|
libflame
revision_anchor
|
Functions | |
| FLA_Error | FLA_Apply_CAQ2_UT_lhfc_blk_var1 (FLA_Obj D, FLA_Obj T, FLA_Obj W1, FLA_Obj C, FLA_Obj E, fla_apcaq2ut_t *cntl) |
| FLA_Error FLA_Apply_CAQ2_UT_lhfc_blk_var1 | ( | FLA_Obj | D, |
| FLA_Obj | T, | ||
| FLA_Obj | W1, | ||
| FLA_Obj | C, | ||
| FLA_Obj | E, | ||
| fla_apcaq2ut_t * | cntl | ||
| ) |
References FLA_Axpy_internal(), FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x1_to_2x1(), FLA_Cont_with_3x3_to_2x2(), FLA_Copy_internal(), FLA_Gemm_internal(), 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_Trmm_internal(), and FLA_Trsm_internal().
Referenced by FLA_Apply_CAQ2_UT_lhfc().
{
FLA_Obj DTL, DTR, D00, D01, D02,
DBL, DBR, D10, D11, D12,
D20, D21, D22;
FLA_Obj TL, TR, T0, T1, T2;
FLA_Obj T1T,
T2B;
FLA_Obj CT, C0,
CB, C1,
C2;
FLA_Obj ET, E0,
EB, E1,
E2;
FLA_Obj W1TL, W1TR,
W1BL, W1BR;
dim_t b_alg, b;
// Query the algorithmic blocksize by inspecting the length of T.
b_alg = FLA_Obj_length( T );
FLA_Part_2x2( D, &DTL, &DTR,
&DBL, &DBR, 0, 0, FLA_TL );
FLA_Part_1x2( T, &TL, &TR, 0, FLA_LEFT );
FLA_Part_2x1( C, &CT,
&CB, 0, FLA_TOP );
FLA_Part_2x1( E, &ET,
&EB, 0, FLA_TOP );
while ( FLA_Obj_width( DBR ) > 0 ){
b = min( b_alg, FLA_Obj_width( DBR ) );
FLA_Repart_2x2_to_3x3( DTL, /**/ DTR, &D00, /**/ &D01, &D02,
/* ************* */ /* ******************** */
&D10, /**/ &D11, &D12,
DBL, /**/ DBR, &D20, /**/ &D21, &D22,
b, b, FLA_BR );
FLA_Repart_1x2_to_1x3( TL, /**/ TR, &T0, /**/ &T1, &T2,
b, FLA_RIGHT );
FLA_Repart_2x1_to_3x1( CT, &C0,
/* ** */ /* ** */
&C1,
CB, &C2, b, FLA_BOTTOM );
FLA_Repart_2x1_to_3x1( ET, &E0,
/* ** */ /* ** */
&E1,
EB, &E2, b, FLA_BOTTOM );
/*------------------------------------------------------------*/
FLA_Part_2x1( T1, &T1T,
&T2B, b, FLA_TOP );
FLA_Part_2x2( W1, &W1TL, &W1TR,
&W1BL, &W1BR, b, FLA_Obj_width( C1 ), FLA_TL );
// W1TL = inv( triu( T1T ) )' * ( C1 + D1' * E );
// = inv( triu( T1T ) )' * ( C1 + D01' * E0 + D11' * E1 );
FLA_Copy_internal( E1, W1TL,
FLA_Cntl_sub_copy( cntl ) );
FLA_Trmm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR,
FLA_CONJ_TRANSPOSE, FLA_NONUNIT_DIAG,
FLA_ONE, D11, W1TL,
FLA_Cntl_sub_trmm1( cntl ) );
FLA_Gemm_internal( FLA_CONJ_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_ONE, D01, E0, FLA_ONE, W1TL,
FLA_Cntl_sub_gemm1( cntl ) );
FLA_Axpy_internal( FLA_ONE, C1, W1TL,
FLA_Cntl_sub_axpy1( cntl ) );
FLA_Trsm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR,
FLA_CONJ_TRANSPOSE, FLA_NONUNIT_DIAG,
FLA_ONE, T1T, W1TL,
FLA_Cntl_sub_trsm( cntl ) );
// C1 = C1 - W1TL;
// E = E - D1 * W1TL;
// => E0 = E0 - D01 * W1TL;
// E1 = E1 - D11 * W1TL;
FLA_Axpy_internal( FLA_MINUS_ONE, W1TL, C1,
FLA_Cntl_sub_axpy2( cntl ) );
FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE,
FLA_MINUS_ONE, D01, W1TL, FLA_ONE, E0,
FLA_Cntl_sub_gemm2( cntl ) );
FLA_Trmm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR,
FLA_NO_TRANSPOSE, FLA_NONUNIT_DIAG,
FLA_ONE, D11, W1TL,
FLA_Cntl_sub_trmm2( cntl ) );
FLA_Axpy_internal( FLA_MINUS_ONE, W1TL, E1,
FLA_Cntl_sub_axpy3( cntl ) );
/*------------------------------------------------------------*/
FLA_Cont_with_3x3_to_2x2( &DTL, /**/ &DTR, D00, D01, /**/ D02,
D10, D11, /**/ D12,
/* ************** */ /* ****************** */
&DBL, /**/ &DBR, D20, D21, /**/ D22,
FLA_TL );
FLA_Cont_with_1x3_to_1x2( &TL, /**/ &TR, T0, T1, /**/ T2,
FLA_LEFT );
FLA_Cont_with_3x1_to_2x1( &CT, C0,
C1,
/* ** */ /* ** */
&CB, C2, FLA_TOP );
FLA_Cont_with_3x1_to_2x1( &ET, E0,
E1,
/* ** */ /* ** */
&EB, E2, FLA_TOP );
}
return FLA_SUCCESS;
}
1.7.6.1