Multiple calls to DGEEV

[Cansey50]

I have not used oneMKL before, and only recently begun experimenting with it. Ultimately I would like to create DLLs for some of the functions including the ones for the eigenvalue/eigenvector calculations. 

I took the dgeev.c example from \examples_core_c\c\lapack\source\ zipped folder in my oneAPI installation folder, and compiled it in a Visual Studio 2019 console project. It runs ok and gives the exact results mentioned at the top of the dgeev.c file.

However, if I call dgeev again the results are different! I have attached the source code in which dgeev is called 3 times and the output of the program. I was expecting to get exactly the same results every single time the dgeev is called as the matrix a has not changed.

Can you please help me understand why the results are different, and what I am doing wrong?

Thanks beforehand.

In case it is needed:

My Processor: Intel(R) Core(TM) i5-2500 CPU

My OS: 64-bit operating system, x64-based processor

[mecej4 · Accepted solution]

Please read the conventions of the Lapack/BLAS routines that are called in your code. For DGEEV, the documentation says:

[in,out]

A

A is DOUBLE PRECISION array, dimension (LDA,N) On entry, the N-by-N matrix A. On exit, A has been overwritten.

If you wish to call DGEEV after the first call, you have to restore the original values in A before the second call (and subsequent ones, if any).

Most of the Lapack and BLAS routines were designed and written at a time when "main memory" was a precious and scarce resource. The computers of the time, such as the CDC6600, had memory measured in kilowords (132 kWords, 60 bits each).

[Cansey50]

Thank you mercej4! I admit I have to read the documentation carefully. 

[ShanmukhS_Intel]

Hi Khosro,

Thanks for posting in Intel Communities.

It’s great to know that the issue has been resolved, in case you run into any other issues please feel free to create a new thread.

Best Regards,

Shanmukh.SS