similar to: Speeding up sum and prod

I'm observing that base::sum(x, na.rm=FALSE) for typeof(x) == "double" is much more time consuming when there are missing values versus when there are not. I'm observing this on both Window and Linux, but it's quite surprising to me. Currently, my main suspect is settings in on how R was built. The second suspect is my brain. I hope that someone can clarify the below

This is a great example how you cannot figure it out after spending two hours troubleshooting, but a few minutes after you post to R-devel, it's just jumps to you (is there a word for this other than "impatient"?); Let me answer my own question. The discrepancy between my sum2() code and the internal code for base::sum() is that the latter uses LDOUBLE = long double (on some system

Full_Name: Nick Efthymiou Version: R1.5.0 and above OS: Red Hat Linux Submission from: (NULL) (162.93.14.73) With the introduction of the functions rowSums(), colSums(), rowMeans() and colMeans() in R1.5.0, function "SEXP do_colsum(SEXP call, SEXP op, SEXP args, SEXP rho)" was added to perform the fast summations. We have an excellent opportunity to improve the accuracy by

I've continued to work on speeding up R, and now have a collection of fourteen patches, some of which speed up particular functions, and some of which reduce general interpretive overhead. The total speed improvement from these patches is substantial. It varies a lot from one R program to the next, of course, and probably from one machine to the next, but speedups of 25% can be expected in

Hi, I'm just following your messages the overhead that the code for dealing with possible NA/NaN values brings. When I was setting up part of the matrixStats package, I've also though about this. I was thinking of having an additional logical argument 'hasNA'/'has.na' where you as a user can specify whether there is NA/NaN:s or not, allowing the code to avoid it or not.

Dear R-Developers, I think I found two bugs in max.col(). Ties between zeros are not broken, which might affect simulations. -Inf and Zero can be treated the same, which can give completely wrong results, e.g. when the second max is sought by replacing all maxs by -Inf. To fix max.col I do offer the C-code behind my function rowMax(), which also handles NAs and seems to be faster. However,

Hi Tomas, Can you share the full code for your benchmark, compiler options, and performance results so that I can try to reproduce them? There are a lot of variables that can affect the results. Private email is fine if it is too much for the mailing list. I am measuring on Knight's Landing (KNL) that was released in November. KNL is not a co-processor so no offload is necessary. R executes

Hi, Please consider the following : > x = as.integer(2^30-1) [1] 1073741823 > sum(c(rep(x, 10000000), rep(-x,9999999))) [1] 1073741824 Tested on 2.15.2 and a recent R-devel (r62132). I'm wondering if s in isum could be LDOUBLE instead of double, like rsum, to fix this edge case? https://svn.r-project.org/R/trunk/src/main/summary.c Thanks, Matthew

Inside of the anyNA() function, it will use the legacy any(is.na()) code if x is an OBJECT(). If x is a vector of POSIXct, it will be an OBJECT(), but it is also TYPEOF(x) == REALSXP. Therefore, it will skip the faster ITERATE_BY_REGION, which is typically 5x faster in my testing. Is the OBJECT() condition really necessary, or could it be moved after the switch() for the individual TYPEOF(x)

I've looked at the code for matrix multiplies in R, and found that it can speeded up quite a bit, for multiplies that are really vector dot products, and for other multiplies in which the result matrix is small. Here's my test program: u <- seq(0,1,length=1000) v <- seq(0,2,length=1000) A2 <- matrix(2.1,2,1000) A5 <- matrix(2.1,5,1000) B3 <- matrix(3.2,1000,3) A4 <-

> Do you have R code (including set.seed(.) if relevant) to show on how to generate > the large square matrices you've mentioned in the beginning? So we get to some > reproducible benchmarks? Hi Martin, Here is the program I used. I only generate 2 random numbers and reuse them to make the benchmark run faster. Let me know if there is something I can do to help--alternate

Hi, I was wondering if I could use the matprod function from array.c in my own C routine. I tried the following as a test /* my_matprod.c */ # include <Rinternals.h> /* for REAL, SEXP etc */ # include <R_ext/Applic.h> /* array.c says need for dgemm */ /* following copied from array.c */ static void matprod(double *x, int nrx, int ncx, double *y, int nry, int ncy, double *z)

I am trying to use R_RegisterCFinalizerEx to ensure some c++ object is properly deleted after use. However, I run into some problems when trying to pass in the third argument which is an Rboolean. The line 'R_RegisterCFinalizerEx(p, finalizer, TRUE);' generates the following error when trying to compile using R CMD SHLIB: g++ -I/usr/share/R/include -DNDEBUG -fpic -O2 -pipe

Dear Tomas, Where do I find these files? Do they contain the code for the sum function? What do you mean exactly with your point on long doubles? Where can I find documentation on this? Cheers, Rampal On Mon, Feb 18, 2019, 15:38 Tomas Kalibera <tomas.kalibera at gmail.com wrote: > See do_summary() in summary.c, rsum() for doubles. R uses long double > type as accumulator on systems

On Thu, 30 Jan 2025 13:07:31 -0800 Michael Chirico <michaelchirico4 at gmail.com> wrote: > There are at least dozens of other cases on CRAN [2],[3]. Some of these involve casting an int to Rboolean. Best case, the int is compared against NA_LOGICAL beforehand, avoiding any mistake (there's at least one like that). Worst case, NA_LOGICAL is not considered before the cast, so NA will

Hello list, I've been working on a package of my own. It works fine on the 32bit linux machines that I've tested it on. Someone using 64bit ubuntu wanted to try it, and surprising segfaults happened. My test code results in no segfault, errors, or leaks from my calls when run under valgrind (I recompiled R with the level 2 valgrind instruments). R and packages are compiled from source,

The good news is that without a C23-enabled compiler, the problem will only happen to source files that #include <stdbool.h>. The bad news is that such a source file will technically disagree with the rest of R about the type of Rboolean, including the prototypes of the API functions that accept Rboolean: #include <stdbool.h> #include <Rinternals.h> typedef void

Hello, The current xfig device lacks the functionality to set the textspecial flag and use the defaultfont in xfig. This is necessary when you want to export to xfig and use interpreted text (e.g., $ \frac{1}{e}$ gets interpreted by latex). The attached patch adds this functionality. Why would you like to do this? - Use math in labels (e.g., name your variables $r_{xy}$, do a

In what way is this a bug in R? It looks like a bug in the package, and as Defn.h is not part of R's API any packge using it is `at risk' (and cannot be installed in a binary-only installation, or even an installed version of R). In particular, Defn,.h depends on config.h, and it seems you installed a binary version of R and used separate sources. I would suggest building R from

On 08/05/2018 2:58 PM, Duncan Murdoch wrote: > On 08/05/2018 1:48 PM, Steven Nydick wrote: >> Reproducible example: >> >> x <- list(list(list(), list())) >> unlist(x) >> >> *> Error in as.character.factor(x) : malformed factor* > > The error comes from the line > > structure(res, levels = lv, names = nm, class = "factor") >