[SIESTA-L] siesta bug on the compaq alpha cluster

[HW Sheng]

Dear Siesta Users,   I have successfully compiled the parallel version of Siesta-1.3 on my compaq alpha cluster (linux 2.4.19) with the lam-mpi mpif77 compiler (fort wrapped with mpif77, lam-7.0.2).  Scalapack and cxml math libraries are used. Due to the peculiarity of the compaq Fort compiler, NetCDF was not included in the makefile, which I guess will not affect the execution of siesta.   During the test run of the H2O system provided by the siesta package, I met the following problem:   After issuing the command like "mpirun -np 2 ./siesta < h2o.fdf", the initiation went on without any problem. but after the MD simulation begins, it shows info like this (typed verbatim): ..................... (snipped) siesta =============          Begin MD step = 1          ============= InitMesh: MESH = 48 x 36 x36 = 62208 InitMesh: Mesh cutoff (required, used) =  100.00  108.450 Ry forrtl: severe (174): SIGSEGV, segmentation fault occurred. ...   the error message virtually indicates a stacksize overflow. However, it hardly seems to be the fault of the stack size since the problem persists even after the stack size is increased to "unlimited".  I suspect there might be a compatibility problem of the lam compiler and the siesta code.  Preliminary debugging test shows that the error occurred in the loop in file cellxc.F (line 684, Do 230 I3=0, NMESH(3) -1 ).    Therefore, I wish to know if anyone of you met similar problems and how you get around this. Any solution would be greatly appreciated.     Thank you,   Howard   Dept. Mat. Sci. Eng. Johns Hopkins Univ.