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_rtcd.h"
12 #include "vpx_dsp/mips/inv_txfm_msa.h"
13
vpx_iwht4x4_16_add_msa(const int16_t * input,uint8_t * dst,int32_t dst_stride)14 void vpx_iwht4x4_16_add_msa(const int16_t *input, uint8_t *dst,
15 int32_t dst_stride) {
16 v8i16 in0, in1, in2, in3;
17 v4i32 in0_r, in1_r, in2_r, in3_r, in4_r;
18
19 /* load vector elements of 4x4 block */
20 LD4x4_SH(input, in0, in2, in3, in1);
21 TRANSPOSE4x4_SH_SH(in0, in2, in3, in1, in0, in2, in3, in1);
22 UNPCK_R_SH_SW(in0, in0_r);
23 UNPCK_R_SH_SW(in2, in2_r);
24 UNPCK_R_SH_SW(in3, in3_r);
25 UNPCK_R_SH_SW(in1, in1_r);
26 SRA_4V(in0_r, in1_r, in2_r, in3_r, UNIT_QUANT_SHIFT);
27
28 in0_r += in2_r;
29 in3_r -= in1_r;
30 in4_r = (in0_r - in3_r) >> 1;
31 in1_r = in4_r - in1_r;
32 in2_r = in4_r - in2_r;
33 in0_r -= in1_r;
34 in3_r += in2_r;
35
36 TRANSPOSE4x4_SW_SW(in0_r, in1_r, in2_r, in3_r, in0_r, in1_r, in2_r, in3_r);
37
38 in0_r += in1_r;
39 in2_r -= in3_r;
40 in4_r = (in0_r - in2_r) >> 1;
41 in3_r = in4_r - in3_r;
42 in1_r = in4_r - in1_r;
43 in0_r -= in3_r;
44 in2_r += in1_r;
45
46 PCKEV_H4_SH(in0_r, in0_r, in1_r, in1_r, in2_r, in2_r, in3_r, in3_r, in0, in1,
47 in2, in3);
48 ADDBLK_ST4x4_UB(in0, in3, in1, in2, dst, dst_stride);
49 }
50
vpx_iwht4x4_1_add_msa(const int16_t * input,uint8_t * dst,int32_t dst_stride)51 void vpx_iwht4x4_1_add_msa(const int16_t *input, uint8_t *dst,
52 int32_t dst_stride) {
53 int16_t a1, e1;
54 v8i16 in1, in0 = { 0 };
55
56 a1 = input[0] >> UNIT_QUANT_SHIFT;
57 e1 = a1 >> 1;
58 a1 -= e1;
59
60 in0 = __msa_insert_h(in0, 0, a1);
61 in0 = __msa_insert_h(in0, 1, e1);
62 in0 = __msa_insert_h(in0, 2, e1);
63 in0 = __msa_insert_h(in0, 3, e1);
64
65 in1 = in0 >> 1;
66 in0 -= in1;
67
68 ADDBLK_ST4x4_UB(in0, in1, in1, in1, dst, dst_stride);
69 }
70
vpx_idct4x4_16_add_msa(const int16_t * input,uint8_t * dst,int32_t dst_stride)71 void vpx_idct4x4_16_add_msa(const int16_t *input, uint8_t *dst,
72 int32_t dst_stride) {
73 v8i16 in0, in1, in2, in3;
74
75 /* load vector elements of 4x4 block */
76 LD4x4_SH(input, in0, in1, in2, in3);
77 /* rows */
78 TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
79 VP9_IDCT4x4(in0, in1, in2, in3, in0, in1, in2, in3);
80 /* columns */
81 TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, in0, in1, in2, in3);
82 VP9_IDCT4x4(in0, in1, in2, in3, in0, in1, in2, in3);
83 /* rounding (add 2^3, divide by 2^4) */
84 SRARI_H4_SH(in0, in1, in2, in3, 4);
85 ADDBLK_ST4x4_UB(in0, in1, in2, in3, dst, dst_stride);
86 }
87
vpx_idct4x4_1_add_msa(const int16_t * input,uint8_t * dst,int32_t dst_stride)88 void vpx_idct4x4_1_add_msa(const int16_t *input, uint8_t *dst,
89 int32_t dst_stride) {
90 int16_t out;
91 v8i16 vec;
92
93 out = ROUND_POWER_OF_TWO((input[0] * cospi_16_64), DCT_CONST_BITS);
94 out = ROUND_POWER_OF_TWO((out * cospi_16_64), DCT_CONST_BITS);
95 out = ROUND_POWER_OF_TWO(out, 4);
96 vec = __msa_fill_h(out);
97
98 ADDBLK_ST4x4_UB(vec, vec, vec, vec, dst, dst_stride);
99 }
100