update easyblocks for Intel tools to support 2016 versions #756
Conversation
… more specific path is not found. This is more in line with how Intel sets up the environment by sourcing the compilervars.sh
* origin/pr/691: removed double modextravars and explicit gccversion = None fixed paths for TBB, debugger. dirmap will now hold original items if more specific path is not found. This is more in line with how Intel sets up the environment by sourcing the compilervars.sh removed make_module_extra Modified according to pull request easybuilders#691 Added support for installing Intel compilers v2016 added support for Intel 2016 compilers
* origin/pr/745: Add mkl 11.3.1.150 support
| @@ -30,6 +30,7 @@ | |||
| @author: Kenneth Hoste (Ghent University) | |||
| @author: Pieter De Baets (Ghent University) | |||
| @author: Jens Timmerman (Ghent University) | |||
| @author: Ward Poelmans (Ghent University) | |||
There was a problem hiding this comment.
@DirkdeDraak You didn't add yourself to the list of authors. What should I add?
|
Easyblocks unit test suite FAILed. See https://jenkins1.ugent.be/job/easybuild-easyblocks-pr-builder/1406/console for more details. Please fix the reported issues by pushing additional commits to the branch corresponding with this pull request; contact @boegel if you're not sure what to do. |
|
|
||
| if LooseVersion(self.version) >= LooseVersion('2016') and not self.cfg['components']: | ||
| self.log.warning("No components specified, falling back to ALL") | ||
| self.cfg.update('components', ['ALL']) |
There was a problem hiding this comment.
or just not specify components at all if components easyconfig parameter is not defined
| mediaconfigpath = os.path.join(self.cfg['start_dir'], 'pset', 'mediaconfig.xml') | ||
| if not os.path.isfile(mediaconfigpath): | ||
| self.log.warning("Could not find pset/mediaconfig.xml in the build directory. Skipping list of components to install") | ||
| return |
There was a problem hiding this comment.
don't return inline please, make it harder to understand the code
fall back to ALL as value for components for icc/ifort 2016.x (+ style fixes)
|
Easyblocks unit test suite PASSed (see https://jenkins1.ugent.be/job/easybuild-easyblocks-pr-builder/1460/console for more details). This pull request is now ready for review/testing. Please try and find someone who can tackle this; contact @boegel if you're not sure what to do. |
sync with develop
|
Easyblocks unit test suite PASSed (see https://jenkins1.ugent.be/job/easybuild-easyblocks-pr-builder/1493/console for more details). This pull request is now ready for review/testing. Please try and find someone who can tackle this; contact @boegel if you're not sure what to do. |
| guesses = { | ||
| 'CLASSPATH': ['daal/lib/daal.jar'], | ||
| # 'include' is deliberately listed last, so it gets precedence over the other subdirs | ||
| 'CPATH': ['daal/include', 'ipp/include', 'mkl/include', 'tbb/include', 'include'], |
There was a problem hiding this comment.
$CPATH was not set before, and it seems like we shouldn't start setting it either...
I think this is causing the problem in discussed at https://github.com/hpcugent/easybuild-easyblocks/pull/772/files#r47450112, i.e. error like this popping up:
lib/types.h(107): error: qualified name is not allowed
typedef std::complex<double> cdouble;
...
lib/types.h(107): error: expected a ";"
typedef std::complex<double> cdouble;
unless the include/c++/4.9.3 subdirectory of the GCC install dir is prepended to $CPATH
this is also mentioned in https://software.intel.com/en-us/forums/intel-c-compiler/topic/338378, to quote:
The problem has been resolved by now. Our system admin, who installed the compiler using the standard Intel installer (only changed the install directory from the default, but nothing else), unnecessarily added the the Intel include paths to the environment variables CPLUS_INCLUDE_PATH, so that #include_next included the same file again (like me, he wasn't aware of the #include_next directive).
…ith C++ std::complex
don't include 'include' icc/ifort subdir in $CPATH, causes problems with C++ std::complex
sync with develop & resolve conflict
|
Easyblocks unit test suite PASSed (see https://jenkins1.ugent.be/job/easybuild-easyblocks-pr-builder/1503/console for more details). This pull request is now ready for review/testing. Please try and find someone who can tackle this; contact @boegel if you're not sure what to do. |
boegel
changed the title
|
I gave this another thorough test using the new easyconfigs for:
and also by reinstalling the full toolchain and rebuilding HPL for I noticed a small issue with the sanity check of Thanks a lot for contributing to this @DirkdeDraak, @jas02 and @wpoely86! |
update easyblocks for Intel tools to support 2016 versions
This PR is based on top of #691 and #745. It adds support for specifying which components to install.