optimizations for stereo 32kbps (81182f04) · Commits · SA4 / Audio / IVAS BASOP

lib_com/fft_fx.c

+88 −136

Original line number	Diff line number	Diff line
		@@ -4727,13 +4727,13 @@ static void fft_len16(
		cmplx t[4];
		cmplx y[16];

		s[0] = CL_shr( x[0], SCALEFACTOR16 ); // Qx
		s[0] = x[0]; // Qx
		move64();
		s[1] = CL_shr( x[4], SCALEFACTOR16 ); // Qx
		s[1] = x[4]; // Qx
		move64();
		s[2] = CL_shr( x[8], SCALEFACTOR16 ); // Qx
		s[2] = x[8]; // Qx
		move64();
		s[3] = CL_shr( x[12], SCALEFACTOR16 ); // Qx
		s[3] = x[12]; // Qx
		move64();

		t[0] = CL_add( s[0], s[2] );
		@@ -4754,13 +4754,13 @@ static void fft_len16(
		y[3] = CL_add( t[1], t[3] );
		move64();

		s[0] = CL_shr( x[1], SCALEFACTOR16 ); // Qx
		s[0] = x[1]; // Qx
		move64();
		s[1] = CL_shr( x[5], SCALEFACTOR16 ); // Qx
		s[1] = x[5]; // Qx
		move64();
		s[2] = CL_shr( x[9], SCALEFACTOR16 ); // Qx
		s[2] = x[9]; // Qx
		move64();
		s[3] = CL_shr( x[13], SCALEFACTOR16 ); // Qx
		s[3] = x[13]; // Qx
		move64();

		t[0] = CL_add( s[0], s[2] );
		@@ -4781,13 +4781,13 @@ static void fft_len16(
		y[7] = CL_add( t[1], t[3] );
		move64();

		s[0] = CL_shr( x[2], SCALEFACTOR16 ); // Qx
		s[0] = x[2]; // Qx
		move64();
		s[1] = CL_shr( x[6], SCALEFACTOR16 ); // Qx
		s[1] = x[6]; // Qx
		move64();
		s[2] = CL_shr( x[10], SCALEFACTOR16 ); // Qx
		s[2] = x[10]; // Qx
		move64();
		s[3] = CL_shr( x[14], SCALEFACTOR16 ); // Qx
		s[3] = x[14]; // Qx
		move64();

		t[0] = CL_add( s[0], s[2] );
		@@ -4810,13 +4810,13 @@ static void fft_len16(
		y[11] = CL_add( t[1], t[3] );
		move64();

		s[0] = CL_shr( x[3], SCALEFACTOR16 ); // Qx
		s[0] = x[3]; // Qx
		move64();
		s[1] = CL_shr( x[7], SCALEFACTOR16 ); // Qx
		s[1] = x[7]; // Qx
		move64();
		s[2] = CL_shr( x[11], SCALEFACTOR16 ); // Qx
		s[2] = x[11]; // Qx
		move64();
		s[3] = CL_shr( x[15], SCALEFACTOR16 ); // Qx
		s[3] = x[15]; // Qx
		move64();

		t[0] = CL_add( s[0], s[2] );
		@@ -4978,15 +4978,15 @@ static void fft_len20_fx(
		cmplx tt[4];
		cmplx y[20];

		xx[0] = CL_shr( x[0], SCALEFACTOR20 ); // Qx
		xx[0] = x[0]; // Qx
		move64();
		xx[1] = CL_shr( x[16], SCALEFACTOR20 ); // Qx
		xx[1] = x[16]; // Qx
		move64();
		xx[2] = CL_shr( x[12], SCALEFACTOR20 ); // Qx
		xx[2] = x[12]; // Qx
		move64();
		xx[3] = CL_shr( x[8], SCALEFACTOR20 ); // Qx
		xx[3] = x[8]; // Qx
		move64();
		xx[4] = CL_shr( x[4], SCALEFACTOR20 ); // Qx
		xx[4] = x[4]; // Qx
		move64();

		s[0] = CL_add( xx[1], xx[4] );
		@@ -5023,15 +5023,15 @@ static void fft_len20_fx(
		y[12] = CL_msu_j( s[2], s[3] );
		move64();

		xx[0] = CL_shr( x[5], SCALEFACTOR20 );
		xx[0] = x[5];
		move64();
		xx[1] = CL_shr( x[1], SCALEFACTOR20 );
		xx[1] = x[1];
		move64();
		xx[2] = CL_shr( x[17], SCALEFACTOR20 );
		xx[2] = x[17];
		move64();
		xx[3] = CL_shr( x[13], SCALEFACTOR20 );
		xx[3] = x[13];
		move64();
		xx[4] = CL_shr( x[9], SCALEFACTOR20 );
		xx[4] = x[9];
		move64();

		s[0] = CL_add( xx[1], xx[4] );
		@@ -5068,15 +5068,15 @@ static void fft_len20_fx(
		y[13] = CL_msu_j( s[2], s[3] );
		move64();

		xx[0] = CL_shr( x[10], SCALEFACTOR20 );
		xx[0] = x[10];
		move64();
		xx[1] = CL_shr( x[6], SCALEFACTOR20 );
		xx[1] = x[6];
		move64();
		xx[2] = CL_shr( x[2], SCALEFACTOR20 );
		xx[2] = x[2];
		move64();
		xx[3] = CL_shr( x[18], SCALEFACTOR20 );
		xx[3] = x[18];
		move64();
		xx[4] = CL_shr( x[14], SCALEFACTOR20 );
		xx[4] = x[14];
		move64();

		s[0] = CL_add( xx[1], xx[4] );
		@@ -5113,15 +5113,15 @@ static void fft_len20_fx(
		y[14] = CL_msu_j( s[2], s[3] );
		move64();

		xx[0] = CL_shr( x[15], SCALEFACTOR20 );
		xx[0] = x[15];
		move64();
		xx[1] = CL_shr( x[11], SCALEFACTOR20 );
		xx[1] = x[11];
		move64();
		xx[2] = CL_shr( x[7], SCALEFACTOR20 );
		xx[2] = x[7];
		move64();
		xx[3] = CL_shr( x[3], SCALEFACTOR20 );
		xx[3] = x[3];
		move64();
		xx[4] = CL_shr( x[19], SCALEFACTOR20 );
		xx[4] = x[19];
		move64();

		s[0] = CL_add( xx[1], xx[4] );
		@@ -6501,84 +6501,66 @@ static void fft_lenN(
		cmplx s[8];
		cmplx y[8];

		test();
		test();
		test();
		test();
		IF( EQ_16( dim1, 30 ) \|\| EQ_16( dim1, 20 ) \|\| EQ_16( dim1, 15 ) \|\| EQ_16( dim1, 10 ) \|\| EQ_16( dim1, 5 ) )
		{
		FOR( i = 0; i < dim1; i++ )
		{
		{
		y[0] = xx[i]; // y[0] = xx[i + 0 * dim1]
		y[1] = xx[1 * dim1];
		move64();
		};
		IF( i == 0 )
		{
		{
		y[1] = xx[( i + ( 1 * dim1 ) )];
		y[2] = xx[2 * dim1];
		move64();
		};
		{
		y[2] = xx[( i + ( 2 * dim1 ) )];
		y[3] = xx[3 * dim1];
		move64();
		};
		{
		y[3] = xx[( i + ( 3 * dim1 ) )];
		y[4] = xx[4 * dim1];
		move64();
		};
		{
		y[4] = xx[( i + ( 4 * dim1 ) )];
		y[5] = xx[5 * dim1];
		move64();
		};
		{
		y[5] = xx[( i + ( 5 * dim1 ) )];
		y[6] = xx[6 * dim1];
		move64();
		};
		{
		y[6] = xx[( i + ( 6 * dim1 ) )];
		y[7] = xx[7 * dim1];
		move64();
		};

		test();
		test();
		IF( EQ_16( dim1, 8 ) \|\| EQ_16( dim1, 16 ) \|\| EQ_16( dim1, 32 ) )
		{
		y[7] = xx[( i + ( 7 * dim1 ) )];
		FOR( i = 0; i < dim1; i++ )
		{
		{
		y[0] = xx[i];
		move64();
		};
		}
		ELSE
		IF( i > 0 )
		{
		{
		y[1] = CL_mac_j( CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 1 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 1 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[1] = CL_mac_j( CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 1 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 1 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[2] = CL_mac_j( CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 2 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 2 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[2] = CL_mac_j( CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 2 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 2 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[3] = CL_mac_j( CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 3 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 3 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[3] = CL_mac_j( CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 3 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 3 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[4] = CL_mac_j( CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 4 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 4 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[4] = CL_mac_j( CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 4 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 4 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[5] = CL_mac_j( CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 5 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 5 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[5] = CL_mac_j( CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 5 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 5 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[6] = CL_mac_j( CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 6 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 6 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[6] = CL_mac_j( CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 6 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 6 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[7] = CL_mac_j( CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 7 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 7 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[7] = CL_mac_j( CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 7 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 7 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		}
		@@ -6625,7 +6607,7 @@ static void fft_lenN(
		s[7] = CL_conjugate( s[7] );
		move64();

		x[i] = CL_add( s[0], s[1] ); /x[add(i + i_mult(0 , dim1)] = CL_add( s[0], s[1] )/
		x[i] = CL_add( s[0], s[1] );
		move64();
		x[( i + ( 1 * dim1 ) )] = CL_add( s[5], s[6] );
		move64();
		@@ -6645,78 +6627,48 @@ static void fft_lenN(
		}
		ELSE
		{

		FOR( i = 0; i < dim1; i++ )
		{
		{
		y[0] = xx[i]; /* y[0] = xx[i + 0 * dim1] */
		move64();
		};
		IF( i == 0 )
		{
		{
		y[1] = xx[( i + ( 1 * dim1 ) )];
		move64();
		};
		{
		y[2] = xx[( i + ( 2 * dim1 ) )];
		move64();
		};
		{
		y[3] = xx[( i + ( 3 * dim1 ) )];
		y[0] = xx[i];
		move64();
		};
		{
		y[4] = xx[( i + ( 4 * dim1 ) )];
		move64();
		};
		{
		y[5] = xx[( i + ( 5 * dim1 ) )];
		move64();
		};
		{
		y[6] = xx[( i + ( 6 * dim1 ) )];
		move64();
		};
		{
		y[7] = xx[( i + ( 7 * dim1 ) )];
		move64();
		};
		}
		ELSE
		IF( i > 0 )
		{
		{
		y[1] = CL_mac_j( CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 1 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 1 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[1] = CL_mac_j( CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 1 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 1 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 1 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[2] = CL_mac_j( CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 2 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 2 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[2] = CL_mac_j( CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 2 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 2 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 2 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[3] = CL_mac_j( CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 3 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 3 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[3] = CL_mac_j( CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 3 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 3 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 3 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[4] = CL_mac_j( CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 4 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 4 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[4] = CL_mac_j( CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 4 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 4 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 4 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[5] = CL_mac_j( CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 5 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 5 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[5] = CL_mac_j( CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 5 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 5 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 5 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[6] = CL_mac_j( CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 6 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 6 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[6] = CL_mac_j( CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 6 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 6 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 6 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		{
		y[7] = CL_mac_j( CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( sc * i ) + ( ( sc * 7 ) * dim1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( sc * 7 ) * dim1 ) ) + 1 ) - Woff )] ) ); // Qx
		y[7] = CL_mac_j( CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( sc * i ) + ( ( ( sc * 7 ) * dim1 ) << 1 ) ) - Woff )] ),
		CL_scale( xx[( i + ( 7 * dim1 ) )], W[( ( ( ( sc * i ) + ( ( ( sc * 7 ) * dim1 ) << 1 ) ) + 1 ) - Woff )] ) ); // Qx
		move64();
		};
		}
		@@ -6763,7 +6715,7 @@ static void fft_lenN(
		s[7] = CL_conjugate( s[7] );
		move64();

		x[i] = CL_add( s[0], s[1] ); /x[i + 0 dim1] = CL_add( s[0], s[1] )*/
		x[i] = CL_add( s[0], s[1] );
		move64();
		x[( i + ( 1 * dim1 ) )] = CL_add( s[5], s[6] );
		move64();
		@@ -7173,7 +7125,7 @@ void rfft_fx(
		move32();
		x[( length - ( i << 1 ) )] = Mpy_32_16_1( L_add( t1, t3 ), 16384 /0.5.Q15/ );
		move32();
		x[( ( length - ( i << 1 ) ) + 1 )] = Mpy_32_16_1( L_negate( L_add( t2, t4 ) ), 16384 /0.5.Q15/ );
		x[( ( length - ( i << 1 ) ) + 1 )] = Mpy_32_16_1( L_add( t2, t4 ), -16384 /0.5.Q15/ );
		move32();
		}

lib_com/hp50_fx.c

+29 −86

Original line number	Diff line number	Diff line
		@@ -469,8 +469,9 @@ void hp20_fx_32(
		{
		Word16 i;
		Word32 a1_fx, a2_fx, b1_fx, b2_fx;
		Word16 Qx0, Qx1, Qx2, Qy1, Qprev_y1, Qy2, Qprev_y2, Qmin;
		Word64 x0_fx64, x1_fx64, x2_fx64, y0_fx64, y1_fx64, y2_fx64, R1, R2, R3, R4, R5;
		Word16 Qy1, Qy2, Qmin;
		Word64 y0_fx64, y1_fx64, y2_fx64;
		Word32 x0, x1, x2;

		IF( EQ_32( Fs, 8000 ) )
		{
		@@ -521,20 +522,22 @@ void hp20_fx_32(
		move32();
		move32();

		Qprev_y1 = extract_l( mem_fx[4] );
		Qprev_y2 = extract_l( mem_fx[5] );
		y1_fx64 = W_deposit32_l( mem_fx[0] );
		y2_fx64 = W_deposit32_l( mem_fx[1] );
		x0_fx64 = W_deposit32_l( mem_fx[2] );
		x1_fx64 = W_deposit32_l( mem_fx[3] );
		y1_fx64 = W_add( W_deposit32_l( mem_fx[0] ), W_deposit32_h( mem_fx[1] ) );
		y2_fx64 = W_add( W_deposit32_l( mem_fx[2] ), W_deposit32_h( mem_fx[3] ) );

		x0 = mem_fx[4];
		move32();
		x1 = mem_fx[5];
		move32();

		FOR( i = 0; i < lg; i++ )
		{
		x2_fx64 = x1_fx64;
		move64();
		x1_fx64 = x0_fx64;
		move64();
		x0_fx64 = W_deposit32_l( signal_fx[i] );
		x2 = x1;
		move32();
		x1 = x0;
		move32();
		x0 = signal_fx[i];
		move32();

		Qy1 = W_norm( y1_fx64 );
		if ( y1_fx64 == 0 )
		@@ -542,9 +545,6 @@ void hp20_fx_32(
		Qy1 = 62;
		move16();
		}
		Qy1 = sub( Qy1, 34 );
		R1 = W_mult0_32_32( W_extract_l( W_shl( y1_fx64, Qy1 ) ), a1_fx );
		Qy1 = add( Qy1, Qprev_y1 );

		Qy2 = W_norm( y2_fx64 );
		if ( y2_fx64 == 0 )
		@@ -552,89 +552,32 @@ void hp20_fx_32(
		Qy2 = 62;
		move16();
		}
		Qy2 = sub( Qy2, 34 );
		R2 = W_mult0_32_32( W_extract_l( W_shl( y2_fx64, Qy2 ) ), a2_fx );
		Qy2 = add( Qy2, Qprev_y2 );

		Qx0 = W_norm( x0_fx64 );
		if ( x0_fx64 == 0 )
		{
		Qx0 = 62;
		move16();
		}
		Qx0 = sub( Qx0, 34 );
		R3 = W_mult0_32_32( W_extract_l( W_shl( x0_fx64, Qx0 ) ), b2_fx );

		Qx1 = W_norm( x1_fx64 );
		if ( x1_fx64 == 0 )
		{
		Qx1 = 62;
		move16();
		}
		Qx1 = sub( Qx1, 34 );
		R4 = W_mult0_32_32( W_extract_l( W_shl( x1_fx64, Qx1 ) ), b1_fx );

		Qx2 = W_norm( x2_fx64 );
		if ( x2_fx64 == 0 )
		{
		Qx2 = 62;
		move16();
		}
		Qx2 = sub( Qx2, 34 );
		R5 = W_mult0_32_32( W_extract_l( W_shl( x2_fx64, Qx2 ) ), b2_fx );

		Qmin = s_min( Qy1, Qy2 );

		y0_fx64 = W_add( W_shr( R1, sub( Qy1, Qmin ) ), W_shr( R2, sub( Qy2, Qmin ) ) );

		Qmin = s_min( Qmin, Qx0 );
		Qmin = s_min( Qmin, Qx1 );
		Qmin = s_min( Qmin, Qx2 );
		Qmin = sub( Qmin, 34 );

		y0_fx64 = W_add( W_shr( y0_fx64, sub( s_min( Qy1, Qy2 ), Qmin ) ), W_add( W_shr( R3, sub( Qx0, Qmin ) ), W_add( W_shr( R4, sub( Qx1, Qmin ) ), W_shr( R5, sub( Qx2, Qmin ) ) ) ) );
		y0_fx64 = W_mac_32_32( W_mult_32_32( W_shl_sat_l( y1_fx64, Qmin ), a1_fx ), W_shl_sat_l( y2_fx64, Qmin ), a2_fx ); // Qmin + Q29 + Q30 + 1

		y0_fx64 = W_shr( y0_fx64, 29 );

		signal_fx[i] = W_extract_l( W_shr( y0_fx64, Qmin ) );
		move32();
		IF( signal_fx[i] < 0 )
		{
		signal_fx[i] = L_add( signal_fx[i], 1 );
		Word64 temp = W_mac_32_32( W_mac_32_32( W_mult_32_32( x2, b2_fx ), x1, b1_fx ), x0, b2_fx ); // Q30
		Word64 y0_fx = W_shr( y0_fx64, add( Qmin, Q30 ) ); // Q30
		y0_fx64 = W_add( temp, y0_fx ); // Q30
		signal_fx[i] = W_shl_sat_l( y0_fx64, -Q30 );
		move32();
		}

		y2_fx64 = y1_fx64;
		y1_fx64 = y0_fx64;
		Qprev_y2 = Qprev_y1;
		Qprev_y1 = Qmin;
		move64();
		y1_fx64 = y0_fx64;
		move64();
		move16();
		move16();
		}

		Qy1 = W_norm( y1_fx64 );
		test();
		IF( y1_fx64 != 0 && LT_16( Qy1, 32 ) )
		{
		y1_fx64 = W_shr( y1_fx64, sub( 32, Qy1 ) );
		Qprev_y1 = sub( Qprev_y1, sub( 32, Qy1 ) );
		}

		Qy2 = W_norm( y2_fx64 );
		test();
		IF( y2_fx64 != 0 && LT_16( Qy2, 32 ) )
		{
		y2_fx64 = W_shr( y2_fx64, sub( 32, Qy2 ) );
		Qprev_y2 = sub( Qprev_y2, sub( 32, Qy2 ) );
		}

		mem_fx[0] = W_extract_l( y1_fx64 );
		mem_fx[1] = W_extract_l( y2_fx64 );
		mem_fx[2] = W_extract_l( x0_fx64 );
		mem_fx[3] = W_extract_l( x1_fx64 );
		mem_fx[4] = Qprev_y1;
		mem_fx[5] = Qprev_y2;
		mem_fx[1] = W_extract_h( y1_fx64 );
		mem_fx[2] = W_extract_l( y2_fx64 );
		mem_fx[3] = W_extract_h( y2_fx64 );
		mem_fx[4] = x0;
		mem_fx[5] = x1;

		move32();
		move32();
		move32();

lib_com/modif_fs.c

+29 −29

File changed.

Preview size limit exceeded, changes collapsed.

lib_com/mslvq_com_fx.c

+34 −26

Original line number	Diff line number	Diff line
		@@ -121,51 +121,59 @@ void init_lvq_fx(
		Word16 no_scales_p[][2] /* o: number of truncations for each LSF subvector at each MSLVQ structure - predictive */
		)
		{
		Word16 i, j;
		Word16 i, j, k;
		/* safety-net mode */
		FOR( i = 0; i < MAX_NO_MODES; i++ )
		{
		j = 0;
		move16();
		test();
		WHILE( ( LT_16( j, MAX_NO_SCALES ) ) && ( no_lead_fx[i][j] > 0 ) )
		FOR( ( j = 0, k = 0 ); j < MAX_NO_SCALES; ( j++, k++ ) )
		{
		if ( ( no_lead_fx[i][j] <= 0 ) )
		{
		j++;
		}
		no_scales[i][0] = j;
		move16();
		j = MAX_NO_SCALES;
		}
		}
		no_scales[i][0] = k;
		move16();
		test();
		WHILE( ( LT_16( j, shl( MAX_NO_SCALES, 1 ) ) ) && ( no_lead_fx[i][j] > 0 ) )

		FOR( k = 0; j < MAX_NO_SCALES << 1; ( j++, k++ ) )
		{
		if ( no_lead_fx[i][j] <= 0 )
		{
		j++;
		j = MAX_NO_SCALES << 1;
		}
		no_scales[i][1] = sub( j, MAX_NO_SCALES );
		}
		no_scales[i][1] = k;
		move16();
		}
		/* predictive mode */


		FOR( i = 0; i < MAX_NO_MODES_p; i++ )
		{
		j = 0;
		move16();
		WHILE( ( LT_16( j, MAX_NO_SCALES ) ) && ( no_lead_p_fx[i][j] > 0 ) )
		FOR( ( j = 0, k = 0 ); j < MAX_NO_SCALES; ( j++, k++ ) )
		{

		if ( ( no_lead_p_fx[i][j] <= 0 ) )
		{
		test();
		j++;
		}
		no_scales_p[i][0] = j;
		move16();
		j = MAX_NO_SCALES;
		}
		}
		no_scales_p[i][0] = k;
		move16();
		WHILE( ( LT_16( j, shl( MAX_NO_SCALES, 1 ) ) ) && ( no_lead_p_fx[i][j] > 0 ) )

		FOR( k = 0; j < MAX_NO_SCALES << 1; ( j++, k++ ) )
		{
		test();
		j++;

		if ( ( no_lead_p_fx[i][j] <= 0 ) )
		{
		j = MAX_NO_SCALES << 1;
		}
		}
		no_scales_p[i][1] = sub( j, MAX_NO_SCALES );

		no_scales_p[i][1] = k;
		move16();
		}


		/* index offsets for each truncation */
		init_offset_fx( offset_scale1, offset_scale2, offset_scale1_p, offset_scale2_p, no_scales, no_scales_p );
		}

lib_com/scale_mem_fx.c

+9 −0

Original line number	Diff line number	Diff line
		@@ -307,8 +307,13 @@ void scale_sig32(
		/* saturation can occur here */
		x[i] = L_shl( x[i], exp0 );
		move32();
		if ( 0 == exp0 )
		{
		i = lg;
		}
		}
		}

		void scale_sig32_r(
		Word32 x[], /* i/o: signal to scale Qx */
		const Word16 lg, /* i : size of x[] Q0 */
		@@ -322,6 +327,10 @@ void scale_sig32_r(
		/* saturation can occur here */
		x[i] = L_shl_r( x[i], exp0 );
		move32();
		if ( 0 == exp0 )
		{
		i = lg;
		}
		}
		}