Many Problems Building w/ MKL 11, ifort, icc; scipy.test() Fails w/ Undefined Symbols
I've tried this about 6 different ways now, and all successful builds result in the same error during scipy.test() (note that NumPy builds fine with same site.cfg): ====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz Here's some info on my environment: Intel Composer XE 2013.5.192, ifort+icc icc flags: -fPIC -DMKL_ILP64 -m64 -fp-model', 'strict', '-fomit-frame-pointer', '-openmp', '-xhost', '-parallel' ifort flags: -fPIC -i8 -xhost -openmp -fp-model strict OS: Ubuntu 12.04 24 Cores Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz Intel LibMKL 11.0.5 site.cfg: [mkl] library_dirs = /opt/intel/mkl/lib/intel64:/opt/intel/lib/intel64 include_dirs = /opt/intel/mkl/include:/opt/intel/include lapack_libs = mkl_lapack95_lp64 mkl_libs = mkl_def, mkl_intel_lp64, mkl_intel_thread, mkl_core, mkl_mc [fftw] library_dirs = /opt/intel/mkl/lib/intel64 include_dirs = /opt/intel/mkl/include libraries = fftw3 [amd] amd_libs = amd library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/ [umfpack] umfpack_libs = umfpack library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/ Built with: python setup.py clean python setup.py -v config --compiler=intelem --fcompiler=intelem build_clib --compiler=intelem --fcompiler=intelem build_ext --compiler=intelem --fcompiler=intelem install FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include'] FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include'] FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include'] ( library_dirs = /usr/local/lib:/usr/lib:/usr/lib/x86_64-linux-gnu ) FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include'] **The problem is here:** ====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz Now, where is mkl_blas_dgemm_blk_info_bdz defined? daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_core.so | grep mkl_blas_dgemm_blk_info_bdz 00000000000c59f0 T mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ Right there ^. I included mkl_core in site.cfg and it seems to be linking it, so I'm not sure why this symbol is still undefined. Note that it *is* undefined in libmkl_intel_thread.so: daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_intel_thread.so | grep mkl_blas_dgemm_blk_info_bdz U mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ I also tried the instructions provided by intel at http://software.intel.com/en-us/articles/numpyscipy-with-intel-mkl, using just mkl_rt. This had the same result, and seems incorrect, as it's not providing any lapack libraries (I'm assuming it would just fall back to "no lapack" and it seems to). This is highly confusing, as all of the MKL numpy/scipy instructions seem to be for older MKLs, before mkl_rt. Could anyone help me debug this? I'm hopelessly frustrated with it at this point and willing to do whatever to help you guys debug. Thanks! What additional info is needed?
On Wed, Aug 14, 2013 at 9:52 PM, Will Buckner <wbuckner@beatsmusic.com>wrote:
I've tried this about 6 different ways now, and all successful builds result in the same error during scipy.test() (note that NumPy builds fine with same site.cfg):
====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz
Here's some info on my environment:
Intel Composer XE 2013.5.192, ifort+icc icc flags: -fPIC -DMKL_ILP64 -m64 -fp-model', 'strict', '-fomit-frame-pointer', '-openmp', '-xhost', '-parallel' ifort flags: -fPIC -i8 -xhost -openmp -fp-model strict OS: Ubuntu 12.04 24 Cores Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz Intel LibMKL 11.0.5
site.cfg:
[mkl] library_dirs = /opt/intel/mkl/lib/intel64:/opt/intel/lib/intel64 include_dirs = /opt/intel/mkl/include:/opt/intel/include lapack_libs = mkl_lapack95_lp64 mkl_libs = mkl_def, mkl_intel_lp64, mkl_intel_thread, mkl_core, mkl_mc
[fftw] library_dirs = /opt/intel/mkl/lib/intel64 include_dirs = /opt/intel/mkl/include libraries = fftw3
[amd] amd_libs = amd library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/
[umfpack] umfpack_libs = umfpack library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/
Built with: python setup.py clean python setup.py -v config --compiler=intelem --fcompiler=intelem build_clib --compiler=intelem --fcompiler=intelem build_ext --compiler=intelem --fcompiler=intelem install
FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
( library_dirs = /usr/local/lib:/usr/lib:/usr/lib/x86_64-linux-gnu ) FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
**The problem is here:**
====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz
Now, where is mkl_blas_dgemm_blk_info_bdz defined?
daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_core.so | grep mkl_blas_dgemm_blk_info_bdz 00000000000c59f0 T mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$
Right there ^.
I included mkl_core in site.cfg and it seems to be linking it, so I'm not sure why this symbol is still undefined. Note that it *is* undefined in libmkl_intel_thread.so:
daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_intel_thread.so | grep mkl_blas_dgemm_blk_info_bdz U mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$
I also tried the instructions provided by intel at http://software.intel.com/en-us/articles/numpyscipy-with-intel-mkl, using just mkl_rt. This had the same result, and seems incorrect, as it's not providing any lapack libraries (I'm assuming it would just fall back to "no lapack" and it seems to). This is highly confusing, as all of the MKL numpy/scipy instructions seem to be for older MKLs, before mkl_rt.
These are fairly recent and mention mkl_rt: http://scipy.org/scipylib/building/linux.html#any-distribution-with-intel-c-.... Did you find those? Ralf
Could anyone help me debug this? I'm hopelessly frustrated with it at this point and willing to do whatever to help you guys debug.
Thanks! What additional info is needed?
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
Hadn't found them until you told me, but I've now carefully reviewed and followed instructions from that doc. It's still not linking mkl_core properly before intel_thread. Using mkl_rt doesn't help at all--and, I'm unclear why that 's not just the default directions if mkl_rt works properly. When using mkl_rt, I still must use LD_PRELOAD=/opt/intel/mkl/lib/intel64/libmkl_core.so:/opt/intel/lib/intel64/libiomp5.so or I get undefined symbol errors when running python. It's as if these two libraries just aren't linked, and LD_PRELOAD is a bandaid. Is there a way to manually edit the link line anywhere? Or where's it being built? If I control+f my terminal output from a verbose build, I see no "-lanything" being passed anywhere. Thanks, Will On Wed, Aug 14, 2013 at 2:19 PM, Ralf Gommers <ralf.gommers@gmail.com>wrote:
On Wed, Aug 14, 2013 at 9:52 PM, Will Buckner <wbuckner@beatsmusic.com>wrote:
I've tried this about 6 different ways now, and all successful builds result in the same error during scipy.test() (note that NumPy builds fine with same site.cfg):
====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz
Here's some info on my environment:
Intel Composer XE 2013.5.192, ifort+icc icc flags: -fPIC -DMKL_ILP64 -m64 -fp-model', 'strict', '-fomit-frame-pointer', '-openmp', '-xhost', '-parallel' ifort flags: -fPIC -i8 -xhost -openmp -fp-model strict OS: Ubuntu 12.04 24 Cores Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz Intel LibMKL 11.0.5
site.cfg:
[mkl] library_dirs = /opt/intel/mkl/lib/intel64:/opt/intel/lib/intel64 include_dirs = /opt/intel/mkl/include:/opt/intel/include lapack_libs = mkl_lapack95_lp64 mkl_libs = mkl_def, mkl_intel_lp64, mkl_intel_thread, mkl_core, mkl_mc
[fftw] library_dirs = /opt/intel/mkl/lib/intel64 include_dirs = /opt/intel/mkl/include libraries = fftw3
[amd] amd_libs = amd library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/
[umfpack] umfpack_libs = umfpack library_dirs = /usr/lib include_dirs = /usr/include/suitesparse/
Built with: python setup.py clean python setup.py -v config --compiler=intelem --fcompiler=intelem build_clib --compiler=intelem --fcompiler=intelem build_ext --compiler=intelem --fcompiler=intelem install
FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
FOUND: libraries = ['mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
( library_dirs = /usr/local/lib:/usr/lib:/usr/lib/x86_64-linux-gnu ) FOUND: libraries = ['mkl_lapack95_lp64', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'mkl_mc', 'pthread'] library_dirs = ['/opt/intel/mkl/lib/intel64'] define_macros = [('SCIPY_MKL_H', None)] include_dirs = ['/opt/intel/mkl/include', '/opt/intel/include']
**The problem is here:**
====================================================================== ERROR: Failure: ImportError (/opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/nose/loader.py", line 413, in loadTestsFromName addr.filename, addr.module) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/local/lib/python2.7/dist-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/usr/local/lib/python2.7/dist-packages/scipy/stats/__init__.py", line 320, in <module> from .stats import * File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 242, in <module> import scipy.linalg as linalg File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/__init__.py", line 147, in <module> from .misc import * File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/misc.py", line 5, in <module> from . import blas File "/usr/local/lib/python2.7/dist-packages/scipy/linalg/blas.py", line 113, in <module> from scipy.linalg import _fblas ImportError: /opt/intel/composer_xe_2013.5.192/mkl/lib/intel64/libmkl_intel_thread.so: undefined symbol: mkl_blas_dgemm_blk_info_bdz
Now, where is mkl_blas_dgemm_blk_info_bdz defined?
daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_core.so | grep mkl_blas_dgemm_blk_info_bdz 00000000000c59f0 T mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$
Right there ^.
I included mkl_core in site.cfg and it seems to be linking it, so I'm not sure why this symbol is still undefined. Note that it *is* undefined in libmkl_intel_thread.so:
daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$ nm libmkl_intel_thread.so | grep mkl_blas_dgemm_blk_info_bdz U mkl_blas_dgemm_blk_info_bdz daisy@speedballcompute01:/opt/intel/mkl/lib/intel64$
I also tried the instructions provided by intel at http://software.intel.com/en-us/articles/numpyscipy-with-intel-mkl, using just mkl_rt. This had the same result, and seems incorrect, as it's not providing any lapack libraries (I'm assuming it would just fall back to "no lapack" and it seems to). This is highly confusing, as all of the MKL numpy/scipy instructions seem to be for older MKLs, before mkl_rt.
These are fairly recent and mention mkl_rt: http://scipy.org/scipylib/building/linux.html#any-distribution-with-intel-c-.... Did you find those?
Ralf
Could anyone help me debug this? I'm hopelessly frustrated with it at this point and willing to do whatever to help you guys debug.
Thanks! What additional info is needed?
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
-- *Will Buckner* Machine Learning & Data Architect (408) 637-8466 (m) wbuckner@beatsmusic.com * * *Beats Music* * * The information contained in or attached to this transmission may contain privileged and confidential information. It is intended only for the use of the person(s) named above. If you are not the intended recipient, you are hereby notified that any review, dissemination, distribution or duplication of this communication is strictly prohibited. If you are not the intended recipient, please contact the sender by reply email and delete all copies of the original message.
Looks to me like you need to update your system's ldconfig files. Without an -rpath / -R linker flag, the lib. directories will need to be known by the dynamic linker during run time, as well as during compile-time linking (which appears to complete successfully). This is accomplished by creating a file in /etc/ldconfig.d/ and putting each of Intel's lib directories on a separate line. Finally, run ldconfig as root. Also, isn't libmkl_mc specific to the mic architecture? I don't think it's necessary to explicitly link against it. Instead, libiomp5, libimf and libirc are libraries I often find are required. KR, and good luck! Alex
Also, isn't libmkl_mc specific to the mic architecture? I don't think it's necessary to explicitly link against it. Instead, libiomp5, libimf and
Thanks, but, I had already added them to ld.so.conf.d/ and ran ldconfig. I eventually got it to run without LD_RUN_PATH somehow by removing everything in build/ (had previously been running python setup.py clean) and finally got everything to work fairly sanely with: [mkl] library_dirs = /opt/intel/mkl/lib/intel64:/opt/intel/lib/intel64 include_dirs = /opt/intel/mkl/include:/opt/intel/include mkl_libs = mkl_def, mkl_intel_lp64, mkl_intel_thread, mkl_core, mkl_mc, iomp5 lapack_libs = libirc are libraries I often find are required. So I know nothing about the MKL stack, or BLAS for that matter. Intel docs say "Kernel library for processors based on the Intel(R) Core(TM) microarchitecture". My CPUs are "Intel(R) Xeon(R) CPU E7- 4860 @ 2.27GHz". Also keep in mind, this is MKL 11 and a lot has changed since 9x, apparently, and it seems most everyone writing docs on MKL w/ SciPy/NumPy is using older MKL. What I really want to figure out is how to get "mkl_rt" working properly. Should I have something in lapack_libs? Thanks! On Thu, Aug 22, 2013 at 3:40 PM, Alex Leach <beamesleach@gmail.com> wrote:
Looks to me like you need to update your system's ldconfig files. Without an -rpath / -R linker flag, the lib. directories will need to be known by the dynamic linker during run time, as well as during compile-time linking (which appears to complete successfully).
This is accomplished by creating a file in /etc/ldconfig.d/ and putting each of Intel's lib directories on a separate line. Finally, run ldconfig as root.
Also, isn't libmkl_mc specific to the mic architecture? I don't think it's necessary to explicitly link against it. Instead, libiomp5, libimf and libirc are libraries I often find are required.
KR, and good luck! Alex
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
mkl_rt are working well for me, here are my options: site.cfg: [mkl] #Vversion=composer_xe_2011_sp1.7.256 #Vversion=composer_xe_2011_sp1.8.273 #Vversion=composer_xe_2011_sp1.9.293 #Vversion=composer_xe_2011_sp1.10.319 #Vversion=composer_xe_2011_sp1.11.339 #Vversion=composer_xe_2013.1.117 Vversion=composer_xe_2013.3.163 library_dirs = /opt/intel/%(Vversion)s/mkl/lib/intel64:/opt/intel/%(Vversion)s/compiler/lib/intel64 include_dirs = /opt/intel/%(Vversion)s/mkl/include:/opt/intel/%(Vversion)s/mkl/include/intel64/lp64 lapack_libs = mkl_blas95_lp64, mkl_lapack95_lp64 mkl_libs = mkl_rt in numpy/distutils/intelccompiler.py: class IntelCCompiler(UnixCCompiler): """ A modified Intel compiler compatible with an gcc built Python.""" compiler_type = 'intel' cc_exe = 'icc' cc_args = 'fPIC' def __init__ (self, verbose=0, dry_run=0, force=0): UnixCCompiler.__init__ (self, verbose,dry_run, force) self.cc_exe = 'icc -fPIC' compiler = self.cc_exe self.set_executables(compiler=compiler, compiler_so=compiler, compiler_cxx=compiler, linker_exe=compiler, linker_so=compiler + ' -shared') class IntelItaniumCCompiler(IntelCCompiler): compiler_type = 'intele' # On Itanium, the Intel Compiler used to be called ecc, let's search for # it (now it's also icc, so ecc is last in the search). for cc_exe in map(find_executable,['icc','ecc']): if cc_exe: break class IntelEM64TCCompiler(UnixCCompiler): """ A modified Intel x86_64 compiler compatible with a 64bit gcc built Python. """ Vversion='composer_xe_2011_sp1.9.293' Vversion='composer_xe_2011_sp1.10.319' Vversion='composer_xe_2011_sp1.11.339' Vversion='composer_xe_2013.1.117' Vversion='composer_xe_2013.3.163' LDir='-L/opt/intel/'+Vversion+'/mkl/lib/intel64 '+'-L/opt/intel/'+Vversion+'/compiler/lib/intel64 ' IDir='-I/opt/intel/'+Vversion+'/mkl/include/intel64/lp64 '+'-I/opt/intel/'+Vversion+'/mkl/include ' compiler_type = 'intelem' cc_exe = 'icc -m64 -O3 -g -fPIC -fp-model strict -fomit-frame-pointer -openmp -xHost -lpthread '+IDir+LDir cc_args = '-m64 -O3 -g -fPIC -fp-model strict -fomit-frame-pointer -openmp -xHost -lpthread '+IDir+LDir def __init__ (self, verbose=0, dry_run=0, force=0): UnixCCompiler.__init__ (self, verbose,dry_run, force) Vversion='composer_xe_2011_sp1.9.293' Vversion='composer_xe_2011_sp1.10.319' Vversion='composer_xe_2011_sp1.11.339' Vversion='composer_xe_2013.1.117' Vversion='composer_xe_2013.3.163' LDir='-L/opt/intel/'+Vversion+'/mkl/lib/intel64 '+'-L/opt/intel/'+Vversion+'/compiler/lib/intel64 ' IDir='-I/opt/intel/'+Vversion+'/mkl/include/intel64/lp64 '+'-I/opt/intel/'+Vversion+'/mkl/include ' self.cc_exe = 'icc -m64 -O3 -g -fPIC -fp-model strict -fomit-frame-pointer -openmp -xHost -lpthread '+IDir+LDir compiler = self.cc_exe self.set_executables(compiler=compiler, compiler_so=compiler, compiler_cxx=compiler, linker_exe=compiler, linker_so=compiler + ' -shared') and in numpy/distutils/fcompiler/intel.py: def get_flags_opt(self): return ['-xhost -m64 -openmp -fp-model strict'] My script to compile everything is: python setup.py config --compiler=intelem --fcompiler=intelem config_fc --fcompiler=intelem build_clib --compiler=intelem --fcompiler=intelem build_ext --compiler=intelem --fcompiler=intelem install --prefix=/opt/Python/Numpy-1.7.1-intel I hope it helps Xavier 2013/8/29 Will Buckner <wbuckner@beatsmusic.com>
Thanks, but, I had already added them to ld.so.conf.d/ and ran ldconfig. I eventually got it to run without LD_RUN_PATH somehow by removing everything in build/ (had previously been running python setup.py clean) and finally got everything to work fairly sanely with:
[mkl] library_dirs = /opt/intel/mkl/lib/intel64:/opt/intel/lib/intel64 include_dirs = /opt/intel/mkl/include:/opt/intel/include mkl_libs = mkl_def, mkl_intel_lp64, mkl_intel_thread, mkl_core, mkl_mc, iomp5 lapack_libs =
Also, isn't libmkl_mc specific to the mic architecture? I don't think it's necessary to explicitly link against it. Instead, libiomp5, libimf and libirc are libraries I often find are required.
So I know nothing about the MKL stack, or BLAS for that matter. Intel docs say "Kernel library for processors based on the Intel(R) Core(TM) microarchitecture". My CPUs are "Intel(R) Xeon(R) CPU E7- 4860 @ 2.27GHz". Also keep in mind, this is MKL 11 and a lot has changed since 9x, apparently, and it seems most everyone writing docs on MKL w/ SciPy/NumPy is using older MKL. What I really want to figure out is how to get "mkl_rt" working properly.
Should I have something in lapack_libs?
Thanks!
On Thu, Aug 22, 2013 at 3:40 PM, Alex Leach <beamesleach@gmail.com> wrote:
Looks to me like you need to update your system's ldconfig files. Without an -rpath / -R linker flag, the lib. directories will need to be known by the dynamic linker during run time, as well as during compile-time linking (which appears to complete successfully).
This is accomplished by creating a file in /etc/ldconfig.d/ and putting each of Intel's lib directories on a separate line. Finally, run ldconfig as root.
Also, isn't libmkl_mc specific to the mic architecture? I don't think it's necessary to explicitly link against it. Instead, libiomp5, libimf and libirc are libraries I often find are required.
KR, and good luck! Alex
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
_______________________________________________ SciPy-Dev mailing list SciPy-Dev@scipy.org http://mail.scipy.org/mailman/listinfo/scipy-dev
-- « Quand le gouvernement viole les droits du peuple, l'insurrection est, pour le peuple et pour chaque portion du peuple, le plus sacré des droits et le plus indispensable des devoirs » Déclaration des droits de l'homme et du citoyen, article 35, 1793
participants (4)
-
Alex Leach -
Ralf Gommers -
Will Buckner -
Xavier Barthelemy