Displaying 1 result from an estimated 1 matches for "mvmultiply".
Did you mean:
multiply
2011 Jul 19
1
Measuring and comparing .C and .Call overhead
...even 10% of the time,
our effective gain will saturate at 10x. In other words, effective
parallelization can only be achieved by moving big chunks of the code to
C/C++ so as to avoid repeated calls from R to C.
I look forward to your comments.
-Alireza
--------------------------------
#code file: mvMultiply.r
#measuring execution time for a matrix-vector multiplication (A%*%x) using
.C and .Call functions
#execution time is measured both in R and inside the C++ functions;
subtracting the two is
#used as an indicator of overhead associated with each call
rm(list = ls())
system("R CMD SHLIB mvMul...