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 <assert.h>
12 
13 #include "vp9/common/vp9_enums.h"
14 #include "vp9/encoder/mips/msa/vp9_fdct_msa.h"
15 
vp9_fwht4x4_msa(const int16_t * input,int16_t * output,int32_t src_stride)16 void vp9_fwht4x4_msa(const int16_t *input, int16_t *output,
17                      int32_t src_stride) {
18   v8i16 in0, in1, in2, in3, in4;
19 
20   LD_SH4(input, src_stride, in0, in1, in2, in3);
21 
22   in0 += in1;
23   in3 -= in2;
24   in4 = (in0 - in3) >> 1;
25   SUB2(in4, in1, in4, in2, in1, in2);
26   in0 -= in2;
27   in3 += in1;
28 
29   TRANSPOSE4x4_SH_SH(in0, in2, in3, in1, in0, in2, in3, in1);
30 
31   in0 += in2;
32   in1 -= in3;
33   in4 = (in0 - in1) >> 1;
34   SUB2(in4, in2, in4, in3, in2, in3);
35   in0 -= in3;
36   in1 += in2;
37 
38   SLLI_4V(in0, in1, in2, in3, 2);
39 
40   TRANSPOSE4x4_SH_SH(in0, in3, in1, in2, in0, in3, in1, in2);
41 
42   ST4x2_UB(in0, output, 4);
43   ST4x2_UB(in3, output + 4, 4);
44   ST4x2_UB(in1, output + 8, 4);
45   ST4x2_UB(in2, output + 12, 4);
46 }
47 
vp9_fht4x4_msa(const int16_t * input,int16_t * output,int32_t stride,int32_t tx_type)48 void vp9_fht4x4_msa(const int16_t *input, int16_t *output, int32_t stride,
49                     int32_t tx_type) {
50   v8i16 in0, in1, in2, in3;
51 
52   LD_SH4(input, stride, in0, in1, in2, in3);
53 
54   /* fdct4 pre-process */
55   {
56     v8i16 temp, mask;
57     v16i8 zero = { 0 };
58     v16i8 one = __msa_ldi_b(1);
59 
60     mask = (v8i16)__msa_sldi_b(zero, one, 15);
61     SLLI_4V(in0, in1, in2, in3, 4);
62     temp = __msa_ceqi_h(in0, 0);
63     temp = (v8i16)__msa_xori_b((v16u8)temp, 255);
64     temp = mask & temp;
65     in0 += temp;
66   }
67 
68   switch (tx_type) {
69     case DCT_DCT:
70       VP9_FDCT4(in0, in1, in2, in3, in0, in1, in2, in3);
71       TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
72       VP9_FDCT4(in0, in1, in2, in3, in0, in1, in2, in3);
73       break;
74     case ADST_DCT:
75       VP9_FADST4(in0, in1, in2, in3, in0, in1, in2, in3);
76       TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
77       VP9_FDCT4(in0, in1, in2, in3, in0, in1, in2, in3);
78       break;
79     case DCT_ADST:
80       VP9_FDCT4(in0, in1, in2, in3, in0, in1, in2, in3);
81       TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
82       VP9_FADST4(in0, in1, in2, in3, in0, in1, in2, in3);
83       break;
84     case ADST_ADST:
85       VP9_FADST4(in0, in1, in2, in3, in0, in1, in2, in3);
86       TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
87       VP9_FADST4(in0, in1, in2, in3, in0, in1, in2, in3);
88       break;
89     default:
90       assert(0);
91       break;
92   }
93 
94   TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
95   ADD4(in0, 1, in1, 1, in2, 1, in3, 1, in0, in1, in2, in3);
96   SRA_4V(in0, in1, in2, in3, 2);
97   PCKEV_D2_SH(in1, in0, in3, in2, in0, in2);
98   ST_SH2(in0, in2, output, 8);
99 }
100