Showing results for

- Intel Community
- Software
- Software Development SDKs and Libraries
- Intel® oneAPI Math Kernel Library
- an accuracy problem between MKL and MATLAB

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

shaopeng_z_

Beginner

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

12-18-2012
04:13 AM

81 Views

an accuracy problem between MKL and MATLAB

Dear all,

rencently,I have run a program using Intel MKL sparse format and MATLAB respectively.I find an accuracy problem in my result. Let me briefly summarize what I did.

At first,I use the MKL libary to compute K*X=KX ,as follows:

the full format symmetric matrix K converts to the sparse format CSC (COPPTR,ROWIND,VALUES or ROWIND,VALUES,pointer_B,pointer_E)

program exe101

implicit none

integer::i,j

integer,parameter::DP=8

integer::nRows,nNonzero

integer,allocatable::COPPTR(:),ROWIND(:),pointer_B(:),pointer_E(:)

real(kind=DP),allocatable::VALUES(:)

real(kind=DP),allocatable::X(:),KX(:)

real(kind=DP)::alpha,beta

character(6):: matdescra

open(unit=10,file='nRows.txt')

read(10,*)nRows

close(10,status="keep")

open(unit=10,file='nNonzero.txt')

read(10,*)nNonzero

close(10,status="keep")

allocate(COPPTR(nRows+1))

open(unit=10,file='COPPTR.txt')

read(10,*)COPPTR

close(10,status="keep")

allocate(ROWIND(nNonzero))

open(unit=10,file='ROWIND.txt')

read(10,*)ROWIND

close(10,status="keep")

allocate(VALUES(nNonzero))

open(unit=10,file='VALUES.txt')

read(10,*)VALUES

close(10,status="keep")

allocate(X(nRows))

allocate(KX(nRows))

X=1.0

alpha=1.0

beta=0.0

matdescra='SLNF'

allocate(pointer_B(nRows))

allocate(pointer_E(nRows))

do i=1,nRows

pointer_B(i)=COPPTR(i)

pointer_E(i)=COPPTR(i+1)

end do

call mkl_dcscmv('N', nRows, nRows, alpha, matdescra, VALUES, ROWIND, pointer_B, pointer_E,X, beta, KX)

open(unit=10,file='KX.txt')

write(10,'(1X,1f40.21)')KX

end

output::

KX=

1999999999.999993562698364257812

-0.000006198883056640625

2000000000.000000000000000000000

0.000000000000000000000

0.000000000000000000000

0.000000000000000000000

0.000002384185791015625

0.000001907348632812500

Secondly,I use MATLAB to compute K*X=KX

output::

1999999999.99999

-6.19888305664063e-06

2000000000.00000

0

0

0

2.14576721191406e-06

1.90734863281250e-06

where

input::

K=

5.414213562373090e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 -2.000000000000000e+009 0.000000000000000e+000 -7.071067811865480e+008 -7.071067811865480e+008

0.000000000000000e+000 3.414213562373090e+009 0.000000000000000e+000 -2.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 -7.071067811865480e+008 -7.071067811865480e+008

0.000000000000000e+000 0.000000000000000e+000 4.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 -2.000000000000000e+009 0.000000000000000e+000

0.000000000000000e+000 -2.000000000000000e+009 0.000000000000000e+000 2.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000

-2.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 2.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000

0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 2.000000000000000e+009 0.000000000000000e+000 -2.000000000000000e+009

-7.071067811865480e+008 -7.071067811865480e+008 -2.000000000000000e+009 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 2.707106781186550e+009 7.071067811865480e+008

-7.071067811865480e+008 -7.071067811865480e+008 0.000000000000000e+000 0.000000000000000e+000 0.000000000000000e+000 -2.000000000000000e+009 7.071067811865480e+008 2.707106781186550e+009

X=

1.0

1.0

1.0

1.0

1.0

1.0

1.0

1.0

I want to know why two results are different and how to improve the accuracy of sparse(dense) matrix- vector.

Thank you,

Link Copied

5 Replies

mecej4

Black Belt

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

12-18-2012
05:36 AM

81 Views

jimdempseyatthecove

Black Belt

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

12-18-2012
07:39 AM

81 Views

mecej4

Black Belt

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

12-18-2012
12:18 PM

81 Views

John_Campbell

New Contributor II

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

12-18-2012
11:31 PM

81 Views

Bernard

Black Belt

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

12-23-2012
10:29 PM

81 Views

- 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.