subject:"\[Wien\] WIEN2k on Xeon 5500 Series"

[Wien] WIEN2k on Xeon 5500 Series

2011-11-27 Thread フォンスポール

I have recompiled Wien2K with the 3.174 version of the compiler (Update 3 of
this year as I recall). Wien2K works fine. As the intel compiler is at least
at update 6, I am curious to know what the problems are in a general sense. Is
Intel's compiler becoming unreliable or might there actually be memory
allocation related errors in the Wien2K code. I ask as it is not clear to me
if the intel compiler will let one run older revisions of the compiler along
side newer versions (e.g. if Wien2K requires and older version, does this mean
I cannot compile other code with a newer version without reinstalling the
compiler? Any idea of what is causing the problems with recent versions of the
Intel compiler and has anyone asked Intel about it (as it might be their
problem?).

Thanks,
Paul Fons

On Nov 24, 2011, at 9:19 AM, Aaron Sutton wrote:

Hi,
I also tried the -heap-arrays suggestion from intel to no avail. I
believe the two packages you've mentioned, including revision 3 of the
2011 intel fortran compiler, should be all you need. As for the C
compiler, I'm using gcc 4.6.1, which I believe is the standard version
on Ubuntu 11.10.

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca

2011/11/23 Paul Fons paul-fons at aist.go.jp:
I am trying to compile Wien2K as well with the latest intel compilers. At
first I thought it might be a stack overflow problem (of course limit
stacksize was set to unlimited) and I tried the suggestion -heap-arrays from
intel
(http://software.intel.com/en-us/articles/determining-root-cause-of-sigsegv-or-sigbus-errors/)
but it did not help either.
Has any progress been made in tracking down the source of this error? The
traceback is essentially worthless. Has anyone corresponded with intel with
regards to the problem?

I am now in the process of downloading an older version of the compiler. I
am a little worried about downloading having different versions of the
compiler as there is also C code in Wien2K. Exactly which versions of the
C, Fortran, and MKL libraries should be used. I have a copy of a Intel
Parallel Studio XE for Linux and don't typically download the individual
packages hence my confusion.
From what I understand, update 3 is the
I am downloading currently:
Fortran
l_fcompxe_2011.3.174.tgz
]MKL
l_mkl_10.3.3.174.tgz

On Nov 11, 2011, at 6:04 AM, Aaron Sutton wrote:

Great, thank you I will try this. I haven't seen this as I'm new to
the mailing list. I actually eliminated the the segmentation fault by
linking to libmkl_mc.so explicitly, this however has caused an error
in SECLR4. Don't think I'm further along, but at least the error is
different.

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca

On Thu, Nov 10, 2011 at 15:48, Gavin Abo gsabo at crimson.ua.edu wrote:

Hi Aaron,

If you have been following some of the recent mailing list posts, I think

this error is because of the intel compiler. I had the same or similar

error with update 6. I never tried update 7, but it sounds like the

compiler does not seem to be compiling the Wien2k code correctly like update

6. Try update 3 of Intel FORTRAN compiler MKL

library, which should be composerxe-2011.3.174.

Gavin

On 11/10/2011 2:31 PM, Aaron Sutton wrote:

Hi,

I am experiencing a segmentation fault when running WIEN2k compiled

using the most recent version of the Intel FORTRAN compiler MKL

library (composer_xe_2011_sp1.7.256). The computer it's being run on

is a 2009 Mac Pro with 2 quad core intel xeon 5500 series processors

at 2.93GHz and 8GB RAM. The computer is natively running Ubuntu Linux

11.10 (though the same issue occurred while running 11.04). WIEN2k is

compiled using the default flags and links provided by the Linux with

Intel FORTRAN v12.0 + MKL option of siteconfig_lapw, no modifications.

The software compiles without error. Upon running the software, either

through the web interface or using run_lapw in the case directory,

lapw0 completes successfully and then I receive:

LAPW0 END

forrtl: severe (174): SIGSEGV, segmentation fault occurred

IMAGE PC

ROUTINE LINE

SOURCE

libmkl_mc3.so 2B0321624A30 Unknown

Unknown Unknown

stop error

According to the intel compiler documentation, the libmkl_mc3 library

is a kernel library specifically for core i7 type processors, which

the Xeon 5500s are. I do not link specifically to libmkl_mc3.so

anywhere as I am

[Wien] WIEN2k on Xeon 5500 Series

2011-11-26 Thread Laurence Marks

It is 99.9% probable that this is a compiler problem. Unfortunately you
cannot simply say to Intel that it does not work and have them
investigate. If someone can do a detailed debugging (ddd?) to isolate the
issue(s) then it can be reported and they will hopefully patch them. What
they would like (having done a little work on other bugs) is a short
Fortran code that reproduces the bug.

I have heard from someone that I trust that Vasp does not work with most of
the release 12 Intel compilers, similar to Wien2k.

N.B., while most of the Intel releases have been very good, they have
produced a bad product before.
---
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
Research is to see what everybody else has seen, and to think what nobody
else has thought
Albert Szent-Gyorgi
On Nov 26, 2011 9:38 AM, ??? paul-fons at aist.go.jp wrote:

I have recompiled Wien2K with the 3.174 version of the compiler (Update 3
of this year as I recall). Wien2K works fine. As the intel compiler is at
least at update 6, I am curious to know what the problems are in a general
sense. Is Intel's compiler becoming unreliable or might there actually be
memory allocation related errors in the Wien2K code. I ask as it is not
clear to me if the intel compiler will let one run older revisions of the
compiler along side newer versions (e.g. if Wien2K requires and older
version, does this mean I cannot compile other code with a newer version
without reinstalling the compiler? Any idea of what is causing the
problems with recent versions of the Intel compiler and has anyone asked
Intel about it (as it might be their problem?).

Thanks,
Paul Fons

On Nov 24, 2011, at 9:19 AM, Aaron Sutton wrote:

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca

2011/11/23 Paul Fons paul-fons at aist.go.jp:

I am trying to compile Wien2K as well with the latest intel compilers. At

first I thought it might be a stack overflow problem (of course limit

stacksize was set to unlimited) and I tried the suggestion -heap-arrays
from

intel

(
http://software.intel.com/en-us/articles/determining-root-cause-of-sigsegv-or-sigbus-errors/
)

but it did not help either.

Has any progress been made in tracking down the source of this error? The

traceback is essentially worthless. Has anyone corresponded with intel
with

regards to the problem?

I am now in the process of downloading an older version of the compiler. I

am a little worried about downloading having different versions of the

compiler as there is also C code in Wien2K. Exactly which versions of the

C, Fortran, and MKL libraries should be used. I have a copy of a Intel

Parallel Studio XE for Linux and don't typically download the individual

packages hence my confusion.

From what I understand, update 3 is the

I am downloading currently:

Fortran

l_fcompxe_2011.3.174.tgz

]MKL

l_mkl_10.3.3.174.tgz

On Nov 11, 2011, at 6:04 AM, Aaron Sutton wrote:

Great, thank you I will try this. I haven't seen this as I'm new to

the mailing list. I actually eliminated the the segmentation fault by

linking to libmkl_mc.so explicitly, this however has caused an error

in SECLR4. Don't think I'm further along, but at least the error is

different.

Aaron Sutton

Ph.D. Candidate | University of Toronto

Office: McLennan MP090 | Phone: +1 416 946 3639

Email: asutton at physics.utoronto.ca

On Thu, Nov 10, 2011 at 15:48, Gavin Abo gsabo at crimson.ua.edu wrote:

Hi Aaron,

If you have been following some of the recent mailing list posts, I think

this error is because of the intel compiler. I had the same or similar

error with update 6. I never tried update 7, but it sounds like the

compiler does not seem to be compiling the Wien2k code correctly like
update

6. Try update 3 of Intel FORTRAN compiler MKL

library, which should be composerxe-2011.3.174.

Gavin

On 11/10/2011 2:31 PM, Aaron Sutton wrote:

Hi,

I am experiencing a segmentation fault when running WIEN2k compiled

using the most recent version of the Intel FORTRAN compiler MKL

library (composer_xe_2011_sp1.7.256). The computer it's being run on

is a 2009 Mac Pro with 2 quad core intel xeon 5500 series processors

at 2.93GHz and 8GB RAM. The computer is natively running Ubuntu Linux

[Wien] WIEN2k on Xeon 5500 Series

2011-11-24 Thread Paul Fons

I am trying to compile Wien2K as well with the latest intel compilers.  At 
first I thought it might be a stack overflow problem (of course limit stacksize 
was set to unlimited) and I tried the suggestion -heap-arrays from intel 
(http://software.intel.com/en-us/articles/determining-root-cause-of-sigsegv-or-sigbus-errors/)
 but it did not help either.  

Has any progress been made in tracking down the source of this error?  The 
traceback is essentially worthless.  Has anyone corresponded with intel with 
regards to the problem?


I am now in the process of downloading an older version of the compiler.  I am 
a little worried about downloading having different versions of the compiler as 
there is also C code in Wien2K.  Exactly which versions of the C, Fortran, and 
MKL libraries should be used.  I have a copy of a Intel Parallel Studio XE for 
Linux and don't typically download the individual packages hence my confusion.

[Wien] WIEN2k on Xeon 5500 Series

2011-11-24 Thread Peter Blaha

Do you have the same error as mentioned in a previous post ?

I'm using composer_xe_2011_sp1.7.256(ifort 12.1.0) and do NOT have any
problems.

However, it runs on an AMD Opteron system, and it could well be that the problem
is specific to using Intel-processors, as the mkl may have special optimizations
for this processor.

FC = ifort
MPF = mpiifort
CC = icc
FOPT = -FR -mp1 -w -prec_div -pc80 -pad -align -DINTEL_VML -traceback -msse3
-fno-alias
FPOPT = $(FOPT)
DParallel = '-DParallel'
FGEN = $(PARALLEL)
LDFLAGS = $(FOPT) -L$(MKLROOT)/lib/intel64
R_LIBS = -lmkl_lapack95_lp64 -lmkl_blas95_lp64 -lmkl_intel_lp64
-lmkl_intel_thread -lmkl_core -openmp -lpthread
C_LIBS = $(R_LIBS)
RP_LIBS = -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 $(R_LIBS)
-L../SRC_lib/fftw/lib -lfftw_mpi -lfftw

On our own Intel based machines we use composerxe-2011.3.174 (ifort
12.0.3) and this works fine.

Am 24.11.2011 00:20, schrieb Paul Fons:
I am trying to compile Wien2K as well with the latest intel compilers. At
first I thought it might be a stack overflow problem (of course limit
stacksize was set to unlimited) and
I tried the suggestion -heap-arrays from intel
(http://software.intel.com/en-us/articles/determining-root-cause-of-sigsegv-or-sigbus-errors/)
but it did not help either.

Has any progress been made in tracking down the source of this error? The
traceback is essentially worthless. Has anyone corresponded with intel with
regards to the problem?

I am now in the process of downloading an older version of the compiler. I am
a little worried about downloading having different versions of the compiler
as there is also C code
in Wien2K. Exactly which versions of the C, Fortran, and MKL libraries should
be used. I have a copy of a Intel Parallel Studio XE for Linux and don't
typically download the
individual packages hence my confusion.

From what I understand, update 3 is the

I am downloading currently:

Fortran
l_fcompxe_2011.3.174.tgz

]MKL
l_mkl_10.3.3.174.tgz

On Nov 11, 2011, at 6:04 AM, Aaron Sutton wrote:

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca mailto:asutton at physics.utoronto.ca

On Thu, Nov 10, 2011 at 15:48, Gavin Abo gsabo at crimson.ua.edu
mailto:gsabo at crimson.ua.edu wrote:
Hi Aaron,

If you have been following some of the recent mailing list posts, I think
this error is because of the intel compiler. I had the same or similar
error with update 6. I never tried update 7, but it sounds like the
compiler does not seem to be compiling the Wien2k code correctly like update
6. Try update 3 of Intel FORTRAN compiler MKL
library, which should be composerxe-2011.3.174.

Gavin

On 11/10/2011 2:31 PM, Aaron Sutton wrote:

Hi,
I am experiencing a segmentation fault when running WIEN2k compiled
using the most recent version of the Intel FORTRAN compiler MKL
library (composer_xe_2011_sp1.7.256). The computer it's being run on
is a 2009 Mac Pro with 2 quad core intel xeon 5500 series processors
at 2.93GHz and 8GB RAM. The computer is natively running Ubuntu Linux
11.10 (though the same issue occurred while running 11.04). WIEN2k is
compiled using the default flags and links provided by the Linux with
Intel FORTRAN v12.0 + MKL option of siteconfig_lapw, no modifications.
The software compiles without error. Upon running the software, either
through the web interface or using run_lapw in the case directory,
lapw0 completes successfully and then I receive:

LAPW0 END
forrtl: severe (174): SIGSEGV, segmentation fault occurred
IMAGE PC
ROUTINE LINE
SOURCE

libmkl_mc3.so 2B0321624A30 Unknown
Unknown Unknown

stop error

According to the intel compiler documentation, the libmkl_mc3 library
is a kernel library specifically for core i7 type processors, which
the Xeon 5500s are. I do not link specifically to libmkl_mc3.so
anywhere as I am using the default options. I can report that this
error does not occur on i3 or i5 processors. Any thoughts on how to
remedy this situation?

Thank you.

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca mailto:asutton at
physics.utoronto.ca
___
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at mailto:Wien at
zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[Wien] WIEN2k on Xeon 5500 Series

2011-11-24 Thread Laurence Marks

Neither the heap array idea (a bad one, look up what it really does) or
ulimit are relevant. In fact the key codes in Wien2k now set ulimit
themselves due to some issues with openmpi.

SIGSEV errors mean in most cases that the code has got corrupted by an
array going out of bounds or an illegal instruction. Almost certainly Intel
has overoptimized something so the machine level code is doing something
wrong. Sometimes turning on every possible debug option then running inside
a debugger (ddd) can work to find it. It is not easy, which is why you do
not have people rushing forward with answers. (I do not have that compiler.)
On Nov 23, 2011 7:20 PM, Aaron Sutton asutton at physics.utoronto.ca wrote:

Aaron Sutton
Ph.D. Candidate | University of Toronto
Office: McLennan MP090 | Phone: +1 416 946 3639
Email: asutton at physics.utoronto.ca

2011/11/23 Paul Fons paul-fons at aist.go.jp:
I am trying to compile Wien2K as well with the latest intel compilers.
At
first I thought it might be a stack overflow problem (of course limit
stacksize was set to unlimited) and I tried the suggestion -heap-arrays
from
intel
(
http://software.intel.com/en-us/articles/determining-root-cause-of-sigsegv-or-sigbus-errors/
)
but it did not help either.
Has any progress been made in tracking down the source of this error?
The
traceback is essentially worthless. Has anyone corresponded with intel
with
regards to the problem?

I am now in the process of downloading an older version of the compiler.
I
am a little worried about downloading having different versions of the
compiler as there is also C code in Wien2K. Exactly which versions of
the
C, Fortran, and MKL libraries should be used. I have a copy of a Intel
Parallel Studio XE for Linux and don't typically download the individual
packages hence my confusion.
From what I understand, update 3 is the
I am downloading currently:
Fortran
l_fcompxe_2011.3.174.tgz
]MKL
l_mkl_10.3.3.174.tgz