xref: /aosp_15_r20/external/libdav1d/src/x86/cdef.h (revision c09093415860a1c2373dacd84c4fde00c507cdfd)
1 /*
2  * Copyright © 2018, VideoLAN and dav1d authors
3  * Copyright © 2018, Two Orioles, LLC
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions are met:
8  *
9  * 1. Redistributions of source code must retain the above copyright notice, this
10  *    list of conditions and the following disclaimer.
11  *
12  * 2. Redistributions in binary form must reproduce the above copyright notice,
13  *    this list of conditions and the following disclaimer in the documentation
14  *    and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19  * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
20  * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
22  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
23  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include "src/cpu.h"
29 #include "src/cdef.h"
30 
31 #define decl_cdef_fns(ext) \
32     decl_cdef_fn(BF(dav1d_cdef_filter_4x4, ext)); \
33     decl_cdef_fn(BF(dav1d_cdef_filter_4x8, ext)); \
34     decl_cdef_fn(BF(dav1d_cdef_filter_8x8, ext))
35 
36 decl_cdef_fns(avx512icl);
37 decl_cdef_fns(avx2);
38 decl_cdef_fns(sse4);
39 decl_cdef_fns(ssse3);
40 decl_cdef_fns(sse2);
41 
42 decl_cdef_dir_fn(BF(dav1d_cdef_dir, avx2));
43 decl_cdef_dir_fn(BF(dav1d_cdef_dir, sse4));
44 decl_cdef_dir_fn(BF(dav1d_cdef_dir, ssse3));
45 
cdef_dsp_init_x86(Dav1dCdefDSPContext * const c)46 static ALWAYS_INLINE void cdef_dsp_init_x86(Dav1dCdefDSPContext *const c) {
47     const unsigned flags = dav1d_get_cpu_flags();
48 
49 #if BITDEPTH == 8
50     if (!(flags & DAV1D_X86_CPU_FLAG_SSE2)) return;
51 
52     c->fb[0] = BF(dav1d_cdef_filter_8x8, sse2);
53     c->fb[1] = BF(dav1d_cdef_filter_4x8, sse2);
54     c->fb[2] = BF(dav1d_cdef_filter_4x4, sse2);
55 #endif
56 
57     if (!(flags & DAV1D_X86_CPU_FLAG_SSSE3)) return;
58 
59     c->dir = BF(dav1d_cdef_dir, ssse3);
60     c->fb[0] = BF(dav1d_cdef_filter_8x8, ssse3);
61     c->fb[1] = BF(dav1d_cdef_filter_4x8, ssse3);
62     c->fb[2] = BF(dav1d_cdef_filter_4x4, ssse3);
63 
64     if (!(flags & DAV1D_X86_CPU_FLAG_SSE41)) return;
65 
66     c->dir = BF(dav1d_cdef_dir, sse4);
67 #if BITDEPTH == 8
68     c->fb[0] = BF(dav1d_cdef_filter_8x8, sse4);
69     c->fb[1] = BF(dav1d_cdef_filter_4x8, sse4);
70     c->fb[2] = BF(dav1d_cdef_filter_4x4, sse4);
71 #endif
72 
73 #if ARCH_X86_64
74     if (!(flags & DAV1D_X86_CPU_FLAG_AVX2)) return;
75 
76     c->dir = BF(dav1d_cdef_dir, avx2);
77     c->fb[0] = BF(dav1d_cdef_filter_8x8, avx2);
78     c->fb[1] = BF(dav1d_cdef_filter_4x8, avx2);
79     c->fb[2] = BF(dav1d_cdef_filter_4x4, avx2);
80 
81     if (!(flags & DAV1D_X86_CPU_FLAG_AVX512ICL)) return;
82 
83     c->fb[0] = BF(dav1d_cdef_filter_8x8, avx512icl);
84     c->fb[1] = BF(dav1d_cdef_filter_4x8, avx512icl);
85     c->fb[2] = BF(dav1d_cdef_filter_4x4, avx512icl);
86 #endif
87 }
88