Commit 44ade344 authored by Fabian Bauer's avatar Fabian Bauer
Browse files

introduce FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig and... 

introduce FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig and FIX_1439_SPEEDUP_sum2_fx_mod , delete FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32 - activate FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig which should be BE
parent e3f013ae

lib_com/prot_fx.h

+7 −13
Original line number Diff line number Diff line
@@ -62,7 +62,13 @@ 
//#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE1

//#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE2

//#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE3

//#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32

#define FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig  // be       // 2    WMOPS

//#define FIX_1439_SPEEDUP_sum2_fx_mod                              // 3liner   // 0.25 WMOPS













// Numbers ALL:     83,2













// Numbers beOnly:  86,162













// NUmbers none:    88,183









































/*----------------------------------------------------------------------------------*















 * Prototypes of global macros










@@ -3262,17 +3268,6 @@ void interp_code_4over2_fx(



























void wb_tbe_extras_reset_synth_fx( Word16 state_lsyn_filt_shb[], Word16 state_lsyn_filt_dwn_shb[], Word16 state_32and48k_WB_upsample[], Word16 state_resamp_HB[] );



























#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













void elliptic_bpf_48k_generic_fx(













    const Word16* input_fx, /* i  : i   signal                              Q_input_fx  */













    Word16 *Q_input_fx,













    Word16 output_fx[],    /* o  : output signal                               */













    Word32 memory_fx[][4], /* i/o: 4 arrays of 4 for memory                   memory_fx_Q */













    Word16 memory_fx_Q[],













    const Word16 full_band_bpf[][5], /* i  : filter coefficients b0,b1,b2,a0,a1,a2  Q13  */













    const Word32 *output32_fx,













    const Word16 *scale_output32_fx);













#else















void elliptic_bpf_48k_generic_fx(















    const Word16 input_fx[], /* i  : i   signal                              Q_input_fx  */















    Word16 *Q_input_fx,









@@ -3281,7 +3276,6 @@ void elliptic_bpf_48k_generic_fx(













    Word16 memory_fx_Q[],















    const Word16 full_band_bpf[][5] /* i  : filter coefficients b0,b1,b2,a0,a1,a2  Q13  */















);













#endif





























void synthesise_fb_high_band_fx(















    const Word16 excitation_in[], /* i  : full band excitation                                */

































lib_com/swb_tbe_com_fx.c



+181
−33




























Original line number
Diff line number
Diff line








@@ -6694,8 +6694,6 @@ void wb_tbe_extras_reset_synth_fx(













 * 18th-order elliptic bandpass filter at 14.0 to 20 kHz sampled at 48 kHz















 * Implemented as 3 fourth order sections cascaded.















 *-------------------------------------------------------------------*/













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













#else















void elliptic_bpf_48k_generic_fx(















    const Word16 input_fx[], /* i  : input signal                             Q_input_fx*/















    Word16 *Q_input_fx,









@@ -6704,12 +6702,12 @@ void elliptic_bpf_48k_generic_fx(













    Word16 memory_fx_Q[],















    const Word16 full_band_bpf_fx[][5] /* i  : filter coefficients b0,b1,b2,a0,a1,a2  Q13  */















)













#endif















{















    Word16 i, j;















    Word16 memory_fx0[4][4], memory_fx[4][4], Q_temp, Q_temp2;













    Word32 L_tmp[L_FRAME48k], L_tmp2[L_FRAME48k], L_output[L_FRAME48k], L_tmpX, memory2_fx[4][4], L_tmpMax;













    Word32 L_tmp[L_FRAME48k], L_tmp2[L_FRAME48k], L_tmpX, memory2_fx[4][4], L_tmpMax;















    Word32 memory2_fx_2[4], memory2_fx_3[4];













    Word32 L_output[L_FRAME48k];































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















    {










@@ -6769,6 +6767,85 @@ 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*/













        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*/













        move32();













        i++;



























        FOR( ; i < L_FRAME48k / 3; )













        {













            // 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*/













            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*/













            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*/













            move32();













            i++;













        }













    }













    ELSE













#else













    {















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















        {















            Word64 W_tmpX;









@@ -6785,7 +6862,80 @@ void elliptic_bpf_48k_generic_fx(













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















            move32();















        }













    }













#endif /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig*/













#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 / 3; )













        {













            // 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















#else













    {















        FOR( i = 4; 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*/









@@ -6799,6 +6949,8 @@ void elliptic_bpf_48k_generic_fx(













            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();















        }













    }













#endif /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_upsampledsig*/















#endif










































@@ -6857,6 +7009,7 @@ void elliptic_bpf_48k_generic_fx(













    L_tmpMax = L_max( L_tmpMax, L_abs( L_tmp2[3] ) );































#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE2





























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















    {















        Word64 W_tmpX;









@@ -6874,7 +7027,8 @@ void elliptic_bpf_48k_generic_fx(













        move32();















        L_tmpMax = L_max( L_tmpMax, L_abs( L_tmp2[i] ) );















    }













#else



























#else /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE2*/















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















    {















        L_tmpX = L_shr( Mult_32_16( L_tmp[i - 4], full_band_bpf_fx[1][4] ), 3 );                              /*Q_input_fx + 11 + 13 - 15 -3*/









@@ -6889,7 +7043,7 @@ void elliptic_bpf_48k_generic_fx(













        move32();















        L_tmpMax = L_max( L_tmpMax, L_abs( L_tmp2[i] ) );















    }













#endif













#endif /*FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_STAGE2*/















    Q_temp = norm_l( L_tmpMax );















    Q_temp = sub( Q_temp, 4 );















    Scale_sig32( L_tmp2, 960, Q_temp );










@@ -7025,12 +7179,15 @@ void elliptic_bpf_48k_generic_fx(













    move16();















    move16();















    Q_temp2 = norm_l( L_tmpMax );



























    {















        Scale_sig32( L_output, 960, Q_temp2 );















        FOR( i = 0; i < 960; i++ )















        {















            output_fx[i] = extract_h( L_output[i] );















            move16();















        }













    }















    *Q_input_fx = sub( add( add( *Q_input_fx, Q_temp ), Q_temp2 ), 15 );















    move16(); /* BASOP_NOGLOB */


























@@ -7066,11 +7223,11 @@ void synthesise_fb_high_band_fx(













{















    Word16 i, j;















    Word16 excitation_in_interp3[L_FRAME48k];













    Word16 tmp[L_FRAME48k];















    Word32 temp1;















    Word32 ratio2;















    Word32 L_tmp;















    Word16 tmp3, tmp1, tmp2, exp, exp2, exp_tmp;













    Word16 tmp[L_FRAME48k];































    push_wmops( "SYNTHESISE_FB_HIGH_BAND PART A" );















    /* Interpolate the white energy shaped gaussian excitation from 16 kHz to 48 kHz with zeros */









@@ -7092,20 +7249,12 @@ void synthesise_fb_high_band_fx(













    IF( EQ_16( L_frame, L_FRAME16k ) )















    {















        /* for 16kHz ACELP core */













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













        elliptic_bpf_48k_generic_fx( excitation_in_interp3, &exp_tmp, NULL, bpf_memory, bpf_memory_Q, full_band_bpf_3_fx, tmp32, scale_tmp32 );













#else















        elliptic_bpf_48k_generic_fx( excitation_in_interp3, &exp_tmp, tmp, bpf_memory, bpf_memory_Q, full_band_bpf_3_fx );













#endif















    }















    ELSE















    {















        /* for 12.8kHz ACELP core */













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













        elliptic_bpf_48k_generic_fx( excitation_in_interp3, &exp_tmp, NULL, bpf_memory, bpf_memory_Q, full_band_bpf_1_fx, tmp32, scale_tmp32 );













#else















        elliptic_bpf_48k_generic_fx( excitation_in_interp3, &exp_tmp, tmp, bpf_memory, bpf_memory_Q, full_band_bpf_1_fx );













#endif































    }















    pop_wmops(); /*push_wmops( "SYNTHESISE_FB_HIGH_BAND PART A.1" );*/









@@ -7113,7 +7262,6 @@ void synthesise_fb_high_band_fx(





























    push_wmops( "SYNTHESISE_FB_HIGH_BAND PART B" );















    temp1 = sum2_fx_mod( tmp, L_FRAME48k );





























    L_tmp = L_max( 1, fb_exc_energy ); /*Q(2*Q_fb_exc + 1)*/















    exp = norm_l( L_tmp );















    tmp3 = extract_h( L_shl( L_tmp, exp ) );

































lib_com/tools_fx.c



+17
−1




























Original line number
Diff line number
Diff line








@@ -446,6 +446,21 @@ Word32 sum2_fx_mod(                    /* o  : sum of all squared vector element













                    const Word16 lvec  /* i  : length of input vector                */















)















{













#ifdef FIX_1439_SPEEDUP_sum2_fx_mod













    Word16 i;













    Word64 W_tmp;













#ifdef BASOP_NOGLOB_DECLARE_LOCAL













    Flag Overflow = 0;













    move32();













#endif



























    W_tmp = W_add(0, 0 );













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













    {













        W_tmp = W_mac_16_16( W_tmp, vec[i], vec[i] );













    }













    return W_sat_l(W_shr(W_tmp, 9));













#else















    Word16 i;















    Word32 L_tmp;















#ifdef BASOP_NOGLOB_DECLARE_LOCAL









@@ -458,8 +473,9 @@ Word32 sum2_fx_mod(                    /* o  : sum of all squared vector element













    {















        L_tmp = L_add_o( L_tmp, L_shr( L_mult_o( vec[i], vec[i], &Overflow ), 9 ), &Overflow );















    }





























    return L_tmp;













#endif





























}















/*-------------------------------------------------------------------*















 * Copy:

































lib_enc/swb_tbe_enc_fx.c



+0
−12




























Original line number
Diff line number
Diff line








@@ -7324,11 +7324,7 @@ void fb_tbe_enc_fx(













    exp_temp = sub( exp_temp, 1 );































    Copy_Scale_sig( new_input, input_fhb, L_FRAME48k, exp_temp );













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













    elliptic_bpf_48k_generic_fx( input_fhb, &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx, NULL, NULL );













#else















    elliptic_bpf_48k_generic_fx( input_fhb, &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx );













#endif















    Sample_Delay_HP = NS2SA( 48000, ACELP_LOOK_NS + DELAY_FD_BWE_ENC_12k8_NS + DELAY_FIR_RESAMPL_NS ) - L_FRAME48k / 2;































    IF( NE_16( st->last_extl, FB_TBE ) )










@@ -7452,19 +7448,11 @@ void fb_tbe_enc_ivas_fx(





























    IF( EQ_16( st->element_mode, IVAS_CPE_DFT ) )















    {













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













        elliptic_bpf_48k_generic_fx( input_fhb_new, &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx, NULL, NULL );













#else















        elliptic_bpf_48k_generic_fx( input_fhb_new, &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx );













#endif















    }















    ELSE















    {













#ifdef FIX_1439_SPEEDUP_elliptic_bpf_48k_generic_OUT32













        elliptic_bpf_48k_generic_fx( input_fhb_new + NS2SA( 48000, DELAY_FIR_RESAMPL_NS ), &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx, NULL, NULL );













#else















        elliptic_bpf_48k_generic_fx( input_fhb_new + NS2SA( 48000, DELAY_FIR_RESAMPL_NS ), &exp_temp, tmp_vec, hBWE_TD->elliptic_bpf_2_48k_mem_fx, hBWE_TD->elliptic_bpf_2_48k_mem_fx_Q, full_band_bpf_2_fx );













#endif















    }































    test();