[Debichem-devel] ACESIII-3.0.7 Availability

[Michael Banck]

Dear Ajith,

thanks for the swift response.

On Wed, Jul 18, 2012 at 07:36:31AM -0400, Ajith perera wrote:
> Dear Michael: I will make a tar file of the prerealese version of 3.0.7 and
> put it in my dropbox so that you can pick it up. 

Great!

> However, it would be nice if you can provide temporary guest access to
> your machine  to sort out the issues that you are having  with your
> installation. My feeling is that it need some attention from one of us
> here. If that is nearly impossible then we will try our best to work
> from distance to help you to get it running.

I am currently travelling, but could make it possible later on.

The first problem seems to be that xaces3 has a BSS size for 800 MB,
which gets allocated on startup:

nighthawk~$ objdump -x /usr/bin/xaces3 | grep -i bss
 24 .bss          300f55c8  083ac380  083ac380  0036436c  2**5

0x300f55c8 is the SIZE, which is roughtly 800 MB.  Due to this, xaces3
simply gets SIGKILLed when not enough memory is available.  Note that
this is even before reading ZMAT or anything.

Second, after freeing some memory, xaces3 spins in xjoda or shortly
thereafter:


                         ACES STATE VARIABLES (STRINGS)
          ------------------------------------------------------------
          BASIS = CC-PVDZ
          OCCUPATION = [ESTIMATED BY SCF]
          ------------------------------------------------------------
 Entering getxyz
         -----------------------------------------------
         Cartesian coordinates corresponding to internal
                 coordinate input (Angstroms)
 ----------------------------------------------------------------
 Z-matrix   Atomic            C o o r d i n a t e s
  Symbol    Number           X              Y              Z
 ----------------------------------------------------------------
     H         1         0.00000000     0.00000000     0.00000000
     H         1         0.00000000     0.00000000     0.52917725
 ----------------------------------------------------------------
   0.00000   0.00000   0.00000
   0.00000   0.00000   1.00000
         -----------------------------------------------
         Cartesian coordinates corresponding to internal
                 coordinate input (Angstroms)
 ----------------------------------------------------------------
 Z-matrix   Atomic            C o o r d i n a t e s
  Symbol    Number           X              Y              Z
 ----------------------------------------------------------------
     H         1         0.00000000     0.00000000     0.00000000
     H         1         0.00000000     0.00000000     0.52917725
 ----------------------------------------------------------------
[...]

At this point, no more output is written, and the xaces3 process spins
at 100% CPU.  If I attach gdb to it and get a backtrace, I get:

0x083197ce in dlamc3_ ()
(gdb) bt
#0  0x083197ce in dlamc3_ ()
#1  0x0831f7b9 in dlamc1_ ()
#2  0x08319294 in dlamc2_ ()
#3  0x083113d9 in dlamch_ ()
#4  0x083103a3 in dsyev_ ()
#5  0x08288046 in eig_ ()
#6  0x0829b01e in symmetry_ ()
#7  0x0828e80e in geopt_ ()
#8  0x0828a62f in aces2_joda_main_ ()
#9  0x080668ce in scf_init_ ()
#10 0x08061851 in MAIN__ ()

Those seem to be lapack symbols, so I am a bit puzzled what could be the
problem here, I get the same whether I use netlib lapack or ATLAS
lapack, MPI is openmpi.  On x86_64, we have no issues.

The only temporary file written is JOBARC, at 6 KB size.

The other weird thing is that if I redirect stdout to a file, only the
variables get dumped ("1:                PRINT =          0 [
0]" and so), the above pasted output is not written to the file.

The ZMAT is:

H2 test of rhf scf energy in an cc-pvdz basis
H   0.0  0.0  0.0
H   0.0  0.0  1.0

*ACES2(CALC=SCF,BASIS=CC-PVDZ
spherical=on
REF=RHF,UNITS=BOHR
COORDINATES=CARTESIAN
SCF_CONV=6
SYMMETRY=NONE)

*SIP
MAXMEM= 200
TIMERS=YES


Best regards,

Michael


Michael