|
libflame
revision_anchor
|
Go to the source code of this file.
Functions | |
| FLA_Error | FLA_LU_piv_blk_var3 (FLA_Obj A, FLA_Obj p, fla_lu_t *cntl) |
| FLA_Error | FLA_LU_piv_blk_var4 (FLA_Obj A, FLA_Obj p, fla_lu_t *cntl) |
| FLA_Error | FLA_LU_piv_blk_var5 (FLA_Obj A, FLA_Obj p, fla_lu_t *cntl) |
| FLA_Error | FLA_LU_piv_unb_var3 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_unb_var3b (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_unb_var4 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_unb_var5 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_opt_var3 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_ops_var3 (int m_A, int n_A, float *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opd_var3 (int m_A, int n_A, double *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opc_var3 (int m_A, int n_A, scomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opz_var3 (int m_A, int n_A, dcomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opt_var4 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_ops_var4 (int m_A, int n_A, float *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opd_var4 (int m_A, int n_A, double *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opc_var4 (int m_A, int n_A, scomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opz_var4 (int m_A, int n_A, dcomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opt_var5 (FLA_Obj A, FLA_Obj p) |
| FLA_Error | FLA_LU_piv_ops_var5 (int m_A, int n_A, float *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opd_var5 (int m_A, int n_A, double *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opc_var5 (int m_A, int n_A, scomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error | FLA_LU_piv_opz_var5 (int m_A, int n_A, dcomplex *buff_A, int rs_A, int cs_A, int *buff_p, int inc_p) |
| FLA_Error FLA_LU_piv_blk_var3 | ( | FLA_Obj | A, |
| FLA_Obj | p, | ||
| fla_lu_t * | cntl | ||
| ) |
References FLA_Apply_pivots_internal(), FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Determine_blocksize(), FLA_Gemm_internal(), FLA_LU_piv_internal(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_1x2(), FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), FLA_Trsm_external(), and FLA_Trsm_internal().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, A01, A02,
ABL, ABR, A10, A11, A12,
A20, A21, A22;
FLA_Obj AL, AR, A0, A1, A2;
FLA_Obj pT, p0,
pB, p1,
p2;
FLA_Obj AB0, AB1;
dim_t b;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_1x2( A, &AL, &AR, 0, FLA_LEFT );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
b = FLA_Determine_blocksize( ABR, FLA_BR, FLA_Cntl_blocksize( cntl ) );
FLA_Repart_2x2_to_3x3( ATL, /**/ ATR, &A00, /**/ &A01, &A02,
/* ************* */ /* ******************** */
&A10, /**/ &A11, &A12,
ABL, /**/ ABR, &A20, /**/ &A21, &A22,
b, b, FLA_BR );
FLA_Repart_1x2_to_1x3( AL, /**/ AR, &A0, /**/ &A1, &A2,
b, FLA_RIGHT );
FLA_Repart_2x1_to_3x1( pT, &p0,
/* ** */ /* ** */
&p1,
pB, &p2, b, FLA_BOTTOM );
/*------------------------------------------------------------*/
// Apply previously computed pivots
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p0, A1,
FLA_Cntl_sub_appiv1( cntl ) );
// A01 = trilu( A00 ) \ A10
FLA_Trsm_internal( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, A00, A01,
FLA_Cntl_sub_trsm1( cntl ) );
// A11 = A11 - A10 * A01
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A10, A01, FLA_ONE, A11,
FLA_Cntl_sub_gemm1( cntl ) );
// A21 = A21 - A20 * A01
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A20, A01, FLA_ONE, A21,
FLA_Cntl_sub_gemm2( cntl ) );
// AB1 = / A11 \
// \ A21 /
FLA_Merge_2x1( A11,
A21, &AB1 );
// AB1, p1 = LU_piv( AB1 )
FLA_LU_piv_internal( AB1, p1,
FLA_Cntl_sub_lu( cntl ) );
// AB0 = / A10 \
// \ A20 /
FLA_Merge_2x1( A10,
A20, &AB0 );
// Apply pivots to previous columns
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p1, AB0,
FLA_Cntl_sub_appiv2( cntl ) );
/*------------------------------------------------------------*/
FLA_Cont_with_3x3_to_2x2( &ATL, /**/ &ATR, A00, A01, /**/ A02,
A10, A11, /**/ A12,
/* ************** */ /* ****************** */
&ABL, /**/ &ABR, A20, A21, /**/ A22,
FLA_TL );
FLA_Cont_with_1x3_to_1x2( &AL, /**/ &AR, A0, A1, /**/ A2,
FLA_LEFT );
FLA_Cont_with_3x1_to_2x1( &pT, p0,
p1,
/* ** */ /* ** */
&pB, p2, FLA_TOP );
}
if ( FLA_Obj_width( ATR ) > 0 )
{
/* Apply pivots to untouched columns */
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR,
FLA_Cntl_sub_appiv1( cntl ) );
/* ATR = trilu( ATL ) \ ATR */
FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, ATL, ATR );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_blk_var4 | ( | FLA_Obj | A, |
| FLA_Obj | p, | ||
| fla_lu_t * | cntl | ||
| ) |
References FLA_Apply_pivots_internal(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Determine_blocksize(), FLA_Gemm_internal(), FLA_LU_piv_internal(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), and FLA_Trsm_internal().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, A01, A02,
ABL, ABR, A10, A11, A12,
A20, A21, A22;
FLA_Obj pT, p0,
pB, p1,
p2;
FLA_Obj AB0, AB1, AB2;
dim_t b;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
b = FLA_Determine_blocksize( ABR, FLA_BR, FLA_Cntl_blocksize( cntl ) );
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( pT, &p0,
/* ** */ /* ** */
&p1,
pB, &p2, b, FLA_BOTTOM );
/*------------------------------------------------------------*/
// A11 = A11 - A10 * A0
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A10, A01, FLA_ONE, A11,
FLA_Cntl_sub_gemm1( cntl ) );
// A21 = A21 - A20 * A01
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A20, A01, FLA_ONE, A21,
FLA_Cntl_sub_gemm3( cntl ) );
// AB1 = / A11 \
// \ A21 /
FLA_Merge_2x1( A11,
A21, &AB1 );
// AB1, p1 = LU_piv( AB1 )
FLA_LU_piv_internal( AB1, p1,
FLA_Cntl_sub_lu( cntl ) );
// AB0 = / A10 \
// \ A20 /
FLA_Merge_2x1( A10,
A20, &AB0 );
// AB2 = / A12 \
// \ A22 /
FLA_Merge_2x1( A12,
A22, &AB2 );
// Apply pivots to remaining columns
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p1, AB0,
FLA_Cntl_sub_appiv1( cntl ) );
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p1, AB2,
FLA_Cntl_sub_appiv1( cntl ) );
// A12 = A12 - A10 * A02
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A10, A02, FLA_ONE, A12,
FLA_Cntl_sub_gemm2( cntl ) );
// A12 = trilu( A11 ) \ A12
FLA_Trsm_internal( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, A11, A12,
FLA_Cntl_sub_trsm1( cntl ) );
/*------------------------------------------------------------*/
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( &pT, p0,
p1,
/* ** */ /* ** */
&pB, p2, FLA_TOP );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_blk_var5 | ( | FLA_Obj | A, |
| FLA_Obj | p, | ||
| fla_lu_t * | cntl | ||
| ) |
References FLA_Apply_pivots_internal(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Determine_blocksize(), FLA_Gemm_internal(), FLA_LU_piv_internal(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), and FLA_Trsm_internal().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, A01, A02,
ABL, ABR, A10, A11, A12,
A20, A21, A22;
FLA_Obj pT, p0,
pB, p1,
p2;
FLA_Obj AB0, AB1, AB2;
dim_t b;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
b = FLA_Determine_blocksize( ABR, FLA_BR, FLA_Cntl_blocksize( cntl ) );
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( pT, &p0,
/* ** */ /* ** */
&p1,
pB, &p2, b, FLA_BOTTOM );
/*------------------------------------------------------------*/
// AB1 = / A11 \
// \ A21 /
FLA_Merge_2x1( A11,
A21, &AB1 );
// AB1, p1 = LU_piv( AB1 )
FLA_LU_piv_internal( AB1, p1,
FLA_Cntl_sub_lu( cntl ) );
// AB0 = / A10 \
// \ A20 /
FLA_Merge_2x1( A10,
A20, &AB0 );
// Apply computed pivots to AB0
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p1, AB0,
FLA_Cntl_sub_appiv1( cntl ) );
// AB2 = / A12 \
// \ A22 /
FLA_Merge_2x1( A12,
A22, &AB2 );
// Apply computed pivots to AB2
FLA_Apply_pivots_internal( FLA_LEFT, FLA_NO_TRANSPOSE, p1, AB2,
FLA_Cntl_sub_appiv1( cntl ) );
// A12 = trilu( A11 ) \ A12
FLA_Trsm_internal( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, A11, A12,
FLA_Cntl_sub_trsm1( cntl ) );
// A22 = A22 - A21 * A12
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, A21, A12, FLA_ONE, A22,
FLA_Cntl_sub_gemm1( cntl ) );
/*------------------------------------------------------------*/
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( &pT, p0,
p1,
/* ** */ /* ** */
&pB, p2, FLA_TOP );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opc_var3 | ( | int | m_A, |
| int | n_A, | ||
| scomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_camax(), bli_cdots(), bli_cgemv(), bli_cinvscalv(), bli_ctrsm(), bli_ctrsv(), BLIS_LEFT, BLIS_LOWER_TRIANGULAR, BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_UNIT_DIAG, FLA_Apply_pivots_ln_opc_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var3().
{
scomplex* buff_1 = FLA_COMPLEX_PTR( FLA_ONE );
scomplex* buff_m1 = FLA_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
scomplex* A00 = buff_A + (0 )*cs_A + (0 )*rs_A;
scomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
scomplex* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
scomplex* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
scomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
scomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
scomplex* a1 = buff_A + (i )*cs_A + (0 )*rs_A;
int* p0 = buff_p;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p0, a1 );
FLA_Apply_pivots_ln_opc_var1( 1,
a1, rs_A, cs_A,
0,
mn_behind - 1,
p0, inc_p );
// FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
bli_ctrsv( BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
mn_behind,
A00, rs_A, cs_A,
a01, rs_A );
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_cdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_cgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_camax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opc_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_cinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opc_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
/*------------------------------------------------------------*/
}
if ( m_A < n_A )
{
scomplex* ATL = buff_A;
scomplex* ATR = buff_A + m_A*cs_A;
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR );
FLA_Apply_pivots_ln_opc_var1( n_A - m_A,
ATR, rs_A, cs_A,
0,
m_A - 1,
buff_p, inc_p );
// FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
// FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
// FLA_ONE, ATL, ATR );
bli_ctrsm( BLIS_LEFT,
BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
m_A,
n_A - m_A,
buff_1,
ATL, rs_A, cs_A,
ATR, rs_A, cs_A );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opc_var4 | ( | int | m_A, |
| int | n_A, | ||
| scomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_camax(), bli_cdots(), bli_cgemv(), bli_cinvscalv(), BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_TRANSPOSE, FLA_Apply_pivots_ln_opc_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var4().
{
scomplex* buff_1 = FLA_COMPLEX_PTR( FLA_ONE );
scomplex* buff_m1 = FLA_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
scomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
scomplex* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
scomplex* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
scomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
scomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
scomplex* A02 = buff_A + (i+1)*cs_A + (0 )*rs_A;
scomplex* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_cdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_cgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_camax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opc_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opc_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a12t,
// A22, &AB2 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
FLA_Apply_pivots_ln_opc_var1( n_ahead,
a12t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Gemv_external( FLA_TRANSPOSE, FLA_MINUS_ONE, A02, a10t, FLA_ONE, a12t );
bli_cgemv( BLIS_TRANSPOSE,
BLIS_NO_CONJUGATE,
mn_behind,
n_ahead,
buff_m1,
A02, rs_A, cs_A,
a10t, cs_A,
buff_1,
a12t, cs_A );
// FLA_Inv_scal_external( alpha11, a21 );
bli_cinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opc_var5 | ( | int | m_A, |
| int | n_A, | ||
| scomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_camax(), bli_cger(), bli_cinvscalv(), BLIS_NO_CONJUGATE, FLA_Apply_pivots_ln_opc_var1(), and FLA_MINUS_ONE.
Referenced by FLA_LU_piv_opt_var5().
{
scomplex* buff_m1 = FLA_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
scomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
scomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
scomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
scomplex* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
scomplex* A22 = buff_A + (i+1)*cs_A + (i+1)*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
/*------------------------------------------------------------*/
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_camax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Merge_1x2( ABL, ABR, &AB );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB );
FLA_Apply_pivots_ln_opc_var1( n_A,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_cinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Ger_external( FLA_MINUS_ONE, a21, a12t, A22 );
bli_cger( BLIS_NO_CONJUGATE,
BLIS_NO_CONJUGATE,
m_ahead,
n_ahead,
buff_m1,
a21, rs_A,
a12t, cs_A,
A22, rs_A, cs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opd_var3 | ( | int | m_A, |
| int | n_A, | ||
| double * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_damax(), bli_ddots(), bli_dgemv(), bli_dinvscalv(), bli_dtrsm(), bli_dtrsv(), BLIS_LEFT, BLIS_LOWER_TRIANGULAR, BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_UNIT_DIAG, FLA_Apply_pivots_ln_opd_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var3().
{
double* buff_1 = FLA_DOUBLE_PTR( FLA_ONE );
double* buff_m1 = FLA_DOUBLE_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
double* A00 = buff_A + (0 )*cs_A + (0 )*rs_A;
double* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
double* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
double* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
double* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
double* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
double* a1 = buff_A + (i )*cs_A + (0 )*rs_A;
int* p0 = buff_p;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p0, a1 );
FLA_Apply_pivots_ln_opd_var1( 1,
a1, rs_A, cs_A,
0,
mn_behind - 1,
p0, inc_p );
// FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
bli_dtrsv( BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
mn_behind,
A00, rs_A, cs_A,
a01, rs_A );
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_ddots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_dgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_damax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opd_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_dinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opd_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
/*------------------------------------------------------------*/
}
if ( m_A < n_A )
{
double* ATL = buff_A;
double* ATR = buff_A + m_A*cs_A;
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR );
FLA_Apply_pivots_ln_opd_var1( n_A - m_A,
ATR, rs_A, cs_A,
0,
m_A - 1,
buff_p, inc_p );
// FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
// FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
// FLA_ONE, ATL, ATR );
bli_dtrsm( BLIS_LEFT,
BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
m_A,
n_A - m_A,
buff_1,
ATL, rs_A, cs_A,
ATR, rs_A, cs_A );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opd_var4 | ( | int | m_A, |
| int | n_A, | ||
| double * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_damax(), bli_ddots(), bli_dgemv(), bli_dinvscalv(), BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_TRANSPOSE, FLA_Apply_pivots_ln_opd_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var4().
{
double* buff_1 = FLA_DOUBLE_PTR( FLA_ONE );
double* buff_m1 = FLA_DOUBLE_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
double* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
double* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
double* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
double* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
double* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
double* A02 = buff_A + (i+1)*cs_A + (0 )*rs_A;
double* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_ddots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_dgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_damax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opd_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opd_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a12t,
// A22, &AB2 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
FLA_Apply_pivots_ln_opd_var1( n_ahead,
a12t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Gemv_external( FLA_TRANSPOSE, FLA_MINUS_ONE, A02, a10t, FLA_ONE, a12t );
bli_dgemv( BLIS_TRANSPOSE,
BLIS_NO_CONJUGATE,
mn_behind,
n_ahead,
buff_m1,
A02, rs_A, cs_A,
a10t, cs_A,
buff_1,
a12t, cs_A );
// FLA_Inv_scal_external( alpha11, a21 );
bli_dinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opd_var5 | ( | int | m_A, |
| int | n_A, | ||
| double * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_damax(), bli_dger(), bli_dinvscalv(), BLIS_NO_CONJUGATE, FLA_Apply_pivots_ln_opd_var1(), and FLA_MINUS_ONE.
Referenced by FLA_LU_piv_opt_var5().
{
double* buff_m1 = FLA_DOUBLE_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
double* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
double* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
double* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
double* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
double* A22 = buff_A + (i+1)*cs_A + (i+1)*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
/*------------------------------------------------------------*/
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_damax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Merge_1x2( ABL, ABR, &AB );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB );
FLA_Apply_pivots_ln_opd_var1( n_A,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_dinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Ger_external( FLA_MINUS_ONE, a21, a12t, A22 );
bli_dger( BLIS_NO_CONJUGATE,
BLIS_NO_CONJUGATE,
m_ahead,
n_ahead,
buff_m1,
a21, rs_A,
a12t, cs_A,
A22, rs_A, cs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_ops_var3 | ( | int | m_A, |
| int | n_A, | ||
| float * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_samax(), bli_sdots(), bli_sgemv(), bli_sinvscalv(), bli_strsm(), bli_strsv(), BLIS_LEFT, BLIS_LOWER_TRIANGULAR, BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_UNIT_DIAG, FLA_Apply_pivots_ln_ops_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var3().
{
float* buff_1 = FLA_FLOAT_PTR( FLA_ONE );
float* buff_m1 = FLA_FLOAT_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
float* A00 = buff_A + (0 )*cs_A + (0 )*rs_A;
float* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
float* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
float* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
float* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
float* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
float* a1 = buff_A + (i )*cs_A + (0 )*rs_A;
int* p0 = buff_p;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p0, a1 );
FLA_Apply_pivots_ln_ops_var1( 1,
a1, rs_A, cs_A,
0,
mn_behind - 1,
p0, inc_p );
// FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
bli_strsv( BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
mn_behind,
A00, rs_A, cs_A,
a01, rs_A );
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_sdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_sgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_samax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_ops_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_sinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_ops_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
/*------------------------------------------------------------*/
}
if ( m_A < n_A )
{
float* ATL = buff_A;
float* ATR = buff_A + m_A*cs_A;
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR );
FLA_Apply_pivots_ln_ops_var1( n_A - m_A,
ATR, rs_A, cs_A,
0,
m_A - 1,
buff_p, inc_p );
// FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
// FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
// FLA_ONE, ATL, ATR );
bli_strsm( BLIS_LEFT,
BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
m_A,
n_A - m_A,
buff_1,
ATL, rs_A, cs_A,
ATR, rs_A, cs_A );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_ops_var4 | ( | int | m_A, |
| int | n_A, | ||
| float * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_samax(), bli_sdots(), bli_sgemv(), bli_sinvscalv(), BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_TRANSPOSE, FLA_Apply_pivots_ln_ops_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var4().
{
float* buff_1 = FLA_FLOAT_PTR( FLA_ONE );
float* buff_m1 = FLA_FLOAT_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
float* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
float* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
float* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
float* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
float* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
float* A02 = buff_A + (i+1)*cs_A + (0 )*rs_A;
float* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_sdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_sgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_samax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_ops_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_ops_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a12t,
// A22, &AB2 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
FLA_Apply_pivots_ln_ops_var1( n_ahead,
a12t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Gemv_external( FLA_TRANSPOSE, FLA_MINUS_ONE, A02, a10t, FLA_ONE, a12t );
bli_sgemv( BLIS_TRANSPOSE,
BLIS_NO_CONJUGATE,
mn_behind,
n_ahead,
buff_m1,
A02, rs_A, cs_A,
a10t, cs_A,
buff_1,
a12t, cs_A );
// FLA_Inv_scal_external( alpha11, a21 );
bli_sinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_ops_var5 | ( | int | m_A, |
| int | n_A, | ||
| float * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_samax(), bli_sger(), bli_sinvscalv(), BLIS_NO_CONJUGATE, FLA_Apply_pivots_ln_ops_var1(), and FLA_MINUS_ONE.
Referenced by FLA_LU_piv_opt_var5().
{
float* buff_m1 = FLA_FLOAT_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
float* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
float* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
float* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
float* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
float* A22 = buff_A + (i+1)*cs_A + (i+1)*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
/*------------------------------------------------------------*/
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_samax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Merge_1x2( ABL, ABR, &AB );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB );
FLA_Apply_pivots_ln_ops_var1( n_A,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_sinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Ger_external( FLA_MINUS_ONE, a21, a12t, A22 );
bli_sger( BLIS_NO_CONJUGATE,
BLIS_NO_CONJUGATE,
m_ahead,
n_ahead,
buff_m1,
a21, rs_A,
a12t, cs_A,
A22, rs_A, cs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opt_var3 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_LU_piv_opc_var3(), FLA_LU_piv_opd_var3(), FLA_LU_piv_ops_var3(), FLA_LU_piv_opz_var3(), FLA_Obj_col_stride(), FLA_Obj_datatype(), FLA_Obj_length(), FLA_Obj_row_stride(), FLA_Obj_vector_inc(), and FLA_Obj_width().
Referenced by FLA_LU_piv_internal().
{
FLA_Datatype datatype;
int m_A, n_A;
int rs_A, cs_A;
int inc_p;
datatype = FLA_Obj_datatype( A );
m_A = FLA_Obj_length( A );
n_A = FLA_Obj_width( A );
rs_A = FLA_Obj_row_stride( A );
cs_A = FLA_Obj_col_stride( A );
inc_p = FLA_Obj_vector_inc( p );
switch ( datatype )
{
case FLA_FLOAT:
{
float* buff_A = FLA_FLOAT_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_ops_var3( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE:
{
double* buff_A = FLA_DOUBLE_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opd_var3( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_COMPLEX:
{
scomplex* buff_A = FLA_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opc_var3( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE_COMPLEX:
{
dcomplex* buff_A = FLA_DOUBLE_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opz_var3( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opt_var4 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_LU_piv_opc_var4(), FLA_LU_piv_opd_var4(), FLA_LU_piv_ops_var4(), FLA_LU_piv_opz_var4(), FLA_Obj_col_stride(), FLA_Obj_datatype(), FLA_Obj_length(), FLA_Obj_row_stride(), FLA_Obj_vector_inc(), and FLA_Obj_width().
Referenced by FLA_LU_piv_internal().
{
FLA_Datatype datatype;
int m_A, n_A;
int rs_A, cs_A;
int inc_p;
datatype = FLA_Obj_datatype( A );
m_A = FLA_Obj_length( A );
n_A = FLA_Obj_width( A );
rs_A = FLA_Obj_row_stride( A );
cs_A = FLA_Obj_col_stride( A );
inc_p = FLA_Obj_vector_inc( p );
switch ( datatype )
{
case FLA_FLOAT:
{
float* buff_A = FLA_FLOAT_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_ops_var4( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE:
{
double* buff_A = FLA_DOUBLE_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opd_var4( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_COMPLEX:
{
scomplex* buff_A = FLA_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opc_var4( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE_COMPLEX:
{
dcomplex* buff_A = FLA_DOUBLE_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opz_var4( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opt_var5 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_LU_piv_opc_var5(), FLA_LU_piv_opd_var5(), FLA_LU_piv_ops_var5(), FLA_LU_piv_opz_var5(), FLA_Obj_col_stride(), FLA_Obj_datatype(), FLA_Obj_length(), FLA_Obj_row_stride(), FLA_Obj_vector_inc(), and FLA_Obj_width().
Referenced by FLA_LU_piv_internal().
{
FLA_Datatype datatype;
int m_A, n_A;
int rs_A, cs_A;
int inc_p;
datatype = FLA_Obj_datatype( A );
m_A = FLA_Obj_length( A );
n_A = FLA_Obj_width( A );
rs_A = FLA_Obj_row_stride( A );
cs_A = FLA_Obj_col_stride( A );
inc_p = FLA_Obj_vector_inc( p );
switch ( datatype )
{
case FLA_FLOAT:
{
float* buff_A = FLA_FLOAT_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_ops_var5( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE:
{
double* buff_A = FLA_DOUBLE_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opd_var5( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_COMPLEX:
{
scomplex* buff_A = FLA_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opc_var5( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
case FLA_DOUBLE_COMPLEX:
{
dcomplex* buff_A = FLA_DOUBLE_COMPLEX_PTR( A );
int* buff_p = FLA_INT_PTR( p );
FLA_LU_piv_opz_var5( m_A,
n_A,
buff_A, rs_A, cs_A,
buff_p, inc_p );
break;
}
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opz_var3 | ( | int | m_A, |
| int | n_A, | ||
| dcomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_zamax(), bli_zdots(), bli_zgemv(), bli_zinvscalv(), bli_ztrsm(), bli_ztrsv(), BLIS_LEFT, BLIS_LOWER_TRIANGULAR, BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_UNIT_DIAG, FLA_Apply_pivots_ln_opz_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var3().
{
dcomplex* buff_1 = FLA_DOUBLE_COMPLEX_PTR( FLA_ONE );
dcomplex* buff_m1 = FLA_DOUBLE_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
dcomplex* A00 = buff_A + (0 )*cs_A + (0 )*rs_A;
dcomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
dcomplex* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
dcomplex* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
dcomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
dcomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
dcomplex* a1 = buff_A + (i )*cs_A + (0 )*rs_A;
int* p0 = buff_p;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p0, a1 );
FLA_Apply_pivots_ln_opz_var1( 1,
a1, rs_A, cs_A,
0,
mn_behind - 1,
p0, inc_p );
// FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
bli_ztrsv( BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
mn_behind,
A00, rs_A, cs_A,
a01, rs_A );
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_zdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_zgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_zamax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opz_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_zinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opz_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
/*------------------------------------------------------------*/
}
if ( m_A < n_A )
{
dcomplex* ATL = buff_A;
dcomplex* ATR = buff_A + m_A*cs_A;
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR );
FLA_Apply_pivots_ln_opz_var1( n_A - m_A,
ATR, rs_A, cs_A,
0,
m_A - 1,
buff_p, inc_p );
// FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
// FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
// FLA_ONE, ATL, ATR );
bli_ztrsm( BLIS_LEFT,
BLIS_LOWER_TRIANGULAR,
BLIS_NO_TRANSPOSE,
BLIS_UNIT_DIAG,
m_A,
n_A - m_A,
buff_1,
ATL, rs_A, cs_A,
ATR, rs_A, cs_A );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opz_var4 | ( | int | m_A, |
| int | n_A, | ||
| dcomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_zamax(), bli_zdots(), bli_zgemv(), bli_zinvscalv(), BLIS_NO_CONJUGATE, BLIS_NO_TRANSPOSE, BLIS_TRANSPOSE, FLA_Apply_pivots_ln_opz_var1(), FLA_MINUS_ONE, and FLA_ONE.
Referenced by FLA_LU_piv_opt_var4().
{
dcomplex* buff_1 = FLA_DOUBLE_COMPLEX_PTR( FLA_ONE );
dcomplex* buff_m1 = FLA_DOUBLE_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
dcomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
dcomplex* A20 = buff_A + (0 )*cs_A + (i+1)*rs_A;
dcomplex* a01 = buff_A + (i )*cs_A + (0 )*rs_A;
dcomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
dcomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
dcomplex* A02 = buff_A + (i+1)*cs_A + (0 )*rs_A;
dcomplex* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
int mn_behind = i;
/*------------------------------------------------------------*/
// FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
bli_zdots( BLIS_NO_CONJUGATE,
mn_behind,
buff_m1,
a10t, cs_A,
a01, rs_A,
buff_1,
alpha11 );
// FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
bli_zgemv( BLIS_NO_TRANSPOSE,
BLIS_NO_CONJUGATE,
m_ahead,
mn_behind,
buff_m1,
A20, rs_A, cs_A,
a01, rs_A,
buff_1,
a21, rs_A );
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_zamax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
FLA_Apply_pivots_ln_opz_var1( 1,
alpha11, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a10t,
// A20, &AB0 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots_ln_opz_var1( mn_behind,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Merge_2x1( a12t,
// A22, &AB2 );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
FLA_Apply_pivots_ln_opz_var1( n_ahead,
a12t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Gemv_external( FLA_TRANSPOSE, FLA_MINUS_ONE, A02, a10t, FLA_ONE, a12t );
bli_zgemv( BLIS_TRANSPOSE,
BLIS_NO_CONJUGATE,
mn_behind,
n_ahead,
buff_m1,
A02, rs_A, cs_A,
a10t, cs_A,
buff_1,
a12t, cs_A );
// FLA_Inv_scal_external( alpha11, a21 );
bli_zinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_opz_var5 | ( | int | m_A, |
| int | n_A, | ||
| dcomplex * | buff_A, | ||
| int | rs_A, | ||
| int | cs_A, | ||
| int * | buff_p, | ||
| int | inc_p | ||
| ) |
References bli_zamax(), bli_zger(), bli_zinvscalv(), BLIS_NO_CONJUGATE, FLA_Apply_pivots_ln_opz_var1(), and FLA_MINUS_ONE.
Referenced by FLA_LU_piv_opt_var5().
{
dcomplex* buff_m1 = FLA_DOUBLE_COMPLEX_PTR( FLA_MINUS_ONE );
int min_m_n = min( m_A, n_A );
int i;
for ( i = 0; i < min_m_n; ++i )
{
dcomplex* a10t = buff_A + (0 )*cs_A + (i )*rs_A;
dcomplex* alpha11 = buff_A + (i )*cs_A + (i )*rs_A;
dcomplex* a21 = buff_A + (i )*cs_A + (i+1)*rs_A;
dcomplex* a12t = buff_A + (i+1)*cs_A + (i )*rs_A;
dcomplex* A22 = buff_A + (i+1)*cs_A + (i+1)*rs_A;
int* pi1 = buff_p + i*inc_p;
int m_ahead = m_A - i - 1;
int n_ahead = n_A - i - 1;
/*------------------------------------------------------------*/
// FLA_Merge_2x1( alpha11,
// a21, &aB1 );
// FLA_Amax_external( aB1, pi1 );
bli_zamax( m_ahead + 1,
alpha11, rs_A,
pi1 );
// FLA_Merge_1x2( ABL, ABR, &AB );
// FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB );
FLA_Apply_pivots_ln_opz_var1( n_A,
a10t, rs_A, cs_A,
0,
0,
pi1, inc_p );
// FLA_Inv_scal_external( alpha11, a21 );
bli_zinvscalv( BLIS_NO_CONJUGATE,
m_ahead,
alpha11,
a21, rs_A );
// FLA_Ger_external( FLA_MINUS_ONE, a21, a12t, A22 );
bli_zger( BLIS_NO_CONJUGATE,
BLIS_NO_CONJUGATE,
m_ahead,
n_ahead,
buff_m1,
a21, rs_A,
a12t, cs_A,
A22, rs_A, cs_A );
/*------------------------------------------------------------*/
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_unb_var3 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_Amax_external(), FLA_Apply_pivots(), FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Dots_external(), FLA_Gemv_external(), FLA_Inv_scal_external(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_1x2(), FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), FLA_Trsm_external(), and FLA_Trsv_external().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, a01, A02,
ABL, ABR, a10t, alpha11, a12t,
A20, a21, A22;
FLA_Obj AL, AR, A0, a1, A2;
FLA_Obj pT, p0,
pB, pi1,
p2;
FLA_Obj AB0, aB1;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_1x2( A, &AL, &AR, 0, FLA_LEFT );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
FLA_Repart_2x2_to_3x3( ATL, /**/ ATR, &A00, /**/ &a01, &A02,
/* ************* */ /* ************************** */
&a10t, /**/ &alpha11, &a12t,
ABL, /**/ ABR, &A20, /**/ &a21, &A22,
1, 1, FLA_BR );
FLA_Repart_1x2_to_1x3( AL, /**/ AR, &A0, /**/ &a1, &A2,
1, FLA_RIGHT );
FLA_Repart_2x1_to_3x1( pT, &p0,
/* ** */ /* *** */
&pi1,
pB, &p2, 1, FLA_BOTTOM );
/*------------------------------------------------------------*/
// Apply previously computed pivots
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p0, a1 );
// a01 = trilu( A00 ) \ a01
FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
// alpha11 = alpha11 - a10t * a01
FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
// a21 = a21 - A20 * a01
FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
// aB1 = / alpha11 \
// \ a21 /
FLA_Merge_2x1( alpha11,
a21, &aB1 );
// Determine pivot index
FLA_Amax_external( aB1, pi1 );
// Apply pivots to current column
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
// a21 = a21 / alpha11
FLA_Inv_scal_external( alpha11, a21 );
// AB0 = / a10t \
// \ A20 /
FLA_Merge_2x1( a10t,
A20, &AB0 );
// Apply pivots to previous columns
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
/*------------------------------------------------------------*/
FLA_Cont_with_3x3_to_2x2( &ATL, /**/ &ATR, A00, a01, /**/ A02,
a10t, alpha11, /**/ a12t,
/* ************** */ /* ************************ */
&ABL, /**/ &ABR, A20, a21, /**/ A22,
FLA_TL );
FLA_Cont_with_1x3_to_1x2( &AL, /**/ &AR, A0, a1, /**/ A2,
FLA_LEFT );
FLA_Cont_with_3x1_to_2x1( &pT, p0,
pi1,
/* ** */ /* *** */
&pB, p2, FLA_TOP );
}
if ( FLA_Obj_width( ATR ) > 0 )
{
/* Apply pivots to untouched columns */
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, p, ATR );
/* ATR = trilu( ATL ) \ ATR */
FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, ATL, ATR );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_unb_var3b | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_Amax_external(), FLA_Apply_pivots(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Dots_external(), FLA_Gemv_external(), FLA_Inv_scal_external(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), FLA_Repart_2x2_to_3x3(), FLA_Trsm_external(), and FLA_Trsv_external().
{
FLA_Obj ATL, ATR, A00, a01, A02,
ABL, ABR, a10t, alpha11, a12t,
A20, a21, A22;
FLA_Obj pT, p0,
pB, pi1,
p2;
FLA_Obj AB0, aB1, AB2;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
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( pT, &p0,
/* ** */ /* *** */
&pi1,
pB, &p2, 1, FLA_BOTTOM );
/*------------------------------------------------------------*/
// a01 = trilu( A00 ) \ a01
FLA_Trsv_external( FLA_LOWER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_UNIT_DIAG, A00, a01 );
// alpha11 = alpha11 - a10t * a01
FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
// a21 = a21 - A20 * a01
FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
// aB1 = / alpha11 \
// \ a21 /
FLA_Merge_2x1( alpha11,
a21, &aB1 );
// Determine pivot index
FLA_Amax_external( aB1, pi1 );
// Apply pivots to current column
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
// AB0 = / a10t \
// \ A20 /
FLA_Merge_2x1( a10t,
A20, &AB0 );
// AB2 = / a12t \
// \ A22 /
FLA_Merge_2x1( a12t,
A22, &AB2 );
// Apply pivots to remaining columns
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
// a21 = a21 / alpha11
FLA_Inv_scal_external( alpha11, a21 );
/*------------------------------------------------------------*/
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( &pT, p0,
pi1,
/* ** */ /* *** */
&pB, p2, FLA_TOP );
}
if ( FLA_Obj_width( ATR ) > 0 )
{
/* ATR = trilu( ATL ) \ ATR */
FLA_Trsm_external( FLA_LEFT, FLA_LOWER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_UNIT_DIAG,
FLA_ONE, ATL, ATR );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_unb_var4 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_Amax_external(), FLA_Apply_pivots(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Dots_external(), FLA_Gemv_external(), FLA_Inv_scal_external(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_ONE, FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), and FLA_Repart_2x2_to_3x3().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, a01, A02,
ABL, ABR, a10t, alpha11, a12t,
A20, a21, A22;
FLA_Obj pT, p0,
pB, pi1,
p2;
FLA_Obj AB0, aB1, AB2;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
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( pT, &p0,
/* ** */ /* *** */
&pi1,
pB, &p2, 1, FLA_BOTTOM );
/*------------------------------------------------------------*/
// alpha11 = alpha11 - a10t * a01
FLA_Dots_external( FLA_MINUS_ONE, a10t, a01, FLA_ONE, alpha11 );
// a21 = a21 - A20 * a01
FLA_Gemv_external( FLA_NO_TRANSPOSE, FLA_MINUS_ONE, A20, a01, FLA_ONE, a21 );
// aB1 = / alpha11 \
// \ a21 /
FLA_Merge_2x1( alpha11,
a21, &aB1 );
// Determine pivot index
FLA_Amax_external( aB1, pi1 );
// Apply pivots to current column
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
// AB0 = / a10t \
// \ A20 /
FLA_Merge_2x1( a10t,
A20, &AB0 );
// AB2 = / a12t \
// \ A22 /
FLA_Merge_2x1( a12t,
A22, &AB2 );
// Apply pivots to remaining columns
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
// a21 = a21 / alpha11
FLA_Inv_scal_external( alpha11, a21 );
// a12t = a12t - a10t * A02
FLA_Gemv_external( FLA_TRANSPOSE, FLA_MINUS_ONE, A02, a10t, FLA_ONE, a12t );
/*------------------------------------------------------------*/
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( &pT, p0,
pi1,
/* ** */ /* *** */
&pB, p2, FLA_TOP );
}
return FLA_SUCCESS;
}
| FLA_Error FLA_LU_piv_unb_var5 | ( | FLA_Obj | A, |
| FLA_Obj | p | ||
| ) |
References FLA_Amax_external(), FLA_Apply_pivots(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Ger_external(), FLA_Inv_scal_external(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_width(), FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_2x1_to_3x1(), and FLA_Repart_2x2_to_3x3().
Referenced by FLA_LU_piv_internal().
{
FLA_Obj ATL, ATR, A00, a01, A02,
ABL, ABR, a10t, alpha11, a12t,
A20, a21, A22;
FLA_Obj pT, p0,
pB, pi1,
p2;
FLA_Obj AB0, aB1, AB2;
FLA_Part_2x2( A, &ATL, &ATR,
&ABL, &ABR, 0, 0, FLA_TL );
FLA_Part_2x1( p, &pT,
&pB, 0, FLA_TOP );
while ( FLA_Obj_length( ATL ) < FLA_Obj_length( A ) &&
FLA_Obj_width( ATL ) < FLA_Obj_width( A )){
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( pT, &p0,
/* ** */ /* *** */
&pi1,
pB, &p2, 1, FLA_BOTTOM );
/*------------------------------------------------------------*/
// aB1 = / alpha11 \
// \ a21 /
FLA_Merge_2x1( alpha11,
a21, &aB1 );
// Determine pivot index
FLA_Amax_external( aB1, pi1 );
// Apply pivots to current column
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, aB1 );
// a21 = a21 / alpha11
FLA_Inv_scal_external( alpha11, a21 );
// AB0 = / a10t \
// \ A20 /
FLA_Merge_2x1( a10t,
A20, &AB0 );
// Apply pivots to previous columns
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB0 );
// AB2 = / a12t \
// \ A22 /
FLA_Merge_2x1( a12t,
A22, &AB2 );
// Apply pivots to remaining columns
FLA_Apply_pivots( FLA_LEFT, FLA_NO_TRANSPOSE, pi1, AB2 );
// A22 = A22 - a21 * a12t
FLA_Ger_external( FLA_MINUS_ONE, a21, a12t, A22 );
/*------------------------------------------------------------*/
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( &pT, p0,
pi1,
/* ** */ /* *** */
&pB, p2, FLA_TOP );
}
return FLA_SUCCESS;
}
1.7.6.1