FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/x86/diracdsp_init.c
Date: 2026-08-26 01:27:24
Exec Total Coverage
Lines: 17 20 85.0%
Functions: 1 6 16.7%
Branches: 4 10 40.0%
Line Branch Exec Source
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 void ff_add_rect_clamped_sse2(uint8_t *dst, const uint16_t *src, ptrdiff_t stride,
26 const int16_t *idwt, ptrdiff_t idwt_stride,
27 int width, int height);
28
29 void ff_add_dirac_obmc8_sse2(uint16_t *dst, const uint8_t *src, ptrdiff_t stride,
30 const uint8_t *obmc_weight, int yblen);
31 void ff_add_dirac_obmc16_sse2(uint16_t *dst, const uint8_t *src, ptrdiff_t stride,
32 const uint8_t *obmc_weight, int yblen);
33 void ff_add_dirac_obmc32_sse2(uint16_t *dst, const uint8_t *src, ptrdiff_t stride,
34 const uint8_t *obmc_weight, int yblen);
35
36 void ff_put_rect_clamped_sse2(uint8_t *dst, ptrdiff_t dst_stride, const int16_t *src,
37 ptrdiff_t src_stride, int width, int height);
38 void ff_put_signed_rect_clamped_sse2(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src,
39 ptrdiff_t src_stride, int width, int height);
40 void ff_put_signed_rect_clamped_10_sse4(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src,
41 ptrdiff_t src_stride, int width, int height);
42
43 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);
44
45 #define HPEL_FILTER(MMSIZE, EXT) \
46 void ff_dirac_hpel_filter_v_ ## EXT(uint8_t *dst, const uint8_t *src, \
47 ptrdiff_t stride, int width); \
48 void ff_dirac_hpel_filter_h_ ## EXT(uint8_t *dst, const uint8_t *src, int width); \
49 \
50 static void dirac_hpel_filter_ ## EXT(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, \
51 const uint8_t *src, ptrdiff_t stride, int width, int height) \
52 { \
53 while( height-- ) \
54 { \
55 ff_dirac_hpel_filter_v_ ## EXT(dstv-MMSIZE, src-MMSIZE, stride, width+MMSIZE+5); \
56 ff_dirac_hpel_filter_h_ ## EXT(dsth, src, width); \
57 ff_dirac_hpel_filter_h_ ## EXT(dstc, dstv, width); \
58 \
59 dsth += stride; \
60 dstv += stride; \
61 dstc += stride; \
62 src += stride; \
63 } \
64 }
65
66 #define DIRAC_PIXOP(OPNAME, EXT)\
67 static void OPNAME ## _dirac_pixels16_ ## EXT(uint8_t *dst, const uint8_t *src[5], \
68 ptrdiff_t stride, int h) \
69 {\
70 if (h&3)\
71 ff_ ## OPNAME ## _dirac_pixels16_c(dst, src, stride, h);\
72 else\
73 ff_ ## OPNAME ## _pixels16_ ## EXT(dst, src[0], stride, h);\
74 }\
75 static void OPNAME ## _dirac_pixels32_ ## EXT(uint8_t *dst, const uint8_t *src[5], \
76 ptrdiff_t stride, int h) \
77 {\
78 if (h&3) {\
79 ff_ ## OPNAME ## _dirac_pixels32_c(dst, src, stride, h);\
80 } else {\
81 ff_ ## OPNAME ## _pixels16_ ## EXT(dst , src[0] , stride, h);\
82 ff_ ## OPNAME ## _pixels16_ ## EXT(dst+16, src[0]+16, stride, h);\
83 }\
84 }
85
86 ✗ DIRAC_PIXOP(put, sse2)
87 ✗ DIRAC_PIXOP(avg, sse2)
88
89 ✗ HPEL_FILTER(16, sse2)
90
91 112 void ff_diracdsp_init_x86(DiracDSPContext* c)
92 {
93 112 int mm_flags = av_get_cpu_flags();
94
95
2/2
✓ Branch 0 taken 30 times.
✓ Branch 1 taken 82 times.
 112 if (EXTERNAL_SSE2(mm_flags)) {
96 30 c->dirac_hpel_filter = dirac_hpel_filter_sse2;
97 30 c->add_rect_clamped = ff_add_rect_clamped_sse2;
98 30 c->put_signed_rect_clamped[0] = ff_put_signed_rect_clamped_sse2;
99
100 30 c->add_dirac_obmc[0] = ff_add_dirac_obmc8_sse2;
101 30 c->add_dirac_obmc[1] = ff_add_dirac_obmc16_sse2;
102 30 c->add_dirac_obmc[2] = ff_add_dirac_obmc32_sse2;
103
104 30 c->put_dirac_pixels_tab[1][0] = put_dirac_pixels16_sse2;
105 30 c->avg_dirac_pixels_tab[1][0] = avg_dirac_pixels16_sse2;
106 30 c->put_dirac_pixels_tab[2][0] = put_dirac_pixels32_sse2;
107 30 c->avg_dirac_pixels_tab[2][0] = avg_dirac_pixels32_sse2;
108 }
109
110
2/2
✓ Branch 0 taken 21 times.
✓ Branch 1 taken 91 times.
 112 if (EXTERNAL_SSE4(mm_flags)) {
111 21 c->dequant_subband[1] = ff_dequant_subband_32_sse4;
112 21 c->put_signed_rect_clamped[1] = ff_put_signed_rect_clamped_10_sse4;
113 }
114 112 }
115