Supercompiler Scp4:Use of Quasi-Distributive Laws in Program Transformation

The author constructed a transformer Scp4 of functional programs. The transformer uses the technology known as Turchin's supercompilation. Scp4 was implemented in a functional language Refal 5. The input language for Scp4 is also Refal 5. In the present paper we give an outline of the supercompiler and consider a number of tools of the transformer in details. The tools can be formally thought of as quasi distributive laws.

Automatic Generation of Very Efficient Programs by Generalized Partial Computation

Generalized Partial Computation (GPC) is a program transformation method utilizing partial information about input data, properties of auxiliary functions and the logical structure of a source program. GPC uses both an inference engine such as a theorem prover and a classical partial evaluator to optimize programs. Therefore, GPC is more powerful than classical partial evaluators but harder to implement and control. We have implemented an experimental GPC system called WSDFU (Waseda Simplify Distribute Fold Unfold). This paper discusses the power of the program transformation system, its theorem prover and future works.

The Least Fixpoint Transformation for Disjunctive Logic Programs

The paradigm of disjunctive logic programming (DLP) enhances greatly the expressive power of normal logic programming (NLP) and many (declarative) semantics have beeu defined for DLP to cope with various problems of knowledge representation in artificial intelligence. However, the expressive ability of the semantics and the soundness of program transformations for DLP have been rarely explored. This paper defines an immediate consequence operator TGP for each disjunctive program and shows that TGP has the least and computable fixpoint Lft(P). Lft is, in fact, a program transformation for DLP which transforms all disjunctive programs into negative programs. It is shown that Lft preserves many key semantics, including the disjunctive stable models, well-founded model, disjunctive argument semantics DAS, three-valued models, etc. This means that every disjunctive program P has a unique canonical form Lft(P) with respect to these semanics. As a result, the work in this paper provides a unifying frameword for studying the expressive ability of various semantics for DLP.On the other hand, the computing of the above semantics for negative programs is just a trivial task, therefore, Lft(P) is also an optimization method for DLP. Another application of Lft is to derive some interesting semantic results for DLP.

Resolving the GPU responsiveness dilemma through program transformations

The emerging integrated CPU-GPU architectures facilitate short computational kernels to utilize GPU acceleration. Evidence has shown that, on such systems, the GPU control responsiveness （how soon the host program finds out about the completion of a GPU kernel） is essential for the overall performance. This study identifies the GPU responsiveness dilemma： host busy polling responds quickly, but at the expense of high energy consumption and interference with co-running CPU programs; interrupt-based notification minimizes energy and CPU interference costs, but suffers from substantial response delay. We present a programlevel solution that wakes up the host program in anticipation of GPU kernel completion. We systematically explore the design space of an anticipatory wakeup scheme through a timerdelayed wakeup or kernel splitting-based pre-completion notification. Experiments show that our proposed technique can achieve the best of both worlds, high responsiveness with low power and CPU costs, for a wide range of GPU workloads.

Abstraction from Constructors and Functional Constructors and Their Applications

Structures using constructors are of ordinary use in functional programming to represent data structures of unbound size. Lack of associativity of constructors, however, hinders program analyses or efficient executions. This paper describes ideas of abstraction toward constructors, and similarly abstraction from constructing functions, which we call functional constructors. We demonstrate our ideas making program analyses easier and enable transformation to efficient execution.

Inverse Computation and the Universal Resolving Algorithm

We survey fundamental concepts for inverse programming and then present the Universal Resolving Algorithm, an algorithm for inverse computation in a first order, functional programming language. We discuss the key concepts of the algorithm, including a three step approach based on the notion of a perfect process tree, and demonstrate our implementation with several examples of inverse computation.

程序变换的若干探讨

本文试用不动点原理研究几种程序变换,讨论了一类二重递归的终止条件,并用之计算91函数与证明3x+1问题(Ⅰ)。

Numerical Simulation of Retrofit Scheme of a Boiler Tail Flue

Aiming at the problem that the total pressure loss of the flue of the electric precipitator of the 350 MW unit of a power plant to the inlet of the draft fan is too large,the numerical simulation software Fluent and the standard k-εmodel was used to simulate the flue,the results show that the main part of the flue mean total pressure loss is derived from the confluence header and elbow.In order to reduce the loss and consider the cost of transformation,the concept of twodimensional feature surface is established,gradually proposed three sets of flue transformation program and analysis of the flue transformation program drag reduction effect,the results show that the total reduction of the flue can be reduced from 486 Pa to 89 Pa and the reduction rate is 81.7%,which is the best solution;The concept of two-dimensional feature plane is helpful for quick condensing of flue;Double V-type structure of the convergence of the box has a better drag reduction effect.

Understanding co-run performance on CPU-GPU integrated processors： observations, insights, directions

Recent years have witnessed a processor develop- ment trend that integrates central processing unit （CPU） and graphic processing unit （GPU） into a single chip. The inte- gration helps to save some host-device data copying that a discrete GPU usually requires, but also introduces deep re- source sharing and possible interference between CPU and GPU. This work investigates the performance implications of independently co-running CPU and GPU programs on these platforms. First, we perform a comprehensive measurement that covers a wide variety of factors, including processor ar- chitectures, operating systems, benchmarks, timing mecha- nisms, inputs, and power management schemes. These mea- surements reveal a number of surprising observations. We an- alyze these observations and produce a list of novel insights, including the important roles of operating system （OS） con- text switching and power management in determining the program performance, and the subtle effect of CPU-GPU data copying. Finally, we confirm those insights through case studies, and point out some promising directions to mitigate anomalous performance degradation on integrated heteroge- neous processors.