1/*
2 * Copyright (c) 2021 Loongson Technology Corporation Limited
3 * Contributed by Xiwei Gu <guxiwei-hf@loongson.cn>
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22#include "libavcodec/bit_depth_template.c"
23#include "h264dsp_lasx.h"
24#include "libavutil/loongarch/loongson_intrinsics.h"
25
26#define H264_LOOP_FILTER_STRENGTH_ITERATION_LASX(edges, step, mask_mv, dir, \
27                                                 d_idx, mask_dir)           \
28do {                                                                        \
29    int b_idx = 0; \
30    int step_x4 = step << 2; \
31    int d_idx_12 = d_idx + 12; \
32    int d_idx_52 = d_idx + 52; \
33    int d_idx_x4 = d_idx << 2; \
34    int d_idx_x4_48 = d_idx_x4 + 48; \
35    int dir_x32  = dir * 32; \
36    uint8_t *ref_t = (uint8_t*)ref; \
37    uint8_t *mv_t  = (uint8_t*)mv; \
38    uint8_t *nnz_t = (uint8_t*)nnz; \
39    uint8_t *bS_t  = (uint8_t*)bS; \
40    mask_mv <<= 3; \
41    for (; b_idx < edges; b_idx += step) { \
42        out &= mask_dir; \
43        if (!(mask_mv & b_idx)) { \
44            if (bidir) { \
45                ref2 = __lasx_xvldx(ref_t, d_idx_12); \
46                ref3 = __lasx_xvldx(ref_t, d_idx_52); \
47                ref0 = __lasx_xvld(ref_t, 12); \
48                ref1 = __lasx_xvld(ref_t, 52); \
49                ref2 = __lasx_xvilvl_w(ref3, ref2); \
50                ref0 = __lasx_xvilvl_w(ref0, ref0); \
51                ref1 = __lasx_xvilvl_w(ref1, ref1); \
52                ref3 = __lasx_xvshuf4i_w(ref2, 0xB1); \
53                ref0 = __lasx_xvsub_b(ref0, ref2); \
54                ref1 = __lasx_xvsub_b(ref1, ref3); \
55                ref0 = __lasx_xvor_v(ref0, ref1); \
56\
57                tmp2 = __lasx_xvldx(mv_t, d_idx_x4_48);   \
58                tmp3 = __lasx_xvld(mv_t, 48); \
59                tmp4 = __lasx_xvld(mv_t, 208); \
60                tmp5 = __lasx_xvld(mv_t + d_idx_x4, 208); \
61                DUP2_ARG3(__lasx_xvpermi_q, tmp2, tmp2, 0x20, tmp5, tmp5, \
62                          0x20, tmp2, tmp5); \
63                tmp3 =  __lasx_xvpermi_q(tmp4, tmp3, 0x20); \
64                tmp2 = __lasx_xvsub_h(tmp2, tmp3); \
65                tmp5 = __lasx_xvsub_h(tmp5, tmp3); \
66                DUP2_ARG2(__lasx_xvsat_h, tmp2, 7, tmp5, 7, tmp2, tmp5); \
67                tmp0 = __lasx_xvpickev_b(tmp5, tmp2); \
68                tmp0 = __lasx_xvpermi_d(tmp0, 0xd8); \
69                tmp0 = __lasx_xvadd_b(tmp0, cnst_1); \
70                tmp0 = __lasx_xvssub_bu(tmp0, cnst_0); \
71                tmp0 = __lasx_xvsat_h(tmp0, 7); \
72                tmp0 = __lasx_xvpickev_b(tmp0, tmp0); \
73                tmp0 = __lasx_xvpermi_d(tmp0, 0xd8); \
74                tmp1 = __lasx_xvpickod_d(tmp0, tmp0); \
75                out = __lasx_xvor_v(ref0, tmp0); \
76                tmp1 = __lasx_xvshuf4i_w(tmp1, 0xB1); \
77                out = __lasx_xvor_v(out, tmp1); \
78                tmp0 = __lasx_xvshuf4i_w(out, 0xB1); \
79                out = __lasx_xvmin_bu(out, tmp0); \
80            } else { \
81                ref0 = __lasx_xvldx(ref_t, d_idx_12); \
82                ref3 = __lasx_xvld(ref_t, 12); \
83                tmp2 = __lasx_xvldx(mv_t, d_idx_x4_48); \
84                tmp3 = __lasx_xvld(mv_t, 48); \
85                tmp4 = __lasx_xvsub_h(tmp3, tmp2); \
86                tmp1 = __lasx_xvsat_h(tmp4, 7); \
87                tmp1 = __lasx_xvpickev_b(tmp1, tmp1); \
88                tmp1 = __lasx_xvadd_b(tmp1, cnst_1); \
89                out = __lasx_xvssub_bu(tmp1, cnst_0); \
90                out = __lasx_xvsat_h(out, 7); \
91                out = __lasx_xvpickev_b(out, out); \
92                ref0 = __lasx_xvsub_b(ref3, ref0); \
93                out = __lasx_xvor_v(out, ref0); \
94            } \
95        } \
96        tmp0 = __lasx_xvld(nnz_t, 12); \
97        tmp1 = __lasx_xvldx(nnz_t, d_idx_12); \
98        tmp0 = __lasx_xvor_v(tmp0, tmp1); \
99        tmp0 = __lasx_xvmin_bu(tmp0, cnst_2); \
100        out  = __lasx_xvmin_bu(out, cnst_2); \
101        tmp0 = __lasx_xvslli_h(tmp0, 1); \
102        tmp0 = __lasx_xvmax_bu(out, tmp0); \
103        tmp0 = __lasx_vext2xv_hu_bu(tmp0); \
104        __lasx_xvstelm_d(tmp0, bS_t + dir_x32, 0, 0); \
105        ref_t += step; \
106        mv_t  += step_x4; \
107        nnz_t += step; \
108        bS_t  += step; \
109    } \
110} while(0)
111
112void ff_h264_loop_filter_strength_lasx(int16_t bS[2][4][4], uint8_t nnz[40],
113                                       int8_t ref[2][40], int16_t mv[2][40][2],
114                                       int bidir, int edges, int step,
115                                       int mask_mv0, int mask_mv1, int field)
116{
117    __m256i out;
118    __m256i ref0, ref1, ref2, ref3;
119    __m256i tmp0, tmp1;
120    __m256i tmp2, tmp3, tmp4, tmp5;
121    __m256i cnst_0, cnst_1, cnst_2;
122    __m256i zero = __lasx_xvldi(0);
123    __m256i one  = __lasx_xvnor_v(zero, zero);
124    int64_t cnst3 = 0x0206020602060206, cnst4 = 0x0103010301030103;
125    if (field) {
126        cnst_0 = __lasx_xvreplgr2vr_d(cnst3);
127        cnst_1 = __lasx_xvreplgr2vr_d(cnst4);
128        cnst_2 = __lasx_xvldi(0x01);
129    } else {
130        DUP2_ARG1(__lasx_xvldi, 0x06, 0x03, cnst_0, cnst_1);
131        cnst_2 = __lasx_xvldi(0x01);
132    }
133    step  <<= 3;
134    edges <<= 3;
135
136    H264_LOOP_FILTER_STRENGTH_ITERATION_LASX(edges, step, mask_mv1,
137                                             1, -8, zero);
138    H264_LOOP_FILTER_STRENGTH_ITERATION_LASX(32, 8, mask_mv0, 0, -1, one);
139
140    DUP2_ARG2(__lasx_xvld, (int8_t*)bS, 0, (int8_t*)bS, 16, tmp0, tmp1);
141    DUP2_ARG2(__lasx_xvilvh_d, tmp0, tmp0, tmp1, tmp1, tmp2, tmp3);
142    LASX_TRANSPOSE4x4_H(tmp0, tmp2, tmp1, tmp3, tmp2, tmp3, tmp4, tmp5);
143    __lasx_xvstelm_d(tmp2, (int8_t*)bS, 0, 0);
144    __lasx_xvstelm_d(tmp3, (int8_t*)bS + 8, 0, 0);
145    __lasx_xvstelm_d(tmp4, (int8_t*)bS + 16, 0, 0);
146    __lasx_xvstelm_d(tmp5, (int8_t*)bS + 24, 0, 0);
147}
148