Bibliothèque Intel® Math Kernel Library

C examples in MKL's ScaLAPACK

I am aware of these examples, but I do not know any Fortran, thus I can not understand much. So finding at least an example with MKL and ScaLAPACK in C would be critical for me. I know there is a C interface. For example p?potrf will be the function I am going to use, for performing a Cholesky factorization.

If someone would reply negatively, that would be also OK, since it would stop me for searching..

mkl_intel_lp64 vs mkl_gf_lp64 and MKL advisor

I wonder if the Advisor get it wrong when suggests:

 -Wl,--no-as-needed -L${MKLROOT}/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_gnu_thread -ldl -lpthread -lm

this bit: -lmkl_intel_lp64

when GNU's bits are chosen when applicable instead of Intel's

reason I'm asking is because I get undefined references to GOMP.

many thanks.

question about number of eigenvalues found w/ dfeast_scsrev

hi guys,

 

just curious about the 'm' integer when using feast.

currently i know what the 'bottom eigenvector' will be (it's the unit vector with eigenvalue 0), so i specify the last eigenvector to ensure the correct solution is found.

in the end, the eigensolver finds exactly one less than the total number of eigenvalues being requested. i.e. if i request 151 eigenvectors, and i've specified the bottom one, it says 150 eigenvectors are found.

PARDISO - Divide by zero exception on second call

The following simple test produces a FPE in PARDISO.

The exception occurs when calling PARDISO for a second time, with a smaller matrix than the first; is this allowed? 

Note that to get the exception the Fortran project settings should be set to abort when dividing by zero (/fpe:0 I think)

Is this a bug in Pardiso, or am I using it incorrectly?

Many thanks!

DFT complex conjugate even data structure

I'm trying to do a comparison between MKLs DFT for a complex --> complex forward DFT vs a real --> real forward DFT. In both cases my starting array is entirely real so the real --> real transform should be able to take advantage of complex conjugate even symmetries. The complex transform is giving me the expected results. However, for the real transform the output seems completely jumbled. It's possible I'm not applying the strides correctly but the code I'm using is practically copied directly from the MKL examples so that seems unlikely.

how to use geqrf to generate unitary matrix Q (matrix size: m*n)

I know ?geqrf to calculate m-by-n A = QR and then ungqr to generate Q. but the Q is m-by-m square matrix, what i want is Q m-by-n.

I really don't know how to deal with this, I found p?geqrf and p?ungqr would generate Q with m-by-n, but I think it's a parallel ?geqrf, it might not help me. could anyone help me, please? thanks very much..

S’abonner à Bibliothèque Intel® Math Kernel Library