[R] R vs Matlab {R in Industry}"

Martin Maechler maechler at stat.math.ethz.ch
Fri Feb 9 09:46:39 CET 2007


>>>>> "Ravi" == Ravi Varadhan <rvaradhan at jhmi.edu>
>>>>>     on Thu, 8 Feb 2007 14:39:41 -0500 writes:

    Ravi> Here is a function to create a Toeplitz matrix of any size, and an example
    Ravi> of a 220 x 220 toeplitz matrix, which was created in almost no time:

Thanks Ravi,
but note two things

- ?toeplitz  tells you that R already has a fast (R-code-only)
  toeplitz() function

- The point of that benchmark is not to measure how fast you can
  build a Toeplitz matrix but simply to exercise a double
  (actually triple) for loop.
  {and the benchmark R script says so as comment in the code}

BTW {not to Ravi, but on the subject}:
 
1) When comparing this (the for-loop benchmark) --- with
   Matlab I would want to be sure that Matlab is not simply
   using an internal short cut since the benchmark is maybe
   too simplistic:
               for(i in 1:N) for(j in 1:N)  b[i,j] <- abs(i - j)

   {and it maybe interesting to see if R's experimental byte
    compiler would speed that up}

2) The above is very fast (IMO) and I cannot say why this could
   be too slow in a realistic situation.

3) The tables I've seen said that Matlab was about a factor of 2
   faster for the above loop benchmark.  That's scarcely a
   reason for downgrading (from R to Matlab).

Martin



More information about the R-help mailing list