Hello again,
thank you for the comments, especially this one:
Prof Brian Ripley wrote:
> My concern would be that there are different cases that fail under
> Fortran compiler X and you are just sweeping the problem under the
> carpet.
It inspired us to go back to search the cause, and we've made some
progress: it's not the compiler, it's the compiler options. Simple, but
it took a while to figure that out since my experience in these things
is limited. When I build the package with default options using INSTALL
--build the dll is built with option -O3 as per R's Makeconfig file. If
I build the dll by hand, using gfortran with no additional options and
dyn.load it, everything works, and also with -O and -Os. (No, I don't
fully understand what the differences between all these are, but that's
another question).
I'm looking at chapter 5.5 in Writing R Extensions and also 6.3.3 in R
Installation and Administration but I can't figure out how to tell
"inside" my package that it is not to be built -O3 but with, say, -O. I
can see how to add flags in the package (and as far as I can tell, if
there are several optimization level flags the last in line is used and
that's the wrong one from my point of view), and also how to override
flags but only on my computer. Am I blind or am I again attempting
something I shouldn't?
Best regards,
Seija S.
Prof Brian Ripley wrote:
On Thu, 14 May 2009, Seija Sirkiä wrote:
Hello all,
it seems my efforts in reading the manuals and help files aren't
enough so here I am. The question is, how would I go about linking a
pre-compiled DLL in to my package? I have previously successfully
built packages with Fortran and C source code, but now I'd like to
take this ready made DLL and call its routines from R.
My collegue was brave enough to simply try and put the DLL in src
folder of the package and proceed with building and installing the
same way as if it had been source code. And it works, on my computer
and his. Clearly this isn't exactly portable, and while that's ok for
the use we have in mind (we both work on Windows XP) I have a nagging
feeling this is somehow criminal or unwise at least.
Unwise, since it is unsupported and might stop working at any time.
Actually, you don't mention the version of R you used and (without
checking) I think this may not work in all recent versions of R.
What you can so is to have an 'inst/libs' directory in your package
sources, and put the DLL you want to use there. Another approach (used
in a few CRAN packages) is to use 'configure.win' to copy the DLL to a
'libs' directory in the installed package. A third approach is to have
a 'src/Makefile.win' that creates the DLL you want (possibly by just
checking it is present in 'src', but also possiby by calling the
mysterious Fortran compiler X you obliquely mention below).
Why I want to do this might explain it a bit further but in fact the
background contains another problem and I welcome anyone to give hints
about that too.
So, we have some Fortran code from way back which deals with fitting
taper curves for tree boles, and some other functions related to that.
We wanted to make these available for use in R and I made a package
with simple wrapper functions for the .Fortran calls, and built it
with the help of the compiler in Rtools. All was fine until my
collegue managed to bump in to a combination of parameters (tree
species, height and breast height diameter) with which the computation
freezes. We traced it to a certain very simple iteration in the
Fortran code that finds the height at which the tree has a given
diameter. I could give more details on that, but the point is that the
very same computation goes through just fine when executed "fully" in
Fortran, with the routines and an interface compiled with another
Fortran compiler. And also when that compiler is used to make a DLL
and that DLL is used within the R package in the way I described above.
In summary, what we seem to have here is a compiler dependent
convergence problem. Possible solutions are to figure out what's wrong
with the computation - and I've pretty much exhausted my skills on
that - or to figure out how to use a working DLL, as was my first
question.
My concern would be that there are different cases that fail under
Fortran compiler X and you are just sweeping the problem under the carpet.
Great big thanks in advance to anyone with advice!
Seija Sirkiä,
senior researcher (statistics), Finnish forestry research institute
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
