2 * MMX optimized DSP utils
3 * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; 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_X86_H
23 #define AVCODEC_X86_DSPUTIL_X86_H
28 #include "libavcodec/avcodec.h"
29 #include "libavcodec/dsputil.h"
30 #include "libavutil/x86/asm.h"
31 #include "constants.h"
33 #define MOVQ_WONE(regd) \
35 "pcmpeqd %%" #regd ", %%" #regd " \n\t" \
36 "psrlw $15, %%" #regd ::)
38 #define JUMPALIGN() __asm__ volatile (".p2align 3"::)
39 #define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::)
41 #define MOVQ_BFE(regd) \
43 "pcmpeqd %%"#regd", %%"#regd" \n\t" \
44 "paddb %%"#regd", %%"#regd" \n\t" ::)
47 #define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo))
49 // for shared library it's better to use this way for accessing constants
51 #define MOVQ_WTWO(regd) \
53 "pcmpeqd %%"#regd", %%"#regd" \n\t" \
54 "psrlw $15, %%"#regd" \n\t" \
55 "psllw $1, %%"#regd" \n\t"::)
59 // using regr as temporary and for the output result
60 // first argument is unmodifed and second is trashed
61 // regfe is supposed to contain 0xfefefefefefefefe
62 #define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \
63 "movq "#rega", "#regr" \n\t" \
64 "pand "#regb", "#regr" \n\t" \
65 "pxor "#rega", "#regb" \n\t" \
66 "pand "#regfe", "#regb" \n\t" \
67 "psrlq $1, "#regb" \n\t" \
68 "paddb "#regb", "#regr" \n\t"
70 #define PAVGB_MMX(rega, regb, regr, regfe) \
71 "movq "#rega", "#regr" \n\t" \
72 "por "#regb", "#regr" \n\t" \
73 "pxor "#rega", "#regb" \n\t" \
74 "pand "#regfe", "#regb" \n\t" \
75 "psrlq $1, "#regb" \n\t" \
76 "psubb "#regb", "#regr" \n\t"
78 // mm6 is supposed to contain 0xfefefefefefefefe
79 #define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \
80 "movq "#rega", "#regr" \n\t" \
81 "movq "#regc", "#regp" \n\t" \
82 "pand "#regb", "#regr" \n\t" \
83 "pand "#regd", "#regp" \n\t" \
84 "pxor "#rega", "#regb" \n\t" \
85 "pxor "#regc", "#regd" \n\t" \
86 "pand %%mm6, "#regb" \n\t" \
87 "pand %%mm6, "#regd" \n\t" \
88 "psrlq $1, "#regb" \n\t" \
89 "psrlq $1, "#regd" \n\t" \
90 "paddb "#regb", "#regr" \n\t" \
91 "paddb "#regd", "#regp" \n\t"
93 #define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \
94 "movq "#rega", "#regr" \n\t" \
95 "movq "#regc", "#regp" \n\t" \
96 "por "#regb", "#regr" \n\t" \
97 "por "#regd", "#regp" \n\t" \
98 "pxor "#rega", "#regb" \n\t" \
99 "pxor "#regc", "#regd" \n\t" \
100 "pand %%mm6, "#regb" \n\t" \
101 "pand %%mm6, "#regd" \n\t" \
102 "psrlq $1, "#regd" \n\t" \
103 "psrlq $1, "#regb" \n\t" \
104 "psubb "#regb", "#regr" \n\t" \
105 "psubb "#regd", "#regp" \n\t"
107 void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx);
108 void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx);
110 void ff_add_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
112 void ff_put_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
114 void ff_put_signed_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels,
117 void ff_clear_block_mmx(int16_t *block);
118 void ff_clear_block_sse(int16_t *block);
119 void ff_clear_blocks_mmx(int16_t *blocks);
120 void ff_clear_blocks_sse(int16_t *blocks);
122 void ff_add_bytes_mmx(uint8_t *dst, uint8_t *src, int w);
124 void ff_add_hfyu_median_prediction_cmov(uint8_t *dst, const uint8_t *top,
125 const uint8_t *diff, int w,
126 int *left, int *left_top);
128 void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
129 int w, int h, int sides);
131 void ff_gmc_mmx(uint8_t *dst, uint8_t *src,
132 int stride, int h, int ox, int oy,
133 int dxx, int dxy, int dyx, int dyy,
134 int shift, int r, int width, int height);
136 void ff_vector_clipf_sse(float *dst, const float *src,
137 float min, float max, int len);
139 void ff_avg_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
140 ptrdiff_t line_size, int h);
141 void ff_avg_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
142 ptrdiff_t line_size, int h);
143 void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
144 ptrdiff_t line_size, int h);
145 void ff_put_pixels16_mmx(uint8_t *block, const uint8_t *pixels,
146 ptrdiff_t line_size, int h);
147 void ff_avg_pixels8_mmxext(uint8_t *block, const uint8_t *pixels,
148 ptrdiff_t line_size, int h);
149 void ff_put_pixels8_mmxext(uint8_t *block, const uint8_t *pixels,
150 ptrdiff_t line_size, int h);
151 void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
152 ptrdiff_t line_size, int h);
153 void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
154 ptrdiff_t line_size, int h);
156 void ff_avg_pixels8_x2_mmx(uint8_t *block, const uint8_t *pixels,
157 ptrdiff_t line_size, int h);
159 void ff_avg_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
160 ptrdiff_t line_size, int h);
161 void ff_avg_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
162 ptrdiff_t line_size, int h);
164 void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels,
165 ptrdiff_t line_size, int h);
166 void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels,
167 ptrdiff_t line_size, int h);
169 void ff_deinterlace_line_mmx(uint8_t *dst,
170 const uint8_t *lum_m4, const uint8_t *lum_m3,
171 const uint8_t *lum_m2, const uint8_t *lum_m1,
175 void ff_deinterlace_line_inplace_mmx(const uint8_t *lum_m4,
176 const uint8_t *lum_m3,
177 const uint8_t *lum_m2,
178 const uint8_t *lum_m1,
179 const uint8_t *lum, int size);
181 #define PIXELS16(STATIC, PFX1, PFX2, TYPE, CPUEXT) \
182 STATIC void PFX1 ## _pixels16 ## TYPE ## CPUEXT(uint8_t *block, \
183 const uint8_t *pixels, \
184 ptrdiff_t line_size, \
187 PFX2 ## PFX1 ## _pixels8 ## TYPE ## CPUEXT(block, pixels, \
189 PFX2 ## PFX1 ## _pixels8 ## TYPE ## CPUEXT(block + 8, pixels + 8, \
193 #endif /* AVCODEC_X86_DSPUTIL_X86_H */