Bulletin of the Seismological Society of America, Vol. 70, No. 4, pp. 1415-1420, August 1980
L E T T E R TO THE E D I T O R RELATIVE COMPUTER SPEEDS FOR SURFACE-WAVE DISPERSION COMPUTATIONS BY L. D. PORTER 1, F. SCHWAB 2' 3, K. K. NAKANISHI 4, I. F. WEEKS 5, G. F. PANZA6, E. MANTOVANI7, D. MCMENAMIN 8, W. D. SMYTHE2, A. H. LIAO2, J. A. LANDONI9, N. N. BISWAS 1°, F.-S. C H A N G 2, S. S. BOR 11, E. G. KAUSEL12, P. GASPERIN113, and J. L. LUTHEY14 INTRODUCTION One of the basic tasks of computational seismology is the calculation of surfacewave dispersion. T h e comparison of the efficiencies of different algorithms for this purpose is sometimes difficult because timing tests are performed on different computers, with different speeds of operation. For any given class of algorithms, it is therefore useful to have a set of conversion factors t h a t allows computation times on one machine to be converted to those on another, t h e r e b y permitting one to compare algorithmic efficiencies independently of computational hardware. Such a set of conversion factors is given here for both Love and Rayleigh waves on multi-, homogenous-layered media. T i m e spent in compilation of source code ( F O R T R A N programs in our case), linkage editing, and any other stages preliminary to execution of the actual machine-language program, has been eliminated from the r e p o r t e d times. In the execution of the machine-language program, we exercised as m u c h care as possible to eliminate preliminary set-up time and i n p u t / o u t p u t time from the conversion factors. In the tabulated results, the "relative characteristic times" refer to times required to treat a single layer of the multi-layered structure, while seeking a root of the dispersion function; this root yields the surface-wave phase velocity at a given frequency (see Schwab and Knopoff, 1972, for details). F r o m the n u m e r o u s times obtained in our measurements, after deleting those t h a t were obviously in error, we have reported the most conservative (slowest). We estimate the errors to be about 5 per cent of any r e p o r t e d value; the errors in the parenthesized values are probably larger. For the type of algorithms t r e a t e d here, practical experience has shown t h a t IBM 360 and 370 computers should be used with double precision (64 bits); for CDC 6400, 6600, and 7600 computers, single precision (60 bits) is satisfactory. T o ensure reliability in this type of calculation, the 72 bits of double-precision computations on the U N I V A ~ 1108 and 1110 computers should probably be used. T h e single' California Department of Conservation, Division of Mines and Geology, Office of Strong Motion Studies, 2811 0 Street, Sacramento, California 95816. z Institute of Geophysics and Planetary Physics, University of California, Los Angeles, California 90024. 3Permanent address: 3245 Ingledale Terrace, Los Angeles, California 90039. 4Lawrence Livermore Laboratory (L-205),P.O. Box 808, Livermore, California 94550. 5Los Alamos Scientific Laboratory, P.O. Box 1663, Los Alamos, New Mexico 87545. Istituto di Geodesia e Geofisica, Universith di Bari, 70100 Bari, Italia. 70sservatorio Geofisico,Universith di Siena, 53100 Siena, Italia. 8Occidental College, 1600 Campus Road, Los Angeles, California 90041. 9General Atomic Company (TO-E 267A), P.O. Box 81608, San Diego, California 92138. ,0 Geophysical Institute, University of Alaska, College, Alaska 99701. l, Geophysics Program, Room 225 ATG Building, AK-50, University of Washington, Seattle, Washington 98195. ,2 Departamento de Geof/sica, Universidad de Chile, Casiila 2777, Santiago, Chile. ,3 Istituto di Geofisica, Universith di Bologna, 40136 Bologna, Italia. ,4 Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, California 91103. 1415
1416
LETTERS
TO
THE
EDITOR
44
~
I
~
~
~
I
x
~
~
I
~
~
~
I
x
~
• I I I I~
•~
I ~
I I I I~1~o
oo
6
Z 0
~
~
~
~
~
~
X
~
~
. . . .
~
~
x
II V
II
0 0 0 0 0 0 0 0 0 0
O0
r~ >
,,.1 >,
~-1 ,
~
..
~
0 r~ e~
II ©
©0
Z
©
I I V ~ l l V
~oo~ q q 4 4 o o
0
r~ N
O0
0
0
0000
0©0
©©
~
~
i
~ i
11 ~
~
r.
0
o
