Memory profiling in R

Thomas Lumley
University of Washington

R is well-known to be slow and able to handle only small data sets. A 
critical part of improving the performance of R is the ability to measure 
performance accurately, and the R profiler has been valuable in improving 
start-up time and S4 method handling. This success motivates extending the 
profiler to measure memory use as well as time. I will explain why 
memory-use profiling is less straightforward than time profiling and 
describe three memory profiling facilities that have been added to R 
2.4.0.