StackOverflow2013

Note that there are some explanatory texts on larger screens.

plurals

POreason for faster matrix allocation in R
text
Body
copied!<p>Posting <a href="https://stackoverflow.com/questions/1745622/best-way-to-allocate-matrix-in-r-null-vs-na">Best way to allocate matrix in R, NULL vs NA?</a> shows that writing your own matrix allocation function in R can be 8 to 10 times faster than using R's built-in matrix() function to pre-allocate a large matrix.</p> <p>Does anyone know why the hand crafted function is so much faster? What is R doing inside matrix() that is so slow? Thanks.</p> <p>Here's the code on my system:</p> <pre><code>create.matrix <- function( nrow, ncol ) { x<-matrix() length(x) <- nrow*ncol dim(x) <- c(nrow,ncol) x } system.time( x <- matrix(nrow=10000, ncol=9999) ) user system elapsed 1.989 0.136 2.127 system.time( y <- create.matrix( 10000, 9999 ) ) user system elapsed 0.192 0.141 0.332 identical(x,y) [1] TRUE </code></pre> <p>I appologize to those who commented thinking that the user-defined function was slower, since what is posted in the answer in the above link is inconsistent. I was looking at the user time, which is about 8 times faster in the above link, and on my system about 10 times faster for the user-defined vs built-in. </p> <p>In response to Joshua's request for session info:</p> <pre><code>> sessionInfo() R version 2.12.1 (2010-12-16) Platform: x86_64-unknown-linux-gnu (64-bit) locale: [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C [3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 [5] LC_MONETARY=C LC_MESSAGES=en_US.UTF-8 [7] LC_PAPER=en_US.UTF-8 LC_NAME=C [9] LC_ADDRESS=C LC_TELEPHONE=C [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C attached base packages: [1] stats graphics grDevices utils datasets methods base loaded via a namespace (and not attached): [1] tools_2.12.1 </code></pre> <p>Also, I tried running Simon's three examples, and the third example that Simon gives as the fastest, turns out for me the slowest:</p> <pre><code>> system.time(matrix(NA, nrow=10000, ncol=9999)) user system elapsed 2.011 0.159 2.171 > system.time({x=NA; length(x)=99990000; dim(x)=c(10000,9999); x}) user system elapsed 0.194 0.137 0.330 > system.time(matrix(logical(0), nrow=10000, ncol=9999)) user system elapsed 4.180 0.200 4.385 </code></pre> <p>I still think however that Simon may be on the right track with the idea that <code>matrix()</code> initially allocates a 1x1 matrix and then copies it. Anyone know of any good documentation on R internals? Thanks.</p>

Querying!

Guidance

An individual column

Larger individual text columns get their own page to allow for proper reading.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload