1 /*
2  *  Copyright (c) 2013 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 #include <stdio.h>
13 
14 #include "./vpx_config.h"
15 #include "./vp9_rtcd.h"
16 #include "vp9/common/vp9_common.h"
17 #include "vp9/common/vp9_blockd.h"
18 #include "vpx_dsp/mips/inv_txfm_dspr2.h"
19 #include "vpx_dsp/txfm_common.h"
20 #include "vpx_ports/mem.h"
21 
22 #if HAVE_DSPR2
vp9_iht8x8_64_add_dspr2(const int16_t * input,uint8_t * dest,int dest_stride,int tx_type)23 void vp9_iht8x8_64_add_dspr2(const int16_t *input, uint8_t *dest,
24                              int dest_stride, int tx_type) {
25   int i, j;
26   DECLARE_ALIGNED(32, int16_t, out[8 * 8]);
27   int16_t *outptr = out;
28   int16_t temp_in[8 * 8], temp_out[8];
29   uint32_t pos = 45;
30 
31   /* bit positon for extract from acc */
32   __asm__ __volatile__ (
33     "wrdsp    %[pos],    1    \n\t"
34     :
35     : [pos] "r" (pos)
36   );
37 
38   switch (tx_type) {
39     case DCT_DCT:     // DCT in both horizontal and vertical
40       idct8_rows_dspr2(input, outptr, 8);
41       idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride);
42       break;
43     case ADST_DCT:    // ADST in vertical, DCT in horizontal
44       idct8_rows_dspr2(input, outptr, 8);
45 
46       for (i = 0; i < 8; ++i) {
47         iadst8_dspr2(&out[i * 8], temp_out);
48 
49         for (j = 0; j < 8; ++j)
50           dest[j * dest_stride + i] =
51                     clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
52                                       + dest[j * dest_stride + i]);
53       }
54       break;
55     case DCT_ADST:    // DCT in vertical, ADST in horizontal
56       for (i = 0; i < 8; ++i) {
57         iadst8_dspr2(input, outptr);
58         input += 8;
59         outptr += 8;
60       }
61 
62       for (i = 0; i < 8; ++i) {
63         for (j = 0; j < 8; ++j) {
64           temp_in[i * 8 + j] = out[j * 8 + i];
65         }
66       }
67       idct8_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride);
68       break;
69     case ADST_ADST:   // ADST in both directions
70       for (i = 0; i < 8; ++i) {
71         iadst8_dspr2(input, outptr);
72         input += 8;
73         outptr += 8;
74       }
75 
76       for (i = 0; i < 8; ++i) {
77         for (j = 0; j < 8; ++j)
78           temp_in[j] = out[j * 8 + i];
79 
80         iadst8_dspr2(temp_in, temp_out);
81 
82         for (j = 0; j < 8; ++j)
83           dest[j * dest_stride + i] =
84                 clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 5)
85                                       + dest[j * dest_stride + i]);
86       }
87       break;
88     default:
89       printf("vp9_short_iht8x8_add_dspr2 : Invalid tx_type\n");
90       break;
91   }
92 }
93 #endif  // #if HAVE_DSPR2
94