Centre Blaise Pascal » HPL sur GPU

/* 

 * -- High Performance Computing Linpack Benchmark (HPL)                

 *    HPL - 2.0 - September 10, 2008                          

 *    Antoine P. Petitet                                                

 *    University of Tennessee, Knoxville                                

 *    Innovative Computing Laboratory                                 

 *    (C) Copyright 2000-2008 All Rights Reserved                       

 *                                                                      

 * -- Copyright notice and Licensing terms:                             

 *                                                                      

 * Redistribution  and  use in  source and binary forms, with or without

 * modification, are  permitted provided  that the following  conditions

 * are met:                                                             

 *                                                                      

 * 1. Redistributions  of  source  code  must retain the above copyright

 * notice, this list of conditions and the following disclaimer.        

 *                                                                      

 * 2. Redistributions in binary form must reproduce  the above copyright

 * notice, this list of conditions,  and the following disclaimer in the

 * documentation and/or other materials provided with the distribution. 

 *                                                                      

 * 3. All  advertising  materials  mentioning  features  or  use of this

 * software must display the following acknowledgement:                 

 * This  product  includes  software  developed  at  the  University  of

 * Tennessee, Knoxville, Innovative Computing Laboratory.             

 *                                                                      

 * 4. The name of the  University,  the name of the  Laboratory,  or the

 * names  of  its  contributors  may  not  be used to endorse or promote

 * products  derived   from   this  software  without  specific  written

 * permission.                                                          

 *                                                                      

 * -- Disclaimer:                                                       

 *                                                                      

 * THIS  SOFTWARE  IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,  INCLUDING,  BUT NOT

 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR

 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY

 * OR  CONTRIBUTORS  BE  LIABLE FOR ANY  DIRECT,  INDIRECT,  INCIDENTAL,

 * SPECIAL,  EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES  (INCLUDING,  BUT NOT

 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,

 * DATA OR PROFITS; OR BUSINESS INTERRUPTION)  HOWEVER CAUSED AND ON ANY

 * THEORY OF LIABILITY, WHETHER IN CONTRACT,  STRICT LIABILITY,  OR TORT

 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE

 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 

 * ---------------------------------------------------------------------

 */ 

/*

 * Include files

 */

#include "hpl.h"

#ifdef STDC_HEADERS

void HPL_pdupdateTN

(

   HPL_T_panel *                    PBCST,

   int *                            IFLAG,

   HPL_T_panel *                    PANEL,

   const int                        NN

)

#else

void HPL_pdupdateTN

( PBCST, IFLAG, PANEL, NN )

   HPL_T_panel *                    PBCST;

   int *                            IFLAG;

   HPL_T_panel *                    PANEL;

   const int                        NN;

#endif

{

/* 

 * Purpose

 * =======

 *

 * HPL_pdupdateTN broadcast - forward the panel PBCST and simultaneously

 * applies the row interchanges and updates part of the trailing  (using

 * the panel PANEL) submatrix.

 *

 * Arguments

 * =========

 *

 * PBCST   (local input/output)          HPL_T_panel *

 *         On entry,  PBCST  points to the data structure containing the

 *         panel (to be broadcast) information.

 *

 * IFLAG   (local output)                int *

 *         On exit,  IFLAG  indicates  whether or not  the broadcast has

 *         been completed when PBCST is not NULL on entry. In that case,

 *         IFLAG is left unchanged.

 *

 * PANEL   (local input/output)          HPL_T_panel *

 *         On entry,  PANEL  points to the data structure containing the

 *         panel (to be updated) information.

 *

 * NN      (local input)                 const int

 *         On entry, NN specifies  the  local  number  of columns of the

 *         trailing  submatrix  to be updated  starting  at the  current

 *         position. NN must be at least zero.

 *

 * ---------------------------------------------------------------------

 */ 

/*

 * .. Local Variables ..

 */

   double                    * Aptr, * L1ptr, * L2ptr, * Uptr, * dpiv;

   int                       * ipiv;

#ifdef HPL_CALL_VSIPL

   vsip_mview_d              * Av0, * Av1, * Lv0, * Lv1, * Uv0, * Uv1;

#endif

   int                       curr, i, iroff, jb, lda, ldl2, mp, n, nb,

                             nq0, nn, test;

   static int                tswap = 0;

   static HPL_T_SWAP         fswap = HPL_NO_SWP;

#define LDU                  jb

/* ..

 * .. Executable Statements ..

 */

#ifdef HPL_DETAILED_TIMING

   HPL_ptimer( HPL_TIMING_UPDATE );

#endif

   nb = PANEL->nb; jb = PANEL->jb; n = PANEL->nq; lda = PANEL->lda;

   if( NN >= 0 ) n = Mmin( NN, n );

/*

 * There is nothing to update, enforce the panel broadcast.

 */

   if( ( n <= 0 ) || ( jb <= 0 ) )

   {

      if( PBCST != NULL )

      {

         do { (void) HPL_bcast( PBCST, IFLAG ); }

         while( *IFLAG != HPL_SUCCESS );

      }

#ifdef HPL_DETAILED_TIMING

      HPL_ptimer( HPL_TIMING_UPDATE );

#endif

      return;

   }

/*

 * Enable/disable the column panel probing mechanism

 */

   (void) HPL_bcast( PBCST, &test );

/*

 * 1 x Q case

 */

   if( PANEL->grid->nprow == 1 )

   {

      Aptr = PANEL->A;       L2ptr = PANEL->L2;   L1ptr = PANEL->L1;

      ldl2 = PANEL->ldl2;    dpiv  = PANEL->DPIV; ipiv  = PANEL->IWORK;

      mp   = PANEL->mp - jb; iroff = PANEL->ii;   nq0   = 0; 

#ifdef HPL_CALL_VSIPL

/*

 * Admit the blocks

 */

      (void) vsip_blockadmit_d(  PANEL->Ablock,  VSIP_TRUE );

      (void) vsip_blockadmit_d(  PANEL->L2block, VSIP_TRUE );

/*

 * Create the matrix views

 */

      Av0 = vsip_mbind_d( PANEL->Ablock,  0, 1, lda,  lda,  PANEL->pmat->nq );

      Lv0 = vsip_mbind_d( PANEL->L2block, 0, 1, ldl2, ldl2,              jb );

/*

 * Create the matrix subviews

 */

      Lv1 = vsip_msubview_d( Lv0, 0, 0, mp, jb );

#endif

      for( i = 0; i < jb; i++ ) { ipiv[i] = (int)(dpiv[i]) - iroff; }

/*

 * So far we have not updated anything -  test availability of the panel

 * to be forwarded - If detected forward it and finish the update in one

 * step.

 */

      while ( test == HPL_KEEP_TESTING )

      {

         nn = n - nq0; nn = Mmin( nb, nn );

/*

 * Update nb columns at a time

 */

#ifdef HPL_DETAILED_TIMING

         HPL_ptimer( HPL_TIMING_LASWP );

         HPL_dlaswp00N( jb, nn, Aptr, lda, ipiv );

         HPL_ptimer( HPL_TIMING_LASWP );

#else

         HPL_dlaswp00N( jb, nn, Aptr, lda, ipiv );

#endif

         HPL_dtrsm( HplColumnMajor, HplLeft, HplUpper, HplTrans,

                    HplUnit, jb, nn, HPL_rone, L1ptr, jb, Aptr, lda );

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

         Uv1 = vsip_msubview_d( Av0, PANEL->ii,    PANEL->jj+nq0, jb, nn );

         Av1 = vsip_msubview_d( Av0, PANEL->ii+jb, PANEL->jj+nq0, mp, nn );

         vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                      HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

         (void) vsip_mdestroy_d( Av1 );

         (void) vsip_mdestroy_d( Uv1 );

#else

         HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                    jb, -HPL_rone, L2ptr, ldl2, Aptr, lda, HPL_rone,

                    Mptr( Aptr, jb, 0, lda ), lda );

#endif

         Aptr = Mptr( Aptr, 0, nn, lda ); nq0 += nn; 

         (void) HPL_bcast( PBCST, &test ); 

      }

/*

 * The panel has been forwarded at that point, finish the update

 */

      if( ( nn = n - nq0 ) > 0 )

      {

#ifdef HPL_DETAILED_TIMING

         HPL_ptimer( HPL_TIMING_LASWP );

         HPL_dlaswp00N( jb, nn, Aptr, lda, ipiv );

         HPL_ptimer( HPL_TIMING_LASWP );

#else

         HPL_dlaswp00N( jb, nn, Aptr, lda, ipiv );

#endif

         HPL_dtrsm( HplColumnMajor, HplLeft, HplUpper, HplTrans,

                    HplUnit, jb, nn, HPL_rone, L1ptr, jb, Aptr, lda );

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

         Uv1 = vsip_msubview_d( Av0, PANEL->ii,    PANEL->jj+nq0, jb, nn );

         Av1 = vsip_msubview_d( Av0, PANEL->ii+jb, PANEL->jj+nq0, mp, nn );

         vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                      HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

         (void) vsip_mdestroy_d( Av1 );

         (void) vsip_mdestroy_d( Uv1 );

#else

         HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                    jb, -HPL_rone, L2ptr, ldl2, Aptr, lda, HPL_rone,

                    Mptr( Aptr, jb, 0, lda ), lda );

#endif

      }

#ifdef HPL_CALL_VSIPL

/*

 * Destroy the matrix subviews

 */

      (void) vsip_mdestroy_d( Lv1 );

/*

 * Release the blocks

 */

      (void) vsip_blockrelease_d( vsip_mgetblock_d( Lv0 ), VSIP_TRUE );

      (void) vsip_blockrelease_d( vsip_mgetblock_d( Av0 ), VSIP_TRUE );

/*

 * Destroy the matrix views

 */

      (void) vsip_mdestroy_d( Lv0 );

      (void) vsip_mdestroy_d( Av0 );

#endif

   }

   else                        /* nprow > 1 ... */

   {

/*

 * Selection of the swapping algorithm - swap:broadcast U.

 */

      if( fswap == HPL_NO_SWP )

      { fswap = PANEL->algo->fswap; tswap = PANEL->algo->fsthr; }

      if( (   fswap == HPL_SWAP01 ) ||

          ( ( fswap == HPL_SW_MIX ) && ( n > tswap ) ) )

      { HPL_pdlaswp01N( PBCST, &test, PANEL, n ); }

      else

      { HPL_pdlaswp00N( PBCST, &test, PANEL, n ); }

/*

 * Compute redundantly row block of U and update trailing submatrix

 */

      nq0 = 0; curr = ( PANEL->grid->myrow == PANEL->prow ? 1 : 0 );

      Aptr = PANEL->A; L2ptr = PANEL->L2;  L1ptr = PANEL->L1;

      Uptr = PANEL->U; ldl2 = PANEL->ldl2;

      mp   = PANEL->mp - ( curr != 0 ? jb : 0 );

#ifdef HPL_CALL_VSIPL

/*

 * Admit the blocks

 */

      (void) vsip_blockadmit_d(  PANEL->Ablock,  VSIP_TRUE );

      (void) vsip_blockadmit_d(  PANEL->L2block, VSIP_TRUE );

      (void) vsip_blockadmit_d(  PANEL->Ublock,  VSIP_TRUE );

/*

 * Create the matrix views

 */

      Av0 = vsip_mbind_d( PANEL->Ablock,  0, 1, lda,  lda,  PANEL->pmat->nq );

      Lv0 = vsip_mbind_d( PANEL->L2block, 0, 1, ldl2, ldl2,              jb );

      Uv0 = vsip_mbind_d( PANEL->Ublock,  0, 1, LDU,  LDU,                n );

/*

 * Create the matrix subviews

 */

      Lv1 = vsip_msubview_d( Lv0, 0, 0, mp, jb );

#endif

/*

 * Broadcast has not occured yet, spliting the computational part

 */

      while ( test == HPL_KEEP_TESTING )

      {

         nn = n - nq0; nn = Mmin( nb, nn );

         HPL_dtrsm( HplColumnMajor, HplLeft,  HplUpper, HplTrans,

                    HplUnit, jb, nn, HPL_rone, L1ptr, jb, Uptr, LDU );

         if( curr != 0 )

         {

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

            Uv1 = vsip_msubview_d( Uv0, 0,            nq0,           jb, nn );

            Av1 = vsip_msubview_d( Av0, PANEL->ii+jb, PANEL->jj+nq0, mp, nn );

            vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                         HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

            (void) vsip_mdestroy_d( Av1 );

            (void) vsip_mdestroy_d( Uv1 );

#else

            HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                       jb, -HPL_rone, L2ptr, ldl2, Uptr, LDU, HPL_rone,

                       Mptr( Aptr, jb, 0, lda ), lda );

#endif

            HPL_dlacpy( jb, nn, Uptr, LDU, Aptr, lda );

         }

         else

         {

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

            Uv1 = vsip_msubview_d( Uv0, 0,            nq0,           jb, nn );

            Av1 = vsip_msubview_d( Av0, PANEL->ii,    PANEL->jj+nq0, mp, nn );

            vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                         HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

            (void) vsip_mdestroy_d( Av1 );

            (void) vsip_mdestroy_d( Uv1 );

#else

            HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                       jb, -HPL_rone, L2ptr, ldl2, Uptr, LDU, HPL_rone,

                       Aptr, lda );

#endif

         }

         Uptr = Mptr( Uptr, 0, nn, LDU );

         Aptr = Mptr( Aptr, 0, nn, lda ); nq0 += nn;

         (void) HPL_bcast( PBCST, &test ); 

      }

/*

 * The panel has been forwarded at that point, finish the update

 */

      if( ( nn = n - nq0 ) > 0 )

      {

         HPL_dtrsm( HplColumnMajor, HplLeft,  HplUpper, HplTrans,

                    HplUnit, jb, nn, HPL_rone, L1ptr, jb, Uptr, LDU );

         if( curr != 0 )

         {

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

            Uv1 = vsip_msubview_d( Uv0, 0,            nq0,           jb, nn );

            Av1 = vsip_msubview_d( Av0, PANEL->ii+jb, PANEL->jj+nq0, mp, nn );

            vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                         HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

            (void) vsip_mdestroy_d( Av1 );

            (void) vsip_mdestroy_d( Uv1 );

#else

            HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                       jb, -HPL_rone, L2ptr, ldl2, Uptr, LDU, HPL_rone,

                       Mptr( Aptr, jb, 0, lda ), lda );

#endif

            HPL_dlacpy( jb, nn, Uptr, LDU, Aptr, lda );

         }

         else

         {

#ifdef HPL_CALL_VSIPL

/*

 * Create the matrix subviews

 */

            Uv1 = vsip_msubview_d( Uv0, 0,            nq0,           jb, nn );

            Av1 = vsip_msubview_d( Av0, PANEL->ii,    PANEL->jj+nq0, mp, nn );

            vsip_gemp_d( -HPL_rone, Lv1, VSIP_MAT_NTRANS, Uv1, VSIP_MAT_NTRANS,

                         HPL_rone, Av1 );

/*

 * Destroy the matrix subviews

 */

            (void) vsip_mdestroy_d( Av1 );

            (void) vsip_mdestroy_d( Uv1 );

#else

            HPL_dgemm( HplColumnMajor, HplNoTrans, HplNoTrans, mp, nn,

                       jb, -HPL_rone, L2ptr, ldl2, Uptr, LDU, HPL_rone,

                       Aptr, lda );

#endif

         }

      }

#ifdef HPL_CALL_VSIPL

/*

 * Destroy the matrix subviews

 */

      (void) vsip_mdestroy_d( Lv1 );

/*

 * Release the blocks

 */

      (void) vsip_blockrelease_d( vsip_mgetblock_d( Uv0 ), VSIP_TRUE );

      (void) vsip_blockrelease_d( vsip_mgetblock_d( Lv0 ), VSIP_TRUE );

      (void) vsip_blockrelease_d( vsip_mgetblock_d( Av0 ), VSIP_TRUE );

/*

 * Destroy the matrix views

 */

      (void) vsip_mdestroy_d( Uv0 );

      (void) vsip_mdestroy_d( Lv0 );

      (void) vsip_mdestroy_d( Av0 );

#endif

   }

   PANEL->A = Mptr( PANEL->A, 0, n, lda ); PANEL->nq -= n; PANEL->jj += n;

/*

 * return the outcome of the probe  (should always be  HPL_SUCCESS,  the

 * panel broadcast is enforced in that routine).

 */

   if( PBCST != NULL ) *IFLAG = test;

#ifdef HPL_DETAILED_TIMING

   HPL_ptimer( HPL_TIMING_UPDATE );

#endif

/*

 * End of HPL_pdupdateTN

 */

}