The preferred compile method is to compile on sasn100 with the cross compilers and then switch over to janus to run 'yod' your program onto the compute nodes. See my .cshrc for a .cshrc which works right for sasn100 and janus, including setting up the cross-compiler environment. To cross-compile, you need to setup the cross-compiler environment.
The cougar f77 compiler is named cif77. The cougar c compiler is cicc. The cougar c++ compiler is ciCC. cif77 invokes the Portland Group F77 compiler (pgf77) with the -cougar switch which sets up the right compile environment for a program that will be run on a cougar compute node.
In this example we assume you are doing the compile on sasn100 and the jump to a window where you are logged into janus.
sasn100 92 > cat tst.f
include 'mpif.h'
integer*4 ierr,num_pe,my_pe
call mpi_init(ierr)
call mpi_comm_size(MPI_COMM_WORLD,num_pe,ierr)
call mpi_comm_rank(MPI_COMM_WORLD,my_pe,ierr)
call mpi_barrier(MPI_COMM_WORLD,ierr)
print *,'hello from node ', my_pe,' of ',num_pe
call mpi_finalize(ierr)
call exit
end
Acknowledgement and Disclaimer
Now to compile it:
sasn100 93 > cif77 tst.f -o tstf -lmpi Linking:Note that the compiler name is pgf77 but a script 'cif77' sets up the correct syntax for calling pgf77. It adds the -cougar switch.
To actually launch a job on the compute nodes, you use the job command. Use 'yod -h' for a quick list of yod options.
We need to login now to janus (above assumes you did the compile on sasn100). So jump to a window where you are logged into janus or do 'ssh janus'.
Now we run the program:
janus 94 > yod -sz 4 tstf & hello from node 0 of 4 hello from node 2 of 4 hello from node 1 of 4 hello from node 3 of 4Now we will run the utility to display the activity on the machine.
janus 94 > showmeshYou will see something like the display below (with janus 'big')
janus ~ 101 > /usr/community/bin/shmesh
Current mesh allocation for janus (last booted Thu Jul 01 12:24:10 1999)
1 2 3 4 5
01234567890123456789012345678901234567890123456789012345
+--------------------------------------------------------+
20|ddddddddddddddddddddddddddddddddddddddddddddddddddddDDDD| 75
96|ddddddddddddddddddddddddddddddddddddddddddddddddddddDDDD| 151
172|dddddddddddddddddddddddddddddddddddddddddddddddddddddddd| 227
248|ddddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 303
324|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 379
400|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 455
476|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 531
552|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 607
628|dddddddddddddddddddddddddddddddd ddddddddddddddDDDDdddd| 683
704|dddddddddddddddddddddddddddddddd ddddddddddddddDDDDdddd| 759
780|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 835
856|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 911
932|dddddddddddddddddddddddddddddddd dddddddddddddddddddddd| 987
1008|ddddddddddddddggffffffffgggggggg ggggggjjjjjjjjjjjjjjjj| 1063
1084|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggggjjjjjjjjjjjjjj| 1139
1160|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggggjjjjjjjjjjjjjj| 1215
1236|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggSDDDDDDDEEEggjjj| 1291
1312|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggSDDDDDgDEEEgjjjj| 1367
1388|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggggiiiiiiiijjjjjj| 1443
1464|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggggiiiiiiiijjjjjj| 1519
1540|eeeeeeeeeeeeeeeeffffffffgggggggg ggggggggiiiiiiiijjjjjj| 1595
1616|eeeeeeeeeeeeeeeegggggggggggggggg jjjjjjjjiiiiiiiijjjjjj| 1671
1692|eeeeeeeeeeeeeeeegggggggggggggggg jjjjjjjjiiiiiiiijjjjjj| 1747
1768|eeeeeeeeeeeeeeeegggggggggggggggg jjjjjjjjiiiiiiiijjjjjj| 1823
1844|eeeeeeeeeeeeeeeegggggggggggggggg jjBDDDSjjjjjjjjjjjjjjj| 1899
1920|eeeeeeeeeeeeeeeegggggggggggggggg kkDDDDSkkkkkcccccaaaaa| 1975
1996|eeeeeeeeeeeeeeeegggggggggggggggg kkSSSSSkkkkkccccccbbbb| 2051
2072|eeeeeeeeeeeeeeeegggggggggggggggg kkkkkkkkkkkkccccccbbbb| 2127
2148|eeeeeeeeeeeeeeeegggggggggggggggg kkkkkkkkkkkkccccccbbbb| 2203
2224|eeeeeeeeeeeeeeeekkkkkkkkkkkkkkkk kkkkkkkkkk||ccccccbbbb| 2279
2300|hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh ||||||||||||ccc:::::::| 2355
2376|hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh ||||||||||||::::::::::| 2431
+--------------------------------------------------------+
Legend:
B boot node D TOS disk node (RAID)
non-existant node S TOS service node
T node booting OS E TOS ethernet node
F other TOS node X failed Cougar node(s)
. free Cougar eagle node : free Cougar kestrel nodes
| free Cougar NQS node(s)
free/total open nodes = 34/140, free/total nqs nodes = 52/3226
Job ID User Base OS Size Start Partition_name/Command line
--- --- --------- ---- ------ ------ --------------- ---------------------------
a 23 wryxxxg 1971 CGR 10 Jul 01 13:56:42 yod -size 10 co
b 29 jfdxxxx 2048 CGR 32 Jul 01 14:43:46 yod -comm 2M -sz 32 cayakiote3_mp.x
c 30 crxxxxx 1966 CGR 64 Jul 01 14:59:11 yod -sz 64 -comm 1M -stack 1M r2d2_and_c3po
d 20 ksxxxxx 20 CGR 1414 Jul 01 12:37:09 yod -sz 1414 my-parallel-exe
e 26 kexxxxx 1084 CGR 512 Jul 01 12:37:15 yod -sz 512 -stack 2M -comm 2M practiko3d
f 27 tvxxxxx 1024 CGR Jul 01 12:37:19 /cougar/bin/yod quest_for_camelot
g 22 hmxxxxx 1022 CGR Jul 01 12:37:11 /cougar/bin/yod -masync -proc 2 -comm 50M mykeep
h 21 fmxxxxx 2300 CGR 128 Jul 01 12:37:11 yod -sz 128 -masync -fyod 1 do_red_dp3
i 25 mkxxxxx 1430 CGR 96 Jul 01 12:37:15 yod -munix -comm 0M -sz 96 btriste
j 28 hmxxxxx 1048 CGR Jul 01 13:10:48 /cougar/bin/yod -masync -proc 2 -comm 50M mykup
k 24 hmxxxxx 1954 CGR Jul 01 14:05:22 /cougar/bin/yod -masync -proc 2 -comm 50M yrkup
The showmesh command still says 'puma' instead of 'cougar' nodes.
Each ':' on the mesh is 2 nodes (one board with 2 nodes. Each node has 2 cpus).
/usr/community/shmesh drops the empty spot when the middle section is
switched to the classified side.
Note that 'call exit' instead of 'stop' gets rid of all the 'FORTRAN STOP' messages.
That's it!
janus 97 > logout