Dear all, I have attached two preliminary ports for the SeqAn library and the official applications developed within the SeqAn project. These are my first OpenBSD ports so I expect some rough edges and am grateful for any and all comments.
Using the ports infrastructure actually resolved the issues I was seeing in my
previous thread "Problems porting bioinformatics library", I suppose this is
due to the gcc4 module doing more than just calling eg++.
Regarding the biology/seqan port I am currently doing a cp -pR for installing
the header files. This is probably not the right thing to do, but I couldn't
figure out what the OpenBSD equivalent to ${COPYTREE_SHARE} of the FreeBSD
ports is. I tried ${INSTALL_DATA} but this only copies files, not directories
and ${INSTALL_DATA_DIR} just creates empty dirs. What would be the right macro
for a recursive copy?
Another thing is, that we build our binaries with -O3 on all other platforms,
i.e. gcc, clang (each multiple versions) and icc; on Linux, Mac and FreeBSD.
We have seen in the past that more agressive inlining and loop vectorization
give measurably better results. I know that OpenBSD is reluctant about -O3 and
users who care about every 1% performance increase will likely pick a
different platform anyway, but I wouldn't want to slow down the binaries if I
don't have to. So, will you be convinced that O3 is worthwhile for the seqan-
apps port if I provide more numbers on this, or is the policy so strict that I
shouldn't bother and just accept it? :)
Regarding the apps a much more important point is OpenMP. I asked about this a
few weeks ago and will try to get back to that now, but as it is rather
independent of the seqan ports, I wanted to send them out already.
Thanks for your feedback,
Hannes
seqan-apps.tar.gz
Description: application/compressed-tar
seqan.tar.gz
Description: application/compressed-tar
