problem compiling pdbg executable on cray with pathscale

garold g.murd... at gmail.com
Sat Feb 6 00:12:05 CET 2010


Teo,

Thank you, I'll compile w/o the bounds-check flag. By the way, you are
correct, I meant 04feb10.

Garold


On Feb 5, 12:07 pm, Laino Teodoro <teodor... at gmail.com> wrote:
> Garold,
>
> it is clearly a problem of the compiler: it is probable that the  
> semantic of that module is too complex to be digested with that  
> specific flag ON.
>
> Teo
> p.s.: just for confirm, your version should be 04feb10.
>
> On 5 Feb 2010, at 21:00, garold wrote:
>
> > Dear cp2k users/developers,
>
> > Yesterday I obtained the 04feb09 cvs version of cp2k and compiled
> > without problems a popt executable using pathscale compiler on a cray
> > (hopper at nersc).  I first set up the environment using commands:
>
> > module swap PrgEnv-pgi PrgEnv-pathscale
> > export GOTO_NUM_THREADS=1
> > export FORT_C_NAME=pathscale
>
> > and then compiled with:
>
> > make ARCH=Linux-x86-64-pathscale VERSION=popt
>
> > Here is the arch file:
>
> > CC       = cc
> > CPP      = cpp
> > FC       = ftn -target=linux
> > LD       = ftn
> > AR       = ar -r
> > DFLAGS   = -D__XT3 -D__FFTSG\
> >            -D__parallel -D__BLACS -D__SCALAPACK
> > CPPFLAGS = -traditional -C $(DFLAGS) -P
> > FCFLAGS  = -O2
> > LDFLAGS  = $(FCFLAGS)
>
> > OBJECTS_ARCHITECTURE = machine_xt3.o
>
> > This produced a correct executable.
>
> > The problem occurred when I then tried to compile a corresponding pdbg
> > executable; using a very similar arch file; simply replacing
>
> > FCFLAGS  = -O2
>
> > with
>
> > FCFLAGS  = -O0 -ffortran-bounds-check -g
>
> > The compilation went fine for a while but then the compiler choked on
> > the module dbcsr_operations.F; here are the relevant lines:
>
> > cpp -traditional -C -D__XT3 -D__FFTSG -D__parallel -D__BLACS -
> > D__SCALAPACK -P -D__COMPILE_ARCH="\"Linux-x86-64-pathscale\"" -
> > D__COMPILE_DATE="\"Thu Feb  4 15:42:09 PST 2010\"" -
> > D__COMPILE_HOST="\"hopper03\"" -D__COMPILE_LASTCVS="\"D\"" -I/scratch/
> > scratchdirs/garold/cp2k_04feb10/makefiles/../src /scratch/scratchdirs/
> > garold/cp2k_04feb10/makefiles/../src/dbcsr_operations.F >
> > dbcsr_operations.f90
> > ftn -target=linux -c -O0 -ffortran-bounds-check -g
> > dbcsr_operations.f90
> > /opt/cray/xt-asyncpe/3.4/bin/ftn: INFO: linux target is being used
> > Signal: Segmentation fault in Front End Parse/Semantic phase.
> > Error: Signal Segmentation fault in phase Front End Parse/Semantic --
> > processing aborted
> > *** Internal stack backtrace:
> > pathf90-3.2 INTERNAL ERROR: /opt/pathscale/lib/3.2/mfef95 died due to
> > signal 4
> > make[1]: *** [dbcsr_operations.o] Error 1
> > make[1]: Leaving directory `/scratch/scratchdirs/garold/cp2k_04feb10/
> > obj/Linux-x86-64-pathscale/pdbg'
> > make: *** [build] Error 2
>
> > If I compile manually the post-processed (.f90) file in the obj
> > directory it does work, provided I remove the flag -ffortran-bounds-
> > check, i.e.:
>
> > ftn -target=linux -c -O0 -g dbcsr_operations.f90
>
> > Previous cp2k versions allowed this flag so is there something strange
> > in this module?  Or is there a problem with the compiler (pathscale
> > version 3.2)?
>
> > Thank you,
> > Garold
>
> > --
> > You received this message because you are subscribed to the Google  
> > Groups "cp2k" group.
> > To post to this group, send email to cp... at googlegroups.com.
> > To unsubscribe from this group, send email to cp2k
> > +unsub... at googlegroups.com.
> > For more options, visit this group athttp://groups.google.com/
> > group/cp2k?hl=en.



More information about the CP2K-user mailing list