sorry, I thought I have answered.
1- thanks for the compiler warning, we have removed it recently. Not a big
deal for your concern: it's just a warning, it concerns a testing file
2- yep we have some failures in the testing. We know. Thanks fro pointing
this out. Once more this should not concern you. Say that the matrices
where our algos fails are really tough and you should not see them in
your daily life.
So enjoy LAPACK and congarts for the successfull install.
On Tue, 31 Oct 2006, Steve Young wrote:
Thank you again Julien,
I went ahead and installed everything: single, complex, double and double
complex. Everything went well with the following exceptions:
One compiler warning:
( cd BLAS/TESTING; make -f Makeblat3 )
make: Entering directory `/m2/pca/LAPACK/BLAS/TESTING'
g77 -funroll-all-loops -fno-f2c -O3 -c sblat3.f
g77 sblat3.o \
../../blas_LINUX.a -o ../xblat3s
g77 -funroll-all-loops -fno-f2c -O3 -c dblat3.f
g77 dblat3.o \
../../blas_LINUX.a -o ../xblat3d
g77 -funroll-all-loops -fno-f2c -O3 -c cblat3.f
g77 cblat3.o \
../../blas_LINUX.a -o ../xblat3c
g77 -funroll-all-loops -fno-f2c -O3 -c zblat3.f
zblat3.f: In subroutine `zchke':
INTRINSIC DCMPLX, MAX, DBLE
ALPHA = DCMPLX( ONE, -ONE )
Intrinsic `dcmplx' referenced implicitly at (2), explicitly at (1) [info -f
g77 M EXPIMP]
g77 zblat3.o \
../../blas_LINUX.a -o ../xblat3z
make: Leaving directory `/m2/pca/LAPACK/BLAS/TESTING'
Several testing failures (files listed below are attached):
cgd.out: CGV drivers: 64 out of 1092 tests failed to pass the threshold
csep.out: CST: 1 out of 4662 tests failed to pass the threshold
dgd.out: DGV drivers: 14 out of 1092 tests failed to pass the threshold
dgd.out: DXV drivers: 200 out of 5000 tests failed to pass the threshold
sgd.out: SGV drivers: 23 out of 1092 tests failed to pass the threshold
sgd.out: SXV drivers: 37 out of 5000 tests failed to pass the threshold
ssep.out: SST drivers: 1 out of 14256 tests failed to pass the
zgd.out: ZGV drivers: 67 out of 1092 tests failed to pass the threshold
zgd.out: ZXV drivers: 24 out of 5000 tests failed to pass the threshold
Is there anything here that I need to be concerned about? The comments in
the code I hope to run that uses LAPACK says:
! Utilizes SSYEV routine from LAPACK to compute eigenvalues
! and eigenvectors of a real symmetric matrix.
On 10/30/06 4:10 PM, "Julien Langou" <langou@Domain.Removed> wrote:
Well, ok it might be our fault here. I have just looked at the output of
the file make.out.tail
(by the way, perfect, all we need in general to help)
So do not try to play with single/double/ etc. unless you really need it.
Here your blas_LINUX.a is missing the double precision routine and when
you are trying to make the executable xblat1d then this does not work.
So just type
make blaslib blas_testing
to do what you are doing.
On Mon, 30 Oct 2006, Steve Young wrote:
Thank you Julien,
I had started to try to compile with pgf77 and then figured I would try g77
instead since I knew that LAPACK has been successfully compiled with g77.
Then I forgot to completely clean out the pgf77 object files.
Now on to another problem - this time from a clean install. I actually
removed all of the files and untarred lapack.tgz again just to be sure.
This time the process made it as far as compiling the BLAS testing routines.
Once again, the linker doesn't seem to be able to find some of the
I'm including the output resulting from the make command in make.out. I'm
also including sgd.out because it shows some failures in the LAPACK testing.
Perhaps the two problems are related?
Thank you for your help.
On 10/30/06 10:10 AM, "Julien Langou" <langou@Domain.Removed> wrote:
you might have messed up with your compilers. You certainly have at least
two Fortran compilers (g77 and pgf77). You certainly have compiled one
part of the object using the Portland compiler and what you are trying to
do now is a link with g77. If this is what you have done, then the error
message is to be expected. g77 can not find the runtime library of pg by
So the best is to start over with a nice
so that all the .o are generated by the same compiler and the linker is
the same as well.
On Thu, 26 Oct 2006, Steve Young wrote:
I am attempting to install LAPACK on a Dual Xeon machine running Linux. I
used the provided make.inc file for Linux and modified the makefile to
install single precision for now. I used the command "make" to do a
complete installation and testing. The process seems to go OK up through
the testing section of the Makefile. The problem seems to occur with the
linking step in LAPACK/TESTING/LIN. I get a whole long list of "undefined
reference" errors and the executable xlintsts (which I believe is the
single precision executable) is not created. It seems that xlintsts
on the 3 libraries that are created first: tmglib_LINUX.a lapack_LINUX.a
and blas_LINUX.a From what I can see, those libraries are created without
any errors. I am attaching copies of the Makefile, make.inc and the
of the make process.
I would appreciate any help you can give or a referral to someone else who
can help. Thank you.