I am trying to migrate simulation code written in R-3.0.3 32 bit from Windows to Linux. But I am not getting matching results even though I used same R version on Linux (Redhat). I am seeing differences in the floating point results in each of the steps
I have used the below command to build R binary in Linux from the source code
./configure --with-readline=no --with-x=no CC="gcc -m32 -mfpmath=387 -ffloat-store" CXX="g++ -m32 -mfpmath=387 -ffloat-store" F77="gfortran -m32 -mfpmath=387 -ffloat-store" FC=${F77} OBJC=${CC} LDFLAGS="-L/usr/local/lib" LIBnn=lib r_arch=i686
I have also tried with the sse compilation options. Even those results were also not matching.
Below are the R configurations on Windows and Linux. In Windows I directly used the exe available.
Linux
R version 3.0.3 (2014-03-06) -- "Warm Puppy" Copyright (C) 2014 The R Foundation for Statistical Computing Platform: x86_64-unknown-linux-gnu/i686 (32-bit)
Windows
R version 3.0.3 (2014-03-06) -- "Warm Puppy" Copyright (C) 2014 The R Foundation for Statistical Computing Platform: i386-w64-mingw32/i386 (32-bit)
I have kept all the version of R libraries used by both Windows and Linux same. Please let me know if anyone has faced this issue. Wondering if any of the compilation option that I would have not tried would fix the issue