2 * MMX optimized DSP utils
3 * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
5 * This file is part of FFmpeg.
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.
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.
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
22 #ifndef AVCODEC_X86_DSPUTIL_MMX_H
23 #define AVCODEC_X86_DSPUTIL_MMX_H
28 #include "libavcodec/dsputil.h"
29 #include "libavutil/x86/asm.h"
30 #include "constants.h"
32 #define MOVQ_WONE(regd) \
34 "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
35 "psrlw $15, %%" #regd ::)
37 #define JUMPALIGN() __asm__ volatile (".p2align 3"::)
38 #define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::)
40 #define MOVQ_BFE(regd) \
42 "pcmpeqd %%"#regd", %%"#regd" \n\t" \
43 "paddb %%"#regd", %%"#regd" \n\t" ::)
46 #define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo))
48 // for shared library it's better to use this way for accessing constants
50 #define MOVQ_WTWO(regd) \
52 "pcmpeqd %%"#regd", %%"#regd" \n\t" \
53 "psrlw $15, %%"#regd" \n\t" \
54 "psllw $1, %%"#regd" \n\t"::)
58 // using regr as temporary and for the output result
59 // first argument is unmodifed and second is trashed
60 // regfe is supposed to contain 0xfefefefefefefefe
61 #define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \
62 "movq "#rega", "#regr" \n\t" \
63 "pand "#regb", "#regr" \n\t" \
64 "pxor "#rega", "#regb" \n\t" \
65 "pand "#regfe", "#regb" \n\t" \
66 "psrlq $1, "#regb" \n\t" \
67 "paddb "#regb", "#regr" \n\t"
69 #define PAVGB_MMX(rega, regb, regr, regfe) \
70 "movq "#rega", "#regr" \n\t" \
71 "por "#regb", "#regr" \n\t" \
72 "pxor "#rega", "#regb" \n\t" \
73 "pand "#regfe", "#regb" \n\t" \
74 "psrlq $1, "#regb" \n\t" \
75 "psubb "#regb", "#regr" \n\t"
77 // mm6 is supposed to contain 0xfefefefefefefefe
78 #define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \
79 "movq "#rega", "#regr" \n\t" \
80 "movq "#regc", "#regp" \n\t" \
81 "pand "#regb", "#regr" \n\t" \
82 "pand "#regd", "#regp" \n\t" \
83 "pxor "#rega", "#regb" \n\t" \
84 "pxor "#regc", "#regd" \n\t" \
85 "pand %%mm6, "#regb" \n\t" \
86 "pand %%mm6, "#regd" \n\t" \
87 "psrlq $1, "#regb" \n\t" \
88 "psrlq $1, "#regd" \n\t" \
89 "paddb "#regb", "#regr" \n\t" \
90 "paddb "#regd", "#regp" \n\t"
92 #define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \
93 "movq "#rega", "#regr" \n\t" \
94 "movq "#regc", "#regp" \n\t" \
95 "por "#regb", "#regr" \n\t" \
96 "por "#regd", "#regp" \n\t" \
97 "pxor "#rega", "#regb" \n\t" \
98 "pxor "#regc", "#regd" \n\t" \
99 "pand %%mm6, "#regb" \n\t" \
100 "pand %%mm6, "#regd" \n\t" \
101 "psrlq $1, "#regd" \n\t" \
102 "psrlq $1, "#regb" \n\t" \
103 "psubb "#regb", "#regr" \n\t" \
104 "psubb "#regd", "#regp" \n\t"
106 void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx);
107 void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
109 void ff_add_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_size);
110 void ff_put_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_size);
111 void ff_put_signed_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_size);
113 void ff_clear_block_mmx(int16_t *block);
114 void ff_clear_block_sse(int16_t *block);
115 void ff_clear_blocks_mmx(int16_t *blocks);
116 void ff_clear_blocks_sse(int16_t *blocks);
118 void ff_add_bytes_mmx(uint8_t *dst, uint8_t *src, int w);
120 void ff_add_hfyu_median_prediction_cmov(uint8_t *dst, const uint8_t *top,
121 const uint8_t *diff, int w,
122 int *left, int *left_top);
124 void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
125 int w, int h, int sides);
127 void ff_gmc_mmx(uint8_t *dst, uint8_t *src,
128 int stride, int h, int ox, int oy,
129 int dxx, int dxy, int dyx, int dyy,
130 int shift, int r, int width, int height);
132 void ff_gmc_sse(uint8_t *dst, uint8_t *src,
133 int stride, int h, int ox, int oy,
134 int dxx, int dxy, int dyx, int dyy,
135 int shift, int r, int width, int height);
137 void ff_vector_clipf_sse(float *dst, const float *src,
138 float min, float max, int len);
140 void ff_avg_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
141 ptrdiff_t line_size, int h);
142 void ff_avg_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
143 ptrdiff_t line_size, int h);
144 void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
145 ptrdiff_t line_size, int h);
146 void ff_put_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
147 ptrdiff_t line_size, int h);
148 void ff_avg_pixels8_mmxext(uint8_t *block, const uint8_t *pixels,
149 ptrdiff_t line_size, int h);
150 void ff_put_pixels8_mmxext(uint8_t *block, const uint8_t *pixels,
151 ptrdiff_t line_size, int h);
152 void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
153 ptrdiff_t line_size, int h);
154 void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
155 ptrdiff_t line_size, int h);
157 void ff_avg_pixels8_x2_mmx(uint8_t *block, const uint8_t *pixels,
158 ptrdiff_t line_size, int h);
160 void ff_avg_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
161 ptrdiff_t line_size, int h);
162 void ff_avg_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
163 ptrdiff_t line_size, int h);
165 void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
166 ptrdiff_t line_size, int h);
167 void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
168 ptrdiff_t line_size, int h);
171 void ff_mmx_idct(int16_t *block);
172 void ff_mmxext_idct(int16_t *block);
174 void ff_deinterlace_line_mmx(uint8_t *dst,
175 const uint8_t *lum_m4, const uint8_t *lum_m3,
176 const uint8_t *lum_m2, const uint8_t *lum_m1,
180 void ff_deinterlace_line_inplace_mmx(const uint8_t *lum_m4,
181 const uint8_t *lum_m3,
182 const uint8_t *lum_m2,
183 const uint8_t *lum_m1,
184 const uint8_t *lum, int size);
186 #define PIXELS16(STATIC, PFX1, PFX2, TYPE, CPUEXT) \
187 STATIC void PFX1 ## _pixels16 ## TYPE ## CPUEXT(uint8_t *block, \
188 const uint8_t *pixels, \
189 ptrdiff_t line_size, \
192 PFX2 ## PFX1 ## _pixels8 ## TYPE ## CPUEXT(block, pixels, \
194 PFX2 ## PFX1 ## _pixels8 ## TYPE ## CPUEXT(block + 8, pixels + 8, \
198 #endif /* AVCODEC_X86_DSPUTIL_MMX_H */