[Nek5000-users] Compilation errors when points in an element become large

[nek5000-users at lists.mcs.anl.gov]

Mani,

You are exceeding local (per processor memory limits).
Are you running this in parallel?

One cure (for some compilers, and assuming you have 
enough memory on your machine) is to add to the makenek
script:

G = "-mcmodel=medium"

which allows one to address the large memory limit.

Another simple thing to do (particularly if running in
serial) is to make certain that lelt and lelv do not
exceed the number of elements in your simulation (determined
by "grep NEL pipe.rea").   Lelt and lelv are the upper 
bounds on number of elements PER PROCESSOR.   If you
are using 8 processors, then you could have

      lelt = lelv >= NEL/8

where lelt and lelv are set in SIZEu (soon to be "SIZE"
with the latest release) and NEL is the value in your
.rea file.

Does this make sense?

Paul


On Sat, 27 Mar 2010, nek5000-users at lists.mcs.anl.gov wrote:

> Dear Nek devs,
>
>    In the example case "pipe", when I put in lx1=16 and lxd=24 with 
> everything else being the same, the following compilation error occurs:
>
> mpif77 -o nek5000  drive.o drive1.o drive2.o plan4.o bdry.o coef.o conduct.o 
> connect1.o connect2.o dssum.o edgec.o eigsolv.o gauss.o genxyz.o navier1.o 
> navier0.o navier2.o navier3.o navier4.o prepost.o speclib.o map2.o turb.o 
> mvmesh.o ic.o ssolv.o planx.o math.o mxm_wrapper.o hmholtz.o gfdm_par.o 
> gfdm_op.o gfdm_solve.o subs1.o subs2.o genbox.o gmres.o hsmg.o convect.o 
> induct.o perturb.o navier5.o navier6.o navier7.o navier8.o fast3d.o fasts.o 
> calcz.o subuser.o isosurf.o byte.o chelpers.o byte_mpi.o postpro.o 
> cvode_driver.o setprop.o qthermal.o makeq.o papi.o ssygv.o dsygv.o mxm_std.o 
> blas.o jl2_gs.o jl2_sort.o jl2_sarray_transfer.o jl2_sarray_sort.o 
> jl2_gs_local.o jl2_crystal.o jl2_comm.o jl2_tensor.o jl2_fail.o 
> jl2_fcrystal.o jl_tuple_list.o jl_transfer.o jl_sort.o jl_fcrystal.o 
> jl_errmem.o jl_crystal.o jl_findpts.o jl_pfindpt.o jl_tensor.o jl_findpt.o 
> jl_poly.o jl2_sparse_cholesky.o jl2_xxt.o jl2_fcrs.o   comm_mpi.o
> drive2.o: In function `a_dmp_':
> drive2.f:(.text+0x61f): relocation truncated to fit: R_X86_64_32S against 
> symbol `scruz_' defined in COMMON section in plan4.o
> drive2.f:(.text+0x631): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x647): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.o: In function `plan3_vol_':
> drive2.f:(.text+0x6b0): relocation truncated to fit: R_X86_64_PC32 against 
> symbol `cvflow_i_' defined in COMMON section in drive2.o
> drive2.f:(.text+0x6de): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x6ee): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x700): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x72f): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x735): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x73a): relocation truncated to fit: R_X86_64_32 against 
> symbol `scrns_' defined in COMMON section in convect.o
> drive2.f:(.text+0x75e): additional relocation overflows omitted from the 
> output
> collect2: ld returned 1 exit status
> make: *** [nek5000] Error 1
>
> These kind of errors seem to come up whenever I increase lx1, lxd or 
> sometimes even lelt and lelg to a large number in other examples too. How do 
> i rectify this?
> Also what are the functions of lelx, lely and lelz? These parameters are not 
> documented on the Nek5000 website.
>
> Regards,
> Mani chandra
>
> _______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>