2 * MMX optimized DSP utils
3 * Copyright (c) 2000, 2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
25 #include "libavutil/attributes.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/x86/asm.h"
28 #include "libavutil/x86/cpu.h"
29 #include "libavcodec/hpeldsp.h"
30 #include "dsputil_x86.h"
32 void ff_put_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,
33 ptrdiff_t line_size, int h);
34 void ff_put_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,
35 ptrdiff_t line_size, int h);
36 void ff_put_pixels16_x2_mmxext(uint8_t *block, const uint8_t *pixels,
37 ptrdiff_t line_size, int h);
38 void ff_put_pixels16_x2_3dnow(uint8_t *block, const uint8_t *pixels,
39 ptrdiff_t line_size, int h);
40 void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,
41 ptrdiff_t line_size, int h);
42 void ff_put_no_rnd_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,
43 ptrdiff_t line_size, int h);
44 void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block,
45 const uint8_t *pixels,
46 ptrdiff_t line_size, int h);
47 void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block,
48 const uint8_t *pixels,
49 ptrdiff_t line_size, int h);
50 void ff_put_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,
51 ptrdiff_t line_size, int h);
52 void ff_put_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,
53 ptrdiff_t line_size, int h);
54 void ff_put_no_rnd_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,
55 ptrdiff_t line_size, int h);
56 void ff_put_no_rnd_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,
57 ptrdiff_t line_size, int h);
58 void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block,
59 const uint8_t *pixels,
60 ptrdiff_t line_size, int h);
61 void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block,
62 const uint8_t *pixels,
63 ptrdiff_t line_size, int h);
64 void ff_avg_pixels8_3dnow(uint8_t *block, const uint8_t *pixels,
65 ptrdiff_t line_size, int h);
66 void ff_avg_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,
67 ptrdiff_t line_size, int h);
68 void ff_avg_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels,
69 ptrdiff_t line_size, int h);
70 void ff_avg_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels,
71 ptrdiff_t line_size, int h);
72 void ff_avg_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels,
73 ptrdiff_t line_size, int h);
74 void ff_avg_pixels8_xy2_mmxext(uint8_t *block, const uint8_t *pixels,
75 ptrdiff_t line_size, int h);
76 void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels,
77 ptrdiff_t line_size, int h);
79 #define avg_pixels8_mmx ff_avg_pixels8_mmx
80 #define avg_pixels8_x2_mmx ff_avg_pixels8_x2_mmx
81 #define avg_pixels16_mmx ff_avg_pixels16_mmx
82 #define avg_pixels8_xy2_mmx ff_avg_pixels8_xy2_mmx
83 #define avg_pixels16_xy2_mmx ff_avg_pixels16_xy2_mmx
84 #define put_pixels8_mmx ff_put_pixels8_mmx
85 #define put_pixels16_mmx ff_put_pixels16_mmx
86 #define put_pixels8_xy2_mmx ff_put_pixels8_xy2_mmx
87 #define put_pixels16_xy2_mmx ff_put_pixels16_xy2_mmx
88 #define avg_no_rnd_pixels16_mmx ff_avg_pixels16_mmx
89 #define put_no_rnd_pixels8_mmx ff_put_pixels8_mmx
90 #define put_no_rnd_pixels16_mmx ff_put_pixels16_mmx
94 /***********************************/
96 #define DEF(x, y) x ## _no_rnd_ ## y ## _mmx
97 #define SET_RND MOVQ_WONE
98 #define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
99 #define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e)
100 #define STATIC static
102 #include "rnd_template.c"
103 #include "hpeldsp_rnd_template.c"
111 PIXELS16(static, avg_no_rnd, , _y2, _mmx)
112 PIXELS16(static, put_no_rnd, , _y2, _mmx)
114 PIXELS16(static, avg_no_rnd, , _xy2, _mmx)
115 PIXELS16(static, put_no_rnd, , _xy2, _mmx)
117 /***********************************/
120 #define DEF(x, y) x ## _ ## y ## _mmx
121 #define SET_RND MOVQ_WTWO
122 #define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f)
123 #define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e)
125 #include "hpeldsp_rnd_template.c"
132 PIXELS16(static, avg, , _y2, _mmx)
133 PIXELS16(static, put, , _y2, _mmx)
135 #endif /* HAVE_INLINE_ASM */
140 #define HPELDSP_AVG_PIXELS16(CPUEXT) \
141 PIXELS16(static, put_no_rnd, ff_, _x2, CPUEXT) \
142 PIXELS16(static, put, ff_, _y2, CPUEXT) \
143 PIXELS16(static, put_no_rnd, ff_, _y2, CPUEXT) \
144 PIXELS16(static, avg, ff_, , CPUEXT) \
145 PIXELS16(static, avg, ff_, _x2, CPUEXT) \
146 PIXELS16(static, avg, ff_, _y2, CPUEXT) \
147 PIXELS16(static, avg, ff_, _xy2, CPUEXT)
149 HPELDSP_AVG_PIXELS16(_3dnow)
150 HPELDSP_AVG_PIXELS16(_mmxext)
152 #endif /* HAVE_YASM */
154 #define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \
156 c->PFX ## _pixels_tab IDX [0] = PFX ## _pixels ## SIZE ## _ ## CPU; \
157 c->PFX ## _pixels_tab IDX [1] = PFX ## _pixels ## SIZE ## _x2_ ## CPU; \
158 c->PFX ## _pixels_tab IDX [2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \
159 c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \
162 static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags)
165 SET_HPEL_FUNCS(put, [0], 16, mmx);
166 SET_HPEL_FUNCS(put_no_rnd, [0], 16, mmx);
167 SET_HPEL_FUNCS(avg, [0], 16, mmx);
168 SET_HPEL_FUNCS(avg_no_rnd, , 16, mmx);
169 SET_HPEL_FUNCS(put, [1], 8, mmx);
170 SET_HPEL_FUNCS(put_no_rnd, [1], 8, mmx);
171 SET_HPEL_FUNCS(avg, [1], 8, mmx);
172 #endif /* HAVE_MMX_INLINE */
175 static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags)
177 #if HAVE_MMXEXT_EXTERNAL
178 c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext;
179 c->put_pixels_tab[0][2] = put_pixels16_y2_mmxext;
181 c->avg_pixels_tab[0][0] = avg_pixels16_mmxext;
182 c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmxext;
183 c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmxext;
185 c->put_pixels_tab[1][1] = ff_put_pixels8_x2_mmxext;
186 c->put_pixels_tab[1][2] = ff_put_pixels8_y2_mmxext;
188 c->avg_pixels_tab[1][0] = ff_avg_pixels8_mmxext;
189 c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_mmxext;
190 c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_mmxext;
192 if (!(flags & CODEC_FLAG_BITEXACT)) {
193 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmxext;
194 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmxext;
195 c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_mmxext;
196 c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_mmxext;
198 c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext;
199 c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmxext;
202 if (flags & CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) {
203 c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext;
204 c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext;
206 #endif /* HAVE_MMXEXT_EXTERNAL */
209 static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags)
211 #if HAVE_AMD3DNOW_EXTERNAL
212 c->put_pixels_tab[0][1] = ff_put_pixels16_x2_3dnow;
213 c->put_pixels_tab[0][2] = put_pixels16_y2_3dnow;
215 c->avg_pixels_tab[0][0] = avg_pixels16_3dnow;
216 c->avg_pixels_tab[0][1] = avg_pixels16_x2_3dnow;
217 c->avg_pixels_tab[0][2] = avg_pixels16_y2_3dnow;
219 c->put_pixels_tab[1][1] = ff_put_pixels8_x2_3dnow;
220 c->put_pixels_tab[1][2] = ff_put_pixels8_y2_3dnow;
222 c->avg_pixels_tab[1][0] = ff_avg_pixels8_3dnow;
223 c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_3dnow;
224 c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_3dnow;
226 if (!(flags & CODEC_FLAG_BITEXACT)){
227 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_3dnow;
228 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_3dnow;
229 c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_3dnow;
230 c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_3dnow;
232 c->avg_pixels_tab[0][3] = avg_pixels16_xy2_3dnow;
233 c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_3dnow;
236 if (flags & CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) {
237 c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow;
238 c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow;
240 #endif /* HAVE_AMD3DNOW_EXTERNAL */
243 static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags)
245 #if HAVE_SSE2_EXTERNAL
246 if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
247 // these functions are slower than mmx on AMD, but faster on Intel
248 c->put_pixels_tab[0][0] = ff_put_pixels16_sse2;
249 c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;
250 c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2;
252 #endif /* HAVE_SSE2_EXTERNAL */
255 av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags)
257 int cpu_flags = av_get_cpu_flags();
259 if (INLINE_MMX(cpu_flags))
260 hpeldsp_init_mmx(c, flags, cpu_flags);
262 if (EXTERNAL_AMD3DNOW(cpu_flags))
263 hpeldsp_init_3dnow(c, flags, cpu_flags);
265 if (EXTERNAL_MMXEXT(cpu_flags))
266 hpeldsp_init_mmxext(c, flags, cpu_flags);
268 if (EXTERNAL_SSE2(cpu_flags))
269 hpeldsp_init_sse2(c, flags, cpu_flags);