1 /*
2 * Copyright (C) 2010 David Conrad
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg 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.
10 *
11 * FFmpeg 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.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavutil/x86/cpu.h"
22 #include "libavcodec/diracdsp.h"
23 #include "fpel.h"
24
25 DECL_DIRAC_PIXOP(put, mmx);
26 DECL_DIRAC_PIXOP(avg, mmx);
27 DECL_DIRAC_PIXOP(avg, mmxext);
28
29 void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);
30 void ff_avg_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);
31 void ff_put_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);
32 void ff_avg_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);
33
34 void ff_add_rect_clamped_mmx(uint8_t *, const uint16_t *, int, const int16_t *, int, int, int);
35 void ff_add_rect_clamped_sse2(uint8_t *, const uint16_t *, int, const int16_t *, int, int, int);
36
37 void ff_add_dirac_obmc8_mmx(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
38 void ff_add_dirac_obmc16_mmx(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
39 void ff_add_dirac_obmc32_mmx(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
40
41 void ff_add_dirac_obmc16_sse2(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
42 void ff_add_dirac_obmc32_sse2(uint16_t *dst, const uint8_t *src, int stride, const uint8_t *obmc_weight, int yblen);
43
44 void ff_put_rect_clamped_mmx(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
45 void ff_put_rect_clamped_sse2(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
46 void ff_put_signed_rect_clamped_mmx(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
47 void ff_put_signed_rect_clamped_sse2(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
48 void ff_put_signed_rect_clamped_10_sse4(uint8_t *dst, int dst_stride, const uint8_t *src, int src_stride, int width, int height);
49
50 void ff_dequant_subband_32_sse4(uint8_t *src, uint8_t *dst, ptrdiff_t stride, const int qf, const int qs, int tot_v, int tot_h);
51
52 #if HAVE_X86ASM
53
54 #define HPEL_FILTER(MMSIZE, EXT) \
55 void ff_dirac_hpel_filter_v_ ## EXT(uint8_t *, const uint8_t *, int, int); \
56 void ff_dirac_hpel_filter_h_ ## EXT(uint8_t *, const uint8_t *, int); \
57 \
58 static void dirac_hpel_filter_ ## EXT(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, \
59 const uint8_t *src, int stride, int width, int height) \
60 { \
61 while( height-- ) \
62 { \
63 ff_dirac_hpel_filter_v_ ## EXT(dstv-MMSIZE, src-MMSIZE, stride, width+MMSIZE+5); \
64 ff_dirac_hpel_filter_h_ ## EXT(dsth, src, width); \
65 ff_dirac_hpel_filter_h_ ## EXT(dstc, dstv, width); \
66 \
67 dsth += stride; \
68 dstv += stride; \
69 dstc += stride; \
70 src += stride; \
71 } \
72 }
73
74 #define PIXFUNC(PFX, IDX, EXT) \
75 /*MMXDISABLEDc->PFX ## _dirac_pixels_tab[0][IDX] = ff_ ## PFX ## _dirac_pixels8_ ## EXT;*/ \
76 c->PFX ## _dirac_pixels_tab[1][IDX] = ff_ ## PFX ## _dirac_pixels16_ ## EXT; \
77 c->PFX ## _dirac_pixels_tab[2][IDX] = ff_ ## PFX ## _dirac_pixels32_ ## EXT
78
79 #define DIRAC_PIXOP(OPNAME2, OPNAME, EXT)\
80 void ff_ ## OPNAME2 ## _dirac_pixels8_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
81 {\
82 if (h&3)\
83 ff_ ## OPNAME2 ## _dirac_pixels8_c(dst, src, stride, h);\
84 else\
85 OPNAME ## _pixels8_ ## EXT(dst, src[0], stride, h);\
86 }\
87 void ff_ ## OPNAME2 ## _dirac_pixels16_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
88 {\
89 if (h&3)\
90 ff_ ## OPNAME2 ## _dirac_pixels16_c(dst, src, stride, h);\
91 else\
92 OPNAME ## _pixels16_ ## EXT(dst, src[0], stride, h);\
93 }\
94 void ff_ ## OPNAME2 ## _dirac_pixels32_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
95 {\
96 if (h&3) {\
97 ff_ ## OPNAME2 ## _dirac_pixels32_c(dst, src, stride, h);\
98 } else {\
99 OPNAME ## _pixels16_ ## EXT(dst , src[0] , stride, h);\
100 OPNAME ## _pixels16_ ## EXT(dst+16, src[0]+16, stride, h);\
101 }\
102 }
103
DIRAC_PIXOP(put,ff_put,mmx)104 DIRAC_PIXOP(put, ff_put, mmx)
105 DIRAC_PIXOP(avg, ff_avg, mmx)
106 DIRAC_PIXOP(avg, ff_avg, mmxext)
107
108 void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
109 {
110 if (h&3)
111 ff_put_dirac_pixels16_c(dst, src, stride, h);
112 else
113 ff_put_pixels16_sse2(dst, src[0], stride, h);
114 }
ff_avg_dirac_pixels16_sse2(uint8_t * dst,const uint8_t * src[5],int stride,int h)115 void ff_avg_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
116 {
117 if (h&3)
118 ff_avg_dirac_pixels16_c(dst, src, stride, h);
119 else
120 ff_avg_pixels16_sse2(dst, src[0], stride, h);
121 }
ff_put_dirac_pixels32_sse2(uint8_t * dst,const uint8_t * src[5],int stride,int h)122 void ff_put_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
123 {
124 if (h&3) {
125 ff_put_dirac_pixels32_c(dst, src, stride, h);
126 } else {
127 ff_put_pixels16_sse2(dst , src[0] , stride, h);
128 ff_put_pixels16_sse2(dst+16, src[0]+16, stride, h);
129 }
130 }
ff_avg_dirac_pixels32_sse2(uint8_t * dst,const uint8_t * src[5],int stride,int h)131 void ff_avg_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
132 {
133 if (h&3) {
134 ff_avg_dirac_pixels32_c(dst, src, stride, h);
135 } else {
136 ff_avg_pixels16_sse2(dst , src[0] , stride, h);
137 ff_avg_pixels16_sse2(dst+16, src[0]+16, stride, h);
138 }
139 }
140
141 #else // HAVE_X86ASM
142
143 #define HPEL_FILTER(MMSIZE, EXT) \
144 void dirac_hpel_filter_ ## EXT(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, \
145 const uint8_t *src, int stride, int width, int height);
146
147 #define PIXFUNC(PFX, IDX, EXT) do {} while (0)
148
149 #endif // HAVE_X86ASM
150
151 #if !ARCH_X86_64
152 HPEL_FILTER(8, mmx)
153 #endif
154 HPEL_FILTER(16, sse2)
155
ff_diracdsp_init_x86(DiracDSPContext * c)156 void ff_diracdsp_init_x86(DiracDSPContext* c)
157 {
158 int mm_flags = av_get_cpu_flags();
159
160 if (EXTERNAL_MMX(mm_flags)) {
161 c->add_dirac_obmc[0] = ff_add_dirac_obmc8_mmx;
162 #if !ARCH_X86_64
163 c->add_dirac_obmc[1] = ff_add_dirac_obmc16_mmx;
164 c->add_dirac_obmc[2] = ff_add_dirac_obmc32_mmx;
165 c->dirac_hpel_filter = dirac_hpel_filter_mmx;
166 c->add_rect_clamped = ff_add_rect_clamped_mmx;
167 c->put_signed_rect_clamped[0] = (void *)ff_put_signed_rect_clamped_mmx;
168 #endif
169 PIXFUNC(put, 0, mmx);
170 PIXFUNC(avg, 0, mmx);
171 }
172
173 if (EXTERNAL_MMXEXT(mm_flags)) {
174 PIXFUNC(avg, 0, mmxext);
175 }
176
177 if (EXTERNAL_SSE2(mm_flags)) {
178 c->dirac_hpel_filter = dirac_hpel_filter_sse2;
179 c->add_rect_clamped = ff_add_rect_clamped_sse2;
180 c->put_signed_rect_clamped[0] = (void *)ff_put_signed_rect_clamped_sse2;
181
182 c->add_dirac_obmc[1] = ff_add_dirac_obmc16_sse2;
183 c->add_dirac_obmc[2] = ff_add_dirac_obmc32_sse2;
184
185 c->put_dirac_pixels_tab[1][0] = ff_put_dirac_pixels16_sse2;
186 c->avg_dirac_pixels_tab[1][0] = ff_avg_dirac_pixels16_sse2;
187 c->put_dirac_pixels_tab[2][0] = ff_put_dirac_pixels32_sse2;
188 c->avg_dirac_pixels_tab[2][0] = ff_avg_dirac_pixels32_sse2;
189 }
190
191 if (EXTERNAL_SSE4(mm_flags)) {
192 c->dequant_subband[1] = ff_dequant_subband_32_sse4;
193 c->put_signed_rect_clamped[1] = ff_put_signed_rect_clamped_10_sse4;
194 }
195 }
196