My first guess would be to lose "-ip" and any other parallel build crap. Intel's compilers are not quite stable w/r to them.
Also AMBER is broken with CUDA 4.0 right now. There's an unofficial fix floating around but that's a use at your own risk solution. CUDA 4.0 is of no benefit to AMBER at this time. Use CUDA 3.2 if you value your sanity...
-----Original Message-----
From: Jacopo Sgrignani [mailto:sgrigna.sissa.it]
Sent: Friday, March 25, 2011 01:48
To: amber.ambermd.org
Subject: [AMBER] compiling with cuda
Dear all
I'm trying to compile pmemd.cuda using intel compilers and the last
cudatoolkit, but the compilation stop whit an error i can not
understand.
Below there are the last lines of my output.
Can you help me?
Thanks
Jacopo
FFTLOADBAL_2PROC -DPUBFFT -DCUDA cuda_info.fpp cuda_info.f90
ifort -ip -O3 -axSTPW -DCUDA -I/scratch/sgrigna/cuda/cuda/include
-IB40C -IB40C/KernelCommon -c cuda_info.f90
icc -ip -O3 -axSTPW -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
-DBINTRAJ -DCUDA -I/scratch/sgrigna/cuda/cuda/include -IB40C
-IB40C/KernelCommon -c gpu.cpp
gputypes.h(1145): error: "pbConstraintSoluteID" has already been
declared in the current scope
GpuBuffer<int>* pbConstraintSoluteID;
// Physical constraint and molecule ID
^
gputypes.h(1146): error: "pbConstraintSoluteAtom" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSoluteAtom;
// Solute constraint atom positions and masses
^
gputypes.h(1147): error: "pbConstraintSolute" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSolute;
// Current and last centers of mass and inverse mass for each solute
constraint molecule
^
gputypes.h(1148): error: "pbConstraintUllSolute" has already been
declared in the current scope
GpuBuffer<PMEUllInt>* pbConstraintUllSolute;
// Current COM and in integer form
^
gputypes.h(1149): error: "pbConstraintSolventAtoms" has already been
declared in the current scope
GpuBuffer<int>* pbConstraintSolventAtoms;
// Number of atoms in each solvent constraint
^
gputypes.h(1150): error: "pbConstraintSolventAtom" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSolventAtom;
// Solvent constraint atom positions and masses, total inverse mass,
and COM
^
gputypes.h(1151): error: "pbConstraintSolventConstraint" has already
been declared in the current scope
GpuBuffer<int4>* pbConstraintSolventConstraint;
// List of solvent constraint constraint IDs
^
gputypes.h(1154): error: "pbConstraintSoluteID" has already been
declared in the current scope
GpuBuffer<int>* pbConstraintSoluteID;
// Physical constraint and molecule ID
^
gputypes.h(1155): error: "pbConstraintSoluteAtom" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSoluteAtom;
// Solute constraint atom positions and masses
^
gputypes.h(1156): error: "pbConstraintSolute" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSolute;
// Current and last centers of mass and inverse mass for each solute
constraint molecule
^
gputypes.h(1157): error: "pbConstraintUllSolute" has already been
declared in the current scope
GpuBuffer<PMEUllInt>* pbConstraintUllSolute;
// Current COM and in integer form
^
gputypes.h(1158): error: "pbConstraintSolventAtoms" has already been
declared in the current scope
GpuBuffer<int>* pbConstraintSolventAtoms;
// Number of atoms in each solvent constraint
^
gputypes.h(1159): error: "pbConstraintSolventAtom" has already been
declared in the current scope
GpuBuffer<PMEDouble>* pbConstraintSolventAtom;
// Solvent constraint atom positions and masses, total inverse mass,
and COM
^
gputypes.h(1160): error: "pbConstraintSolventConstraint" has already
been declared in the current scope
GpuBuffer<int4>* pbConstraintSolventConstraint;
// List of solvent constraint constraint IDs
^
compilation aborted for gpu.cpp (code 2)
make[3]: *** [gpu.o] Error 2
make[3]: Leaving directory `/scratch/sgrigna/amber11/src/pmemd/src/cuda'
make[2]: *** [-L/scratch/sgrigna/cuda/cuda/lib64] Error 2
make[2]: Leaving directory `/scratch/sgrigna/amber11/src/pmemd/src'
make[1]: *** [cuda] Error 2
make[1]: Leaving directory `/scratch/sgrigna/amber11/src/pmemd'
make: *** [cuda] Error 2
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Mar 25 2011 - 09:00:03 PDT
