2 * Copyright (C) 2009 Loren Merritt <lorenm@u.washignton.edu>
4 * This file is part of Libav.
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/attributes.h"
22 #include "libavutil/cpu.h"
23 #include "libavutil/mem.h"
24 #include "libavutil/x86/asm.h"
25 #include "libavfilter/gradfun.h"
29 DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F};
30 DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF};
32 #if HAVE_MMXEXT_INLINE
33 static void gradfun_filter_line_mmx2(uint8_t *dst, uint8_t *src, uint16_t *dc, int width, int thresh, const uint16_t *dithers)
38 ff_gradfun_filter_line_c(dst + x, src + x, dc + x / 2, width - x, thresh, dithers);
44 "pxor %%mm7, %%mm7 \n"
45 "pshufw $0, %%mm5, %%mm5 \n"
49 "movd (%2,%0), %%mm0 \n"
50 "movd (%3,%0), %%mm1 \n"
51 "punpcklbw %%mm7, %%mm0 \n"
52 "punpcklwd %%mm1, %%mm1 \n"
54 "pxor %%mm2, %%mm2 \n"
55 "psubw %%mm0, %%mm1 \n" // delta = dc - pix
56 "psubw %%mm1, %%mm2 \n"
57 "pmaxsw %%mm1, %%mm2 \n"
58 "pmulhuw %%mm5, %%mm2 \n" // m = abs(delta) * thresh >> 16
59 "psubw %%mm6, %%mm2 \n"
60 "pminsw %%mm7, %%mm2 \n" // m = -max(0, 127-m)
61 "pmullw %%mm2, %%mm2 \n"
62 "paddw %%mm4, %%mm0 \n" // pix += dither
63 "pmulhw %%mm2, %%mm1 \n"
64 "psllw $2, %%mm1 \n" // m = m*m*delta >> 14
65 "paddw %%mm1, %%mm0 \n" // pix += m
67 "packuswb %%mm0, %%mm0 \n"
68 "movd %%mm0, (%1,%0) \n" // dst = clip(pix>>7)
73 :"r"(dst+width), "r"(src+width), "r"(dc+width/2),
74 "rm"(thresh), "m"(*dithers), "m"(*pw_7f)
81 static void gradfun_filter_line_ssse3(uint8_t *dst, uint8_t *src, uint16_t *dc, int width, int thresh, const uint16_t *dithers)
85 // could be 10% faster if I somehow eliminated this
87 ff_gradfun_filter_line_c(dst + x, src + x, dc + x / 2, width - x, thresh, dithers);
93 "pxor %%xmm7, %%xmm7 \n"
94 "pshuflw $0,%%xmm5, %%xmm5 \n"
95 "movdqa %6, %%xmm6 \n"
96 "punpcklqdq %%xmm5, %%xmm5 \n"
97 "movdqa %5, %%xmm4 \n"
99 "movq (%2,%0), %%xmm0 \n"
100 "movq (%3,%0), %%xmm1 \n"
101 "punpcklbw %%xmm7, %%xmm0 \n"
102 "punpcklwd %%xmm1, %%xmm1 \n"
103 "psllw $7, %%xmm0 \n"
104 "psubw %%xmm0, %%xmm1 \n" // delta = dc - pix
105 "pabsw %%xmm1, %%xmm2 \n"
106 "pmulhuw %%xmm5, %%xmm2 \n" // m = abs(delta) * thresh >> 16
107 "psubw %%xmm6, %%xmm2 \n"
108 "pminsw %%xmm7, %%xmm2 \n" // m = -max(0, 127-m)
109 "pmullw %%xmm2, %%xmm2 \n"
110 "psllw $1, %%xmm2 \n"
111 "paddw %%xmm4, %%xmm0 \n" // pix += dither
112 "pmulhrsw %%xmm2, %%xmm1 \n" // m = m*m*delta >> 14
113 "paddw %%xmm1, %%xmm0 \n" // pix += m
114 "psraw $7, %%xmm0 \n"
115 "packuswb %%xmm0, %%xmm0 \n"
116 "movq %%xmm0, (%1,%0) \n" // dst = clip(pix>>7)
120 :"r"(dst+width), "r"(src+width), "r"(dc+width/2),
121 "rm"(thresh), "m"(*dithers), "m"(*pw_7f)
125 #endif /* HAVE_SSSE3_INLINE */
128 static void gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, uint16_t *buf1, uint8_t *src, int src_linesize, int width)
131 intptr_t x = -2*width;\
133 "movdqa %6, %%xmm7 \n"\
135 load" (%4,%0), %%xmm0 \n"\
136 load" (%5,%0), %%xmm1 \n"\
137 "movdqa %%xmm0, %%xmm2 \n"\
138 "movdqa %%xmm1, %%xmm3 \n"\
139 "psrlw $8, %%xmm0 \n"\
140 "psrlw $8, %%xmm1 \n"\
141 "pand %%xmm7, %%xmm2 \n"\
142 "pand %%xmm7, %%xmm3 \n"\
143 "paddw %%xmm1, %%xmm0 \n"\
144 "paddw %%xmm3, %%xmm2 \n"\
145 "paddw %%xmm2, %%xmm0 \n"\
146 "paddw (%2,%0), %%xmm0 \n"\
147 "movdqa (%1,%0), %%xmm1 \n"\
148 "movdqa %%xmm0, (%1,%0) \n"\
149 "psubw %%xmm1, %%xmm0 \n"\
150 "movdqa %%xmm0, (%3,%0) \n"\
158 "r"(src+width*2+src_linesize),\
162 if (((intptr_t) src | src_linesize) & 15) {
168 #endif /* HAVE_SSE2_INLINE */
170 #endif /* HAVE_INLINE_ASM */
172 av_cold void ff_gradfun_init_x86(GradFunContext *gf)
174 int cpu_flags = av_get_cpu_flags();
176 #if HAVE_MMXEXT_INLINE
177 if (cpu_flags & AV_CPU_FLAG_MMXEXT)
178 gf->filter_line = gradfun_filter_line_mmx2;
180 #if HAVE_SSSE3_INLINE
181 if (cpu_flags & AV_CPU_FLAG_SSSE3)
182 gf->filter_line = gradfun_filter_line_ssse3;
185 if (cpu_flags & AV_CPU_FLAG_SSE2)
186 gf->blur_line = gradfun_blur_line_sse2;