AMBER Archive (2009)

Subject: [AMBER] array out of bound in Amber10 tests

From: Shan-ho Tsai (
Date: Tue Feb 10 2009 - 10:19:42 CST

Hello all,

I built Amber10 on a 64-bit Linux machine (RHEL4) using pgf90
7.2 and mpich. Both the serial version and the parallel version
compiled fine and all tests passed. However, when I recompiled
the parallel version with a compiler option to trap array out
of bound (which for pgf90 it is -C), several tests failed to
run, because they got an array out of bound trap. Here is an
example of a test that failed to run:

cd test
make test.parallel.MM < /dev/null

cd 4096wat && ./Run.pure_wat_nmr_temp_reg
0: Subscript out of range for array tgtvec (_nmr.f: 5633)
     subscript=2, lower bound=1, upper bound=1, dimension=1
p1_26462: p4_error: net_recv read: probable EOF on socket: 1
rm_l_1_26464: (0.289062) net_send: could not write to fd=5, errno = 32
   ./Run.pure_wat_nmr_temp_reg: Program error
make[1]: *** [test.sander.BASIC] Error 1
make[1]: Leaving directory `/usr/local/src/amber10/test'
make: *** [test.sander.BASIC.MPI] Error 2

Similarly when I built parallel Amber10 on an AIX box using
xlf90 9.1, all tests ran and most passed, but rem_hybrid
generated incorrect results. This motivated me to compile
parallel Amber10 with the array out of bound option on
(which is also -C for xlf90), and then several tests
failed to run, with a trap like

ERROR: 0031-250 task 0: Trace/BPT trap
ERROR: 0031-250 task 1: Terminated

indicating array out of bound. The failed tests include:

        cd 4096wat && ./Run.pure_wat_nmr_temp_reg
        cd rdc && ./Run.dip
        cd rem_hybrid && ./Run.rem

(when I compiled serial Amber10 with the -C option included,
several tests invoked with "cd test; gmake test" also received
traps and failed to run)

I am doing a quite straightfoward installation, especially
on the Linux box, so I am wondering what I might be doing
wrong that is causing these array out of bound traps for
the tests.

Has anyone else compiled with the array out of bound check
turned on and encountered this problem?

Any suggestions how I might be able to correct this problem?

Thank you very much in advance!

Shan-Ho Tsai
University of Georgia, Athens, GA

AMBER mailing list