Yasumasa Kanada

Density-functional study of liquid methanol

Abstract Properties of liquid methanol as well as of a molecule are studied by the combination of gradient-corrected density-functional theory, norm-conserving pseudopotentials, and the finite-element basis set. The structural and dynamical quantities are calculated through molecular dynamics simulations, and are compared with experiments. Satisfactory agreement is found, which makes these techniques more reliable for hydrogen-bonded liquids.

High-Performance Radix-2, 3 and 5 Parallel 1-D Complex FFT Algorithms for Distributed-Memory Parallel Computers

In this paper, we propose high-performance radix-2, 3 and 5 parallel 1-D complex FFT algorithms for distributed-memory parallel computers. We use the four-step or six-step FFT algorithms to implement the radix-2, 3 and 5 parallel 1-D complex FFT algorithms. In our parallel FFT algorithms, since we use cyclic distribution, all-to-all communication takes place only once. Moreover, the input data and output data are both in natural order.We also show that the suitability of a parallel FFT algorithm is machine-dependent because of the differences in the architecture of the processor elements in distributed-memory parallel computers. Experimental results of 2p3q5r point FFTs on distributed-memory parallel computers, HITACHI SR2201 and IBM SP2 are reported. We succeeded to get performances of about 130 GFLOPS on a 1024PE HITACHI SR2201 and about 1.25 GFLOPS on a 32PE IBM SP2.

Monte Carlo simulation of the Ising model and random number generation on the vector processor

The Monte Carlo simulation of the Ising model on the vector processor is discussed with particular attention given to the development of the vectorizable algorithm and the implementation, tuning, and tests necessary for the large-scale computations. The machine is the HITAC S820/80. The maximum simulation speed is 0.93*10/sup 9/ spins per second for the ferromagnetic Ising models on a simple cubic lattice. The random number generation algorithms are also explained. The present generator can produce 0.62*10/sup 9/ 32-bit random integers per second. The results of randomness tests which use the two-dimensional Ising model are presented.<<ETX>>

An Efficient Implementation of Parallel Eigenvalue Computation for Massively Parallel Processing

Abstract This paper describes an efficient implementation and evaluation of a parallel eigensolver for computing all eigenvalues of dense symmetric matrices. Our eigensolver uses a Householder tridiagonalization method, which has higher parallelism and performance than conventional methods when problem size is relatively small, e.g., the order of 10,000. This is very important for relevant practical applications, where many diagonalizations for such matrices are required so often. The routine was evaluated on the 1024 processors HITACHI SR2201, and giving speedup ratios of about 2–5 times as compared to the ScaLAPACK library on 1024 processors of the HITACHI SR2201.

Determination of the Critical Points of Antiferromagnetic Ising Model with Next Nearest Neighbour Interactions on the Triangular Lattice

Positions of critical temperatures of antiferromagnetic Ising model with ferromagnetic next nearest neighbour interactions on the triangular lattice are investigated by making use of their properties of the Kosterlitz-Thouless type of phase transition. In particular, finite-size scaling method and Monte Carlo Renormalization Group methods (MCRG) with flow diagrams of the anomalous dimension η show the fixed line structure where the correlation length diverges. Another MCRG with flow diagram of the stiffness constant also shows a critical property clearly. The temperature dependence of the long range order is also discussed in order to locate the lower critical point.

Knowledge Discovery in Auto-tuning Parallel Numerical Library

This paper proposes the parallel numerical library called ILIB which realises auto-tuning facilities with selectable calculation kernels, communication methods between processors, and various number of unrolling for loop expansion. This auto-tuning methodology has advantage not only in usability of library but also in performance of library. In fact, results of the performance evaluation show that the auto-tuning or auto-correction feature for the parameters is a crucial technique to attain high performance. A set of parameters which are auto-selected by this auto-tuning methodology gives us several kinds of important knowledge for highly efficient program production. These kinds of knowledge will help us to develop some other high-performance programs, in general.

Vectorization of multiple-precision arithmetic program and 201,326,000 decimal digits of pi calculation

More than 200 million decimal places of pi were calculated using an arithmetic geometric mean formula independently discovered by E. Salamin and R.P. Brent in 1976. Correctness of the calculation was verified through Borweins quartic convergent formula developed in 1983. The computation took CPU times of 5 hours 57 minutes for the main calculation and 7 hours 30 minutes for the verification calculation on the HITAC S-820 model 80 supercomputer. Two programs generated values up to 3*2/sup 26/, about 201 million. The two results agreed except for the last 21 digits. The results also agree with the 133,554,000-place calculation of pi that was done by the author in January 1987. Compared to the record in 1987, 50% more decimal digits were calculated with about 1/6 of CPU time. The computation was performed with a real-arithmetic-based vectorized fast Fourier transform (FFT) multiplier and vectorized multiple-precision add, subtract, and (single-word) constant multiplication programs. Vectorizations for the later cases were realized through first order linear recurrence vector instruction on the S-820.<<ETX>>

Extraction of Authors' Characteristics from Japanese Modern Sentences via N-gram Distribution

Objects of many studies of authorship attribution have been text data in which boundaries between words are obvious [1] [2]. When we apply these studies to languages in which sentences could not be divided obviously into words, such as Japanese or Chinese, preliminary processing of text data such as morphological analysis is required and may influence the final results. The methods which make use of characteristics of particular languages or particular compositions also have limited coverage [3]. Extracting authors’ characteristics from sentences is generally an unsolved problem. Therefore, we propose a method for authorship attribution based on distribution of n-grams of characters in sentences. The proposed method can analyze sentences without any additional information, i.e. preliminary analyses. The experiments, where 3-grams to represent author’s characteristics were educed on the basis of their distributions, are also reported in the following.

Bonding in selected normal (octet), hypervalent (decet) and super-hypervalent (duodecet) sulfur compounds

Abstract A series of sulfur compounds with eight valence electrons (octet) such as SF 2 and S(CH 3 ) 2 , 10 valence electrons (decet) such as SF 4 , SF 2 (CH 3 ) 2 and S(CH 3 ) 4 and 12 valence electrons (duodecet) such as SF 6 , SF 4 (CH 3 ) 2 , SF 2 (CH 3 ) 4 and S(CH 3 ) 6 were subjected to ab initio SCF-MO computations. Conclusions were drawn from the numerical results with respect to the relative stabilities and ease of syntheses for the super-hypervalent (duodecet) sulfur compounds.

Fast Quadruple Precision Arithmetic Library on Parallel Computer SR11000/J2

In this paper, the fast quadruple precision arithmetic of four kinds of basic operations and multiply-add operations are introduced. The proposed methods provide a maximum speed-up factor of 5 times to gcc 4.1.1 with POWER 5+ processor used on parallel computer SR11000/J2. We also developed the fast quadruple precision vector library optimized on POWER 5 architecture. Quadruple precision numbers, which is 128 bit long double data type, are emulated with a pair of 64 bit double data type on POWER 5+ prosessor used on SR11000/J2 with Hitachi Optimizing Compiler and gcc 4.1.1. To avoid rounding errors in computing quadruple precision arithmetic operations, emulation needs high computational cost. The proposed methods focus on optimizing the number of registers and instruction latency.