1/*
2 *  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
3 *
4 *  Use of this source code is governed by a BSD-style license
5 *  that can be found in the LICENSE file in the root of the source
6 *  tree. An additional intellectual property rights grant can be found
7 *  in the file PATENTS.  All contributing project authors may
8 *  be found in the AUTHORS file in the root of the source tree.
9 */
10
11#include "vpx_dsp/mips/common_dspr2.h"
12
13#if HAVE_DSPR2
14void vpx_h_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
15                               const uint8_t *above, const uint8_t *left) {
16  int32_t  tmp1, tmp2, tmp3, tmp4;
17
18  __asm__ __volatile__ (
19      "lb         %[tmp1],      (%[left])                    \n\t"
20      "lb         %[tmp2],      1(%[left])                   \n\t"
21      "lb         %[tmp3],      2(%[left])                   \n\t"
22      "lb         %[tmp4],      3(%[left])                   \n\t"
23      "replv.qb   %[tmp1],      %[tmp1]                      \n\t"
24      "replv.qb   %[tmp2],      %[tmp2]                      \n\t"
25      "replv.qb   %[tmp3],      %[tmp3]                      \n\t"
26      "replv.qb   %[tmp4],      %[tmp4]                      \n\t"
27      "sw         %[tmp1],      (%[dst])                     \n\t"
28      "add        %[dst],       %[dst],         %[stride]    \n\t"
29      "sw         %[tmp2],      (%[dst])                     \n\t"
30      "add        %[dst],       %[dst],         %[stride]    \n\t"
31      "sw         %[tmp3],      (%[dst])                     \n\t"
32      "add        %[dst],       %[dst],         %[stride]    \n\t"
33      "sw         %[tmp4],      (%[dst])                     \n\t"
34
35      : [tmp1] "=&r" (tmp1), [tmp2] "=&r" (tmp2),
36        [tmp3] "=&r" (tmp3), [tmp4] "=&r" (tmp4)
37      : [left] "r" (left), [dst] "r" (dst), [stride] "r" (stride)
38  );
39}
40
41void vpx_dc_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
42                                const uint8_t *above, const uint8_t *left) {
43  int32_t  expected_dc;
44  int32_t  average;
45  int32_t  tmp, above_c, above_l, above_r, left_c, left_r, left_l;
46
47  __asm__ __volatile__ (
48      "lw              %[above_c],         (%[above])                    \n\t"
49      "lw              %[left_c],          (%[left])                     \n\t"
50
51      "preceu.ph.qbl   %[above_l],         %[above_c]                    \n\t"
52      "preceu.ph.qbr   %[above_r],         %[above_c]                    \n\t"
53      "preceu.ph.qbl   %[left_l],          %[left_c]                     \n\t"
54      "preceu.ph.qbr   %[left_r],          %[left_c]                     \n\t"
55
56      "addu.ph         %[average],         %[above_r],       %[above_l]  \n\t"
57      "addu.ph         %[average],         %[average],       %[left_l]   \n\t"
58      "addu.ph         %[average],         %[average],       %[left_r]   \n\t"
59      "addiu           %[average],         %[average],       4           \n\t"
60      "srl             %[tmp],             %[average],       16          \n\t"
61      "addu.ph         %[average],         %[tmp],           %[average]  \n\t"
62      "srl             %[expected_dc],     %[average],       3           \n\t"
63      "replv.qb        %[expected_dc],     %[expected_dc]                \n\t"
64
65      "sw              %[expected_dc],     (%[dst])                      \n\t"
66      "add             %[dst],              %[dst],          %[stride]   \n\t"
67      "sw              %[expected_dc],     (%[dst])                      \n\t"
68      "add             %[dst],              %[dst],          %[stride]   \n\t"
69      "sw              %[expected_dc],     (%[dst])                      \n\t"
70      "add             %[dst],              %[dst],          %[stride]   \n\t"
71      "sw              %[expected_dc],     (%[dst])                      \n\t"
72
73      : [above_c] "=&r" (above_c), [above_l] "=&r" (above_l),
74        [above_r] "=&r" (above_r), [left_c] "=&r" (left_c),
75        [left_l] "=&r" (left_l), [left_r] "=&r" (left_r),
76        [average] "=&r" (average), [tmp] "=&r" (tmp),
77        [expected_dc] "=&r" (expected_dc)
78      : [above] "r" (above), [left] "r" (left),
79        [dst] "r" (dst), [stride] "r" (stride)
80  );
81}
82
83void vpx_tm_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
84                                const uint8_t *above, const uint8_t *left) {
85  int32_t  abovel, abover;
86  int32_t  left0, left1, left2, left3;
87  int32_t  res0, res1;
88  int32_t  resl;
89  int32_t  resr;
90  int32_t  top_left;
91  uint8_t  *cm = vpx_ff_cropTbl;
92
93  __asm__ __volatile__ (
94      "ulw             %[resl],       (%[above])                         \n\t"
95
96      "lbu             %[left0],       (%[left])                         \n\t"
97      "lbu             %[left1],       1(%[left])                        \n\t"
98      "lbu             %[left2],       2(%[left])                        \n\t"
99      "lbu             %[left3],       3(%[left])                        \n\t"
100
101      "lbu             %[top_left],    -1(%[above])                      \n\t"
102
103      "preceu.ph.qbl   %[abovel],      %[resl]                           \n\t"
104      "preceu.ph.qbr   %[abover],      %[resl]                           \n\t"
105
106      "replv.ph        %[left0],       %[left0]                          \n\t"
107      "replv.ph        %[left1],       %[left1]                          \n\t"
108      "replv.ph        %[left2],       %[left2]                          \n\t"
109      "replv.ph        %[left3],       %[left3]                          \n\t"
110
111      "replv.ph        %[top_left],    %[top_left]                       \n\t"
112
113      "addu.ph         %[resl],        %[abovel],         %[left0]       \n\t"
114      "subu.ph         %[resl],        %[resl],           %[top_left]    \n\t"
115
116      "addu.ph         %[resr],        %[abover],         %[left0]       \n\t"
117      "subu.ph         %[resr],        %[resr],           %[top_left]    \n\t"
118
119      "sll             %[res0],        %[resr],           16             \n\t"
120      "sra             %[res0],        %[res0],           16             \n\t"
121      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
122
123      "sra             %[res1],        %[resr],           16             \n\t"
124      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
125      "sb              %[res0],        (%[dst])                          \n\t"
126
127      "sll             %[res0],        %[resl],           16             \n\t"
128      "sra             %[res0],        %[res0],           16             \n\t"
129      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
130      "sb              %[res1],        1(%[dst])                         \n\t"
131
132      "sra             %[res1],        %[resl],           16             \n\t"
133      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
134
135      "addu.ph         %[resl],        %[abovel],         %[left1]       \n\t"
136      "subu.ph         %[resl],        %[resl],           %[top_left]    \n\t"
137
138      "addu.ph         %[resr],        %[abover],         %[left1]       \n\t"
139      "subu.ph         %[resr],        %[resr],           %[top_left]    \n\t"
140
141      "sb              %[res0],        2(%[dst])                         \n\t"
142      "sb              %[res1],        3(%[dst])                         \n\t"
143
144      "add             %[dst],          %[dst],           %[stride]      \n\t"
145
146      "sll             %[res0],        %[resr],           16             \n\t"
147      "sra             %[res0],        %[res0],           16             \n\t"
148      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
149
150      "sra             %[res1],        %[resr],           16             \n\t"
151      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
152      "sb              %[res0],        (%[dst])                          \n\t"
153
154      "sll             %[res0],        %[resl],           16             \n\t"
155      "sra             %[res0],        %[res0],           16             \n\t"
156      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
157
158      "sb              %[res1],        1(%[dst])                         \n\t"
159      "sra             %[res1],        %[resl],           16             \n\t"
160      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
161
162      "addu.ph         %[resl],        %[abovel],         %[left2]       \n\t"
163      "subu.ph         %[resl],        %[resl],           %[top_left]    \n\t"
164
165      "addu.ph         %[resr],        %[abover],         %[left2]       \n\t"
166      "subu.ph         %[resr],        %[resr],           %[top_left]    \n\t"
167
168      "sb              %[res0],        2(%[dst])                         \n\t"
169      "sb              %[res1],        3(%[dst])                         \n\t"
170
171      "add             %[dst],          %[dst],           %[stride]      \n\t"
172
173      "sll             %[res0],        %[resr],           16             \n\t"
174      "sra             %[res0],        %[res0],           16             \n\t"
175      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
176
177
178      "sra             %[res1],        %[resr],           16             \n\t"
179      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
180      "sb              %[res0],        (%[dst])                          \n\t"
181
182      "sll             %[res0],        %[resl],           16             \n\t"
183      "sra             %[res0],        %[res0],           16             \n\t"
184      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
185
186
187      "sb              %[res1],        1(%[dst])                         \n\t"
188      "sra             %[res1],        %[resl],           16             \n\t"
189      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
190
191      "addu.ph         %[resl],        %[abovel],        %[left3]        \n\t"
192      "subu.ph         %[resl],        %[resl],          %[top_left]     \n\t"
193
194      "addu.ph         %[resr],        %[abover],        %[left3]        \n\t"
195      "subu.ph         %[resr],        %[resr],          %[top_left]     \n\t"
196
197      "sb              %[res0],        2(%[dst])                         \n\t"
198      "sb              %[res1],        3(%[dst])                         \n\t"
199
200      "add             %[dst],          %[dst],          %[stride]       \n\t"
201
202      "sll             %[res0],        %[resr],           16             \n\t"
203      "sra             %[res0],        %[res0],           16             \n\t"
204      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
205
206      "sra             %[res1],        %[resr],           16             \n\t"
207      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
208      "sb              %[res0],        (%[dst])                          \n\t"
209
210      "sll             %[res0],        %[resl],           16             \n\t"
211      "sra             %[res0],        %[res0],           16             \n\t"
212      "lbux            %[res0],        %[res0](%[cm])                    \n\t"
213      "sb              %[res1],        1(%[dst])                         \n\t"
214
215      "sra             %[res1],        %[resl],           16             \n\t"
216      "lbux            %[res1],        %[res1](%[cm])                    \n\t"
217
218      "sb              %[res0],        2(%[dst])                         \n\t"
219      "sb              %[res1],        3(%[dst])                         \n\t"
220
221      : [abovel] "=&r" (abovel), [abover] "=&r" (abover),
222        [left0] "=&r" (left0), [left1] "=&r" (left1), [left2] "=&r" (left2),
223        [res0] "=&r" (res0), [res1] "=&r" (res1), [left3] "=&r" (left3),
224        [resl] "=&r" (resl), [resr] "=&r" (resr), [top_left] "=&r" (top_left)
225      : [above] "r" (above), [left] "r" (left),
226        [dst] "r" (dst), [stride] "r" (stride), [cm] "r" (cm)
227  );
228}
229#endif  // #if HAVE_DSPR2
230