FLA_Accum_T_UT_fc.h File Reference

(r)

Go to the source code of this file.

Functions
FLA_Error	FLA_Accum_T_UT_fc_unb_var1 (FLA_Obj A, FLA_Obj t, FLA_Obj T)
FLA_Error	FLA_Accum_T_UT_fc_blk_var2 (FLA_Obj A, FLA_Obj t, FLA_Obj T)
FLA_Error	FLA_Accum_T_UT_fc_opt_var1 (FLA_Obj A, FLA_Obj t, FLA_Obj T)
FLA_Error	FLA_Accum_T_UT_fc_ops_var1 (int m_A, int n_AT, float *A, int rs_A, int cs_A, int m_t, float *t, int inc_t, float *T, int rs_T, int cs_T)
FLA_Error	FLA_Accum_T_UT_fc_opd_var1 (int m_A, int n_AT, double *A, int rs_A, int cs_A, int m_t, double *t, int inc_t, double *T, int rs_T, int cs_T)
FLA_Error	FLA_Accum_T_UT_fc_opc_var1 (int m_A, int n_AT, scomplex *A, int rs_A, int cs_A, int m_t, scomplex *t, int inc_t, scomplex *T, int rs_T, int cs_T)
FLA_Error	FLA_Accum_T_UT_fc_opz_var1 (int m_A, int n_AT, dcomplex *A, int rs_A, int cs_A, int m_t, dcomplex *t, int inc_t, dcomplex *T, int rs_T, int cs_T)

Function Documentation

FLA_Accum_T_UT_fc_blk_var2()

FLA_Error FLA_Accum_T_UT_fc_blk_var2	(	FLA_Obj	A,
		FLA_Obj	t,
		FLA_Obj	T
	)

{
  FLA_Obj ATL,   ATR,      A00, A01, A02, 
          ABL,   ABR,      A10, A11, A12,
                           A20, A21, A22;
 
  FLA_Obj tT,              t0,
          tB,              t1,
                           t2;
 
  FLA_Obj TL,    TR,       T0,  T1,  T2;
 
  FLA_Obj AB1;
 
  dim_t b_alg, b;
 
  b_alg = FLA_Obj_length( T );
 
  FLA_Part_2x2( A,    &ATL, &ATR,
                      &ABL, &ABR,     0, 0, FLA_TL );
 
  FLA_Part_2x1( t,    &tT, 
                      &tB,            0, FLA_TOP );
 
  FLA_Part_1x2( T,    &TL,  &TR,      0, FLA_LEFT );
 
  while ( FLA_Obj_length( tB ) > 0 ) {
 
    b = min( FLA_Obj_length( tB ), b_alg );
 
    FLA_Repart_2x2_to_3x3( ATL, /**/ ATR,       &A00, /**/ &A01, &A02,
                        /* ************* */   /* ******************** */
                                                &A10, /**/ &A11, &A12,
                           ABL, /**/ ABR,       &A20, /**/ &A21, &A22,
                           b, b, FLA_BR );
 
    FLA_Repart_2x1_to_3x1( tT,                &t0, 
                        /* ** */            /* ** */
                                              &t1, 
                           tB,                &t2,        b, FLA_BOTTOM );
 
    FLA_Repart_1x2_to_1x3( TL,  /**/ TR,        &T0, /**/ &T1, &T2,
                           b, FLA_RIGHT );
 
    /*------------------------------------------------------------*/
 
    FLA_Merge_2x1( A11,
                   A21,   &AB1 );
 
    //FLA_Accum_T_UT_fc_unb_var1( AB1, t1, T1 );
    FLA_Accum_T_UT_fc_opt_var1( AB1, t1, T1 );
 
    /*------------------------------------------------------------*/
 
    FLA_Cont_with_3x3_to_2x2( &ATL, /**/ &ATR,       A00, A01, /**/ A02,
                                                     A10, A11, /**/ A12,
                            /* ************** */  /* ****************** */
                              &ABL, /**/ &ABR,       A20, A21, /**/ A22,
                              FLA_TL );
 
    FLA_Cont_with_3x1_to_2x1( &tT,                t0, 
                                                  t1, 
                            /* ** */           /* ** */
                              &tB,                t2,     FLA_TOP );
 
    FLA_Cont_with_1x3_to_1x2( &TL,  /**/ &TR,        T0, T1, /**/ T2,
                              FLA_LEFT );
 
  }
 
  return FLA_SUCCESS;
}

References FLA_Accum_T_UT_fc_opt_var1(), FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Merge_2x1(), FLA_Obj_length(), FLA_Part_1x2(), FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), and i.

Referenced by FLA_Accum_T_UT_internal().

FLA_Accum_T_UT_fc_opc_var1()

FLA_Error FLA_Accum_T_UT_fc_opc_var1	(	int	m_A,
		int	n_AT,
		scomplex *	A,
		int	rs_A,
		int	cs_A,
		int	m_t,
		scomplex *	t,
		int	inc_t,
		scomplex *	T,
		int	rs_T,
		int	cs_T
	)

{
  scomplex* buff_1   = FLA_COMPLEX_PTR( FLA_ONE );
  int       i;
 
  for ( i = 0; i < m_t; ++i )
  {
    scomplex* a10t      = buff_A + (0  )*cs_A + (i  )*rs_A;
    scomplex* A20       = buff_A + (0  )*cs_A + (i+1)*rs_A;
    scomplex* a21       = buff_A + (i  )*cs_A + (i+1)*rs_A;
 
    scomplex* tau1      = buff_t + (i  )*inc_t;
 
    scomplex* tau11     = buff_T + (i  )*cs_T + (i  )*rs_T;
    scomplex* t01       = buff_T + (i  )*cs_T + (0  )*rs_T;
 
    int       m_ahead   = m_A  - i - 1;
    int       n_behind  = i;
 
    /*------------------------------------------------------------*/
 
    // FLA_Copy_external( tau1, tau11 );
    *tau11 = *tau1;
 
    // FLA_Copyt_external( FLA_CONJ_TRANSPOSE, a10t, t01 );
    bl1_ccopyv( BLIS1_CONJUGATE,
                n_behind,
                a10t, cs_A,
                t01, rs_T );
 
    // FLA_Gemv_external( FLA_CONJ_TRANSPOSE, FLA_ONE, A20, a21, FLA_ONE, t01 );
    bl1_cgemv( BLIS1_CONJ_TRANSPOSE,
               BLIS1_NO_CONJUGATE,
               m_ahead, n_behind,
               buff_1,
               A20, rs_A, cs_A,
               a21, rs_A,
               buff_1,
               t01, rs_T );
 
    /*------------------------------------------------------------*/
 
  }
 
  return FLA_SUCCESS;
}

References bl1_ccopyv(), bl1_cgemv(), BLIS1_CONJ_TRANSPOSE, BLIS1_CONJUGATE, BLIS1_NO_CONJUGATE, FLA_ONE, and i.

Referenced by FLA_Accum_T_UT_fc_opt_var1().

FLA_Accum_T_UT_fc_opd_var1()

FLA_Error FLA_Accum_T_UT_fc_opd_var1	(	int	m_A,
		int	n_AT,
		double *	A,
		int	rs_A,
		int	cs_A,
		int	m_t,
		double *	t,
		int	inc_t,
		double *	T,
		int	rs_T,
		int	cs_T
	)

{
  double* buff_1   = FLA_DOUBLE_PTR( FLA_ONE );
  int     i;
 
  for ( i = 0; i < m_t; ++i )
  {
    double* a10t      = buff_A + (0  )*cs_A + (i  )*rs_A;
    double* A20       = buff_A + (0  )*cs_A + (i+1)*rs_A;
    double* a21       = buff_A + (i  )*cs_A + (i+1)*rs_A;
 
    double* tau1      = buff_t + (i  )*inc_t;
 
    double* tau11     = buff_T + (i  )*cs_T + (i  )*rs_T;
    double* t01       = buff_T + (i  )*cs_T + (0  )*rs_T;
 
    int     m_ahead   = m_A  - i - 1;
    int     n_behind  = i;
 
    /*------------------------------------------------------------*/
 
    // FLA_Copy_external( tau1, tau11 );
    *tau11 = *tau1;
 
    // FLA_Copyt_external( FLA_CONJ_TRANSPOSE, a10t, t01 );
    bl1_dcopyv( BLIS1_CONJUGATE,
                n_behind,
                a10t, cs_A,
                t01, rs_T );
 
    // FLA_Gemv_external( FLA_CONJ_TRANSPOSE, FLA_ONE, A20, a21, FLA_ONE, t01 );
    bl1_dgemv( BLIS1_CONJ_TRANSPOSE,
               BLIS1_NO_CONJUGATE,
               m_ahead,
               n_behind,
               buff_1,
               A20, rs_A, cs_A,
               a21, rs_A,
               buff_1,
               t01, rs_T );
 
    /*------------------------------------------------------------*/
 
  }
 
  return FLA_SUCCESS;
}

References bl1_dcopyv(), bl1_dgemv(), BLIS1_CONJ_TRANSPOSE, BLIS1_CONJUGATE, BLIS1_NO_CONJUGATE, FLA_ONE, and i.

Referenced by FLA_Accum_T_UT_fc_opt_var1().

FLA_Accum_T_UT_fc_ops_var1()

FLA_Error FLA_Accum_T_UT_fc_ops_var1	(	int	m_A,
		int	n_AT,
		float *	A,
		int	rs_A,
		int	cs_A,
		int	m_t,
		float *	t,
		int	inc_t,
		float *	T,
		int	rs_T,
		int	cs_T
	)

{
  float* buff_1   = FLA_FLOAT_PTR( FLA_ONE );
  int    i;
 
  for ( i = 0; i < m_t; ++i )
  {
    float* a10t     = buff_A + (0  )*cs_A + (i  )*rs_A;
    float* A20      = buff_A + (0  )*cs_A + (i+1)*rs_A;
    float* a21      = buff_A + (i  )*cs_A + (i+1)*rs_A;
 
    float* tau1     = buff_t + (i  )*inc_t;
 
    float* tau11    = buff_T + (i  )*cs_T + (i  )*rs_T;
    float* t01      = buff_T + (i  )*cs_T + (0  )*rs_T;
 
    int    m_ahead  = m_A  - i - 1;
    int    n_behind = i;
 
    /*------------------------------------------------------------*/
 
    // FLA_Copy_external( tau1, tau11 );
    *tau11 = *tau1;
 
    // FLA_Copyt_external( FLA_CONJ_TRANSPOSE, a10t, t01 );
    bl1_scopyv( BLIS1_CONJUGATE,
                n_behind,
                a10t, cs_A,
                t01, rs_T );
 
    // FLA_Gemv_external( FLA_CONJ_TRANSPOSE, FLA_ONE, A20, a21, FLA_ONE, t01 );
    bl1_sgemv( BLIS1_CONJ_TRANSPOSE,
               BLIS1_NO_CONJUGATE,
               m_ahead,
               n_behind,
               buff_1,
               A20, rs_A, cs_A,
               a21, rs_A,
               buff_1,
               t01, rs_T );
 
    /*------------------------------------------------------------*/
 
  }
 
  return FLA_SUCCESS;
}

References bl1_scopyv(), bl1_sgemv(), BLIS1_CONJ_TRANSPOSE, BLIS1_CONJUGATE, BLIS1_NO_CONJUGATE, FLA_ONE, and i.

Referenced by FLA_Accum_T_UT_fc_opt_var1().

FLA_Accum_T_UT_fc_opt_var1()

FLA_Error FLA_Accum_T_UT_fc_opt_var1	(	FLA_Obj	A,
		FLA_Obj	t,
		FLA_Obj	T
	)

{
  FLA_Datatype datatype;
  int          m_A, n_AT;
  int          rs_A, cs_A;
  int          m_t, inc_t;
  int          rs_T, cs_T;
 
  datatype = FLA_Obj_datatype( A );
 
  m_A      = FLA_Obj_length( A );
  n_AT     = FLA_Obj_width( A );
  rs_A     = FLA_Obj_row_stride( A );
  cs_A     = FLA_Obj_col_stride( A );
 
  m_t      = FLA_Obj_vector_dim( t );
  inc_t    = FLA_Obj_vector_inc( t );
 
  rs_T     = FLA_Obj_row_stride( T );
  cs_T     = FLA_Obj_col_stride( T );
 
  switch ( datatype )
  {
    case FLA_FLOAT:
    {
      float* buff_A = FLA_FLOAT_PTR( A );
      float* buff_t = FLA_FLOAT_PTR( t );
      float* buff_T = FLA_FLOAT_PTR( T );
 
      FLA_Accum_T_UT_fc_ops_var1( m_A,
                                  n_AT,
                                  buff_A, rs_A, cs_A,
                                  m_t, 
                                  buff_t, inc_t,
                                  buff_T, rs_T, cs_T );
 
      break;
    }
 
    case FLA_DOUBLE:
    {
      double* buff_A = FLA_DOUBLE_PTR( A );
      double* buff_t = FLA_DOUBLE_PTR( t );
      double* buff_T = FLA_DOUBLE_PTR( T );
 
      FLA_Accum_T_UT_fc_opd_var1( m_A,
                                  n_AT,
                                  buff_A, rs_A, cs_A,
                                  m_t, 
                                  buff_t, inc_t,
                                  buff_T, rs_T, cs_T );
 
      break;
    }
 
    case FLA_COMPLEX:
    {
      scomplex* buff_A = FLA_COMPLEX_PTR( A );
      scomplex* buff_t = FLA_COMPLEX_PTR( t );
      scomplex* buff_T = FLA_COMPLEX_PTR( T );
 
      FLA_Accum_T_UT_fc_opc_var1( m_A,
                                  n_AT,
                                  buff_A, rs_A, cs_A,
                                  m_t, 
                                  buff_t, inc_t,
                                  buff_T, rs_T, cs_T );
 
      break;
    }
 
    case FLA_DOUBLE_COMPLEX:
    {
      dcomplex* buff_A = FLA_DOUBLE_COMPLEX_PTR( A );
      dcomplex* buff_t = FLA_DOUBLE_COMPLEX_PTR( t );
      dcomplex* buff_T = FLA_DOUBLE_COMPLEX_PTR( T );
 
      FLA_Accum_T_UT_fc_opz_var1( m_A,
                                  n_AT,
                                  buff_A, rs_A, cs_A,
                                  m_t,
                                  buff_t, inc_t,
                                  buff_T, rs_T, cs_T );
 
      break;
    }
  }
 
  return FLA_SUCCESS;
}

References FLA_Accum_T_UT_fc_opc_var1(), FLA_Accum_T_UT_fc_opd_var1(), FLA_Accum_T_UT_fc_ops_var1(), FLA_Accum_T_UT_fc_opz_var1(), FLA_Obj_col_stride(), FLA_Obj_datatype(), FLA_Obj_length(), FLA_Obj_row_stride(), FLA_Obj_vector_dim(), FLA_Obj_vector_inc(), FLA_Obj_width(), and i.

Referenced by FLA_Accum_T_UT_fc_blk_var2().

FLA_Accum_T_UT_fc_opz_var1()

FLA_Error FLA_Accum_T_UT_fc_opz_var1	(	int	m_A,
		int	n_AT,
		dcomplex *	A,
		int	rs_A,
		int	cs_A,
		int	m_t,
		dcomplex *	t,
		int	inc_t,
		dcomplex *	T,
		int	rs_T,
		int	cs_T
	)

{
  dcomplex* buff_1   = FLA_DOUBLE_COMPLEX_PTR( FLA_ONE );
  int       i;
 
  for ( i = 0; i < m_t; ++i )
  {
    dcomplex* a10t      = buff_A + (0  )*cs_A + (i  )*rs_A;
    dcomplex* A20       = buff_A + (0  )*cs_A + (i+1)*rs_A;
    dcomplex* a21       = buff_A + (i  )*cs_A + (i+1)*rs_A;
 
    dcomplex* tau1      = buff_t + (i  )*inc_t;
 
    dcomplex* tau11     = buff_T + (i  )*cs_T + (i  )*rs_T;
    dcomplex* t01       = buff_T + (i  )*cs_T + (0  )*rs_T;
 
    int       m_ahead   = m_A  - i - 1;
    int       n_behind  = i;
 
    /*------------------------------------------------------------*/
 
    // FLA_Copy_external( tau1, tau11 );
    *tau11 = *tau1;
 
    // FLA_Copyt_external( FLA_CONJ_TRANSPOSE, a10t, t01 );
    bl1_zcopyv( BLIS1_CONJUGATE,
                n_behind,
                a10t, cs_A,
                t01, rs_T );
 
    // FLA_Gemv_external( FLA_CONJ_TRANSPOSE, FLA_ONE, A20, a21, FLA_ONE, t01 );
    bl1_zgemv( BLIS1_CONJ_TRANSPOSE,
               BLIS1_NO_CONJUGATE,
               m_ahead, n_behind,
               buff_1,
               A20, rs_A, cs_A,
               a21, rs_A,
               buff_1,
               t01, rs_T );
 
    /*------------------------------------------------------------*/
 
  }
 
  return FLA_SUCCESS;
}

References bl1_zcopyv(), bl1_zgemv(), BLIS1_CONJ_TRANSPOSE, BLIS1_CONJUGATE, BLIS1_NO_CONJUGATE, FLA_ONE, and i.

Referenced by FLA_Accum_T_UT_fc_opt_var1().

FLA_Accum_T_UT_fc_unb_var1()

FLA_Error FLA_Accum_T_UT_fc_unb_var1	(	FLA_Obj	A,
		FLA_Obj	t,
		FLA_Obj	T
	)

{
  FLA_Obj ATL,   ATR,      A00,  a01,     A02, 
          ABL,   ABR,      a10t, alpha11, a12t,
                           A20,  a21,     A22;
 
  FLA_Obj tauT,            tau0,
          tauB,            tau1,
                           tau2;
 
  FLA_Obj TTL,   TTR,      T00,  t01,   T02, 
          TBL,   TBR,      t10t, tau11, t12t,
                           T20,  t21,   T22;
 
 
  FLA_Part_2x2( A,    &ATL, &ATR,
                      &ABL, &ABR,     0, 0, FLA_TL );
 
  FLA_Part_2x1( tau,  &tauT, 
                      &tauB,          0, FLA_TOP );
 
  FLA_Part_2x2( T,    &TTL, &TTR,
                      &TBL, &TBR,     0, 0, FLA_TL );
 
  while ( FLA_Obj_length( tauB ) > 0 ) {
 
    FLA_Repart_2x2_to_3x3( ATL, /**/ ATR,       &A00,  /**/ &a01,     &A02,
                        /* ************* */   /* ************************** */
                                                &a10t, /**/ &alpha11, &a12t,
                           ABL, /**/ ABR,       &A20,  /**/ &a21,     &A22,
                           1, 1, FLA_BR );
 
    FLA_Repart_2x1_to_3x1( tauT,                &tau0, 
                        /* ** */              /* ** */
                                                &tau1, 
                           tauB,                &tau2,        1, FLA_BOTTOM );
 
    FLA_Repart_2x2_to_3x3( TTL, /**/ TTR,       &T00,  /**/ &t01,   &T02,
                        /* ************* */   /* ************************ */
                                                &t10t, /**/ &tau11, &t12t,
                           TBL, /**/ TBR,       &T20,  /**/ &t21,   &T22,
                           1, 1, FLA_BR );
 
    /*------------------------------------------------------------*/
 
    // tau11 = tau1;
    FLA_Copy_external( tau1, tau11 );
 
    // t01 = a10t' + A20' * a21;
    FLA_Copyt_external( FLA_CONJ_TRANSPOSE, a10t, t01 );
    FLA_Gemv_external( FLA_CONJ_TRANSPOSE, FLA_ONE, A20, a21, FLA_ONE, t01 );
 
    /*------------------------------------------------------------*/
 
    FLA_Cont_with_3x3_to_2x2( &ATL, /**/ &ATR,       A00,  a01,     /**/ A02,
                                                     a10t, alpha11, /**/ a12t,
                            /* ************** */  /* ************************ */
                              &ABL, /**/ &ABR,       A20,  a21,     /**/ A22,
                              FLA_TL );
 
    FLA_Cont_with_3x1_to_2x1( &tauT,                tau0, 
                                                    tau1, 
                            /* ** */              /* ** */
                              &tauB,                tau2,     FLA_TOP );
 
    FLA_Cont_with_3x3_to_2x2( &TTL, /**/ &TTR,       T00,  t01,   /**/ T02,
                                                     t10t, tau11, /**/ t12t,
                            /* ************** */  /* ********************** */
                              &TBL, /**/ &TBR,       T20,  t21,   /**/ T22,
                              FLA_TL );
 
  }
 
  return FLA_SUCCESS;
}

References FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Copy_external(), FLA_Copyt_external(), FLA_Gemv_external(), FLA_Obj_length(), FLA_ONE, FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), and i.