deactivate STAGE2 and func1 - implement func1 (d41a00b7) · Commits · SA4 / Audio / IVAS BASOP

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -87,5 +87,6 @@

		#define FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic /FhG: reduces maintenance complexity & reduces WMOPS & prepares STAGE2 patch/
		//#define FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2 /FhG: reduces WMOPS /
		//#define FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1

		#endif

lib_com/swb_tbe_com_fx.c

+127 −6

Original line number	Diff line number	Diff line
		@@ -6687,6 +6687,112 @@ void wb_tbe_extras_reset_synth_fx(
		return;
		}

		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1
		inline static void elliptic_bpf_48k_generic_func1( Word32 input_fx, Word32 L_tmp, const Word16 full_band_bpf_fx[][5], Word16 IsUpsampled3, Word32 L_tmpMax )
		{
		Word32 L_tmpX;
		Word16 i;
		IF( !IsUpsampled3 )
		{
		#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_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_shr( W_add( W_tmpX, W_shl( W_tmpY, 2 - 16 + 3 ) ), 3 ) ); /Q_input_fx + 11/
		move32();
		}
		#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();
		}
		#endif
		} /IsUpsampled3/
		ELSE
		{ /IsUpsampled3/
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		FOR( i = 0; i < L_FRAME48k; )
		{
		W_tmpX = W_mac_16_16( 0, 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_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 2], full_band_bpf_fx[3][2] );
		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_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_tmpY = W_msu_32_16( 0, L_tmp[i - 1], full_band_bpf_fx[3][1] );
		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_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/
		move32();
		i++;

		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_tmpY = W_msu_32_16( W_tmpY, L_tmp[i - 3], full_band_bpf_fx[3][3] );
		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/
		move32();
		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();
		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();
		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();
		i++;
		}
		#endif /#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2/
		} /IsUpsampled3/
		}
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1/

		/-------------------------------------------------------------------
		* elliptic_bpf_48k_generic()
		*
		@@ -6715,7 +6821,7 @@ void elliptic_bpf_48k_generic_fx(
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic
		Word16 memory_fx0, Q_temp, Q_temp2;
		Word32 L_tmp_buffer[L_FRAME48k + 4], L_tmp2_buffer[L_FRAME48k + 4], L_output_buffer[L_FRAME48k + 4], L_tmpX;
		Word32 L_tmpMax = L_add( 0, 0 );
		Word32 L_tmpMax;
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		Word64 W_tmpX;
		Word64 W_tmpY;
		@@ -6750,6 +6856,10 @@ 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 /use subfunc/
		elliptic_bpf_48k_generic_func1( input_fx, L_tmp, &full_band_bpf_fx[0], IsUpsampled3, 0 );
		#else
		IF( !IsUpsampled3 )
		{
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		@@ -6783,7 +6893,7 @@ void elliptic_bpf_48k_generic_fx(
		move32();
		}
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2/
		}
		} /IsUpsampled3/
		ELSE
		{
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		@@ -6848,7 +6958,9 @@ void elliptic_bpf_48k_generic_fx(
		i++;
		}
		#endif /#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2/
		}
		} /IsUpsampled3/
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1/

		#else /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic/
		L_tmpX = L_shr( L_mult( memory_fx[0][0], full_band_bpf_fx[0][4] ), 3 ); /Q_input_fx + 13 + 1 - 3/
		L_tmpX = L_add( L_shr( L_mult( memory_fx[0][1], full_band_bpf_fx[0][3] ), 3 ), L_tmpX ); /Q_input_fx + 13 + 1 - 3/
		@@ -6921,6 +7033,10 @@ void elliptic_bpf_48k_generic_fx(
		move32();

		#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 /use subfunc/
		elliptic_bpf_48k_generic_func1( L_tmp, L_tmp2, &full_band_bpf_fx[1], 0, &L_tmpMax );
		#else
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		FOR( i = 0; i < L_FRAME48k; i++ )
		{
		@@ -6953,8 +7069,8 @@ void elliptic_bpf_48k_generic_fx(
		L_tmpMax = L_max( L_tmpMax, L_abs( L_tmp2[i] ) );
		}
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2/

		#else
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1/
		#else /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic/
		L_tmpX = L_shr( Mult_32_16( memory2_fx[1][0], full_band_bpf_fx[1][4] ), 3 ); /Q_input_fx + 11 + 13 - 15 -3/
		L_tmpX = L_add( L_shr( Mult_32_16( memory2_fx[1][1], full_band_bpf_fx[1][3] ), 3 ), L_tmpX ); /Q_input_fx + 11 + 13 - 15 -3/
		L_tmpX = L_add( L_shr( Mult_32_16( memory2_fx[1][2], full_band_bpf_fx[1][2] ), 3 ), L_tmpX ); /Q_input_fx + 11 + 13 - 15 -3/
		@@ -7038,6 +7154,10 @@ void elliptic_bpf_48k_generic_fx(
		move32();
		}
		#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 /use subfunc/
		elliptic_bpf_48k_generic_func1( L_tmp2, L_output, &full_band_bpf_fx[2], 0, &L_tmpMax );
		#else
		#ifdef FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2
		FOR( i = 0; i < L_FRAME48k; i++ )
		{
		@@ -7073,7 +7193,8 @@ void elliptic_bpf_48k_generic_fx(
		L_tmpMax = L_max( L_tmpMax, L_abs( L_output[i] ) );
		}
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_STAGE2/
		#else
		#endif /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic_func1/
		#else /FIX_1439_SPEEDUP_SIMPLIFY_elliptic_bpf_48k_generic/
		L_tmpX = L_shr( Mult_32_16( memory2_fx_2[0], full_band_bpf_fx[2][4] ), 3 ); /* Q_input_fx+6 +Q_temp +13 -15 -3 /
		L_tmpX = L_add_sat( L_shr( Mult_32_16( memory2_fx_2[1], full_band_bpf_fx[2][3] ), 3 ), L_tmpX ); /Q_input_fx + 6 +Q_temp+13 -15 -3/
		L_tmpX = L_add_sat( L_shr( Mult_32_16( memory2_fx_2[2], full_band_bpf_fx[2][2] ), 3 ), L_tmpX ); /Q_input_fx + 6 +Q_temp+13 -15 -3/