Sparse Hermitian matrix and vector multiplication

[flygaoonly]

In MKL, are there functions or subroutines that can performcomplex sparse Hermitian matrix and vector multiplication? I only saw the one for sparse symmetrical matrix.

[Sergey_K_Intel1]

MKL has got sparse matrix-vector multiply several routines for Hermitian matrices and one of them is

call mkl_zcsrmv(transa, m, k, alpha, matdescra, val, indx, pntrb, pntre, x, beta, y)

In the case of Hermitian matrix, the first element of the parameter

matdescra

CHARACTER. Array of six elements, specifies properties of the matrix used for operation.

must be H or h.

All the best
Sergey

[flygaoonly]

I got it. Thank you very much.

[flygaoonly]

Another thing. If I want to compile my program (bdg.f90 which uses mkl_ccoomv) on a Linux system, can I use the following command:

ifort bdg.f90 -L /opt/intel/mkl/10.0.1.014/lib/em64t -I /opt/intel/mkl/10.0.1.014/include -lmkl_intel -lmkl_intel_thread -lmkl_core -liomp5 -lpthread

If not, what command should I use? Thank you.

[flygaoonly]

After reading the User's Guide, I use the following command to compile my program bdg.f90, in which I called mkl_ccoomv.

ifort bdg.f90 -L/opt/intel/mkl/10.0.1.014/lib/em64t -I/opt/intel/mkl/10.0.1.014/include -Wl,--start-group /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_em64t.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_blas95.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_lp64.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_thread.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_core.a -Wl,--end-group -lmkl -lguide -lpthread

But I still got the following error:

undefined reference to 'mkl_ccoomv'

What's wrong with it, can you give me some advice? Thanks.

[barragan_villanueva_]

Hi,

Please use Intel MKL Link Line Advisor
E.g., for MKL 10.0 it should be:

-lmkl_blas95_lp64 $(MKLROOT)/lib/em64t/libmkl_solver_lp64.a -Wl,--start-group $(MKLROOT)/lib/em64t/libmkl_intel_lp64.a $(MKLROOT)/lib/em64t/libmkl_intel_thread.a $(MKLROOT)/lib/em64t/libmkl_core.a -Wl,--end-group -openmp -lpthread

[flygaoonly]

Thank you for your advice. However, as you suggested, I use the following command:

ifort bdg.f90 -lmkl_blas95_lp64 /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_solver_lp64.a -Wl,--start-group /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_lp64.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_thread.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_core.a -Wl,--end-group -openmp -lpthread

There is an error:

ld: cannot find -lmkl_blas95_lp64

The same thing happans if I use

ifort bdg.f90 -I/opt/intel/include/em64t/lp64 -lmkl_blas95_lp64 /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_solver_lp64.a -Wl,--start-group /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_lp64.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_thread.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_core.a -Wl,--end-group -openmp -lpthread

So what's the problem?

[barragan_villanueva_]

Sorry,

libmkl_blas95_lp64.a was introduced in MKL 10.2 only but you use MKL 10.0. MKL Link Advisor is to be fixed.
So please try to link your program without it.

[flygaoonly]

Without it, I then use

ifort bdg.f90 /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_solver_lp64.a -Wl,--start-group /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_lp64.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_intel_thread.a /opt/intel/mkl/10.0.1.014/lib/em64t/libmkl_core.a -Wl,--end-group -openmp -lpthread

but still got "undefined reference to mkl_ccoomv".

[barragan_villanueva_]

Hi,

This function was introduced in MKL 10.1 and therefore is missing in MKL 10.0

[flygaoonly]

So in this case, is there any other sparse matrix vector multiplication subroutine I can use?

[Customer__Intel4]

Hi

In case of using this function with hermitian or symmetric matrix, do we store only the upper / lower triangle or the whole matrix ?

Thanks

[Konstantin_A_Intel]

Quoting tarek.e

Hi

In case of using this function with hermitian or symmetric matrix, do we store only the upper / lower triangle or the whole matrix ?

Thanks

Yes, sure.