[ This might not be the right mailing list to ask, but I want to start
somewhere - google-ing ifunc and pie didn't bring me any answers ]
I have made a home-build glibc-2.17 (on a core-avx system). It works
great - linking against it (instead of using the current
Debian-Testing's eglibc-2.13) brought the wall-clock time of my weather
forecasting job down from 3:35 hours to 2:45 (mostly due to a more
efficient implementation of powf, expf and logf).
However, if I try to link another weather model against that libm.a, I
get an error saying that you cannot link an executable with 'ifunc'
functions like 'cos'; however, the following "just works":
double precision x
read*,x
print*,exp(x)
print*,log(x)
print*,sin(x)
print*,cos(x)
print*,tan(x)
print*,asin(x)
print*,acos(x)
print*,atan(x)
end
and also if I replace "double precision" with "real".
So perhaps it's the special way the weather model's executable is linked:
gfortran -g -Wl,-M -Wl,-export-dynamic master.o C_code.o _odb_glue.o
-Wl,--start-group
-L/scratch/harmonie/hm_home/trunk/lib/linux.gfortran.mpi/src
-lxrd -lxla -ltfl -ltal -lsur -larp -lald -lbla -lmpa -lmse -lsurfex
-lbip -lsat -luti -lodb -lodbport -lCCMA -lECMA -Wl,--end-group dummies.o \
-L/scratch/harmonie/hm_home/trunk/lib/linux.gfortran.mpi/src
-lodbdummy -L/scratch/harmonie/hm_home/trunk/lib/linux.gfortran.mpi/src
-lrgb -lgribex -lbufr -L/home/harmonie/grib_api -lgrib_api_f90
-lgrib_api -lnetcdff -lnetcdf -llapack -lblas -lmpi_f90 -lmpi_f77 -lmpi
-lpthread -o
/scratch/harmonie/hm_home/trunk/lib/linux.gfortran.mpi/src/MASTERODB >
/scratch/harmonie/hm_home/trunk/lib/linux.gfortran.mpi/src/load.map
The error occurs if I put my glibc-2.17's libm.a between "-lpthread" and
"-o".
Anyone have a suggestion (or a mailing list better suited for this
question) ?
Thanks in advance,
--
Toon Moene - e-mail: toon@xxxxxxxxx - phone: +31 346 214290
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
At home: http://moene.org/~toon/; weather: http://moene.org/~hirlam/
Progress of GNU Fortran: http://gcc.gnu.org/wiki/GFortran#news
