Commit cfeceee1 authored by Fabian Bauer's avatar Fabian Bauer
Browse files

synched FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic with be branch, changed func1

parent 4cf3a559

lib_com/swb_tbe_com_fx.c

+37 −90
Original line number Diff line number Diff line
@@ -6697,9 +6697,14 @@ static inline Word64 wmac_3216( Word64 x1, Word32 x2, Word16 x3 ) 
    return W_mac_32_16( x1, x2, x3 );

}



static inline Word64 wmsu_3216( Word64 x1, Word32 x2, Word16 x3 )

static inline Word64 finalSat16( Word64 W_tmpx, Word64 W_tmpy )

{

    return W_msu_32_16( x1, x2, x3 );

    return W_sat_l( W_shr( W_add( W_tmpX, W_shl( W_tmpY, 2 - 16 + 3 ) ), 3 ) ); /*Q_input_fx + 11*/

}



static inline Word64 finalSat32( Word64 W_tmpx, Word64 W_tmpy )

{

    return W_sat_l( W_shr( W_add( W_tmpX, W_shl( W_tmpY, 2 - 16 + 3 + 16 ) ), 3 + 16 ) );

}



inline static void elliptic_bpf_48k_generic_func1( Word16 *input16_fx, Word32 *input32_fx, Word32 *L_tmp, const Word16 full_band_bpf_fx[][5], Word16 IsUpsampled3, Word32 *L_tmpMax )
@@ -6713,12 +6718,16 @@ inline static void elliptic_bpf_48k_generic_func1( Word16 *input16_fx, Word32 *i 
    if ( input16_fx > 0 )

    {

        Word64 ( *wmac )( Word64, Word16, Word16 );

        Word64 ( *wmsu )( Word64, Word32, Word16 );

        Word64 ( *finalSat )( Word64, Word64 );

        wmac = wmac_1616;

        finalSat = finalSat16;

    }

    if ( input32_fx > 0 )

    {

        Word64 ( *wmac )( Word64, Word32, Word16 );

        Word64 ( *wmsu )( Word64, Word32, Word16 );

        Word64 ( *finalSat )( Word64, Word64 );

        wmac = wmac_3216;

        finalSat = finalSat32;

    }



    IF( !IsUpsampled3 )
@@ -6726,16 +6735,16 @@ inline static void elliptic_bpf_48k_generic_func1( Word16 *input16_fx, Word32 *i 
#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2

        FOR( i = 0; i < L_FRAME48k; i++ )

        {

            W_tmpX = W_mac_16_16( 0, input_fx[i - 4], full_band_bpf_fx[0][4] );

            W_tmpX = W_mac_16_16( W_tmpX, input_fx[i - 3], full_band_bpf_fx[0][3] );

            W_tmpX = wmac( 0, input_fx[i - 4], full_band_bpf_fx[0][4] );

            W_tmpX = wmac( W_tmpX, input_fx[i - 3], full_band_bpf_fx[0][3] );

            W_tmpY = W_msu_32_16( 0, L_tmp[i - 1], full_band_bpf_fx[3][1] );

            W_tmpX = W_mac_16_16( W_tmpX, input_fx[i - 2], full_band_bpf_fx[0][2] );

            W_tmpX = wmac( W_tmpX, input_fx[i - 2], full_band_bpf_fx[0][2] );

            W_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 2], full_band_bpf_fx[3][2] );

            W_tmpX = W_mac_16_16( W_tmpX, input_fx[i - 1], full_band_bpf_fx[0][1] );

            W_tmpX = wmac( W_tmpX, input_fx[i - 1], full_band_bpf_fx[0][1] );

            W_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 3], full_band_bpf_fx[3][3] );

            W_tmpX = W_mac_16_16( W_tmpX, input_fx[i], full_band_bpf_fx[0][0] );

            W_tmpX = wmac( W_tmpX, input_fx[i], full_band_bpf_fx[0][0] );

            W_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 4], full_band_bpf_fx[3][4] );

            L_tmp[i] = W_sat_l( W_shr( W_add( W_tmpX, W_shl( W_tmpY, 2 - 16 + 3 ) ), 3 ) ); /*Q_input_fx + 11*/

            L_tmp[i] = finalSat( W_tmpX, W_tmpY ); /*Q_input_fx + 11*/

            move32();

            if ( *L_tmpMax > 0 )

            {
@@ -6747,27 +6756,7 @@ inline static void elliptic_bpf_48k_generic_func1( Word16 *input16_fx, Word32 *i 
            }

        }

#else

        FOR( i = 0; i < L_FRAME48k; i++ )

        {

            L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 );                             /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 3], full_band_bpf_fx[0][3] ), 3 ), L_tmpX );        /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 1], full_band_bpf_fx[3][1] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 2], full_band_bpf_fx[0][2] ), 3 ), L_tmpX );        /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 2], full_band_bpf_fx[3][2] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 1], full_band_bpf_fx[0][1] ), 3 ), L_tmpX );        /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 3], full_band_bpf_fx[3][3] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i], full_band_bpf_fx[0][0] ), 3 ), L_tmpX );            /*Q_input_fx + 13 + 1 - 3*/

            L_tmp[i] = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 4], full_band_bpf_fx[3][4] ), 2 ) ); /*Q_input_fx + 11 + 13  -15 +2*/

            move32();

            if ( *L_tmpMax > 0 )

            {

                L_tmpAbs = L_abs( L_tmp[i] );

            }

            if ( *L_tmpMax > 0 )

            {

                L_tmpMax2 = L_max( L_tmpMax2, L_tmpAbs );

            }

        }

        assert( 0 ); /*kein bock*/

#endif

    } /*IsUpsampled3*/

    ELSE
@@ -6829,58 +6818,7 @@ inline static void elliptic_bpf_48k_generic_func1( Word16 *input16_fx, Word32 *i 
            i++;

        }

#else

        FOR( i = 0; i < L_FRAME48k; )

        {

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 3], full_band_bpf_fx[0][3] ), 3 ), 0 );             /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 1], full_band_bpf_fx[3][1] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 2], full_band_bpf_fx[3][2] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 3], full_band_bpf_fx[3][3] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i], full_band_bpf_fx[0][0] ), 3 ), L_tmpX );            /*Q_input_fx + 13 + 1 - 3*/

            L_tmp[i] = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 4], full_band_bpf_fx[3][4] ), 2 ) ); /*Q_input_fx + 11 + 13  -15 +2*/

            move32();

            if ( *L_tmpMax > 0 )

            {

                L_tmpAbs = L_abs( L_tmp[i] );

            }

            if ( *L_tmpMax > 0 )

            {

                L_tmpMax2 = L_max( L_tmpMax2, L_tmpAbs );

            }

            i++;



            L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 );                             /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 1], full_band_bpf_fx[3][1] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 2], full_band_bpf_fx[3][2] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 1], full_band_bpf_fx[0][1] ), 3 ), L_tmpX );        /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 3], full_band_bpf_fx[3][3] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmp[i] = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 4], full_band_bpf_fx[3][4] ), 2 ) ); /*Q_input_fx + 11 + 13  -15 +2*/

            move32();

            if ( *L_tmpMax > 0 )

            {

                L_tmpAbs = L_abs( L_tmp[i] );

            }

            if ( *L_tmpMax > 0 )

            {

                L_tmpMax2 = L_max( L_tmpMax2, L_tmpAbs );

            }

            i++;



            L_tmpX = L_sub_sat( 0, L_shl_sat( Mult_32_16( L_tmp[i - 1], full_band_bpf_fx[3][1] ), 2 ) );        /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_add_sat( L_shr( L_mult( input_fx[i - 2], full_band_bpf_fx[0][2] ), 3 ), L_tmpX );        /*Q_input_fx + 13 + 1 - 3*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 2], full_band_bpf_fx[3][2] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmpX = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 3], full_band_bpf_fx[3][3] ), 2 ) );   /*Q_input_fx + 11 + 13  -15 +2*/

            L_tmp[i] = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( L_tmp[i - 4], full_band_bpf_fx[3][4] ), 2 ) ); /*Q_input_fx + 11 + 13  -15 +2*/

            move32();

            if ( *L_tmpMax > 0 )

            {

                L_tmpAbs = L_abs( L_tmp[i] );

            }

            if ( *L_tmpMax > 0 )

            {

                L_tmpMax2 = L_max( L_tmpMax2, L_tmpAbs );

            }

            i++;

        }

        assert( 0 ); /*kein bock*/

#endif /*#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2*/

    }  /*IsUpsampled3*/

    *L_tmpMax = L_tmpMax2;
@@ -6932,32 +6870,41 @@ void elliptic_bpf_48k_generic_fx( 
    Word32 memory2_fx_2[4], memory2_fx_3[4];

#endif



    FOR( i = 0; i < 4; i++ )

    {

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic

    {

    FOR( i = 0; i < 4; i++ )

        memory_fx0 = extract_l( memory_fx2[0][i] );

        input_fx[i - 4] = shl_sat( memory_fx0, sub( *Q_input_fx, memory_fx_Q[0] ) );

        L_tmp[i - 4] = L_shl_sat( memory_fx2[1][i], sub( add( *Q_input_fx, 11 ), memory_fx_Q[1] ) );

        L_tmp2[i - 4] = L_shl_sat( memory_fx2[2][i], sub( add( *Q_input_fx, 6 ), memory_fx_Q[2] ) );

        // memory2_fx[3][i] = L_shl_sat( memory_fx2[3][i], sub( add( *Q_input_fx, 1 ), memory_fx_Q[3] ) );

        move32();

        move32();

        move32();

        move32();

        move32();

    }

#else

    FOR( i = 0; i < 4; i++ )

    {

        memory_fx0[0][i] = extract_l( memory_fx2[0][i] );

        memory_fx[0][i] = shl_sat( memory_fx0[0][i], sub( *Q_input_fx, memory_fx_Q[0] ) );

        memory2_fx[1][i] = L_shl_sat( memory_fx2[1][i], sub( add( *Q_input_fx, 11 ), memory_fx_Q[1] ) );

        memory2_fx[2][i] = L_shl_sat( memory_fx2[2][i], sub( add( *Q_input_fx, 6 ), memory_fx_Q[2] ) );

        memory2_fx[3][i] = L_shl_sat( memory_fx2[3][i], sub( add( *Q_input_fx, 1 ), memory_fx_Q[3] ) );

#endif

        move32();

        move32();

        move32();

        move32();

        move32();

    }

    }

#endif



#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic

    L_tmpMax = L_add( 0, 0 );

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1

    elliptic_bpf_48k_generic_func1( input_fx, L_tmp, &full_band_bpf_fx[0], IsUpsampled3, 0 );

    elliptic_bpf_48k_generic_func1( input_fx, 0, L_tmp, &full_band_bpf_fx[0], IsUpsampled3, 0 );

#else /*FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1*/

    IF( !IsUpsampled3 )

    {
@@ -7141,7 +7088,7 @@ void elliptic_bpf_48k_generic_fx( 
#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic

    L_tmpMax = L_add( 0, 0 );

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1

    elliptic_bpf_48k_generic_func1( L_tmp, L_tmp2, &full_band_bpf_fx[1], 0, &L_tmpMax );

    elliptic_bpf_48k_generic_func1( 0, L_tmp, L_tmp2, &full_band_bpf_fx[1], 0, &L_tmpMax );

#else /*FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1*/

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2

    IF( isIVAS )
@@ -7270,7 +7217,7 @@ void elliptic_bpf_48k_generic_fx( 


    L_tmpMax = L_add( 0, 0 );

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1 /*use subfunc*/

    elliptic_bpf_48k_generic_func1( L_tmp2, L_output, &full_band_bpf_fx[2], 0, &L_tmpMax );

    elliptic_bpf_48k_generic_func1( 0, L_tmp2, L_output, &full_band_bpf_fx[2], 0, &L_tmpMax );

#else /*FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1*/

#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2

    IF( isIVAS )