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

Showing results for

- Intel Community
- Software
- Software Development SDKs and Libraries
- Intel® oneAPI Math Kernel Library
- Preconditioned CG(RCI) initial residual computation

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

nityar

Beginner

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

05-07-2020
09:18 AM

75 Views

Preconditioned CG(RCI) initial residual computation

I am using RCI implementation of CG in MKL using dcg* API.

Can anybody explain how the initial residual is computed in preconditioned CG version. In non preconditioned version i can confirm it is r0 = square sum of (AX0 - b), if X0=0, is simply square sum of initial b . For preconditioned, how do we compute the value reported in parameter dpar_[2].

Also, in FGMRES the initial residue is norm2, i.e square root of residue reported in CG, any reason for this inconsistency?

thanks,

-Nitya

Link Copied

2 Replies

nityar

Beginner

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

05-11-2020
07:55 AM

75 Views

Can anybody please help on this?

Kirill_V_Intel

Employee

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

05-16-2020
04:32 PM

75 Views

Hello Nitya,

I think it should be the squared norm of the preconditioned residual: || C^{-1} r0, r0 || where r0 is the initial residual r0 = (Ax - b) and C is the preconditioner. Let me know if it happens not to be so for you (I haven't done a full check).

Best,

Kirill

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

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