2 * This file is part of FFmpeg.
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 #include "libavutil/attributes.h"
20 #include "libavutil/avassert.h"
21 #include "libavutil/cpu.h"
22 #include "libavutil/x86/cpu.h"
23 #include "libavcodec/avcodec.h"
24 #include "libavcodec/mpegvideoencdsp.h"
26 int ff_pix_sum16_mmx(uint8_t *pix, int line_size);
27 int ff_pix_sum16_mmxext(uint8_t *pix, int line_size);
28 int ff_pix_sum16_sse2(uint8_t *pix, int line_size);
29 int ff_pix_sum16_xop(uint8_t *pix, int line_size);
30 int ff_pix_norm1_mmx(uint8_t *pix, int line_size);
31 int ff_pix_norm1_sse2(uint8_t *pix, int line_size);
35 #define PHADDD(a, t) \
36 "movq " #a ", " #t " \n\t" \
37 "psrlq $32, " #a " \n\t" \
38 "paddd " #t ", " #a " \n\t"
41 * pmulhw: dst[0 - 15] = (src[0 - 15] * dst[0 - 15])[16 - 31]
42 * pmulhrw: dst[0 - 15] = (src[0 - 15] * dst[0 - 15] + 0x8000)[16 - 31]
43 * pmulhrsw: dst[0 - 15] = (src[0 - 15] * dst[0 - 15] + 0x4000)[15 - 30]
45 #define PMULHRW(x, y, s, o) \
46 "pmulhw " #s ", " #x " \n\t" \
47 "pmulhw " #s ", " #y " \n\t" \
48 "paddw " #o ", " #x " \n\t" \
49 "paddw " #o ", " #y " \n\t" \
50 "psraw $1, " #x " \n\t" \
51 "psraw $1, " #y " \n\t"
52 #define DEF(x) x ## _mmx
53 #define SET_RND MOVQ_WONE
54 #define SCALE_OFFSET 1
56 #include "mpegvideoenc_qns_template.c"
63 #define DEF(x) x ## _3dnow
65 #define SCALE_OFFSET 0
66 #define PMULHRW(x, y, s, o) \
67 "pmulhrw " #s ", " #x " \n\t" \
68 "pmulhrw " #s ", " #y " \n\t"
70 #include "mpegvideoenc_qns_template.c"
79 #define DEF(x) x ## _ssse3
81 #define SCALE_OFFSET -1
83 #define PHADDD(a, t) \
84 "pshufw $0x0E, " #a ", " #t " \n\t" \
85 /* faster than phaddd on core2 */ \
86 "paddd " #t ", " #a " \n\t"
88 #define PMULHRW(x, y, s, o) \
89 "pmulhrsw " #s ", " #x " \n\t" \
90 "pmulhrsw " #s ", " #y " \n\t"
92 #include "mpegvideoenc_qns_template.c"
99 #endif /* HAVE_SSSE3_INLINE */
101 /* Draw the edges of width 'w' of an image of size width, height
102 * this MMX version can only handle w == 8 || w == 16. */
103 static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
104 int w, int h, int sides)
106 uint8_t *ptr, *last_line;
109 last_line = buf + (height - 1) * wrap;
115 "movd (%0), %%mm0 \n\t"
116 "punpcklbw %%mm0, %%mm0 \n\t"
117 "punpcklwd %%mm0, %%mm0 \n\t"
118 "punpckldq %%mm0, %%mm0 \n\t"
119 "movq %%mm0, -8(%0) \n\t"
120 "movq -8(%0, %2), %%mm1 \n\t"
121 "punpckhbw %%mm1, %%mm1 \n\t"
122 "punpckhwd %%mm1, %%mm1 \n\t"
123 "punpckhdq %%mm1, %%mm1 \n\t"
124 "movq %%mm1, (%0, %2) \n\t"
129 : "r" ((x86_reg) wrap), "r" ((x86_reg) width),
130 "r" (ptr + wrap * height));
131 } else if (w == 16) {
134 "movd (%0), %%mm0 \n\t"
135 "punpcklbw %%mm0, %%mm0 \n\t"
136 "punpcklwd %%mm0, %%mm0 \n\t"
137 "punpckldq %%mm0, %%mm0 \n\t"
138 "movq %%mm0, -8(%0) \n\t"
139 "movq %%mm0, -16(%0) \n\t"
140 "movq -8(%0, %2), %%mm1 \n\t"
141 "punpckhbw %%mm1, %%mm1 \n\t"
142 "punpckhwd %%mm1, %%mm1 \n\t"
143 "punpckhdq %%mm1, %%mm1 \n\t"
144 "movq %%mm1, (%0, %2) \n\t"
145 "movq %%mm1, 8(%0, %2) \n\t"
150 : "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
156 "movd (%0), %%mm0 \n\t"
157 "punpcklbw %%mm0, %%mm0 \n\t"
158 "punpcklwd %%mm0, %%mm0 \n\t"
159 "movd %%mm0, -4(%0) \n\t"
160 "movd -4(%0, %2), %%mm1 \n\t"
161 "punpcklbw %%mm1, %%mm1 \n\t"
162 "punpckhwd %%mm1, %%mm1 \n\t"
163 "punpckhdq %%mm1, %%mm1 \n\t"
164 "movd %%mm1, (%0, %2) \n\t"
169 : "r" ((x86_reg) wrap), "r" ((x86_reg) width),
170 "r" (ptr + wrap * height));
173 /* top and bottom (and hopefully also the corners) */
174 if (sides & EDGE_TOP) {
175 for (i = 0; i < h; i += 4) {
176 ptr = buf - (i + 1) * wrap - w;
179 "movq (%1, %0), %%mm0 \n\t"
180 "movq %%mm0, (%0) \n\t"
181 "movq %%mm0, (%0, %2) \n\t"
182 "movq %%mm0, (%0, %2, 2) \n\t"
183 "movq %%mm0, (%0, %3) \n\t"
188 : "r" ((x86_reg) buf - (x86_reg) ptr - w),
189 "r" ((x86_reg) - wrap), "r" ((x86_reg) - wrap * 3),
190 "r" (ptr + width + 2 * w));
194 if (sides & EDGE_BOTTOM) {
195 for (i = 0; i < h; i += 4) {
196 ptr = last_line + (i + 1) * wrap - w;
199 "movq (%1, %0), %%mm0 \n\t"
200 "movq %%mm0, (%0) \n\t"
201 "movq %%mm0, (%0, %2) \n\t"
202 "movq %%mm0, (%0, %2, 2) \n\t"
203 "movq %%mm0, (%0, %3) \n\t"
208 : "r" ((x86_reg) last_line - (x86_reg) ptr - w),
209 "r" ((x86_reg) wrap), "r" ((x86_reg) wrap * 3),
210 "r" (ptr + width + 2 * w));
215 #endif /* HAVE_INLINE_ASM */
217 av_cold void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
218 AVCodecContext *avctx)
220 int cpu_flags = av_get_cpu_flags();
223 if (EXTERNAL_MMX(cpu_flags)) {
224 c->pix_sum = ff_pix_sum16_mmx;
225 c->pix_norm1 = ff_pix_norm1_mmx;
228 if (EXTERNAL_MMXEXT(cpu_flags)) {
229 c->pix_sum = ff_pix_sum16_mmxext;
233 if (EXTERNAL_SSE2(cpu_flags)) {
234 c->pix_sum = ff_pix_sum16_sse2;
235 c->pix_norm1 = ff_pix_norm1_sse2;
238 if (EXTERNAL_XOP(cpu_flags)) {
239 c->pix_sum = ff_pix_sum16_xop;
244 if (INLINE_MMX(cpu_flags)) {
245 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
246 c->try_8x8basis = try_8x8basis_mmx;
248 c->add_8x8basis = add_8x8basis_mmx;
250 if (avctx->bits_per_raw_sample <= 8) {
251 c->draw_edges = draw_edges_mmx;
255 if (INLINE_AMD3DNOW(cpu_flags)) {
256 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
257 c->try_8x8basis = try_8x8basis_3dnow;
259 c->add_8x8basis = add_8x8basis_3dnow;
262 #if HAVE_SSSE3_INLINE
263 if (INLINE_SSSE3(cpu_flags)) {
264 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
265 c->try_8x8basis = try_8x8basis_ssse3;
267 c->add_8x8basis = add_8x8basis_ssse3;
269 #endif /* HAVE_SSSE3_INLINE */
271 #endif /* HAVE_INLINE_ASM */