Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

MKL bug in function dgesvd?

jeff33
Principiante
919 Vistas
A call of dgesvd with a random matrix results in a segmentation fault. I work with MKL 8.0.2 on a Pention IV with Linux. A call of dgesvd with the same matrix using LAPACK leads to no error.

My program reads:

program ftest

integer M
integer N
integer Q
integer LWORK

parameter (M=300)
parameter (N=400)
parameter (Q=300)
parameter (LWORK=100000)

double precision A(M,N)

double precision U(M,M)
double precision VT(N,N)
double precision S(Q)

double precision WORK(LWORK)

integer INFO

do i=1,M
do j=1,N
A(i,j) = rand(0)
end do
end do

call dgesvd('A','A',M,N,A,M,S,U,M,VT,N,WORK,LWORK,INFO)

stop

end

I compile it with the command:

ifort ftest.f -lmkl_lapack -lmkl -lguide -o ftest

Is this a MKL-bug or a wrong usage of the function dgesvd?

Thanks for your interest!
0 kudos
3 Respuestas
hajek
Principiante
919 Vistas
Works fine with MKL 7.2 (Cluster)
max57
Principiante
919 Vistas
Segmentation fault also with MKL 8.0 on an Intel Xeon 64bit. According to the debugger, the error is in function mkl_blas_dgemm_copybt() in libmkl_p4n.so.
sragner
Principiante
919 Vistas

http://softwareforums.intel.com/en-us/forums//topic/51912

I have a similar problem 8.0.1 with matrix sized 350x320. Couldn't find any solution.

Laszlo

Responder