[libav.git] / libavcodec / faandct.c

/*
 * Floating point AAN DCT
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
 * this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * The AAN DCT in this file except ff_faandct248() can also be used under the
 * new (3 clause) BSD license.
 */

/**
 * @file faandct.c
 * @brief
 *     Floating point AAN DCT
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

#include "dsputil.h"
#include "faandct.h"

#define FLOAT float
#ifdef FAAN_POSTSCALE
#    define SCALE(x) postscale[x]
#else
#    define SCALE(x) 1
#endif

//numbers generated by simple c code (not as accurate as they could be)
/*
for(i=0; i<8; i++){
    printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
}
*/
#define B0 1.00000000000000000000
#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1


#define A1 0.70710678118654752438 // cos(pi*4/16)
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
#define A5 0.38268343236508977170 // cos(pi*6/16)
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)

static FLOAT postscale[64]={
B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7,
B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7,
B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7,
B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7,
B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7,
B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7,
B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
};

static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
{
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
    FLOAT tmp10, tmp11, tmp12, tmp13;
    FLOAT z1, z2, z3, z4, z5, z11, z13;
    int i;

    for (i=0; i<8*8; i+=8) {
        tmp0= data[0 + i] + data[7 + i];
        tmp7= data[0 + i] - data[7 + i];
        tmp1= data[1 + i] + data[6 + i];
        tmp6= data[1 + i] - data[6 + i];
        tmp2= data[2 + i] + data[5 + i];
        tmp5= data[2 + i] - data[5 + i];
        tmp3= data[3 + i] + data[4 + i];
        tmp4= data[3 + i] - data[4 + i];

        tmp10= tmp0 + tmp3;
        tmp13= tmp0 - tmp3;
        tmp11= tmp1 + tmp2;
        tmp12= tmp1 - tmp2;

        temp[0 + i]= tmp10 + tmp11;
        temp[4 + i]= tmp10 - tmp11;

        z1= (tmp12 + tmp13)*A1;
        temp[2 + i]= tmp13 + z1;
        temp[6 + i]= tmp13 - z1;

        tmp10= tmp4 + tmp5;
        tmp11= tmp5 + tmp6;
        tmp12= tmp6 + tmp7;

        z5= (tmp10 - tmp12) * A5;
        z2= tmp10*A2 + z5;
        z4= tmp12*A4 + z5;
        z3= tmp11*A1;

        z11= tmp7 + z3;
        z13= tmp7 - z3;

        temp[5 + i]= z13 + z2;
        temp[3 + i]= z13 - z2;
        temp[1 + i]= z11 + z4;
        temp[7 + i]= z11 - z4;
    }
}

void ff_faandct(DCTELEM * data)
{
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
    FLOAT tmp10, tmp11, tmp12, tmp13;
    FLOAT z1, z2, z3, z4, z5, z11, z13;
    FLOAT temp[64];
    int i;

    emms_c();

    row_fdct(temp, data);

    for (i=0; i<8; i++) {
        tmp0= temp[8*0 + i] + temp[8*7 + i];
        tmp7= temp[8*0 + i] - temp[8*7 + i];
        tmp1= temp[8*1 + i] + temp[8*6 + i];
        tmp6= temp[8*1 + i] - temp[8*6 + i];
        tmp2= temp[8*2 + i] + temp[8*5 + i];
        tmp5= temp[8*2 + i] - temp[8*5 + i];
        tmp3= temp[8*3 + i] + temp[8*4 + i];
        tmp4= temp[8*3 + i] - temp[8*4 + i];

        tmp10= tmp0 + tmp3;
        tmp13= tmp0 - tmp3;
        tmp11= tmp1 + tmp2;
        tmp12= tmp1 - tmp2;

        data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
        data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));

        z1= (tmp12 + tmp13)* A1;
        data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + z1));
        data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - z1));

        tmp10= tmp4 + tmp5;
        tmp11= tmp5 + tmp6;
        tmp12= tmp6 + tmp7;

        z5= (tmp10 - tmp12) * A5;
        z2= tmp10*A2 + z5;
        z4= tmp12*A4 + z5;
        z3= tmp11*A1;

        z11= tmp7 + z3;
        z13= tmp7 - z3;

        data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
        data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
        data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
        data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
    }
}

void ff_faandct248(DCTELEM * data)
{
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
    FLOAT tmp10, tmp11, tmp12, tmp13;
    FLOAT z1;
    FLOAT temp[64];
    int i;

    emms_c();

    row_fdct(temp, data);

    for (i=0; i<8; i++) {
        tmp0 = temp[8*0 + i] + temp[8*1 + i];
        tmp1 = temp[8*2 + i] + temp[8*3 + i];
        tmp2 = temp[8*4 + i] + temp[8*5 + i];
        tmp3 = temp[8*6 + i] + temp[8*7 + i];
        tmp4 = temp[8*0 + i] - temp[8*1 + i];
        tmp5 = temp[8*2 + i] - temp[8*3 + i];
        tmp6 = temp[8*4 + i] - temp[8*5 + i];
        tmp7 = temp[8*6 + i] - temp[8*7 + i];

        tmp10 = tmp0 + tmp3;
        tmp11 = tmp1 + tmp2;
        tmp12 = tmp1 - tmp2;
        tmp13 = tmp0 - tmp3;

        data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
        data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));

        z1 = (tmp12 + tmp13)* A1;
        data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
        data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));

        tmp10 = tmp4 + tmp7;
        tmp11 = tmp5 + tmp6;
        tmp12 = tmp5 - tmp6;
        tmp13 = tmp4 - tmp7;

        data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
        data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));

        z1 = (tmp12 + tmp13)* A1;
        data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
        data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
    }
}
Commit	Line	Data
65e4c8c9 MN	1	/*
	2	* Floating point AAN DCT
	3	* Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
	4	*
7b94177e DB	5	* this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
7b94177e DB	6	*
b78e7197 DB	7	* This file is part of FFmpeg.
	8	*
	9	* FFmpeg is free software; you can redistribute it and/or
65e4c8c9 MN	10	* modify it under the terms of the GNU Lesser General Public
65e4c8c9 MN	11	* License as published by the Free Software Foundation; either
b78e7197	12	* version 2.1 of the License, or (at your option) any later version.
65e4c8c9	13	*
b78e7197	14	* FFmpeg is distributed in the hope that it will be useful,
65e4c8c9 MN	15	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	16	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	17	* Lesser General Public License for more details.
	18	*
	19	* You should have received a copy of the GNU Lesser General Public
b78e7197	20	* License along with FFmpeg; if not, write to the Free Software
5509bffa	21	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
5b0d04d3 MN	22	*
5b0d04d3 MN	23	* The AAN DCT in this file except ff_faandct248() can also be used under the
062942e7	24	* new (3 clause) BSD license.
65e4c8c9 MN	25	*/
	26
	27	/**
	28	* @file faandct.c
115329f1	29	* @brief
65e4c8c9 MN	30	* Floating point AAN DCT
	31	* @author Michael Niedermayer <michaelni@gmx.at>
	32	*/
	33
65e4c8c9 MN	34	#include "dsputil.h"
	35	#include "faandct.h"
	36
	37	#define FLOAT float
b4c3816c MN	38	#ifdef FAAN_POSTSCALE
	39	# define SCALE(x) postscale[x]
	40	#else
	41	# define SCALE(x) 1
	42	#endif
65e4c8c9 MN	43
	44	//numbers generated by simple c code (not as accurate as they could be)
	45	/*
	46	for(i=0; i<8; i++){
	47	printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(iacosl(-1.0)/(long double)16.0)sqrtl(2)));
	48	}
	49	*/
	50	#define B0 1.00000000000000000000
	51	#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
	52	#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
	53	#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
	54	#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
	55	#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
	56	#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
	57	#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
	58
	59
	60	#define A1 0.70710678118654752438 // cos(pi*4/16)
	61	#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
	62	#define A5 0.38268343236508977170 // cos(pi*6/16)
	63	#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
	64
	65	static FLOAT postscale[64]={
	66	B0B0, B0B1, B0B2, B0B3, B0B4, B0B5, B0B6, B0B7,
	67	B1B0, B1B1, B1B2, B1B3, B1B4, B1B5, B1B6, B1B7,
	68	B2B0, B2B1, B2B2, B2B3, B2B4, B2B5, B2B6, B2B7,
	69	B3B0, B3B1, B3B2, B3B3, B3B4, B3B5, B3B6, B3B7,
	70	B4B0, B4B1, B4B2, B4B3, B4B4, B4B5, B4B6, B4B7,
	71	B5B0, B5B1, B5B2, B5B3, B5B4, B5B5, B5B6, B5B7,
	72	B6B0, B6B1, B6B2, B6B3, B6B4, B6B5, B6B6, B6B7,
	73	B7B0, B7B1, B7B2, B7B3, B7B4, B7B5, B7B6, B7B7,
	74	};
	75
849f1035	76	static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
65e4c8c9 MN	77	{
	78	FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
	79	FLOAT tmp10, tmp11, tmp12, tmp13;
	80	FLOAT z1, z2, z3, z4, z5, z11, z13;
65e4c8c9 MN	81	int i;
65e4c8c9 MN	82
65e4c8c9 MN	83	for (i=0; i<8*8; i+=8) {
	84	tmp0= data[0 + i] + data[7 + i];
	85	tmp7= data[0 + i] - data[7 + i];
	86	tmp1= data[1 + i] + data[6 + i];
	87	tmp6= data[1 + i] - data[6 + i];
	88	tmp2= data[2 + i] + data[5 + i];
	89	tmp5= data[2 + i] - data[5 + i];
	90	tmp3= data[3 + i] + data[4 + i];
	91	tmp4= data[3 + i] - data[4 + i];
115329f1	92
65e4c8c9 MN	93	tmp10= tmp0 + tmp3;
	94	tmp13= tmp0 - tmp3;
	95	tmp11= tmp1 + tmp2;
	96	tmp12= tmp1 - tmp2;
115329f1	97
65e4c8c9 MN	98	temp[0 + i]= tmp10 + tmp11;
65e4c8c9 MN	99	temp[4 + i]= tmp10 - tmp11;
115329f1	100
65e4c8c9 MN	101	z1= (tmp12 + tmp13)*A1;
	102	temp[2 + i]= tmp13 + z1;
	103	temp[6 + i]= tmp13 - z1;
115329f1	104
65e4c8c9 MN	105	tmp10= tmp4 + tmp5;
	106	tmp11= tmp5 + tmp6;
	107	tmp12= tmp6 + tmp7;
	108
	109	z5= (tmp10 - tmp12) * A5;
	110	z2= tmp10*A2 + z5;
	111	z4= tmp12*A4 + z5;
	112	z3= tmp11*A1;
	113
	114	z11= tmp7 + z3;
	115	z13= tmp7 - z3;
	116
	117	temp[5 + i]= z13 + z2;
	118	temp[3 + i]= z13 - z2;
	119	temp[1 + i]= z11 + z4;
	120	temp[7 + i]= z11 - z4;
115329f1	121	}
d43fb4e8 MN	122	}
	123
	124	void ff_faandct(DCTELEM * data)
	125	{
	126	FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
	127	FLOAT tmp10, tmp11, tmp12, tmp13;
	128	FLOAT z1, z2, z3, z4, z5, z11, z13;
	129	FLOAT temp[64];
	130	int i;
	131
	132	emms_c();
	133
	134	row_fdct(temp, data);
65e4c8c9 MN	135
	136	for (i=0; i<8; i++) {
	137	tmp0= temp[80 + i] + temp[87 + i];
	138	tmp7= temp[80 + i] - temp[87 + i];
	139	tmp1= temp[81 + i] + temp[86 + i];
	140	tmp6= temp[81 + i] - temp[86 + i];
	141	tmp2= temp[82 + i] + temp[85 + i];
	142	tmp5= temp[82 + i] - temp[85 + i];
	143	tmp3= temp[83 + i] + temp[84 + i];
	144	tmp4= temp[83 + i] - temp[84 + i];
115329f1	145
65e4c8c9 MN	146	tmp10= tmp0 + tmp3;
	147	tmp13= tmp0 - tmp3;
	148	tmp11= tmp1 + tmp2;
	149	tmp12= tmp1 - tmp2;
115329f1	150
8d00c562 MN	151	data[80 + i]= lrintf(SCALE(80 + i) * (tmp10 + tmp11));
8d00c562 MN	152	data[84 + i]= lrintf(SCALE(84 + i) * (tmp10 - tmp11));
115329f1	153
65e4c8c9	154	z1= (tmp12 + tmp13)* A1;
8d00c562 MN	155	data[82 + i]= lrintf(SCALE(82 + i) * (tmp13 + z1));
8d00c562 MN	156	data[86 + i]= lrintf(SCALE(86 + i) * (tmp13 - z1));
115329f1	157
65e4c8c9 MN	158	tmp10= tmp4 + tmp5;
	159	tmp11= tmp5 + tmp6;
	160	tmp12= tmp6 + tmp7;
	161
	162	z5= (tmp10 - tmp12) * A5;
	163	z2= tmp10*A2 + z5;
	164	z4= tmp12*A4 + z5;
	165	z3= tmp11*A1;
	166
	167	z11= tmp7 + z3;
	168	z13= tmp7 - z3;
	169
8d00c562 MN	170	data[85 + i]= lrintf(SCALE(85 + i) * (z13 + z2));
	171	data[83 + i]= lrintf(SCALE(83 + i) * (z13 - z2));
	172	data[81 + i]= lrintf(SCALE(81 + i) * (z11 + z4));
	173	data[87 + i]= lrintf(SCALE(87 + i) * (z11 - z4));
65e4c8c9 MN	174	}
65e4c8c9 MN	175	}
48b1f800 RS	176
	177	void ff_faandct248(DCTELEM * data)
	178	{
	179	FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
	180	FLOAT tmp10, tmp11, tmp12, tmp13;
d43fb4e8	181	FLOAT z1;
48b1f800 RS	182	FLOAT temp[64];
	183	int i;
	184
	185	emms_c();
	186
d43fb4e8	187	row_fdct(temp, data);
48b1f800 RS	188
	189	for (i=0; i<8; i++) {
	190	tmp0 = temp[80 + i] + temp[81 + i];
	191	tmp1 = temp[82 + i] + temp[83 + i];
	192	tmp2 = temp[84 + i] + temp[85 + i];
	193	tmp3 = temp[86 + i] + temp[87 + i];
	194	tmp4 = temp[80 + i] - temp[81 + i];
	195	tmp5 = temp[82 + i] - temp[83 + i];
	196	tmp6 = temp[84 + i] - temp[85 + i];
	197	tmp7 = temp[86 + i] - temp[87 + i];
115329f1	198
48b1f800 RS	199	tmp10 = tmp0 + tmp3;
	200	tmp11 = tmp1 + tmp2;
	201	tmp12 = tmp1 - tmp2;
	202	tmp13 = tmp0 - tmp3;
115329f1	203
48b1f800 RS	204	data[80 + i] = lrintf(SCALE(80 + i) * (tmp10 + tmp11));
48b1f800 RS	205	data[84 + i] = lrintf(SCALE(84 + i) * (tmp10 - tmp11));
115329f1	206
48b1f800 RS	207	z1 = (tmp12 + tmp13)* A1;
	208	data[82 + i] = lrintf(SCALE(82 + i) * (tmp13 + z1));
	209	data[86 + i] = lrintf(SCALE(86 + i) * (tmp13 - z1));
115329f1	210
48b1f800	211	tmp10 = tmp4 + tmp7;
bb270c08 DB	212	tmp11 = tmp5 + tmp6;
	213	tmp12 = tmp5 - tmp6;
	214	tmp13 = tmp4 - tmp7;
48b1f800	215
bb270c08 DB	216	data[81 + i] = lrintf(SCALE(80 + i) * (tmp10 + tmp11));
bb270c08 DB	217	data[85 + i] = lrintf(SCALE(84 + i) * (tmp10 - tmp11));
48b1f800	218
bb270c08 DB	219	z1 = (tmp12 + tmp13)* A1;
	220	data[83 + i] = lrintf(SCALE(82 + i) * (tmp13 + z1));
	221	data[87 + i] = lrintf(SCALE(86 + i) * (tmp13 - z1));
48b1f800 RS	222	}
48b1f800 RS	223	}