Unexpected results of LAPACKE_dspevd

Open discussion regarding features, bugs, issues, vendors, etc.

Unexpected results of LAPACKE_dspevd

Postby pysmile » Wed Aug 31, 2016 5:23 am


I am using LAPACKE_dspevd subroutine, but I got some unexpected results with it. So I wrote a test to compare it with LAPACK fortran code. The initial packed matrices are same for both and parameters are set like this:
matrix_layout = LAPACK_ROW_MAJOR, uplo = 'U' for LAPACKE_dspevd.
column major, uplo = 'L' for dspevd.

With this setting, they represent same matrices and the expected output of eigenvalues should be identical. However, I got different results, listed below:
LAPACKE_dspevd Example Program Results
-4.627763 -0.955899 6.451161 19.132501
dspevd Example Program Results
-7.947764 1.256506 4.176978 19.514279

I used octave and verified that the second result is correct. Could you please help on this? Thanks. PS: The attachment is source code of my test.
(1.75 KiB) Downloaded 378 times
Posts: 2
Joined: Thu Jun 23, 2016 12:44 am

Return to User Discussion

Who is online

Users browsing this forum: Google [Bot] and 4 guests