forked from KuangLab-Harvard/SAM_SRCv6.11
-
Notifications
You must be signed in to change notification settings - Fork 0
/
task_exchange.f90
178 lines (131 loc) · 3.1 KB
/
task_exchange.f90
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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
! Non-blocking receives before blocking sends for efficiency on IBM SP.
subroutine task_exchange(f,dimx1,dimx2,dimy1,dimy2,dimz,i_1, i_2, j_1, j_2, id)
! sends and receives the boundary messages
use grid
use advection, only: NADVS
implicit none
integer dimx1, dimx2, dimy1, dimy2, dimz
integer i_1, i_2, j_1, j_2
real f(dimx1:dimx2, dimy1:dimy2, dimz)
integer bufflen
parameter (bufflen = max(nx,ny)*(3+NADVS)*nz)
real buff_send(bufflen) ! buff to send data
real buff_recv(bufflen,8) ! buff to receive data
integer reqs_in(8), ids(8), count
integer i_start(8), i_end(8), j_start(8), j_end(8), ranks(8)
logical flag(8)
integer id ! id of the sent field
integer i, j, k, n, rf, tag, m, mend
integer i1, i2, j1, j2, mask
i1 = i_1 - 1
i2 = i_2 - 1
j1 = j_1 - 1
j2 = j_2 - 1
i_start(1) = nx-i1
i_end (1) = nx
j_start(1) = 1
j_end (1) = ny
i_start(2) = 1
i_end (2) = 1+i2
j_start(2) = 1
j_end (2) = ny
i_start(7) = 1
i_end (7) = nx
j_start(7) = ny-j1
j_end (7) = ny
i_start(8) = nx-i1
i_end (8) = nx
j_start(8) = ny-j1
j_end (8) = ny
i_start(3) = nx-i1
i_end (3) = nx
j_start(3) = 1
j_end (3) = 1+j2
i_start(4) = 1
i_end (4) = nx
j_start(4) = 1
j_end (4) = 1+j2
i_start(5) = 1
i_end (5) = 1+i2
j_start(5) = 1
j_end (5) = 1+j2
i_start(6) = 1
i_end (6) = 1+i2
j_start(6) = ny-j1
j_end (6) = ny
ranks(1) = rankee
ranks(2) = rankww
ranks(7) = ranknn
ranks(8) = rankne
ranks(3) = rankse
ranks(4) = rankss
ranks(5) = ranksw
ranks(6) = ranknw
mask=id*100000+i1*10000+i2*1000+j1*100+j2*10
ids(1) = mask+7
ids(2) = mask+3
ids(7) = mask+5
ids(8) = mask+6
ids(3) = mask+8
ids(4) = mask+1
ids(5) = mask+2
ids(6) = mask+4
!----------------------------------------------------------------------
! Send/receive buffs to/from neighbors
!----------------------------------------------------------------------
if(RUN3D) then
mend = 8
else
mend = 2
endif
! Non-blocking receives first:
do m = 1,mend
if(rank.ne.ranks(m)) then
call task_receive_float(buff_recv(1,m),bufflen,reqs_in(m))
flag(m) = .false.
else
flag(m) = .true.
end if
end do ! m
! Blocking sends:
do m = 1,mend
n=0
do k=1,dimz
do j=j_start(m),j_end(m)
do i=i_start(m),i_end(m)
n = n+1
buff_send(n) = f(i,j,k)
end do
end do
end do
if(rank.ne.ranks(m)) then
call task_bsend_float(ranks(m),buff_send,n,ids(m))
else
do i=1,n
buff_recv(i,m) = buff_send(i)
end do
end if
end do ! m
! Fill the data from the buffers that have the same rank (were not sent):
count = 0
do m = 1,mend
if(flag(m)) then
count = count+1
tag = ids(m)
call task_dispatch(buff_recv(1,m),tag)
end if
end do
! Monitor the progress of receiving; fill the data:
do while (count .lt. mend)
do m = 1,mend
if(.not.flag(m)) then
call task_test(reqs_in(m),flag(m),rf,tag)
if(flag(m)) then
count=count+1
call task_dispatch(buff_recv(1,m),tag)
endif
endif
end do
end do
call task_barrier()
end