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 #ifndef VPX_VP9_ENCODER_MIPS_MSA_VP9_FDCT_MSA_H_
12 #define VPX_VP9_ENCODER_MIPS_MSA_VP9_FDCT_MSA_H_
13 
14 #include "vpx_dsp/mips/fwd_txfm_msa.h"
15 #include "vpx_dsp/mips/txfm_macros_msa.h"
16 #include "vpx_ports/mem.h"
17 
18 #define VP9_ADST8(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2,  \
19                   out3, out4, out5, out6, out7)                              \
20   {                                                                          \
21     v8i16 cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst4_m;                       \
22     v8i16 vec0_m, vec1_m, vec2_m, vec3_m, s0_m, s1_m;                        \
23     v8i16 coeff0_m = { cospi_2_64,  cospi_6_64,  cospi_10_64, cospi_14_64,   \
24                        cospi_18_64, cospi_22_64, cospi_26_64, cospi_30_64 }; \
25     v8i16 coeff1_m = { cospi_8_64,  -cospi_8_64,  cospi_16_64, -cospi_16_64, \
26                        cospi_24_64, -cospi_24_64, 0,           0 };          \
27                                                                              \
28     SPLATI_H2_SH(coeff0_m, 0, 7, cnst0_m, cnst1_m);                          \
29     cnst2_m = -cnst0_m;                                                      \
30     ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m);       \
31     SPLATI_H2_SH(coeff0_m, 4, 3, cnst2_m, cnst3_m);                          \
32     cnst4_m = -cnst2_m;                                                      \
33     ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m);       \
34                                                                              \
35     ILVRL_H2_SH(in0, in7, vec1_m, vec0_m);                                   \
36     ILVRL_H2_SH(in4, in3, vec3_m, vec2_m);                                   \
37     DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst1_m,  \
38                           cnst2_m, cnst3_m, in7, in0, in4, in3);             \
39                                                                              \
40     SPLATI_H2_SH(coeff0_m, 2, 5, cnst0_m, cnst1_m);                          \
41     cnst2_m = -cnst0_m;                                                      \
42     ILVEV_H2_SH(cnst0_m, cnst1_m, cnst1_m, cnst2_m, cnst0_m, cnst1_m);       \
43     SPLATI_H2_SH(coeff0_m, 6, 1, cnst2_m, cnst3_m);                          \
44     cnst4_m = -cnst2_m;                                                      \
45     ILVEV_H2_SH(cnst2_m, cnst3_m, cnst3_m, cnst4_m, cnst2_m, cnst3_m);       \
46                                                                              \
47     ILVRL_H2_SH(in2, in5, vec1_m, vec0_m);                                   \
48     ILVRL_H2_SH(in6, in1, vec3_m, vec2_m);                                   \
49                                                                              \
50     DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst1_m,  \
51                           cnst2_m, cnst3_m, in5, in2, in6, in1);             \
52     BUTTERFLY_4(in7, in0, in2, in5, s1_m, s0_m, in2, in5);                   \
53     out7 = -s0_m;                                                            \
54     out0 = s1_m;                                                             \
55                                                                              \
56     SPLATI_H4_SH(coeff1_m, 0, 4, 1, 5, cnst0_m, cnst1_m, cnst2_m, cnst3_m);  \
57                                                                              \
58     ILVEV_H2_SH(cnst3_m, cnst0_m, cnst1_m, cnst2_m, cnst3_m, cnst2_m);       \
59     cnst0_m = __msa_ilvev_h(cnst1_m, cnst0_m);                               \
60     cnst1_m = cnst0_m;                                                       \
61                                                                              \
62     ILVRL_H2_SH(in4, in3, vec1_m, vec0_m);                                   \
63     ILVRL_H2_SH(in6, in1, vec3_m, vec2_m);                                   \
64     DOT_ADD_SUB_SRARI_PCK(vec0_m, vec1_m, vec2_m, vec3_m, cnst0_m, cnst2_m,  \
65                           cnst3_m, cnst1_m, out1, out6, s0_m, s1_m);         \
66                                                                              \
67     SPLATI_H2_SH(coeff1_m, 2, 3, cnst0_m, cnst1_m);                          \
68     cnst1_m = __msa_ilvev_h(cnst1_m, cnst0_m);                               \
69                                                                              \
70     ILVRL_H2_SH(in2, in5, vec1_m, vec0_m);                                   \
71     ILVRL_H2_SH(s0_m, s1_m, vec3_m, vec2_m);                                 \
72     out3 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst0_m);                   \
73     out4 = DOT_SHIFT_RIGHT_PCK_H(vec0_m, vec1_m, cnst1_m);                   \
74     out2 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst0_m);                   \
75     out5 = DOT_SHIFT_RIGHT_PCK_H(vec2_m, vec3_m, cnst1_m);                   \
76                                                                              \
77     out1 = -out1;                                                            \
78     out3 = -out3;                                                            \
79     out5 = -out5;                                                            \
80   }
81 
82 #define VP9_FADST4(in0, in1, in2, in3, out0, out1, out2, out3)              \
83   {                                                                         \
84     v4i32 s0_m, s1_m, s2_m, s3_m, constant_m;                               \
85     v4i32 in0_r_m, in1_r_m, in2_r_m, in3_r_m;                               \
86                                                                             \
87     UNPCK_R_SH_SW(in0, in0_r_m);                                            \
88     UNPCK_R_SH_SW(in1, in1_r_m);                                            \
89     UNPCK_R_SH_SW(in2, in2_r_m);                                            \
90     UNPCK_R_SH_SW(in3, in3_r_m);                                            \
91                                                                             \
92     constant_m = __msa_fill_w(sinpi_4_9);                                   \
93     MUL2(in0_r_m, constant_m, in3_r_m, constant_m, s1_m, s0_m);             \
94                                                                             \
95     constant_m = __msa_fill_w(sinpi_1_9);                                   \
96     s0_m += in0_r_m * constant_m;                                           \
97     s1_m -= in1_r_m * constant_m;                                           \
98                                                                             \
99     constant_m = __msa_fill_w(sinpi_2_9);                                   \
100     s0_m += in1_r_m * constant_m;                                           \
101     s1_m += in3_r_m * constant_m;                                           \
102                                                                             \
103     s2_m = in0_r_m + in1_r_m - in3_r_m;                                     \
104                                                                             \
105     constant_m = __msa_fill_w(sinpi_3_9);                                   \
106     MUL2(in2_r_m, constant_m, s2_m, constant_m, s3_m, in1_r_m);             \
107                                                                             \
108     in0_r_m = s0_m + s3_m;                                                  \
109     s2_m = s1_m - s3_m;                                                     \
110     s3_m = s1_m - s0_m + s3_m;                                              \
111                                                                             \
112     SRARI_W4_SW(in0_r_m, in1_r_m, s2_m, s3_m, DCT_CONST_BITS);              \
113     PCKEV_H4_SH(in0_r_m, in0_r_m, in1_r_m, in1_r_m, s2_m, s2_m, s3_m, s3_m, \
114                 out0, out1, out2, out3);                                    \
115   }
116 #endif  // VPX_VP9_ENCODER_MIPS_MSA_VP9_FDCT_MSA_H_
117