Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Romain Francois
Le 02/12/2013 21:43, Douglas Bates a écrit : On Mon, Dec 2, 2013 at 2:20 PM, Dirk Eddelbuettel mailto:e...@debian.org>> wrote: On 2 December 2013 at 13:52, Douglas Bates wrote: | The important thing is to use the -g0 flag. Even though RcppEigen is a Right. And for the op

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Douglas Bates
On Mon, Dec 2, 2013 at 2:20 PM, Dirk Eddelbuettel wrote: > > On 2 December 2013 at 13:52, Douglas Bates wrote: > | The important thing is to use the -g0 flag. Even though RcppEigen is a > > Right. > > And for the opencpu server deployment, you may want to edit the -g out > /etc/R/Makeconf as we

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Dirk Eddelbuettel
On 2 December 2013 at 13:52, Douglas Bates wrote: | The important thing is to use the -g0 flag.  Even though RcppEigen is  a Right. And for the opencpu server deployment, you may want to edit the -g out /etc/R/Makeconf as well. As I recall, there was an r-devel thread in which the desire to ove

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Douglas Bates
The important thing is to use the -g0 flag. Even though RcppEigen is a header-only package we include an example R function fastlm. If you leave the symbols in the DLL file you get a massive library size whereas stripping the symbols provides you with a much smaller file size. And because packag

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Dirk Eddelbuettel
On 2 December 2013 at 11:21, Kevin Ushey wrote: | Try removing -pipe: see e.g. | http://gcc.gnu.org/onlinedocs/gcc/Overall-Options.html#Overall-Options, | at -pipe. Very good suggestion. That is "my fault" too via the default Debian/Ubuntu setup: edd@max:~$ grep pipe /etc/R/Makeconf # configur

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Dirk Eddelbuettel
On 2 December 2013 at 11:17, Jeroen Ooms wrote: | I noticed that RcppEigen fails to install on my servers because it requires | more than 1gb of memory to compile. Is this expected? Are there any flags or Yes. It also falls over on R-Forge. | options I could set to trade of some memory for cpu?

Re: [Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Kevin Ushey
Try removing -pipe: see e.g. http://gcc.gnu.org/onlinedocs/gcc/Overall-Options.html#Overall-Options, at -pipe. On Mon, Dec 2, 2013 at 11:17 AM, Jeroen Ooms wrote: > I noticed that RcppEigen fails to install on my servers because it requires > more than 1gb of memory to compile. Is this expected?

[Rcpp-devel] RcppEigen needs 1GB memory to compile

2013-12-02 Thread Jeroen Ooms
I noticed that RcppEigen fails to install on my servers because it requires more than 1gb of memory to compile. Is this expected? Are there any flags or options I could set to trade of some memory for cpu? A short simulation: > library(RAppArmor) > rlimit_as(1e9) $hardlim [1] 1e+09 $softlim [1] 1

Re: [Rcpp-devel] Building RcppBDT fails on Win 7 64-bit

2013-12-02 Thread Dino Veritas
Dirk, Thank you for prompt reply. I did try the -std=c++11 flag but it was not valid for my g++ version. Of course I never bothered with -std=c++0x which built the SVN version just fine! With that said, I will follow your advice and stick with the CRAN version for now; it built for me without any

Re: [Rcpp-devel] Building RcppBDT fails on Win 7 64-bit

2013-12-02 Thread Dirk Eddelbuettel
Dino, Also, if you want another example of 'long long int': RcppCNPy needs to enable -std=c++11 to the 'long long int' beheaviour to support int64 as needed in the NumPy headers for ints. If -std=c++11 is not enable, RcppCNPy does not support integers, otherwise it does. You could building tha

Re: [Rcpp-devel] Building RcppBDT fails on Win 7 64-bit

2013-12-02 Thread Dirk Eddelbuettel
Hi Dino, On 2 December 2013 at 12:13, Dino Veritas wrote: | Hello Rcpp-devel community, | | I am trying to build the RcppBDT package on Windows 7 64-bit using the >From SVN, or from CRAN? I left SVN in a less-than-ideal state which a set of not-really-finished changes at HEAD. If you use the

Re: [Rcpp-devel] Rcpp / g++ ignoring TMPDIR

2013-12-02 Thread Dirk Eddelbuettel
On 2 December 2013 at 09:19, Jeroen Ooms wrote: | On Mon, Dec 2, 2013 at 4:25 AM, Dirk Eddelbuettel wrote: | | As I said when we hashed this out some more in private mail, I'd try | setting | TMPDIR in /etc/profile or alike. Or just try to have a working /tmp | directory. | S

Re: [Rcpp-devel] Rcpp / g++ ignoring TMPDIR

2013-12-02 Thread Jeroen Ooms
On Mon, Dec 2, 2013 at 4:25 AM, Dirk Eddelbuettel wrote: > As I said when we hashed this out some more in private mail, I'd try > setting > TMPDIR in /etc/profile or alike. Or just try to have a working /tmp > directory. > Surely you are not the only AppArmor user with that issue? > Well it's n

[Rcpp-devel] Building RcppBDT fails on Win 7 64-bit

2013-12-02 Thread Dino Veritas
Hello Rcpp-devel community, I am trying to build the RcppBDT package on Windows 7 64-bit using the latest version of Rtools. This version of Rtools comes with gcc-4.6.3. The process fails when building the 64-bit library for RcppBDT, complaining about an invalid conversion from 'SEXP' to 'long lon

Re: [Rcpp-devel] Rcpp / g++ ignoring TMPDIR

2013-12-02 Thread Dirk Eddelbuettel
On 1 December 2013 at 22:38, Jeroen Ooms wrote: | On Sun, Dec 1, 2013 at 6:33 PM, Dirk Eddelbuettel wrote: | | | I don't think this is us, and I don't think something changed. See below. | | |  I'll ask on r-devel if anyone is aware of install.packages unsetting TMPDIR | when calling g++.