Uisng MKL_lapack function with TBB

Uisng MKL_lapack function with TBB

Dear Support Provider

I working on an application which uses MKL_LAPACK routine from inside of a C++ program on MAC-Ox 10.6. using icpc

My data is complex type 3D data of A[4x4xn] and B[4x1xn] dimensions;

{n ranges upto 2^48};

I have to find out least square estimation for equation B[4x4xi] = A[4x4xi].X;

I am using zgelss function for performing this routine;

now I have directed by MKL support provider as the task size of matrises are small so internal optimization of routines can not be used

So I have to paralleize the data over the routine and gave example openMP for doing so

{Details: http://software.intel.com/en-us/forums/showthread.php?t=85278}

I have not used openMP or TBB before. I spent some time reading openMP and find out it is not possible to safely parrallelize a routine which is used during link time.
{Please comment: right/wrong/other}

I belive My sitution is best for using Data Parallel model.

the sequential statement is

for (int loop = 0 ; loop < 1000000; loop++){
zgelss_( &m, &n, &nrhs, a, &lda, b, &ldb, s,
&rcond, &rank, work, &lwork, rwork, &info );

My query --> is it possible to parallelize the statement using TBB.
If possible How it can be done;


1 post / 0 new
For more complete information about compiler optimizations, see our Optimization Notice.