Automatic reply from Jenkins: Can I test this?

Jenkins: ok to test

even though we've configured Jenkins to run tests automagically for your PRs, it doesn't seem to agree on that... weird

@boegel I just copied it from the app makefile. You are right, this should be left for EB. I will fix it

@boegel please take a look at the updated ebconfig

is there any way to setup CPPFLAGS="-mfpmath=sse -msse -mmmx" using toolchainopts?

Setting machine specific code generation options is IMHO site specific customization and there is no builtin approach for this in EB as of yet. One way would be versionsuffix. The RaxML configs do that (also to differentiate between OpenMP+MPI, OpenMP, MPI builds).

This is a good case of hard coded CPPFLAGS et al which is quite common in a lot of bioinformatics packages where the authors evidently expect users to adjust the options to local conditions. I think setting CFLAGS or CPPFLAGS in such a case is actually a good pattern for MakeCP based builds that have machine specific compiler flags set - makes it much easier to edit and change this way.

There might be people using EB on ARM or POWER some day :)

Also looks good to me, and builds fine.

Actually, I don't feel this is site-specific. It's specific to the compiler, that's true, but those flags were hardcoded in already anyway.

For a GCC-based toolchain, EasyBuild will use --march=native on both Intel and AMD systems, see https://github.com/hpcugent/easybuild-framework/blob/master/easybuild/toolchains/compiler/gcc.py#L69 .

@pescobar: I think --march=native already includes to flag you mention? Except maybe for -mfpmath=sse, I feel it's OK to add that (x86-specific stuff is currently used all over the place anyway).

@boegel it is for sure specific to the machine, by setting compiler code generation options that are dependent on a specific feature set of the CPU. Just imagine that the code is setting -mcpu=power8 -maltivec instead.

@pescobar: No need to add mfpmath=sse yourself, it's already enabled by default in GCC 4.x (see http://gcc.gnu.org/onlinedocs/gcc-4.0.0/gcc/i386-and-x86_002d64-Options.html).

The other two should be covered by march=native

thanks for clarification @boegel

then I think the easyconfig is fine now.

Jenkins: test this please

SUCCESS
Build succeeded for 1 out of 1
Linux debian 6.0.9, Intel(R) Xeon(R) CPU L5640 @ 2.27GHz, Python 2.6.6
See https://gist.github.com/86ac85d8cc47eafcb910 for a full test report.

looks good - and PR worked, too!

Test report by @boegel
SUCCESS
Build succeeded for 1 out of 1
Linux SL 6.5, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.6.6
See https://gist.github.com/426bc1bfed618db48e86 for a full test report.

Test report by @boegel
SUCCESS
Build succeeded for 1 out of 1
Linux SL 6.5, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.6.6
See https://gist.github.com/e4ce68b0f16d64e6477e for a full test report.

Test report by @fgeorgatos
SUCCESS
Build succeeded for 1 out of 1
Linux debian 6.0.10, Intel(R) Xeon(R) CPU L5640 @ 2.27GHz, Python 2.6.6
See https://gist.github.com/18a03833cdfd01ee73be for a full test report.

Test PASSed.

Test report by @boegel
SUCCESS
Build succeeded for 1 out of 1 (1 easyconfigs in this PR)
Linux SL 6.5, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.6.6
See https://gist.github.com/72b9c3563777bff5a503 for a full test report.

Going in, thanks @pescobar!