References

Baron, E. and Hauschildt, P. H.
Parallel implementation of the PHOENIX generalized stellar atmosphere program. II: Wavelength parallelization.
ApJ 495, (1998) 370.

10

Message Passing Interface Forum.
MPI: A Message-Passing Interface Standard, Version 1.1 (Univ. of Tennessee, Knoxville, TN, 1995).

11

Mihalas, D.
Stellar Atmospheres (W. H. Freeman, New York, 1978), second edn.

12

Magnan, C.
JQSRT 10, (1970) 1.

13

Caroff, L. J., Noerdlinger, P. D., and Scargle, J. D.
ApJ 176, (1972) 439.

14

Auer, L. H. and Blerkom, D. V.
ApJ 178, (1972) 175.

15

Castor, J. I.
MNRAS 149, (1970) 111.

16

Mihalas, D., Kunasz, P., and Hummer, D.
ApJ 202, (1975) 465.

17

Hauschildt, P. H. and Wehrse, R.
JQSRT 46, (1991) 81.

18

Hauschildt, P. H., Best, M., and Wehrse, R.
A&A 247, (1991) L21.

19

Cannon, C. J.
JQSRT 13, (1973) 627.

20

Scharmer, G. B.
In W. Kalkofen, ed., Methods in Radiative Transfer (Cambridge Univ. Press, Cambridge, 1984) p. 173, p. 173.

21

Olson, G. L., Auer, L. H., and Buchler, J. R.
JQSRT 38, (1987) 431.

22

Hamann, W.-R.
In Kalkofen [36] p. 35, p. 35.

23

Werner, K.
In Kalkofen [36] p. 67, p. 67.

24

Olson, G. L. and Kunasz, P. B.
JQSRT 38, (1987) 325.

25

Hauschildt, P. H., Störzer, H., and Baron, E.
JQSRT 51, (1994) 875.

26

Mihalas, D. and Mihalas, B. W.
Foundations of Radiation Hydrodynamics (Oxford University, Oxford, 1984).

27

Mihalas, D.
ApJ 237, (1980) 574.

28

Golub, G. H. and Loan, C. V.
Matrix computations (The Johns Hopkins University Press, 1989).

29

Turek, S.
preprint, (1993).

30

Ng, K. C.
J. Chem. Phys. 61, (1974) 2680.

31

Vinsome, P. W.
In Proc. of the 4th Symp on Reservoir Simulation . Soc. of Petroleum Engineers, (1976) p. 149, p. 149.

32

Auer, L.
In L. Crivellari, I. Hubeny, and D. G. Hummer, eds., Stellar Atmospheres: Beyond Classical Models (NATO ASI Series, Kluwer, Dordrecht, 1991) p. 9, p. 9.

33

Rybicki, G. B. and Hummer, D. G.
A&A 245, (1991) 171.

34

Mihalas, D.
Stellar Atmospheres (W. H. Freeman, New York, 1970), first edn.

35

Unsöld, A.
Physik der Sternatmosphären (Springer Verlag, Heidelberg, 1968), second edn.

36

Kalkofen, W., ed.
Cambridge, (1987). Cambridge Univ. Press.

**Figure 1:** Relation between (some of) the physical and mathematical blocks that describe the physics of a stellar atmosphere. In order to calculate a model atmosphere, a set of value of the physical variables, e.g., temperatures, densities, population densities and the radiation field, must be found that satisfies all constraints simultaneously.
$\begin{figure} \psfig {file=flow1.ps,clip=,width=\hsize} \vskip -2cm\end{figure}$

**Figure 2:** Relations between the main types of variables represented by blocks are indicated. The labels name the equations that relate the block to each other.
$\begin{figure} \psfig {file=synspec.ps,clip=,width=\hsize} \end{figure}$

**Figure 3:** The basic ``torus'' design of the wavelength-parallelized version of `PHOENIX`: groups of processors are divided up into wavelength clusters which will work on individual wavelength points, each wavelength cluster is further divided into a set of worker nodes, where each worker node is assigned a set of specific tasks, e.g., it will work on the LTE background line opacity for a set of radial points. Our design requires that each worker node on all wavelength clusters work on exactly the same set of tasks, although additional inherently serial operations can be assigned to one particular master worker, or master wavelength cluster. This reduces communication between clusters to its absolute minimum and allows the maximum speedup.
$\begin{figure} \psfig {file=fig2.ps,angle=270,clip=,width=\hsize}\end{figure}$

**Figure 4:** Scalability of the Supernova model atmosphere test run as function of the number of nodes (processing elements or nodes) used. The y-axis gives the speedup obtained relative to the serial run. The different symbols show the results for different numbers of worker tasks for each wavelength cluster.
$\begin{figure} \psfig {file=sn.ps,angle=90,clip=,width=\hsize}\end{figure}$

Next: About this document ... Up: Numerical Solution of the Previous: Discussion and Conclusions

Peter H. Hauschildt
8/20/1998