1 /*
2 * The simplest mpeg encoder (well, it was the simplest!)
3 * Copyright (c) 2000,2001 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/x86/asm.h"
25 #include "libavutil/x86/cpu.h"
26 #include "libavcodec/avcodec.h"
27 #include "libavcodec/dct.h"
28 #include "libavcodec/mpegvideo.h"
29
30 /* not permutated inverse zigzag_direct + 1 for MMX quantizer */
31 DECLARE_ALIGNED(16, static const uint16_t, inv_zigzag_direct16)[64] = {
32 1, 2, 6, 7, 15, 16, 28, 29,
33 3, 5, 8, 14, 17, 27, 30, 43,
34 4, 9, 13, 18, 26, 31, 42, 44,
35 10, 12, 19, 25, 32, 41, 45, 54,
36 11, 20, 24, 33, 40, 46, 53, 55,
37 21, 23, 34, 39, 47, 52, 56, 61,
38 22, 35, 38, 48, 51, 57, 60, 62,
39 36, 37, 49, 50, 58, 59, 63, 64,
40 };
41
42 #if HAVE_6REGS
43
44 #if HAVE_MMX_INLINE
45 #define COMPILE_TEMPLATE_MMXEXT 0
46 #define COMPILE_TEMPLATE_SSE2 0
47 #define COMPILE_TEMPLATE_SSSE3 0
48 #define RENAME(a) a ## _mmx
49 #define RENAME_FDCT(a) a ## _mmx
50 #include "mpegvideoenc_template.c"
51 #endif /* HAVE_MMX_INLINE */
52
53 #if HAVE_MMXEXT_INLINE
54 #undef COMPILE_TEMPLATE_SSSE3
55 #undef COMPILE_TEMPLATE_SSE2
56 #undef COMPILE_TEMPLATE_MMXEXT
57 #define COMPILE_TEMPLATE_MMXEXT 1
58 #define COMPILE_TEMPLATE_SSE2 0
59 #define COMPILE_TEMPLATE_SSSE3 0
60 #undef RENAME
61 #undef RENAME_FDCT
62 #define RENAME(a) a ## _mmxext
63 #define RENAME_FDCT(a) a ## _mmxext
64 #include "mpegvideoenc_template.c"
65 #endif /* HAVE_MMXEXT_INLINE */
66
67 #if HAVE_SSE2_INLINE
68 #undef COMPILE_TEMPLATE_MMXEXT
69 #undef COMPILE_TEMPLATE_SSE2
70 #undef COMPILE_TEMPLATE_SSSE3
71 #define COMPILE_TEMPLATE_MMXEXT 0
72 #define COMPILE_TEMPLATE_SSE2 1
73 #define COMPILE_TEMPLATE_SSSE3 0
74 #undef RENAME
75 #undef RENAME_FDCT
76 #define RENAME(a) a ## _sse2
77 #define RENAME_FDCT(a) a ## _sse2
78 #include "mpegvideoenc_template.c"
79 #endif /* HAVE_SSE2_INLINE */
80
81 #if HAVE_SSSE3_INLINE
82 #undef COMPILE_TEMPLATE_MMXEXT
83 #undef COMPILE_TEMPLATE_SSE2
84 #undef COMPILE_TEMPLATE_SSSE3
85 #define COMPILE_TEMPLATE_MMXEXT 0
86 #define COMPILE_TEMPLATE_SSE2 1
87 #define COMPILE_TEMPLATE_SSSE3 1
88 #undef RENAME
89 #undef RENAME_FDCT
90 #define RENAME(a) a ## _ssse3
91 #define RENAME_FDCT(a) a ## _sse2
92 #include "mpegvideoenc_template.c"
93 #endif /* HAVE_SSSE3_INLINE */
94
95 #endif /* HAVE_6REGS */
96
97 #if HAVE_INLINE_ASM
98 #if HAVE_MMX_INLINE
denoise_dct_mmx(MpegEncContext * s,int16_t * block)99 static void denoise_dct_mmx(MpegEncContext *s, int16_t *block){
100 const int intra= s->mb_intra;
101 int *sum= s->dct_error_sum[intra];
102 uint16_t *offset= s->dct_offset[intra];
103
104 s->dct_count[intra]++;
105
106 __asm__ volatile(
107 "pxor %%mm7, %%mm7 \n\t"
108 "1: \n\t"
109 "pxor %%mm0, %%mm0 \n\t"
110 "pxor %%mm1, %%mm1 \n\t"
111 "movq (%0), %%mm2 \n\t"
112 "movq 8(%0), %%mm3 \n\t"
113 "pcmpgtw %%mm2, %%mm0 \n\t"
114 "pcmpgtw %%mm3, %%mm1 \n\t"
115 "pxor %%mm0, %%mm2 \n\t"
116 "pxor %%mm1, %%mm3 \n\t"
117 "psubw %%mm0, %%mm2 \n\t"
118 "psubw %%mm1, %%mm3 \n\t"
119 "movq %%mm2, %%mm4 \n\t"
120 "movq %%mm3, %%mm5 \n\t"
121 "psubusw (%2), %%mm2 \n\t"
122 "psubusw 8(%2), %%mm3 \n\t"
123 "pxor %%mm0, %%mm2 \n\t"
124 "pxor %%mm1, %%mm3 \n\t"
125 "psubw %%mm0, %%mm2 \n\t"
126 "psubw %%mm1, %%mm3 \n\t"
127 "movq %%mm2, (%0) \n\t"
128 "movq %%mm3, 8(%0) \n\t"
129 "movq %%mm4, %%mm2 \n\t"
130 "movq %%mm5, %%mm3 \n\t"
131 "punpcklwd %%mm7, %%mm4 \n\t"
132 "punpckhwd %%mm7, %%mm2 \n\t"
133 "punpcklwd %%mm7, %%mm5 \n\t"
134 "punpckhwd %%mm7, %%mm3 \n\t"
135 "paddd (%1), %%mm4 \n\t"
136 "paddd 8(%1), %%mm2 \n\t"
137 "paddd 16(%1), %%mm5 \n\t"
138 "paddd 24(%1), %%mm3 \n\t"
139 "movq %%mm4, (%1) \n\t"
140 "movq %%mm2, 8(%1) \n\t"
141 "movq %%mm5, 16(%1) \n\t"
142 "movq %%mm3, 24(%1) \n\t"
143 "add $16, %0 \n\t"
144 "add $32, %1 \n\t"
145 "add $16, %2 \n\t"
146 "cmp %3, %0 \n\t"
147 " jb 1b \n\t"
148 : "+r" (block), "+r" (sum), "+r" (offset)
149 : "r"(block+64)
150 );
151 }
152 #endif /* HAVE_MMX_INLINE */
153
154 #if HAVE_SSE2_INLINE
denoise_dct_sse2(MpegEncContext * s,int16_t * block)155 static void denoise_dct_sse2(MpegEncContext *s, int16_t *block){
156 const int intra= s->mb_intra;
157 int *sum= s->dct_error_sum[intra];
158 uint16_t *offset= s->dct_offset[intra];
159
160 s->dct_count[intra]++;
161
162 __asm__ volatile(
163 "pxor %%xmm7, %%xmm7 \n\t"
164 "1: \n\t"
165 "pxor %%xmm0, %%xmm0 \n\t"
166 "pxor %%xmm1, %%xmm1 \n\t"
167 "movdqa (%0), %%xmm2 \n\t"
168 "movdqa 16(%0), %%xmm3 \n\t"
169 "pcmpgtw %%xmm2, %%xmm0 \n\t"
170 "pcmpgtw %%xmm3, %%xmm1 \n\t"
171 "pxor %%xmm0, %%xmm2 \n\t"
172 "pxor %%xmm1, %%xmm3 \n\t"
173 "psubw %%xmm0, %%xmm2 \n\t"
174 "psubw %%xmm1, %%xmm3 \n\t"
175 "movdqa %%xmm2, %%xmm4 \n\t"
176 "movdqa %%xmm3, %%xmm5 \n\t"
177 "psubusw (%2), %%xmm2 \n\t"
178 "psubusw 16(%2), %%xmm3 \n\t"
179 "pxor %%xmm0, %%xmm2 \n\t"
180 "pxor %%xmm1, %%xmm3 \n\t"
181 "psubw %%xmm0, %%xmm2 \n\t"
182 "psubw %%xmm1, %%xmm3 \n\t"
183 "movdqa %%xmm2, (%0) \n\t"
184 "movdqa %%xmm3, 16(%0) \n\t"
185 "movdqa %%xmm4, %%xmm6 \n\t"
186 "movdqa %%xmm5, %%xmm0 \n\t"
187 "punpcklwd %%xmm7, %%xmm4 \n\t"
188 "punpckhwd %%xmm7, %%xmm6 \n\t"
189 "punpcklwd %%xmm7, %%xmm5 \n\t"
190 "punpckhwd %%xmm7, %%xmm0 \n\t"
191 "paddd (%1), %%xmm4 \n\t"
192 "paddd 16(%1), %%xmm6 \n\t"
193 "paddd 32(%1), %%xmm5 \n\t"
194 "paddd 48(%1), %%xmm0 \n\t"
195 "movdqa %%xmm4, (%1) \n\t"
196 "movdqa %%xmm6, 16(%1) \n\t"
197 "movdqa %%xmm5, 32(%1) \n\t"
198 "movdqa %%xmm0, 48(%1) \n\t"
199 "add $32, %0 \n\t"
200 "add $64, %1 \n\t"
201 "add $32, %2 \n\t"
202 "cmp %3, %0 \n\t"
203 " jb 1b \n\t"
204 : "+r" (block), "+r" (sum), "+r" (offset)
205 : "r"(block+64)
206 XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3",
207 "%xmm4", "%xmm5", "%xmm6", "%xmm7")
208 );
209 }
210 #endif /* HAVE_SSE2_INLINE */
211 #endif /* HAVE_INLINE_ASM */
212
ff_dct_encode_init_x86(MpegEncContext * s)213 av_cold void ff_dct_encode_init_x86(MpegEncContext *s)
214 {
215 const int dct_algo = s->avctx->dct_algo;
216
217 if (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX) {
218 #if HAVE_MMX_INLINE
219 int cpu_flags = av_get_cpu_flags();
220 if (INLINE_MMX(cpu_flags)) {
221 #if HAVE_6REGS
222 s->dct_quantize = dct_quantize_mmx;
223 #endif
224 s->denoise_dct = denoise_dct_mmx;
225 }
226 #endif
227 #if HAVE_6REGS && HAVE_MMXEXT_INLINE
228 if (INLINE_MMXEXT(cpu_flags))
229 s->dct_quantize = dct_quantize_mmxext;
230 #endif
231 #if HAVE_SSE2_INLINE
232 if (INLINE_SSE2(cpu_flags)) {
233 #if HAVE_6REGS
234 s->dct_quantize = dct_quantize_sse2;
235 #endif
236 s->denoise_dct = denoise_dct_sse2;
237 }
238 #endif
239 #if HAVE_6REGS && HAVE_SSSE3_INLINE
240 if (INLINE_SSSE3(cpu_flags))
241 s->dct_quantize = dct_quantize_ssse3;
242 #endif
243 }
244 }
245