiler Support

In Speculative Multithreading, data dependence that limits the speedup of speculative parallelization needs to be resolved to achieve a high performance. This paper designs a synchronization execution model, with the support of compiler, to synchronize store and load instructions that frequently have data dependence on each other. We use hardware profiler to gather dependence violation information of memory data, and the profiler information is fed back to the compiler. The compiler analyzes the synchronization efficiency, select store/load pairs of great synchronization potential, and inserts synchronization instructions using insertion algorithm. Loop threads and non-loop threads can both be synchronized. The hardware support is also given in the paper. The experimental results show that the synchronization under the compiler support can effectively resolve some memory data dependence and improve the performance of the speculative execution.