Jan,
Can you elaborate on how you invoke these settings for a R-devel
build? Do you pass these along in a Terminal session, assuming BASH
shell, via a series of export commands? I'd like to build this myself
using llvm and see how it turns out.
Certainly, speeding up compile times is a nice benefit. Have you
tested any R commands to compare performance of this build versus
gcc-4.2? I wonder if the gains are passed along to application
performance as well.
Thanks,
John
On Oct 26, 2009, at 11:18 PM, Jan de Leeuw wrote:
The following setup on SL compiles R-devel and passes check.
C compiler: /Developer/usr/bin/llvm-gcc -std=gnu99 -
m64 -O4 -Wall -fopenmp -mtune=native
Fortran 77 compiler: /usr/local/bin/gfortran-4.5 -m64 -O4 -
Wall -fopenmp -mtune=native
C++ compiler: /Developer/usr/bin/llvm-g++ -m64 -O4 -
Wall -fopenmp -mtune=native
Fortran 90/95 compiler: /usr/local/bin/gfortran-4.5 -m64 -O4 -
Wall -fopenmp -mtune=native
Obj-C compiler: /Developer/usr/bin/llvm-gcc -m64 -O4 -Wall -
fopenmp -mtune=native
This compiles at breakneck speed (compared to gcc-4.2 or gcc-4.5).
In fact, with recommended packages,
for all of R-devel:
time make -j8
real 2m33.028s
user 3m59.076s
sys 0m33.248s
NB: I still have problems with CC and OBJC equal to clang (from Apple)
or to clang (from the llvm svn), even without the -fopenmp flag.
===============================================================
Jan de Leeuw, 11667 Steinhoff Rd, Frazier Park, CA 93225
home 661-245-1725 mobile 661-231-5416 work 310-825-9550
.mac: jdeleeuw +++ aim: deleeuwjan +++ skype: j_deleeuw
===============================================================
We are always the same age inside.
Gertrude Stein
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
