2 * Copyright (C) 2010 David Conrad
4 * This file is part of FFmpeg.
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.
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.
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
21 #include "libavutil/x86/cpu.h"
22 #include "dsputil_x86.h"
23 #include "diracdsp_mmx.h"
26 void ff_put_rect_clamped_mmx(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
27 void ff_put_rect_clamped_sse2(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
28 void ff_put_signed_rect_clamped_mmx(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
29 void ff_put_signed_rect_clamped_sse2(uint8_t *dst, int dst_stride, const int16_t *src, int src_stride, int width, int height);
31 #define HPEL_FILTER(MMSIZE, EXT) \
32 void ff_dirac_hpel_filter_v_ ## EXT(uint8_t *, const uint8_t *, int, int); \
33 void ff_dirac_hpel_filter_h_ ## EXT(uint8_t *, const uint8_t *, int); \
35 static void dirac_hpel_filter_ ## EXT(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, \
36 const uint8_t *src, int stride, int width, int height) \
40 ff_dirac_hpel_filter_v_ ## EXT(dstv-MMSIZE, src-MMSIZE, stride, width+MMSIZE+5); \
41 ff_dirac_hpel_filter_h_ ## EXT(dsth, src, width); \
42 ff_dirac_hpel_filter_h_ ## EXT(dstc, dstv, width); \
56 #define PIXFUNC(PFX, IDX, EXT) \
57 /*MMXDISABLEDc->PFX ## _dirac_pixels_tab[0][IDX] = ff_ ## PFX ## _dirac_pixels8_ ## EXT;*/ \
58 c->PFX ## _dirac_pixels_tab[1][IDX] = ff_ ## PFX ## _dirac_pixels16_ ## EXT; \
59 c->PFX ## _dirac_pixels_tab[2][IDX] = ff_ ## PFX ## _dirac_pixels32_ ## EXT
61 #define DIRAC_PIXOP(OPNAME2, OPNAME, EXT)\
62 void ff_ ## OPNAME2 ## _dirac_pixels8_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
65 ff_ ## OPNAME2 ## _dirac_pixels8_c(dst, src, stride, h);\
67 OPNAME ## _pixels8_ ## EXT(dst, src[0], stride, h);\
69 void ff_ ## OPNAME2 ## _dirac_pixels16_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
72 ff_ ## OPNAME2 ## _dirac_pixels16_c(dst, src, stride, h);\
74 OPNAME ## _pixels16_ ## EXT(dst, src[0], stride, h);\
76 void ff_ ## OPNAME2 ## _dirac_pixels32_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
79 ff_ ## OPNAME2 ## _dirac_pixels32_c(dst, src, stride, h);\
81 OPNAME ## _pixels16_ ## EXT(dst , src[0] , stride, h);\
82 OPNAME ## _pixels16_ ## EXT(dst+16, src[0]+16, stride, h);\
86 DIRAC_PIXOP(put, ff_put, mmx)
87 DIRAC_PIXOP(avg, ff_avg, mmx)
88 DIRAC_PIXOP(avg, ff_avg, mmxext)
90 void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
93 ff_put_dirac_pixels16_c(dst, src, stride, h);
95 ff_put_pixels16_sse2(dst, src[0], stride, h);
97 void ff_avg_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
100 ff_avg_dirac_pixels16_c(dst, src, stride, h);
102 ff_avg_pixels16_sse2(dst, src[0], stride, h);
104 void ff_put_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
107 ff_put_dirac_pixels32_c(dst, src, stride, h);
109 ff_put_pixels16_sse2(dst , src[0] , stride, h);
110 ff_put_pixels16_sse2(dst+16, src[0]+16, stride, h);
113 void ff_avg_dirac_pixels32_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
116 ff_avg_dirac_pixels32_c(dst, src, stride, h);
118 ff_avg_pixels16_sse2(dst , src[0] , stride, h);
119 ff_avg_pixels16_sse2(dst+16, src[0]+16, stride, h);
123 void ff_diracdsp_init_mmx(DiracDSPContext* c)
125 int mm_flags = av_get_cpu_flags();
127 if (EXTERNAL_MMX(mm_flags)) {
128 c->add_dirac_obmc[0] = ff_add_dirac_obmc8_mmx;
130 c->add_dirac_obmc[1] = ff_add_dirac_obmc16_mmx;
131 c->add_dirac_obmc[2] = ff_add_dirac_obmc32_mmx;
132 c->dirac_hpel_filter = dirac_hpel_filter_mmx;
133 c->add_rect_clamped = ff_add_rect_clamped_mmx;
134 c->put_signed_rect_clamped = ff_put_signed_rect_clamped_mmx;
136 PIXFUNC(put, 0, mmx);
137 PIXFUNC(avg, 0, mmx);
140 if (EXTERNAL_MMXEXT(mm_flags)) {
141 PIXFUNC(avg, 0, mmxext);
144 if (EXTERNAL_SSE2(mm_flags)) {
145 c->dirac_hpel_filter = dirac_hpel_filter_sse2;
146 c->add_rect_clamped = ff_add_rect_clamped_sse2;
147 c->put_signed_rect_clamped = ff_put_signed_rect_clamped_sse2;
149 c->add_dirac_obmc[1] = ff_add_dirac_obmc16_sse2;
150 c->add_dirac_obmc[2] = ff_add_dirac_obmc32_sse2;
152 c->put_dirac_pixels_tab[1][0] = ff_put_dirac_pixels16_sse2;
153 c->avg_dirac_pixels_tab[1][0] = ff_avg_dirac_pixels16_sse2;
154 c->put_dirac_pixels_tab[2][0] = ff_put_dirac_pixels32_sse2;
155 c->avg_dirac_pixels_tab[2][0] = ff_avg_dirac_pixels32_sse2;