Commit 1e85b773 authored by Fabian Bauer's avatar Fabian Bauer
Browse files

remove all BE changes

parent 83bf110a

lib_com/options.h

+1 −6
Original line number Diff line number Diff line
@@ -81,12 +81,7 @@ 
#define HARM_ENC_INIT

//#define HARM_SCE_INIT



// new speedups - BE ones already in another branch

//#define FIX_1439_SPEEDUP_Copy_Scale_sig_16_32_no_sat            /*FhG: reduces WMOPS - bit-exact*/ //   |

//#define FIX_1439_SPEEDUP_stereo_icBWE_dec_fx                    /*FhG: reduces WMOPS - bit-exact*/ //   |   2.4 WMOPS

//#define FIX_1439_SPEEDUP_ivas_swb_tbe_dec_fx                    /*FhG: reduces WMOPS - bit-exact*/ //   |

//#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig  /*FhG: reduces WMOPS - bit-exact*/ //   |

//#define FIX_1439_SPEEDUP_synthesise_fb_high_band_fx                                              // |   0.4 WMOPS - BE? Need for a manual test as long as be tests dont work





#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE1       // nonbe                          //   |

#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE2       // nonbe                          //   |   3.1 WMOPS, pipe testing https://forge.3gpp.org/rep/sa4/audio/ivas-basop/-/pipelines/50562

lib_com/swb_tbe_com_fx.c

+0 −195
Original line number Diff line number Diff line
@@ -6767,89 +6767,6 @@ void elliptic_bpf_48k_generic_fx( 
    L_tmp[3] = L_sub_sat( L_tmpX, L_shl_sat( Mult_32_16( memory2_fx[1][3], full_band_bpf_fx[3][4] ), 2 ) ); /*Q_input_fx + 11 + 13  -15 +2*/

    move32();

#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE1

#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig

    test();

    IF( full_band_bpf_fx == full_band_bpf_3_fx || full_band_bpf_fx == full_band_bpf_1_fx )

    {

        Word64 W_tmpX;

        Word64 W_tmpY;

        i = 4;



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

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

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

        move32();

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

        W_tmpX = W_mac_16_16( 0, 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_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_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 4], full_band_bpf_fx[3][4] );

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

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

        move32();

        i++;



        FOR( ; i < L_FRAME48k; )

        {

            // W_tmpX = W_mac_16_16( 0, input_fx[i - 4], full_band_bpf_fx[0][4] );         //2

            W_tmpX = W_mac_16_16( 0, input_fx[i - 3], full_band_bpf_fx[0][3] ); // 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] );    //4

            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] );    //5

            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] ); // 6

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

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

            move32();

            i++;



            W_tmpX = W_mac_16_16( 0, input_fx[i - 4], full_band_bpf_fx[0][4] ); // 3

            // W_tmpX = W_mac_16_16( W_tmpX, input_fx[i - 3], full_band_bpf_fx[0][3] );      //4

            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] );      //5

            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] ); // 6

            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] );          //7

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

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

            move32();

            i++;



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

            // W_tmpX = W_mac_16_16( W_tmpX, input_fx[i - 3], full_band_bpf_fx[0][3] );    //5

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

            W_tmpX = W_mac_16_16( 0, input_fx[i - 2], full_band_bpf_fx[0][2] ); // 6

            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] );    //7

            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] );        //8

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

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

            move32();

            i++;

        }

    }

    ELSE

#endif /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig*/

    {

        FOR( i = 4; i < L_FRAME48k; i++ )

        {
@@ -6871,76 +6788,6 @@ void elliptic_bpf_48k_generic_fx( 
    }



#else /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE1*/

#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig

    test();

    IF( full_band_bpf_fx == full_band_bpf_3_fx || full_band_bpf_fx == full_band_bpf_1_fx )

    {

        i = 4;

        L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 ); // 0 /*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 );        // 1 /*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 );        // 2 /*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 );      // 3 /*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 );            // 4 /*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();

        i++;



        // L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 );                             // 1 /*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 );        // 2 /*Q_input_fx + 13 + 1 - 3*/

        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 );      // 3 /*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 );        // 4 /*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 );            // 5 /*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();

        i++;



        FOR( ; i < L_FRAME48k; )

        {

            // L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 );                             // 2 /*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 ), 0 );           // 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_add_sat( L_shr( L_mult( input_fx[i - 2], full_band_bpf_fx[0][2] ), 3 ), L_tmpX );        // 4 /*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 );        // 5 /*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 );            // 6 /*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();

            i++;



            L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 ); // 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 );        // 4 /*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 );        // 5 /*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 );      // 6 /*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 );            // 7 /*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();

            i++;



            // L_tmpX = L_shr( L_mult( input_fx[i - 4], full_band_bpf_fx[0][4] ), 3 );                             // 4 /*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 );        // 5 /*Q_input_fx + 13 + 1 - 3*/

            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 );      // 6 /*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 );        // 7 /*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 );            // 8 /*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();

            i++;

        }

    }

    ELSE

#endif /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig*/

    {

        FOR( i = 4; i < L_FRAME48k; i++ )

        {
@@ -7307,47 +7154,6 @@ void synthesise_fb_high_band_fx( 
    tmp3 = add( sub( Qout, add( sub( 1, exp ), exp_tmp ) ), 16 ); /*Qout - (1 -exp +exp_tmp) + 16 */

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

    {

#ifdef FIX_1439_SPEEDUP_synthesise_fb_high_band_fx

        L_tmp = Mult_32_16( ratio2, tmp[i] ); /* Q(16-exp+exp_tmp-15 = 1-exp+exp_tmp) */

        Word32 L_tmp32;

        Word16 tmp16;



        if ( L_tmp < 0 )

        {

            L_tmp32 = L_negate( L_tmp );

        }

        if ( L_tmp < 0 )

        {

            L_tmp32 = L_shl_sat( L_tmp32, tmp3 );

        }

        if ( L_tmp < 0 )

        {

            tmp16 = extract_h( L_tmp32 );

        }

        if ( L_tmp < 0 )

        {

            tmp16 = negate( tmp16 );

        }



        if ( L_tmp == 0 )

        {

            tmp16 = 0;

            move16();

        }



        if ( L_tmp > 0 )

        {

            L_tmp32 = L_shl_sat( L_tmp, tmp3 );

        }

        if ( L_tmp > 0 )

        {

            tmp16 = extract_h( L_tmp32 );

        }



        output[i] = tmp16;

        move16();



#else

        L_tmp = Mult_32_16( ratio2, tmp[i] ); /* Q(16-exp+exp_tmp-15 = 1-exp+exp_tmp) */

        IF( L_tmp < 0 )

        {
@@ -7359,7 +7165,6 @@ void synthesise_fb_high_band_fx( 
            output[i] = extract_h( L_shl_sat( L_tmp, tmp3 ) ); /*Qout*/

            move16();

        }

#endif

    }

    pop_wmops(); /*push_wmops( "SYNTHESISE_FB_HIGH_BAND PART B" );*/

    return;

lib_com/tools_fx.c

+0 −24
Original line number Diff line number Diff line
@@ -794,37 +794,13 @@ void Copy_Scale_sig_16_32_no_sat( 
        }

        return;

    }

#ifdef FIX_1439_SPEEDUP_Copy_Scale_sig_16_32_no_sat

    L_tmp = L_shl_o( 1, exp0 - 1, &Overflow );



    IF( L_tmp >= 0x7FFF )

    {

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

    {

        // y[i] = L_mult0(x[i], L_tmp);

        y[i] = W_extract_l( W_mult_32_16( L_tmp, x[i] ) );

        move32(); /* Overflow can occur here */

    }

        return;

    }

    // ELSE

    {

        Word16 tmp = extract_l( L_tmp );

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

        {

            y[i] = L_mult( x[i], tmp );

            move32();

        }

    }

#else

    L_tmp = L_shl_o( 1, exp0 - 1, &Overflow );

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

    {

        // y[i] = L_mult0(x[i], L_tmp);

        y[i] = W_extract_l( W_mult_32_16( L_tmp, x[i] ) );

        move32(); /* Overflow can occur here */

    }

#endif

}



void Copy_Scale_sig_32_16(

lib_dec/ivas_stereo_icbwe_dec_fx.c

+0 −10
Original line number Diff line number Diff line
@@ -904,9 +904,6 @@ void stereo_icBWE_dec_fx( 
    winSlope_fx = div_s( 1, winLen_fx );                       /* Q15 */

    alpha_fx = winSlope_fx;                                    /* Q15 */

    move16();

#ifdef FIX_1439_SPEEDUP_stereo_icBWE_dec_fx

    Word16 winSlope_fx_ = sub( 32767 /* 1.0 in Q15*/, winSlope_fx );

#endif

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

    {

        L_tmp = L_mult0( alpha_fx, icbweM2Ref_fx );                                                        /* Q29 */
@@ -914,17 +911,10 @@ void stereo_icBWE_dec_fx( 
        tmp = shl( round_fx( L_tmp ), 1 );                                                                 /* Q14 */

        synthRef_fx[i] = Mpy_32_16_1( synthRef_fx[i], tmp );                                               /* Qsyn - 1 */

        move32();

#ifdef FIX_1439_SPEEDUP_stereo_icBWE_dec_fx

        if ( LE_16( alpha_fx, winSlope_fx_ ) )

        {

            alpha_fx = add( alpha_fx, winSlope_fx ); /* Q15 */

        }

#else

        IF( LE_16( alpha_fx, sub( 32767 /* 1.0 in Q15*/, winSlope_fx ) ) )

        {

            alpha_fx = add( alpha_fx, winSlope_fx ); /* Q15 */

        }

#endif

    }



    FOR( ; i < NS2SA_FX2( st->output_Fs, FRAME_SIZE_NS ); i++ )

lib_dec/swb_tbe_dec_fx.c

+0 −5
Original line number Diff line number Diff line
@@ -7037,17 +7037,12 @@ void ivas_swb_tbe_dec_fx( 


            FOR( i = 0; i < L_FRAME16k; i++ )

            {

#ifndef FIX_1439_SPEEDUP_ivas_swb_tbe_dec_fx

                Word16 idx = 0;

                move16();

                IF( i != 0 )

                {

                    idx = idiv1616( i_mult( NUM_SHB_SUBFR, i ), L_FRAME16k );

                }

#else

                Word16 idx;

                idx = extract_h( imult3216( idx32, i ) ); /*Q0*/

#endif

                L_tmp1 = Mult_32_16( L_tmp, GainShape_fx[idx] );                           /* Q : 18 + tmp +15 -15*/

                White_exc16k_fx[i] = round_fx( Mult_32_16( L_tmp1, White_exc16k_fx[i] ) ); /* 18 + tmp +*Q_white_exc -15 -16 */

                move16();