R devel - Aug 2008 - config.log told me to report these (PR#12035)

Hello r-bugs,

I am running openSuse 11.0, with my new Intel 8500 (Penryn) C2D, and I want to
use icc and ifort.
I was able to get a working Goto BLAS, following Dr. Goto's instructions.
That seems to work OK with gcc and gfortran.
I get about 10% faster computations by using -O3 rather than -O2.
I get faster results with Goto BLAS rather than the R BLAS.
Now I want to make icc and ifort work...

So I tried this:

./configure CC=icc CFLAGS='-O3 -mp'
CPPFLAGS=-I/opt/intel/cce/10.1.015/include/ F77=ifort CXX=icc CXXFLAGS='-O3
-mp' FFLAGS='-C90 -w90 -w95 -mp' CPICFLAGS=-shared
CXXPICFLAGS=-shared FPICFLAGS=-shared SHLIB_LDFLAGS=-shared
SHLIB_CXXLDFLAGS=-shared --with-blas="-L/usr/lib64/GotoBLAS
-lgoto_penrynp-r1.26 -lpthread"

I get six purported "bugs" in the config.log file (attached), but I
hope there is something I've overlooked in the past week of agony.
I have had similar log files with about 100 different combinations of options.
I hope that you can help.

Thank you so much.

Tom Richards

This looks like a thoroughly screwed-up compilation system, not an R 
problem.  E.g.

/usr/include/limits.h(125): error #5: could not open source file 
"limits.h"

configure:5465: icc -E -traditional-cpp -I/opt/intel/cce/10.1.015/include/ 
conftest.c
<command-line>:11: warning: "__GNUC_PATCHLEVEL__" redefined
<built-in>:3: warning: this is the location of the previous definition
In file included from /usr/include/features.h:331,
                  from /usr/include/assert.h:38,
                  from conftest.c:20:
/usr/include/sys/cdefs.h:32: error: #error "You need a ISO C conforming 
compiler to use the glibc headers"

/usr/include/limits.h(125): catastrophic error: could not open source file 
"limits.h"

conftest.c(74): catastrophic error: could not open source file "dl.h"

conftest.c(77): catastrophic error: could not open source file 
"floatingpoint.h"

./conftest: error while loading shared libraries: libimf.so: cannot open 
shared object file: No such file or directory


On Thu, 7 Aug 2008, RichardsTJ2 at UPMC.EDU wrote:
> This is a multi-part message in MIME format.
>
> ------_=_NextPart_001_01C8EAD7.27542985
> Content-Type: text/plain;
> 	charset="iso-8859-1"
> Content-Transfer-Encoding: quoted-printable
>
> Hello r-bugs,
>
> I am running openSuse 11.0, with my new Intel 8500 (Penryn) C2D, and I >
want to use icc and ifort.
> I was able to get a working Goto BLAS, following Dr. Goto's >
instructions.
> That seems to work OK with gcc and gfortran.
> I get about 10% faster computations by using -O3 rather than -O2.
> I get faster results with Goto BLAS rather than the R BLAS.
> Now I want to make icc and ifort work...
>
> So I tried this:
>
> ./configure CC=3Dicc CFLAGS=3D'-O3 -mp' >
CPPFLAGS=3D-I/opt/intel/cce/10.1.015/include/ F77=3Difort CXX=3Dicc >
CXXFLAGS=3D'-O3 -mp' FFLAGS=3D'-C90 -w90 -w95 -mp'
CPICFLAGS=3D-shared > CXXPICFLAGS=3D-shared FPICFLAGS=3D-shared
SHLIB_LDFLAGS=3D-shared > SHLIB_CXXLDFLAGS=3D-shared
--with-blas=3D"-L/usr/lib64/GotoBLAS > -lgoto_penrynp-r1.26
-lpthread"
>
> I get six purported "bugs" in the config.log file (attached), but
I hope > there is something I've overlooked in the past week of agony.
> I have had similar log files with about 100 different combinations of >
options.
> I hope that you can help.
>
> Thank you so much.
>
> Tom Richards
>
>
> ------_=_NextPart_001_01C8EAD7.27542985
> Content-Type: text/x-log;
> 	name="config.log"
> Content-Transfer-Encoding: base64
> Content-Description: config.log
> Content-Disposition: attachment;
> 	filename="config.log"
[needed unpacking]