[Rd] rpvm arch check
ostrouchovg at ornl.gov
Thu Jan 24 16:14:32 CET 2013
Is there a reason that you choose to use PVM rather than MPI? MPI is the
current standard and PVM is no longer supported. There are R packages
pbdMPI and Rmpi that use MPI. There is also additional infrastructure
for SPMD style programming in other pbd packages on CRAN.
The best instructions on installing MPI on a Mac are at
Note that you might need to install Xcode and GNU compilers beforehand.
On 1/17/13 11:07 AM, Simon Urbanek wrote:
> On Jan 17, 2013, at 10:52 AM, Michael Li wrote:
>> OK. I will upload the updated version to CRAN and see if there is any problem with the checking.
>> One additional question, if a user decides to build the library from source he may still encounter the problem. Is there a way (though configure.in file maybe) to detect the architecture of the external library (libpvm.a) and use the correct architecture automatically when compiling rpvm code?
> That is automatic - multi-lib compilers will only pick libraries for the architecture that is currently being compiled.
> If you are using autoconf [properly], it will detect the libraries for the architecture being compiled at the time. There is really nothing you as an author need to do as long as you are aware of the multi-arch setup of R.
> It seems that rpvm has been removed from CRAN so I cannot comment on the rpvm package in particular.
>> On Tue, Jan 15, 2013 at 8:53 PM, Simon Urbanek<simon.urbanek at r-project.org> wrote:
>> On Jan 15, 2013, at 2:24 PM, Michael Li wrote:
>>> I've trying to fix the check error in rpvm package. Right now on my Mac, I use darwinport to install PVM and I can get rpvm to pass the check using 'R -arch x86_64 CMD check'. However the default 'R CMD check' uses i386 which fails since PVM is built for x86_64 only.
>>> So is there a way to ensure proper architecture is used that matches that of the external library like PVM? What does CRAN use when it checks packages?
>> We use static builds which have all necessary architectures (so users don't need to install them and can use any architecture).
> R-devel at r-project.org mailing list
George Ostrouchov, Ph.D.
Scientific Data Group
Computer Science and Mathematics Division
Oak Ridge National Laboratory
More information about the R-devel