What is the difference betwen Microsoft R Open (MRO) and R?

Question

I'm really want to know what is the main different betwen MRO and R. Then i found information that MRO is faster than R because it has multithreading.

i'm really interesting, so i want to test it by myself. First, i am install MRO + MKL, then i will run some script code in MRO and R then calculate the time.

here's my example (matrix multiplication) code:

start.time <- Sys.time()

d <- numeric(5); 
res <- replicate(5, {for(i in 1:3500){d[i] <- print(i)} ; d})
rex=res*res^0.6*res^-1
rex

end.time <- Sys.time()
time.taken <- end.time - start.time
time.taken

But the time result is not different (nearly same). i try another calculation like gwr (500 data), read and write all SHP data, but got same result. anybody know why the result isnt different? or my code isn't complicated enough to test the difference?, if so, can you give me some example?

Answer 1

Three differences:

• MRO comes with Intel MKL as the backend for linear algebra operations (BLAS nad LAPACK), while base R (for Windows) ships an unoptimized replacement of BLAS and LAPACK which will run much slower. This impacts things like matrix multiplications, solving linear systems, eigenvalues, and such (search for BLAS benchmarks), but not loops or models like decision trees. If platforms other than windows it won't make a difference, because then R will link to the system's default BLAS (typically MKL or OpenBLAS depending on the specific OS).
• MRO has its own CRAN mirror, so it will lag behind on updates to packages and will likely be slower to downloaded packages from there. This can be disabled if you know what you're doing. For some packages (particularly dealing with multi-threading) they might offer some non-default configurations too.
• MRO will auto-generate a default initialization script for R sessions which prints annoying messages and changes some configuration options such as where to download repositories from.

Tip: it's possible to get the same optimized BLAS and LAPACK libraries (e.g. MKL, OpenBLAS) in the default R install for windows by substituting the .dll files (Rblas.dll and Rlapack.dll) with the correct .dll, which can be taken from the same MRO install or from NumPy's windows builds (something like openblas.dll and openblas.lib, the former of which should be copied twice and renamed).