1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
|
\documentclass{article}
\usepackage{epsf}
\topmargin -0.5in
\textheight 9.9in
\oddsidemargin -0.3in
\leftmargin -1in
\textwidth 7in
\begin{document}
\title{Getting Three Timelevels of FMR Initial Data}
\author{Scott Hawley}
\date{June 6, 2002}
\maketitle
For parabolic interpolation in time, one needs three timelevels worth
of data on the coarse grid in order to provide boundary data on the
finer grid. Here we describe one scheme for obtaining these extra
timelevels of data at the initial time. Essentially, we evolve
each grid backwards in time by two steps, but we do so rather carefully.
The following diagram shows the scheme. We start with initial
data for all grids being specified at the current time (Step 1, not
shown). We then evolve the coarse grid forward one step (Step 2) by
the timestep $\Delta t$;
in so doing we first move the coarse grid storage from the current
timelevel (denoted by `` '' on the right of the first pane in the
diagram, because variables at the current time receive no suffix
in Cactus) to the previous timelevel (denoted by ``\_p'') before
evolving. We then ``flip'' all the timelevels for the coarse grid (l=0),
meaning
that we exchange the data on the `` '' and \_p\_p timelevels,
thus turning the picture upside down.
We then evolve the coarse grid ``backward'' --- which is really
just forward but with a timestep of $-\Delta t$. Then we move on
to the finer grid, evolving and flipping...
Take a look at the diagram. (We use $\tau$ instead of $t$ to denote
the time, simply because it was easier to do that in the
drawing program.) After the diagram we give a
listing of pseudo-code which implements this scheme.
\begin{center}
\epsfxsize=5.5in
\epsffile{threelev_initdata.eps}
\end{center}
\begin{center}
\epsfxsize=5.5in
\epsffile{threelev_initdata_2.eps}
\end{center}
Diagram finished.
\pagebreak
Here's some pseudo-code:
\begin{verbatim}
int time_dir = 1; //Positive = forward (+t), Negative = backward (-t)
int lev;
// At this point we assume that we have initial data given on
// one timelevel, and we want to get data on the other timelevels
do lev = 0 to MaxLevels-1
// Evolve "forward" (which may be backward for lev=1,3,5,7...)
Evolve(lev, time_dir*dt(lev))
flip_timelevels_on_lev_and_coarser(lev)
// Keep track of which direction (in time) we're integrating
time_dir = -time_dir
// Evolve in the opposite time-direction
Evolve(lev, time_dir*dt(lev))
end do
// Make sure we're pointed backwards, in order to get 2 "previous"
// timelevels. We could change the if statement to
// "if (mod(MaxLevels,2) == 0)", but I prefer to check time_dir
// explicitly, because it's easier to follow and I don't have to
// worry about having made a mistake
if (time_dir > 0) then
flip_timelevels_on_lev_and_coarser(MaxLevels-1)
time_dir = -time_dir
end if
// Evolve each level backwards one more timestep
do lev = MaxLevels-1 to 0
Evolve(lev,-dt(lev))
if (lev>0) then
Restrict(lev -> lev-1)
end if
end do
// Here's where a user can add in extra evolution code if they want
// to be extra-careful ("anal"), but remember that you'll be
// overwriting the t=0 data so you'll need to re-load it if you do
// this.
// Finally, one last flip to point us forward again
flip_all_timelevels()
time_dir = -time_dir
\end{verbatim}
\end{document}
|
