FFmpeg
hevc_mc_uniw_lsx.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2022 Loongson Technology Corporation Limited
3  * Contributed by Lu Wang <wanglu@loongson.cn>
4  * Hao Chen <chenhao@loongson.cn>
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
24 #include "hevcdsp_lsx.h"
25 
26 static const uint8_t ff_hevc_mask_arr[16 * 2] __attribute__((aligned(0x40))) = {
27  /* 8 width cases */
28  0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
29  /* 4 width cases */
30  0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20
31 };
32 
33 static av_always_inline
34 void hevc_hv_8t_8x2_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
35  int32_t dst_stride, const int8_t *filter_x,
36  const int8_t *filter_y, int32_t height, int32_t weight,
38 {
39  uint32_t loop_cnt, cnt;
40  const uint8_t *src_tmp;
41  uint8_t *dst_tmp;
42  const int32_t src_stride_2x = (src_stride << 1);
43  const int32_t dst_stride_2x = (dst_stride << 1);
44  const int32_t src_stride_4x = (src_stride << 2);
45  const int32_t src_stride_3x = src_stride_2x + src_stride;
46 
47  __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8;
48  __m128i filt0, filt1, filt2, filt3;
49  __m128i filt_h0, filt_h1, filt_h2, filt_h3;
50  __m128i mask1, mask2, mask3;
51  __m128i filter_vec;
52  __m128i vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
53  __m128i vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
54  __m128i dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst8;
55  __m128i dst0_r, dst0_l, dst1_r, dst1_l;
56  __m128i dst10_r, dst32_r, dst54_r, dst76_r;
57  __m128i dst10_l, dst32_l, dst54_l, dst76_l;
58  __m128i dst21_r, dst43_r, dst65_r, dst87_r;
59  __m128i dst21_l, dst43_l, dst65_l, dst87_l;
60  __m128i weight_vec, offset_vec, rnd_vec;
61  __m128i mask0 = __lsx_vld(ff_hevc_mask_arr, 0);
62 
63  src -= (src_stride_3x + 3);
64  weight_vec = __lsx_vreplgr2vr_w(weight);
65  offset_vec = __lsx_vreplgr2vr_w(offset);
66  rnd_vec = __lsx_vreplgr2vr_w(rnd_val);
67 
68  DUP4_ARG2(__lsx_vldrepl_h, filter_x, 0, filter_x, 2, filter_x, 4,
69  filter_x, 6, filt0, filt1, filt2, filt3);
70  filter_vec = __lsx_vld(filter_y, 0);
71  filter_vec = __lsx_vsllwil_h_b(filter_vec, 0);
72  DUP4_ARG2(__lsx_vreplvei_w, filter_vec, 0, filter_vec, 1, filter_vec, 2,
73  filter_vec, 3, filt_h0, filt_h1, filt_h2, filt_h3);
74  DUP2_ARG2(__lsx_vaddi_bu, mask0, 2, mask0, 4, mask1, mask2);
75  mask3 = __lsx_vaddi_bu(mask0, 6);
76 
77  for (cnt = width >> 3; cnt--;) {
78  src_tmp = src;
79  dst_tmp = dst;
80 
81  src0 = __lsx_vld(src_tmp, 0);
82  DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x,
83  src1, src2);
84  src3 = __lsx_vldx(src_tmp, src_stride_3x);
85  src_tmp += src_stride_4x;
86  src4 = __lsx_vld(src_tmp, 0);
87  DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x,
88  src5, src6);
89  src_tmp += src_stride_3x;
90 
91  DUP4_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, src0,
92  src0, mask2, src0, src0, mask3, vec0, vec1, vec2, vec3);
93  DUP4_ARG3(__lsx_vshuf_b, src1, src1, mask0, src1, src1, mask1, src1,
94  src1, mask2, src1, src1, mask3, vec4, vec5, vec6, vec7);
95  DUP4_ARG3(__lsx_vshuf_b, src2, src2, mask0, src2, src2, mask1, src2,
96  src2, mask2, src2, src2, mask3, vec8, vec9, vec10, vec11);
97  DUP4_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, src3,
98  src3, mask2, src3, src3, mask3, vec12, vec13, vec14, vec15);
99  DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec4, filt0, vec8, filt0,
100  vec12, filt0, dst0, dst1, dst2, dst3);
101  DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec1, filt1, dst1, vec5, filt1,
102  dst2, vec9, filt1, dst3, vec13, filt1, dst0, dst1, dst2, dst3);
103  DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec2, filt2, dst1, vec6, filt2,
104  dst2, vec10, filt2, dst3, vec14, filt2, dst0, dst1, dst2, dst3);
105  DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec3, filt3, dst1, vec7, filt3,
106  dst2, vec11, filt3, dst3, vec15, filt3, dst0, dst1, dst2, dst3);
107  DUP4_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, src4,
108  src4, mask2, src4, src4, mask3, vec0, vec1, vec2, vec3);
109  DUP4_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, src5,
110  src5, mask2, src5, src5, mask3, vec4, vec5, vec6, vec7);
111  DUP4_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, src6,
112  src6, mask2, src6, src6, mask3, vec8, vec9, vec10, vec11);
113  DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec4, filt0, dst4, dst5);
114  dst6 = __lsx_vdp2_h_bu_b(vec8, filt0);
115  DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec1, filt1, dst5, vec5, filt1,
116  dst6, vec9, filt1, dst4, vec2, filt2, dst4, dst5, dst6, dst4);
117  DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst5, vec6, filt2, dst6, vec10, filt2,
118  dst4, vec3, filt3, dst5, vec7, filt3, dst5, dst6, dst4, dst5);
119  dst6 = __lsx_vdp2add_h_bu_b(dst6, vec11, filt3);
120 
121  DUP4_ARG2(__lsx_vilvl_h, dst1, dst0, dst3, dst2, dst5, dst4, dst2,
122  dst1, dst10_r, dst32_r, dst54_r, dst21_r);
123  DUP2_ARG2(__lsx_vilvl_h, dst4, dst3, dst6, dst5, dst43_r, dst65_r);
124  DUP4_ARG2(__lsx_vilvh_h, dst1, dst0, dst3, dst2, dst5, dst4, dst2,
125  dst1, dst10_l, dst32_l, dst54_l, dst21_l);
126  DUP2_ARG2(__lsx_vilvh_h, dst4, dst3, dst6, dst5, dst43_l, dst65_l);
127 
128  for (loop_cnt = height >> 1; loop_cnt--;) {
129  src7 = __lsx_vld(src_tmp, 0);
130  src8 = __lsx_vldx(src_tmp, src_stride);
131  src_tmp += src_stride_2x;
132  DUP4_ARG3(__lsx_vshuf_b, src7, src7, mask0, src7, src7, mask1, src7,
133  src7, mask2, src7, src7, mask3, vec0, vec1, vec2, vec3);
134  dst7 = __lsx_vdp2_h_bu_b(vec0, filt0);
135  DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst7, vec1, filt1, dst7, vec2,
136  filt2, dst7, dst7);
137  dst7 = __lsx_vdp2add_h_bu_b(dst7, vec3, filt3);
138  dst76_r = __lsx_vilvl_h(dst7, dst6);
139  dst76_l = __lsx_vilvh_h(dst7, dst6);
140  DUP2_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst10_l, filt_h0,
141  dst0_r, dst0_l);
142  DUP4_ARG3(__lsx_vdp2add_w_h, dst0_r, dst32_r, filt_h1, dst0_l,
143  dst32_l, filt_h1, dst0_r, dst54_r, filt_h2, dst0_l,
144  dst54_l, filt_h2, dst0_r, dst0_l, dst0_r, dst0_l);
145  DUP2_ARG3(__lsx_vdp2add_w_h, dst0_r, dst76_r, filt_h3, dst0_l,
146  dst76_l, filt_h3, dst0_r, dst0_l);
147  DUP2_ARG2(__lsx_vsrai_w, dst0_r, 6, dst0_l, 6, dst0_r, dst0_l);
148 
149  /* row 8 */
150  DUP4_ARG3(__lsx_vshuf_b, src8, src8, mask0, src8, src8, mask1, src8,
151  src8, mask2, src8, src8, mask3, vec0, vec1, vec2, vec3);
152  dst8 = __lsx_vdp2_h_bu_b(vec0, filt0);
153  DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst8, vec1, filt1, dst8, vec2,
154  filt2, dst8, dst8);
155  dst8 = __lsx_vdp2add_h_bu_b(dst8, vec3, filt3);
156 
157  dst87_r = __lsx_vilvl_h(dst8, dst7);
158  dst87_l = __lsx_vilvh_h(dst8, dst7);
159  DUP2_ARG2(__lsx_vdp2_w_h, dst21_r, filt_h0, dst21_l, filt_h0,
160  dst1_r, dst1_l);
161  DUP4_ARG3(__lsx_vdp2add_w_h, dst1_r, dst43_r, filt_h1, dst1_l,
162  dst43_l, filt_h1, dst1_r, dst65_r, filt_h2, dst1_l,
163  dst65_l, filt_h2, dst1_r, dst1_l, dst1_r, dst1_l);
164  DUP2_ARG3(__lsx_vdp2add_w_h, dst1_r, dst87_r, filt_h3, dst1_l,
165  dst87_l, filt_h3, dst1_r, dst1_l);
166  DUP2_ARG2(__lsx_vsrai_w, dst1_r, 6, dst1_l, 6, dst1_r, dst1_l);
167 
168  DUP2_ARG2(__lsx_vmul_w, dst0_r, weight_vec, dst0_l, weight_vec,
169  dst0_r, dst0_l);
170  DUP2_ARG2(__lsx_vmul_w, dst1_r, weight_vec, dst1_l, weight_vec,
171  dst1_r, dst1_l);
172  DUP4_ARG2(__lsx_vsrar_w, dst0_r, rnd_vec, dst1_r, rnd_vec, dst0_l,
173  rnd_vec, dst1_l, rnd_vec, dst0_r, dst1_r, dst0_l, dst1_l);
174 
175  DUP2_ARG2(__lsx_vadd_w, dst0_r, offset_vec, dst0_l, offset_vec,
176  dst0_r, dst0_l);
177  DUP2_ARG2(__lsx_vadd_w, dst1_r, offset_vec, dst1_l, offset_vec,
178  dst1_r, dst1_l);
179  DUP4_ARG1(__lsx_vclip255_w, dst0_r, dst1_r, dst0_l, dst1_l, dst0_r,
180  dst1_r, dst0_l, dst1_l);
181  DUP2_ARG2(__lsx_vpickev_h, dst0_l, dst0_r, dst1_l, dst1_r,
182  dst0_r, dst1_r);
183  dst0_r = __lsx_vpickev_b(dst1_r, dst0_r);
184 
185  __lsx_vstelm_d(dst0_r, dst_tmp, 0, 0);
186  __lsx_vstelm_d(dst0_r, dst_tmp + dst_stride, 0, 1);
187  dst_tmp += dst_stride_2x;
188 
189  dst10_r = dst32_r;
190  dst32_r = dst54_r;
191  dst54_r = dst76_r;
192  dst10_l = dst32_l;
193  dst32_l = dst54_l;
194  dst54_l = dst76_l;
195  dst21_r = dst43_r;
196  dst43_r = dst65_r;
197  dst65_r = dst87_r;
198  dst21_l = dst43_l;
199  dst43_l = dst65_l;
200  dst65_l = dst87_l;
201  dst6 = dst8;
202  }
203 
204  src += 8;
205  dst += 8;
206  }
207 }
208 
209 static
210 void hevc_hv_8t_8w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
211  int32_t dst_stride, const int8_t *filter_x,
212  const int8_t *filter_y, int32_t height, int32_t weight,
213  int32_t offset, int32_t rnd_val)
214 {
215  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
216  filter_y, height, weight, offset, rnd_val, 8);
217 }
218 
219 static
220 void hevc_hv_8t_16w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
221  int32_t dst_stride, const int8_t *filter_x,
222  const int8_t *filter_y, int32_t height, int32_t weight,
223  int32_t offset, int32_t rnd_val)
224 {
225  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
226  filter_y, height, weight, offset, rnd_val, 16);
227 }
228 
229 static
230 void hevc_hv_8t_24w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
231  int32_t dst_stride, const int8_t *filter_x,
232  const int8_t *filter_y, int32_t height, int32_t weight,
233  int32_t offset, int32_t rnd_val)
234 {
235  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
236  filter_y, height, weight, offset, rnd_val, 24);
237 }
238 
239 static
240 void hevc_hv_8t_32w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
241  int32_t dst_stride, const int8_t *filter_x,
242  const int8_t *filter_y, int32_t height, int32_t weight,
243  int32_t offset, int32_t rnd_val)
244 {
245  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
246  filter_y, height, weight, offset, rnd_val, 32);
247 }
248 
249 static
250 void hevc_hv_8t_48w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
251  int32_t dst_stride, const int8_t *filter_x,
252  const int8_t *filter_y, int32_t height, int32_t weight,
253  int32_t offset, int32_t rnd_val)
254 {
255  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
256  filter_y, height, weight, offset, rnd_val, 48);
257 }
258 
259 static
260 void hevc_hv_8t_64w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst,
261  int32_t dst_stride, const int8_t *filter_x,
262  const int8_t *filter_y, int32_t height, int32_t weight,
263  int32_t offset, int32_t rnd_val)
264 {
265  hevc_hv_8t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x,
266  filter_y, height, weight, offset, rnd_val, 64);
267 }
268 
269 
270 #define UNI_W_MC_HV(PEL, WIDTH, TAP) \
271 void ff_hevc_put_hevc_uni_w_##PEL##_hv##WIDTH##_8_lsx(uint8_t *dst, \
272  ptrdiff_t dst_stride, \
273  const uint8_t *src, \
274  ptrdiff_t src_stride, \
275  int height, \
276  int denom, \
277  int weight, \
278  int offset, \
279  intptr_t mx, \
280  intptr_t my, \
281  int width) \
282 { \
283  const int8_t *filter_x = ff_hevc_##PEL##_filters[mx - 1]; \
284  const int8_t *filter_y = ff_hevc_##PEL##_filters[my - 1]; \
285  int shift = denom + 14 - 8; \
286  \
287  hevc_hv_##TAP##t_##WIDTH##w_lsx(src, src_stride, dst, dst_stride, filter_x,\
288  filter_y, height, weight, offset, shift); \
289 }
290 
291 UNI_W_MC_HV(qpel, 8, 8);
292 UNI_W_MC_HV(qpel, 16, 8);
293 UNI_W_MC_HV(qpel, 24, 8);
294 UNI_W_MC_HV(qpel, 32, 8);
295 UNI_W_MC_HV(qpel, 48, 8);
296 UNI_W_MC_HV(qpel, 64, 8);
297 
298 #undef UNI_W_MC_HV
src1
const pixel * src1
Definition: h264pred_template.c:421
hevc_hv_8t_32w_lsx
static void hevc_hv_8t_32w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:240
DUP2_ARG2
#define DUP2_ARG2(_INS, _IN0, _IN1, _IN2, _IN3, _OUT0, _OUT1)
Definition: loongson_intrinsics.h:58
UNI_W_MC_HV
#define UNI_W_MC_HV(PEL, WIDTH, TAP)
Definition: hevc_mc_uniw_lsx.c:270
hevc_hv_8t_24w_lsx
static void hevc_hv_8t_24w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:230
aligned
static int aligned(int val)
Definition: dashdec.c:168
DUP4_ARG2
#define DUP4_ARG2(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _IN6, _IN7, _OUT0, _OUT1, _OUT2, _OUT3)
Definition: loongson_intrinsics.h:76
width
#define width
hevc_hv_8t_8x2_lsx
static av_always_inline void hevc_hv_8t_8x2_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val, int32_t width)
Definition: hevc_mc_uniw_lsx.c:34
weight
static int weight(int i, int blen, int offset)
Definition: diracdec.c:1561
DUP4_ARG1
#define DUP4_ARG1(_INS, _IN0, _IN1, _IN2, _IN3, _OUT0, _OUT1, _OUT2, _OUT3)
Definition: loongson_intrinsics.h:70
DUP2_ARG3
#define DUP2_ARG3(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _OUT0, _OUT1)
Definition: loongson_intrinsics.h:64
height
#define height
offset
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf offset
Definition: writing_filters.txt:86
hevc_hv_8t_48w_lsx
static void hevc_hv_8t_48w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:250
hevc_hv_8t_16w_lsx
static void hevc_hv_8t_16w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:220
src2
const pixel * src2
Definition: h264pred_template.c:422
av_always_inline
#define av_always_inline
Definition: attributes.h:49
ff_hevc_mask_arr
static const uint8_t ff_hevc_mask_arr[16 *2]
Definition: hevc_mc_uniw_lsx.c:26
src0
const pixel *const src0
Definition: h264pred_template.c:420
loongson_intrinsics.h
hevc_hv_8t_8w_lsx
static void hevc_hv_8t_8w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:210
src
INIT_CLIP pixel * src
Definition: h264pred_template.c:418
int32_t
int32_t
Definition: audioconvert.c:56
hevcdsp_lsx.h
hevc_hv_8t_64w_lsx
static void hevc_hv_8t_64w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height, int32_t weight, int32_t offset, int32_t rnd_val)
Definition: hevc_mc_uniw_lsx.c:260
DUP4_ARG3
#define DUP4_ARG3(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _IN6, _IN7, _IN8, _IN9, _IN10, _IN11, _OUT0, _OUT1, _OUT2, _OUT3)
Definition: loongson_intrinsics.h:83