Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- Intel Community
- Software Development SDKs and Libraries
- Intel® oneAPI Math Kernel Library & Intel® Math Kernel Library
- Why a SYMMETRIC type option for the Extended Eigensolver?

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Mute
- Printer Friendly Page

Highlighted
##

Vishnu

Novice

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-26-2020
10:17 AM

156 Views

Why a SYMMETRIC type option for the Extended Eigensolver?

The extended eigensolver routines: `mkl_sparse_?_ev` allow for specifying the type of the input matrix as SYMMETRIC or GENERAL. But since the routine only accepts symmetric matrices any way, what, if any difference does either option make?

(The documentation says that the matrix 'A' is symmetric)

I tried both and get roughly the same time on both.

Is it that I can use an input sparse matrix with half the amount of data if I set it to the SYMMETRIC type?

4 Replies

Highlighted
##

Gennady_F_Intel

Moderator

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-26-2020
10:24 PM

147 Views

yes, you can use an input sparse matrix with half the amount of data by setting

sparse_matrix_type_t `type == SPARSE_MATRIX_TYPE_SYMMETRIC`

Highlighted
##

Vishnu

Novice

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-26-2020
10:47 PM

142 Views

Okay, thanks for that! It would be nice if this information were present in the documentation, so that users know when to use the option.

So then does this come with a computational cost? I ask because my current priority is speed and not space. Also because examining some back-traces seem to show the usage of some 'transpose' and 'symmetric-to-general' conversion routines when using the SYMMETRIC option, and I'm assuming that is not cheap (especially for large matrices).

Highlighted
##

IntelSupport

Community Manager

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-29-2020
01:46 PM

122 Views

Hello Vishnu,

Thank you for posting your question on this Intel® Community.

To better assist you, please submit your question on the Intel® Developer Zone.

https://software.intel.com/content/www/us/en/develop/home.html

You can also submit tickets through the customer portal at http://www.intel.com/supporttickets for support.

Wanner G.

Intel Customer Support Technician

Highlighted
##

Vishnu

Novice

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-29-2020
10:55 PM

109 Views

> To better assist you, please submit your question on the Intel® Developer Zone.

> https://software.intel.com/content/www/us/en/develop/home.html

I am not sure where exactly you mean. That link does not take me to anywhere I can make a post.

For more complete information about compiler optimizations, see our Optimization Notice.