libflame
revision_anchor
|
Functions | |
FLA_Error | FLA_CAQR2_UT_blk_var1 (FLA_Obj U, FLA_Obj D, FLA_Obj T, fla_caqr2ut_t *cntl) |
FLA_Error FLA_CAQR2_UT_blk_var1 | ( | FLA_Obj | U, |
FLA_Obj | D, | ||
FLA_Obj | T, | ||
fla_caqr2ut_t * | cntl | ||
) |
References FLA_Axpy_internal(), FLA_CAQR2_UT_internal(), FLA_Cont_with_1x3_to_1x2(), FLA_Cont_with_3x3_to_2x2(), FLA_Copy_internal(), FLA_Gemm_internal(), FLA_Merge_2x1(), FLA_MINUS_ONE, FLA_Obj_length(), FLA_Obj_min_dim(), FLA_Obj_width(), FLA_ONE, FLA_Part_1x2(), FLA_Part_2x1(), FLA_Part_2x2(), FLA_Repart_1x2_to_1x3(), FLA_Repart_2x2_to_3x3(), FLA_Trmm_internal(), and FLA_Trsm_internal().
Referenced by FLA_CAQR2_UT_internal().
{ FLA_Obj UTL, UTR, U00, U01, U02, UBL, UBR, U10, U11, U12, U20, U21, U22; FLA_Obj DTL, DTR, D00, D01, D02, DBL, DBR, D10, D11, D12, D20, D21, D22; FLA_Obj TL, TR, T0, T1, W12; FLA_Obj D1; FLA_Obj W12T, W12B; FLA_Obj T1T, T2B; dim_t b_alg, b; dim_t m_DT; // Query the algorithmic blocksize by inspecting the length of T. b_alg = FLA_Obj_length( T ); // Begin partitioning diagonally through D with m - n rows above // the diagonal. m_DT = FLA_Obj_length( D ) - FLA_Obj_width( D ); FLA_Part_2x2( U, &UTL, &UTR, &UBL, &UBR, 0, 0, FLA_TL ); FLA_Part_2x2( D, &DTL, &DTR, &DBL, &DBR, m_DT, 0, FLA_TL ); FLA_Part_1x2( T, &TL, &TR, 0, FLA_LEFT ); while ( FLA_Obj_min_dim( UBR ) > 0 ){ b = min( b_alg, FLA_Obj_min_dim( UBR ) ); FLA_Repart_2x2_to_3x3( UTL, /**/ UTR, &U00, /**/ &U01, &U02, /* ************* */ /* ******************** */ &U10, /**/ &U11, &U12, UBL, /**/ UBR, &U20, /**/ &U21, &U22, b, b, FLA_BR ); 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, &W12, b, FLA_RIGHT ); /*------------------------------------------------------------*/ // T1T = FLA_Top_part( T1, b ); FLA_Part_2x1( T1, &T1T, &T2B, b, FLA_TOP ); FLA_Merge_2x1( D01, D11, &D1 ); // [ U11, ... // D1, T1 ] = FLA_CAQR2_UT( U11 // D1, T1T ); FLA_CAQR2_UT_internal( U11, D1, T1T, FLA_Cntl_sub_caqr2ut( cntl ) ); if ( FLA_Obj_width( U12 ) > 0 ) { // W12T = FLA_Top_part( W12, b ); FLA_Part_2x1( W12, &W12T, &W12B, b, FLA_TOP ); // W12T = inv( triu( T1T ) )' * ( U12 + D1' * D2 ); // = inv( triu( T1T ) )' * ( U12 + D01' * D02 + D11' * D12 ); FLA_Copy_internal( D12, W12T, FLA_Cntl_sub_copy( cntl ) ); FLA_Trmm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR, FLA_CONJ_TRANSPOSE, FLA_NONUNIT_DIAG, FLA_ONE, D11, W12T, FLA_Cntl_sub_trmm1( cntl ) ); FLA_Gemm_internal( FLA_CONJ_TRANSPOSE, FLA_NO_TRANSPOSE, FLA_ONE, D01, D02, FLA_ONE, W12T, FLA_Cntl_sub_gemm1( cntl ) ); FLA_Axpy_internal( FLA_ONE, U12, W12T, FLA_Cntl_sub_axpy1( cntl ) ); FLA_Trsm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR, FLA_CONJ_TRANSPOSE, FLA_NONUNIT_DIAG, FLA_ONE, T1T, W12T, FLA_Cntl_sub_trsm( cntl ) ); // U12 = U12 - W12T; // D2 = D2 - D1 * W12T; // => D02 = D02 - D01 * W12T; // D12 = D12 - D11 * W12T; FLA_Axpy_internal( FLA_MINUS_ONE, W12T, U12, FLA_Cntl_sub_axpy2( cntl ) ); FLA_Gemm_internal( FLA_NO_TRANSPOSE, FLA_NO_TRANSPOSE, FLA_MINUS_ONE, D01, W12T, FLA_ONE, D02, FLA_Cntl_sub_gemm2( cntl ) ); FLA_Trmm_internal( FLA_LEFT, FLA_UPPER_TRIANGULAR, FLA_NO_TRANSPOSE, FLA_NONUNIT_DIAG, FLA_ONE, D11, W12T, FLA_Cntl_sub_trmm2( cntl ) ); FLA_Axpy_internal( FLA_MINUS_ONE, W12T, D12, FLA_Cntl_sub_axpy3( cntl ) ); } /*------------------------------------------------------------*/ FLA_Cont_with_3x3_to_2x2( &UTL, /**/ &UTR, U00, U01, /**/ U02, U10, U11, /**/ U12, /* ************** */ /* ****************** */ &UBL, /**/ &UBR, U20, U21, /**/ U22, FLA_TL ); 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, /**/ W12, FLA_LEFT ); } return FLA_SUCCESS; }