Doug Bates writes:>... Also, the search to resolve a name is a linear search over the >available names. ...Does anyone have an idea of how series a problem this is with large packages. R still runs slower than Splus 3.3 for me, and I wonder how big a part of the problem this is? Paul Gilbert -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Fri, 26 Mar 1999, Paul Gilbert wrote:> Doug Bates writes: > >... Also, the search to resolve a name is a linear search over the > >available names. ... > > Does anyone have an idea of how series a problem this is with large packages. R > still runs slower than Splus 3.3 for me, and I wonder how big a part of the > problem this is?A very limited profiling experiment about a year ago (so out of date as well as limited) suggested that perhaps 15-30% of R's time is spent in the linear search functions when running the survival4 test suite. We do intend to implement hashing for name lookup fairly soon. It should make a noticeable difference, especially with huge libraries like lme (over 500 functions) loaded. The relative speed of R and S-PLUS is very problem-dependent, so it's hard to make any reliable predictions. -thomas -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
> Paul Gilbert writes: > > Does anyone have an idea of how series a problem this is with large > packages. R still runs slower than Splus 3.3 for me, and I wonder how > big a part of the problem this is?I suspect that it is a very large part of your problem. A substantial fraction (30% - 50% ?) of R's time is spent doing name lookup. I hope hashing will speed up name lookup by one or two orders of magnitude. I am part-way through making the changes now. Ross -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._