/src/ffmpeg/libavcodec/rv60dsp.c

Source
/*
 * RV60 dsp routines
 *
 * 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
 */

#include "rv60dsp.h"
#include "libavutil/common.h"

void ff_rv60_idct4x4_add(const int16_t * block, uint8_t * dst, int dst_stride)
{
    int tmp[16];
#define IDCT4X4(src, src_stride, src_step, dst, dst_stride, dst_step) \
    for (int y = 0; y < 4; y++) { \
        int a = src[y*src_stride + 0*src_step]; \
        int b = src[y*src_stride + 1*src_step]; \
        int c = src[y*src_stride + 2*src_step]; \
        int d = src[y*src_stride + 3*src_step]; \
        int t0 = 13 * (a + c); \
        int t1 = 13 * (a - c); \
        int t2 = 7 * b - 17 * d; \
        int t3 = 7 * d + 17 * b; \
        STORE(dst[y*dst_stride + 0*dst_step], (t0 + t3 + 16) >> 5); \
        STORE(dst[y*dst_stride + 1*dst_step], (t1 + t2 + 16) >> 5); \
        STORE(dst[y*dst_stride + 2*dst_step], (t1 - t2 + 16) >> 5); \
        STORE(dst[y*dst_stride + 3*dst_step], (t0 - t3 + 16) >> 5); \
    }
#define STORE(a, b) a = b
    IDCT4X4(block, 1, 4, tmp, 1, 4)
#undef STORE
#define STORE(a, b) a = av_clip_uint8(a + (b))
    IDCT4X4(tmp, 4, 1, dst, dst_stride, 1)
#undef STORE
}

void ff_rv60_idct8x8_add(const int16_t * block, uint8_t * dst, int dst_stride)
{
    int tmp[64];
#define IDCT8X8(src, src_stride, src_step, dst, dst_stride, dst_step) \
    for (int y = 0; y < 8; y++) { \
        int a = src[y*src_stride + 0*src_step]; \
        int b = src[y*src_stride + 1*src_step]; \
        int c = src[y*src_stride + 2*src_step]; \
        int d = src[y*src_stride + 3*src_step]; \
        int e = src[y*src_stride + 4*src_step]; \
        int f = src[y*src_stride + 5*src_step]; \
        int g = src[y*src_stride + 6*src_step]; \
        int h = src[y*src_stride + 7*src_step]; \
        int t0 = 37 * (a + e); \
        int t1 = 37 * (a - e); \
        int t2 = 48 * c + 20 * g; \
        int t3 = 20 * c - 48 * g; \
        int t4 = t0 + t2; \
        int t5 = t0 - t2; \
        int t6 = t1 + t3; \
        int t7 = t1 - t3;  \
        int t8 = 51 * b + 43 * d + 29 * f + 10 * h; \
        int t9 = 43 * b - 10 * d - 51 * f - 29 * h; \
        int ta = 29 * b - 51 * d + 10 * f + 43 * h; \
        int tb = 10 * b - 29 * d + 43 * f - 51 * h; \
        STORE(dst[y*dst_stride + 0*dst_step], (t4 + t8 + 64) >> 7); \
        STORE(dst[y*dst_stride + 1*dst_step], (t6 + t9 + 64) >> 7); \
        STORE(dst[y*dst_stride + 2*dst_step], (t7 + ta + 64) >> 7); \
        STORE(dst[y*dst_stride + 3*dst_step], (t5 + tb + 64) >> 7); \
        STORE(dst[y*dst_stride + 4*dst_step], (t5 - tb + 64) >> 7); \
        STORE(dst[y*dst_stride + 5*dst_step], (t7 - ta + 64) >> 7); \
        STORE(dst[y*dst_stride + 6*dst_step], (t6 - t9 + 64) >> 7); \
        STORE(dst[y*dst_stride + 7*dst_step], (t4 - t8 + 64) >> 7); \
    }
#define STORE(a, b) a = b
    IDCT8X8(block, 1, 8, tmp, 1, 8)
#undef STORE
#define STORE(a, b) a = av_clip_uint8(a + (b))
    IDCT8X8(tmp, 8, 1, dst, dst_stride, 1)
#undef STORE
}

void ff_rv60_idct16x16_add(const int16_t * block, uint8_t * dst, int dst_stride)
{
    int16_t tmp[256];
#define IDCT16X16(src, src_stride, src_step, dst, dst_stride, dst_step) \
    for (int y = 0; y < 16; y++) { \
        int a = src[y*src_stride +  0*src_step]; \
        int b = src[y*src_stride +  1*src_step]; \
        int c = src[y*src_stride +  2*src_step]; \
        int d = src[y*src_stride +  3*src_step]; \
        int e = src[y*src_stride +  4*src_step]; \
        int f = src[y*src_stride +  5*src_step]; \
        int g = src[y*src_stride +  6*src_step]; \
        int h = src[y*src_stride +  7*src_step]; \
        int i = src[y*src_stride +  8*src_step]; \
        int j = src[y*src_stride +  9*src_step]; \
        int k = src[y*src_stride + 10*src_step]; \
        int l = src[y*src_stride + 11*src_step]; \
        int m = src[y*src_stride + 12*src_step]; \
        int n = src[y*src_stride + 13*src_step]; \
        int o = src[y*src_stride + 14*src_step]; \
        int p = src[y*src_stride + 15*src_step]; \
        int t0 = 26 * (a + i); \
        int t1 = 26 * (a - i); \
        int t2 = 14 * e - 34 * m; \
        int t3 = 34 * e + 14 * m; \
        int t4 = t0 + t3; \
        int t5 = t0 - t3; \
        int t6 = t1 + t2; \
        int t7 = t1 - t2; \
        int tmp00 = 31 * c -  7 * g - 36 * k - 20 * o; \
        int tmp01 = 36 * c + 31 * g + 20 * k +  7 * o; \
        int tmp02 = 20 * c - 36 * g +  7 * k + 31 * o; \
        int tmp03 =  7 * c - 20 * g + 31 * k - 36 * o; \
        int tm0 = t4 + tmp01; \
        int tm1 = t4 - tmp01; \
        int tm2 = t5 + tmp03; \
        int tm3 = t5 - tmp03; \
        int tm4 = t6 + tmp00; \
        int tm5 = t6 - tmp00; \
        int tm6 = t7 + tmp02; \
        int tm7 = t7 - tmp02; \
        int tt0 = 37 * b + 35 * d + 32 * f + 28 * h + 23 * j + 17 * l + 11 * n +  4 * p; \
        int tt1 = 35 * b + 23 * d +  4 * f - 17 * h - 32 * j - 37 * l - 28 * n - 11 * p; \
        int tt2 = 32 * b +  4 * d - 28 * f - 35 * h - 11 * j + 23 * l + 37 * n + 17 * p; \
        int tt3 = 28 * b - 17 * d - 35 * f +  4 * h + 37 * j + 11 * l - 32 * n - 23 * p; \
        int tt4 = 23 * b - 32 * d - 11 * f + 37 * h -  4 * j - 35 * l + 17 * n + 28 * p; \
        int tt5 = 17 * b - 37 * d + 23 * f + 11 * h - 35 * j + 28 * l +  4 * n - 32 * p; \
        int tt6 = 11 * b - 28 * d + 37 * f - 32 * h + 17 * j +  4 * l - 23 * n + 35 * p; \
        int tt7 =  4 * b - 11 * d + 17 * f - 23 * h + 28 * j - 32 * l + 35 * n - 37 * p; \
        STORE(dst[y*dst_stride+  0*dst_step], (tm0 + tt0 + 64) >> 7); \
        STORE(dst[y*dst_stride+  1*dst_step], (tm4 + tt1 + 64) >> 7); \
        STORE(dst[y*dst_stride+  2*dst_step], (tm6 + tt2 + 64) >> 7); \
        STORE(dst[y*dst_stride+  3*dst_step], (tm2 + tt3 + 64) >> 7); \
        STORE(dst[y*dst_stride+  4*dst_step], (tm3 + tt4 + 64) >> 7); \
        STORE(dst[y*dst_stride+  5*dst_step], (tm7 + tt5 + 64) >> 7); \
        STORE(dst[y*dst_stride+  6*dst_step], (tm5 + tt6 + 64) >> 7); \
        STORE(dst[y*dst_stride+  7*dst_step], (tm1 + tt7 + 64) >> 7); \
        STORE(dst[y*dst_stride+  8*dst_step], (tm1 - tt7 + 64) >> 7); \
        STORE(dst[y*dst_stride+  9*dst_step], (tm5 - tt6 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 10*dst_step], (tm7 - tt5 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 11*dst_step], (tm3 - tt4 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 12*dst_step], (tm2 - tt3 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 13*dst_step], (tm6 - tt2 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 14*dst_step], (tm4 - tt1 + 64) >> 7); \
        STORE(dst[y*dst_stride+ 15*dst_step], (tm0 - tt0 + 64) >> 7); \
    }
#define STORE(a, x) a = av_clip_intp2(x, 15)
    IDCT16X16(block, 1, 16, tmp, 1, 16)
#undef STORE
#define STORE(a, x) a = av_clip_uint8(a + (x))
    IDCT16X16(tmp, 16, 1, dst, dst_stride, 1)
#undef STORE
}

Line	Count	Source
1		/*
2		* RV60 dsp routines
3		*
4		* This file is part of FFmpeg.
5		*
6		* FFmpeg is free software; you can redistribute it and/or
7		* modify it under the terms of the GNU Lesser General Public
8		* License as published by the Free Software Foundation; either
9		* version 2.1 of the License, or (at your option) any later version.
10		*
11		* FFmpeg is distributed in the hope that it will be useful,
12		* but WITHOUT ANY WARRANTY; without even the implied warranty of
13		* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14		* Lesser General Public License for more details.
15		*
16		* You should have received a copy of the GNU Lesser General Public
17		* License along with FFmpeg; if not, write to the Free Software
18		* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19		*/
20
21		#include "rv60dsp.h"
22		#include "libavutil/common.h"
23
24		void ff_rv60_idct4x4_add(const int16_t * block, uint8_t * dst, int dst_stride)
25	19.2M	{
26	19.2M	int tmp[16];
27	19.2M	#define IDCT4X4(src, src_stride, src_step, dst, dst_stride, dst_step) \
28	192M	for (int y = 0; y < 4; y++) { \
29	154M	int a = src[ysrc_stride + 0src_step]; \
30	154M	int b = src[ysrc_stride + 1src_step]; \
31	154M	int c = src[ysrc_stride + 2src_step]; \
32	154M	int d = src[ysrc_stride + 3src_step]; \
33	154M	int t0 = 13 * (a + c); \
34	154M	int t1 = 13 * (a - c); \
35	154M	int t2 = 7 * b - 17 * d; \
36	154M	int t3 = 7 * d + 17 * b; \
37	154M	STORE(dst[ydst_stride + 0dst_step], (t0 + t3 + 16) >> 5); \
38	154M	STORE(dst[ydst_stride + 1dst_step], (t1 + t2 + 16) >> 5); \
39	154M	STORE(dst[ydst_stride + 2dst_step], (t1 - t2 + 16) >> 5); \
40	154M	STORE(dst[ydst_stride + 3dst_step], (t0 - t3 + 16) >> 5); \
41	154M	}
42	308M	#define STORE(a, b) a = b
43	19.2M	IDCT4X4(block, 1, 4, tmp, 1, 4)
44	19.2M	#undef STORE
45	308M	#define STORE(a, b) a = av_clip_uint8(a + (b))
46	19.2M	IDCT4X4(tmp, 4, 1, dst, dst_stride, 1)
47	19.2M	#undef STORE
48	19.2M	}
49
50		void ff_rv60_idct8x8_add(const int16_t * block, uint8_t * dst, int dst_stride)
51	8.81M	{
52	8.81M	int tmp[64];
53	8.81M	#define IDCT8X8(src, src_stride, src_step, dst, dst_stride, dst_step) \
54	158M	for (int y = 0; y < 8; y++) { \
55	141M	int a = src[ysrc_stride + 0src_step]; \
56	141M	int b = src[ysrc_stride + 1src_step]; \
57	141M	int c = src[ysrc_stride + 2src_step]; \
58	141M	int d = src[ysrc_stride + 3src_step]; \
59	141M	int e = src[ysrc_stride + 4src_step]; \
60	141M	int f = src[ysrc_stride + 5src_step]; \
61	141M	int g = src[ysrc_stride + 6src_step]; \
62	141M	int h = src[ysrc_stride + 7src_step]; \
63	141M	int t0 = 37 * (a + e); \
64	141M	int t1 = 37 * (a - e); \
65	141M	int t2 = 48 * c + 20 * g; \
66	141M	int t3 = 20 * c - 48 * g; \
67	141M	int t4 = t0 + t2; \
68	141M	int t5 = t0 - t2; \
69	141M	int t6 = t1 + t3; \
70	141M	int t7 = t1 - t3; \
71	141M	int t8 = 51 * b + 43 * d + 29 * f + 10 * h; \
72	141M	int t9 = 43 * b - 10 * d - 51 * f - 29 * h; \
73	141M	int ta = 29 * b - 51 * d + 10 * f + 43 * h; \
74	141M	int tb = 10 * b - 29 * d + 43 * f - 51 * h; \
75	141M	STORE(dst[ydst_stride + 0dst_step], (t4 + t8 + 64) >> 7); \
76	141M	STORE(dst[ydst_stride + 1dst_step], (t6 + t9 + 64) >> 7); \
77	141M	STORE(dst[ydst_stride + 2dst_step], (t7 + ta + 64) >> 7); \
78	141M	STORE(dst[ydst_stride + 3dst_step], (t5 + tb + 64) >> 7); \
79	141M	STORE(dst[ydst_stride + 4dst_step], (t5 - tb + 64) >> 7); \
80	141M	STORE(dst[ydst_stride + 5dst_step], (t7 - ta + 64) >> 7); \
81	141M	STORE(dst[ydst_stride + 6dst_step], (t6 - t9 + 64) >> 7); \
82	141M	STORE(dst[ydst_stride + 7dst_step], (t4 - t8 + 64) >> 7); \
83	141M	}
84	564M	#define STORE(a, b) a = b
85	8.81M	IDCT8X8(block, 1, 8, tmp, 1, 8)
86	8.81M	#undef STORE
87	564M	#define STORE(a, b) a = av_clip_uint8(a + (b))
88	8.81M	IDCT8X8(tmp, 8, 1, dst, dst_stride, 1)
89	8.81M	#undef STORE
90	8.81M	}
91
92		void ff_rv60_idct16x16_add(const int16_t * block, uint8_t * dst, int dst_stride)
93	2.39M	{
94	2.39M	int16_t tmp[256];
95	2.39M	#define IDCT16X16(src, src_stride, src_step, dst, dst_stride, dst_step) \
96	81.3M	for (int y = 0; y < 16; y++) { \
97	76.5M	int a = src[ysrc_stride + 0src_step]; \
98	76.5M	int b = src[ysrc_stride + 1src_step]; \
99	76.5M	int c = src[ysrc_stride + 2src_step]; \
100	76.5M	int d = src[ysrc_stride + 3src_step]; \
101	76.5M	int e = src[ysrc_stride + 4src_step]; \
102	76.5M	int f = src[ysrc_stride + 5src_step]; \
103	76.5M	int g = src[ysrc_stride + 6src_step]; \
104	76.5M	int h = src[ysrc_stride + 7src_step]; \
105	76.5M	int i = src[ysrc_stride + 8src_step]; \
106	76.5M	int j = src[ysrc_stride + 9src_step]; \
107	76.5M	int k = src[ysrc_stride + 10src_step]; \
108	76.5M	int l = src[ysrc_stride + 11src_step]; \
109	76.5M	int m = src[ysrc_stride + 12src_step]; \
110	76.5M	int n = src[ysrc_stride + 13src_step]; \
111	76.5M	int o = src[ysrc_stride + 14src_step]; \
112	76.5M	int p = src[ysrc_stride + 15src_step]; \
113	76.5M	int t0 = 26 * (a + i); \
114	76.5M	int t1 = 26 * (a - i); \
115	76.5M	int t2 = 14 * e - 34 * m; \
116	76.5M	int t3 = 34 * e + 14 * m; \
117	76.5M	int t4 = t0 + t3; \
118	76.5M	int t5 = t0 - t3; \
119	76.5M	int t6 = t1 + t2; \
120	76.5M	int t7 = t1 - t2; \
121	76.5M	int tmp00 = 31 * c - 7 * g - 36 * k - 20 * o; \
122	76.5M	int tmp01 = 36 * c + 31 * g + 20 * k + 7 * o; \
123	76.5M	int tmp02 = 20 * c - 36 * g + 7 * k + 31 * o; \
124	76.5M	int tmp03 = 7 * c - 20 * g + 31 * k - 36 * o; \
125	76.5M	int tm0 = t4 + tmp01; \
126	76.5M	int tm1 = t4 - tmp01; \
127	76.5M	int tm2 = t5 + tmp03; \
128	76.5M	int tm3 = t5 - tmp03; \
129	76.5M	int tm4 = t6 + tmp00; \
130	76.5M	int tm5 = t6 - tmp00; \
131	76.5M	int tm6 = t7 + tmp02; \
132	76.5M	int tm7 = t7 - tmp02; \
133	76.5M	int tt0 = 37 * b + 35 * d + 32 * f + 28 * h + 23 * j + 17 * l + 11 * n + 4 * p; \
134	76.5M	int tt1 = 35 * b + 23 * d + 4 * f - 17 * h - 32 * j - 37 * l - 28 * n - 11 * p; \
135	76.5M	int tt2 = 32 * b + 4 * d - 28 * f - 35 * h - 11 * j + 23 * l + 37 * n + 17 * p; \
136	76.5M	int tt3 = 28 * b - 17 * d - 35 * f + 4 * h + 37 * j + 11 * l - 32 * n - 23 * p; \
137	76.5M	int tt4 = 23 * b - 32 * d - 11 * f + 37 * h - 4 * j - 35 * l + 17 * n + 28 * p; \
138	76.5M	int tt5 = 17 * b - 37 * d + 23 * f + 11 * h - 35 * j + 28 * l + 4 * n - 32 * p; \
139	76.5M	int tt6 = 11 * b - 28 * d + 37 * f - 32 * h + 17 * j + 4 * l - 23 * n + 35 * p; \
140	76.5M	int tt7 = 4 * b - 11 * d + 17 * f - 23 * h + 28 * j - 32 * l + 35 * n - 37 * p; \
141	76.5M	STORE(dst[ydst_stride+ 0dst_step], (tm0 + tt0 + 64) >> 7); \
142	76.5M	STORE(dst[ydst_stride+ 1dst_step], (tm4 + tt1 + 64) >> 7); \
143	76.5M	STORE(dst[ydst_stride+ 2dst_step], (tm6 + tt2 + 64) >> 7); \
144	76.5M	STORE(dst[ydst_stride+ 3dst_step], (tm2 + tt3 + 64) >> 7); \
145	76.5M	STORE(dst[ydst_stride+ 4dst_step], (tm3 + tt4 + 64) >> 7); \
146	76.5M	STORE(dst[ydst_stride+ 5dst_step], (tm7 + tt5 + 64) >> 7); \
147	76.5M	STORE(dst[ydst_stride+ 6dst_step], (tm5 + tt6 + 64) >> 7); \
148	76.5M	STORE(dst[ydst_stride+ 7dst_step], (tm1 + tt7 + 64) >> 7); \
149	76.5M	STORE(dst[ydst_stride+ 8dst_step], (tm1 - tt7 + 64) >> 7); \
150	76.5M	STORE(dst[ydst_stride+ 9dst_step], (tm5 - tt6 + 64) >> 7); \
151	76.5M	STORE(dst[ydst_stride+ 10dst_step], (tm7 - tt5 + 64) >> 7); \
152	76.5M	STORE(dst[ydst_stride+ 11dst_step], (tm3 - tt4 + 64) >> 7); \
153	76.5M	STORE(dst[ydst_stride+ 12dst_step], (tm2 - tt3 + 64) >> 7); \
154	76.5M	STORE(dst[ydst_stride+ 13dst_step], (tm6 - tt2 + 64) >> 7); \
155	76.5M	STORE(dst[ydst_stride+ 14dst_step], (tm4 - tt1 + 64) >> 7); \
156	76.5M	STORE(dst[ydst_stride+ 15dst_step], (tm0 - tt0 + 64) >> 7); \
157	76.5M	}
158	612M	#define STORE(a, x) a = av_clip_intp2(x, 15)
159	2.39M	IDCT16X16(block, 1, 16, tmp, 1, 16)
160	2.39M	#undef STORE
161	612M	#define STORE(a, x) a = av_clip_uint8(a + (x))
162	2.39M	IDCT16X16(tmp, 16, 1, dst, dst_stride, 1)
163	2.39M	#undef STORE
164	2.39M	}

Coverage Report

Created: 2025-11-16 07:20