static int
RunTest(int *iparam, float *dparam, real_Double_t *t_)
{
float *A, *Acpy = NULL, *L, *b, *x;
real_Double_t t;
int *piv;
int n = iparam[TIMING_N];
int nrhs = iparam[TIMING_NRHS];
int check = iparam[TIMING_CHECK];
int lda = n;
int ldb = n;
/* Allocate Data */
A = (float *)malloc(lda*n*sizeof(float));
/* Check if unable to allocate memory */
if ( !A ){
printf("Out of Memory \n ");
exit(0);
}
/* Initialize Plasma */
PLASMA_Init( iparam[TIMING_THRDNBR] );
if ( iparam[TIMING_SCHEDULER] )
PLASMA_Set(PLASMA_SCHEDULING_MODE, PLASMA_DYNAMIC_SCHEDULING );
else
PLASMA_Set(PLASMA_SCHEDULING_MODE, PLASMA_STATIC_SCHEDULING );
/*if ( !iparam[TIMING_AUTOTUNING] ) {*/
PLASMA_Disable(PLASMA_AUTOTUNING);
PLASMA_Set(PLASMA_TILE_SIZE, iparam[TIMING_NB] );
PLASMA_Set(PLASMA_INNER_BLOCK_SIZE, iparam[TIMING_IB] );
/* } else { */
/* PLASMA_Get(PLASMA_TILE_SIZE, &iparam[TIMING_NB] ); */
/* PLASMA_Get(PLASMA_INNER_BLOCK_SIZE, &iparam[TIMING_IB] ); */
/* } */
/* Initialiaze Data */
LAPACKE_slarnv_work(1, ISEED, n*lda, A);
/* Allocate Workspace */
PLASMA_Alloc_Workspace_sgesv_incpiv(n, &L, &piv);
/* Save AT in lapack layout for check */
if ( check ) {
Acpy = (float *)malloc(lda*n*sizeof(float));
LAPACKE_slacpy_work(LAPACK_COL_MAJOR,' ', n, n, A, lda, Acpy, lda);
}
t = -cWtime();
PLASMA_sgetrf_incpiv( n, n, A, lda, L, piv );
t += cWtime();
*t_ = t;
/* Check the solution */
if ( check )
{
b = (float *)malloc(ldb*nrhs *sizeof(float));
x = (float *)malloc(ldb*nrhs *sizeof(float));
LAPACKE_slarnv_work(1, ISEED, ldb*nrhs, x);
LAPACKE_slacpy_work(LAPACK_COL_MAJOR,' ', n, nrhs, x, ldb, b, ldb);
PLASMA_sgetrs_incpiv( PlasmaNoTrans, n, nrhs, A, lda, L, piv, x, ldb );
dparam[TIMING_RES] = s_check_solution(n, n, nrhs, Acpy, lda, b, x, ldb,
&(dparam[TIMING_ANORM]), &(dparam[TIMING_BNORM]),
&(dparam[TIMING_XNORM]));
free( Acpy ); free( b ); free( x );
}
free( A );
free( L );
free( piv );
PLASMA_Finalize();
return 0;
}
int main ()
{
int cores = 2;
int N = 10;
int LDA = 10;
int NRHS = 5;
int LDB = 10;
int info;
int info_solution;
int i,j;
int LDAxN = LDA*N;
int LDBxNRHS = LDB*NRHS;
float *A1 = (float *)malloc(LDA*N*(sizeof*A1));
float *A2 = (float *)malloc(LDA*N*(sizeof*A2));
float *B1 = (float *)malloc(LDB*NRHS*(sizeof*B1));
float *B2 = (float *)malloc(LDB*NRHS*(sizeof*B2));
float *L;
int *IPIV;
/* Check if unable to allocate memory */
if ((!A1)||(!A2)||(!B1)||(!B2)){
printf("Out of Memory \n ");
exit(0);
}
/*Plasma Initialize*/
PLASMA_Init(cores);
printf("-- PLASMA is initialized to run on %d cores. \n",cores);
/* Initialize A1 and A2 Matrix */
LAPACKE_slarnv_work(IONE, ISEED, LDAxN, A1);
for ( i = 0; i < N; i++)
for ( j = 0; j < N; j++)
A2[LDA*j+i] = A1[LDA*j+i];
/* Initialize B1 and B2 */
LAPACKE_slarnv_work(IONE, ISEED, LDBxNRHS, B1);
for ( i = 0; i < N; i++)
for ( j = 0; j < NRHS; j++)
B2[LDB*j+i] = B1[LDB*j+i];
/* Allocate L and IPIV */
info = PLASMA_Alloc_Workspace_sgetrf_incpiv(N, N, &L, &IPIV);
/* LU factorization of the matrix A */
info = PLASMA_sgetrf_incpiv(N, N, A2, LDA, L, IPIV);
/* Solve the problem */
info = PLASMA_sgetrs_incpiv(PlasmaNoTrans, N, NRHS, A2, LDA, L, IPIV, B2, LDB);
/* Check the solution */
info_solution = check_solution(N, NRHS, A1, LDA, B1, B2, LDB);
if ((info_solution != 0)|(info != 0))
printf("-- Error in SGETRS example ! \n");
else
printf("-- Run of SGETRS example successful ! \n");
free(A1); free(A2); free(B1); free(B2); free(IPIV); free(L);
PLASMA_Finalize();
exit(0);
}
