annotate bfin/fdct_bfin.S @ 5746:55ed6dc5d476 libavcodec

Remove const vector macro indirection that is useless and obfuscating now that the Metrowerks workarounds are gone.
author diego
date Mon, 01 Oct 2007 14:23:36 +0000
parents 75bf61c6c385
children 78aa57eba353
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
4765
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
1 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
2 * fdct BlackFin
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
3 *
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
4 * Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com>
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
5 *
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
6 * This file is part of FFmpeg.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
7 *
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
8 * FFmpeg is free software; you can redistribute it and/or
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
9 * modify it under the terms of the GNU Lesser General Public
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
10 * License as published by the Free Software Foundation; either
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
11 * version 2.1 of the License, or (at your option) any later version.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
12 *
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
13 * FFmpeg is distributed in the hope that it will be useful,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
16 * Lesser General Public License for more details.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
17 *
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
18 * You should have received a copy of the GNU Lesser General Public
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
19 * License along with FFmpeg; if not, write to the Free Software
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
21 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
22 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
23 void ff_bfin_fdct (DCTELEM *buf);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
24
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
25 This implementation works only for 8x8 input. The range of input
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
26 must be -256 to 255 i.e. 8bit input represented in a 16bit data
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
27 word. The original data must be sign extended into the 16bit data
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
28 words.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
29
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
30
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
31 Chen factorization of
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
32
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
33 8
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
34 X(m) = sum (x(n) * cos ((2n+1)*m*pi/16))
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
35 n=0
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
36
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
37 C4
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
38 0 --*-------------*0+7---*-----*0+3-------*-*-------------------> 0
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
39 \ / \ / X S4,S4
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
40 1 --*-\---------/-*1+6---*-\-/-*1+2-------*-*-------------------> 4
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
41 \ / \ -C4 C3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
42 2 --*---\-----/---*2+5---*-/-\-*1-2---------------*-*-----------> 2
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
43 \ / / \ X S3,-S3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
44 3 --*-----\-/-----*3+4---*-----*0-3---------------*-*-----------> 6
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
45 / C7 C3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
46 4 --*-----/-\-----*3-4------------*-*4+5--*-----*---------------> 1
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
47 / \ -C4 X \ /S7 C3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
48 5 --*---/-----\---*2-5---*-*------*=*4-5----\-/------*-*--------> 5
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
49 / \ X S4,S4 / X S3,-S3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
50 6 --*-/---------\-*1-6---*-*------*=*7-6----/-\------*-*--------> 3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
51 / \ C4 X / \-S7 C3
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
52 --*-------------*0-7------------*-*7+6--*-----*---------------> 7
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
53 C7
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
54
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
55 Notation
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
56 Cn = cos(n*pi/8) used throughout the code.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
57
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
58
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
59 Registers used:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
60 R0, R1, R2, R3, R4, R5, R6,R7, P0, P1, P2, P3, P4, P5, A0, A1.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
61 Other registers used:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
62 I0, I1, I2, I3, B0, B2, B3, M0, M1, L3 registers and LC0.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
63
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
64 Input - r0 - pointer to start of DCTELEM *block
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
65
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
66 Output - The DCT output coefficients in the DCTELEM *block
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
67
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
68 Register constraint:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
69 This code is called from jpeg_encode.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
70 R6, R5, R4 if modified should be stored and restored.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
71
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
72
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
73 Performance: (Timer version 0.6.33)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
74 Code Size : 240 Bytes.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
75 Memory Required :
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
76 Input Matrix : 8 * 8 * 2 Bytes.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
77 Coefficients : 16 Bytes
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
78 Temporary matrix: 8 * 8 * 2 Bytes.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
79 Cycle Count :26+{18+8*(14+2S)}*2 where S -> Stalls
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
80 (7.45 c/pel)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
81 -----------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
82 | Size | Forward DCT | Inverse DCT |
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
83 -----------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
84 | 8x8 | 284 Cycles | 311 Cycles |
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
85 -----------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
86
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
87 Ck = int16(cos(k/16*pi)*32767+.5)/2
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
88 #define C4 23170
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
89 #define C3 13623
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
90 #define C6 6270
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
91 #define C7 3196
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
92
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
93 Sk = int16(sin(k/16*pi)*32767+.5)/2
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
94 #define S4 11585
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
95 #define S3 9102
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
96 #define S6 15137
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
97 #define S7 16069
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
98
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
99 the coefficients are ordered as follows:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
100 short dct_coef[]
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
101 C4,S4,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
102 C6,S6,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
103 C7,S7,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
104 S3,C3,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
105
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
106 -----------------------------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
107 FFMPEG conformance testing results
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
108 -----------------------------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
109 dct-test: modified with the following
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
110 dct_error("BFINfdct", 0, ff_bfin_fdct, fdct, test);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
111 produces the following output:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
112
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
113 root:/u/ffmpeg/bhead/libavcodec> ./dct-test
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
114 ffmpeg DCT/IDCT test
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
115
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
116 2 -131 -6 -48 -36 33 -83 24
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
117 34 52 -24 -15 5 92 57 143
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
118 -67 -43 -1 74 -16 5 -71 32
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
119 -78 106 92 -34 -38 81 20 -18
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
120 7 -62 40 2 -15 90 -62 -83
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
121 -83 1 -104 -13 43 -19 7 11
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
122 -63 31 12 -29 83 72 21 10
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
123 -17 -63 -15 73 50 -91 159 -14
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
124 DCT BFINfdct: err_inf=2 err2=0.16425938 syserr=0.00795000 maxout=2098 blockSumErr=27
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
125 DCT BFINfdct: 92.1 kdct/s
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
126 root:/u/ffmpeg/bhead/libavcodec>
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
127
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
128 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
129
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
130 #include "config_bfin.h"
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
131
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
132 .section .l1.data.B,"aw",@progbits
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
133 .align 4;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
134 dct_coeff:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
135 .short 0x5a82, 0x2d41, 0x187e, 0x3b21, 0x0c7c, 0x3ec5, 0x238e, 0x3537;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
136
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
137 .section .l1.data.A,"aw",@progbits
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
138 .align 4
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
139 vtmp: .space 128
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
140
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
141 DEFUN(fdct,mL1,
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
142 (DCTELEM *block)):
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
143 [--SP] = (R7:4, P5:3); // Push the registers onto the stack.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
144
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
145 b0 = r0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
146 r0 = [P3+dct_coeff@GOT17M4];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
147 b3 = r0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
148 r0 = [P3+vtmp@GOT17M4];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
149 b2 = r0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
150
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
151 L3 = 16; // L3 is set to 16 to make the coefficient
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
152 // array Circular.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
153
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
154
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
155 //----------------------------------------------------------------------------
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
156
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
157 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
158 * I0, I1, and I2 registers are used to read the input data. I3 register is used
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
159 * to read the coefficients. P0 and P1 registers are used for writing the output
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
160 * data.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
161 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
162 M0 = 12 (X); // All these initializations are used in the
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
163 M1 = 16 (X); // modification of address offsets.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
164
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
165 M2 = 128 (X);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
166
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
167 P2 = 16;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
168 P3 = 32 (X);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
169 P4 = -110 (X);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
170 P5 = -62 (X);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
171 P0 = 2(X);
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
172
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
173
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
174 // Prescale the input to get the correct precision.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
175 i0=b0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
176 i1=b0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
177
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
178 lsetup (.0, .1) LC0 = P3;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
179 r0=[i0++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
180 .0: r1=r0<<3 (v) || r0=[i0++] ;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
181 .1: [i1++]=r1;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
182
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
183 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
184 * B0 points to the "in" buffer.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
185 * B2 points to "temp" buffer in the first iteration.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
186 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
187
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
188 lsetup (.2, .3) LC0 = P0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
189 .2:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
190 I0 = B0; // I0 points to Input Element (0, 0).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
191 I1 = B0; // Element 1 and 0 is read in R0.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
192 I1 += M0 || R0 = [I0++]; // I1 points to Input Element (0, 6).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
193 I2 = I1; // Element 6 is read into R3.H.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
194 I2 -= 4 || R3.H = W[I1++]; // I2 points to Input Element (0, 4).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
195
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
196 I3 = B3; // I3 points to Coefficients.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
197 P0 = B2; // P0 points to temporary array Element
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
198 // (0, 0).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
199 P1 = B2; // P1 points to temporary array.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
200 R7 = [P1++P2] || R2 = [I2++]; // P1 points to temporary array
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
201 // Element (1, 0).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
202 // R7 is a dummy read. X4,X5
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
203 // are read into R2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
204 R3.L = W[I1--]; // X7 is read into R3.L.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
205 R1.H = W[I0++]; // X2 is read into R1.H.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
206
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
207
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
208 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
209 * X0 = (X0 + X7) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
210 * X1 = (X1 + X6) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
211 * X6 = (X1 - X6) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
212 * X7 = (X0 - X7) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
213 * It reads the data 3 in R1.L.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
214 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
215
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
216 R0 = R0 +|+ R3, R3 = R0 -|- R3 || R1.L = W[I0++] || NOP;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
217
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
218 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
219 * X2 = (X2 + X5) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
220 * X3 = (X3 + X4) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
221 * X4 = (X3 - X4) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
222 * X5 = (X2 - X5) / 2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
223 * R7 = C4 = cos(4*pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
224 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
225
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
226 R1 = R1 +|+ R2, R2 = R1 -|- R2 (CO) || NOP || R7 = [I3++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
227
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
228 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
229 * At the end of stage 1 R0 has (1,0), R1 has (2,3), R2 has (4, 5) and
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
230 * R3 has (6,7).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
231 * Where the notation (x, y) represents uper/lower half pairs.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
232 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
233
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
234 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
235 * X0 = X0 + X3.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
236 * X1 = X1 + X2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
237 * X2 = X1 - X2.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
238 * X3 = X0 - X3.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
239 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
240 R0 = R0 +|+ R1, R1 = R0 -|- R1;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
241
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
242 lsetup (.row0, .row1) LC1 = P2 >> 1; // 1d dct, loops 8x
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
243 .row0:
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
244
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
245 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
246 * This is part 2 computation continued.....
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
247 * A1 = X6 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
248 * A0 = X6 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
249 * A1 = A1 - X5 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
250 * A0 = A0 + X5 * cos(pi/4).
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
251 * The instruction W[I0] = R3.L is used for packing it to R2.L.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
252 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
253
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
254 A1=R3.H*R7.l, A0=R3.H*R7.l || I1+=M1 || W[I0] = R3.L;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
255 R4.H=(A1-=R2.L*R7.l), R4.L=(A0+=R2.L*R7.l) || I2+=M0 || NOP;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
256
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
257 /* R0 = (X1,X0) R1 = (X2,X3) R4 = (X5, X6). */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
258
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
259 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
260 * A1 = X0 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
261 * A0 = X0 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
262 * A1 = A1 - X1 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
263 * A0 = A0 + X1 * cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
264 * R7 = (C2,C6)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
265 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
266 A1=R0.L*R7.h, A0=R0.L*R7.h || NOP || R3.H=W[I1++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
267 R5.H=(A1-=R0.H*R7.h),R5.L=(A0+=R0.H*R7.h) || R7=[I3++] || NOP;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
268
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
269 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
270 * A1 = X2 * cos(3pi/8)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
271 * A0 = X3 * cos(3pi/8)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
272 * A1 = A1 + X3 * cos(pi/8)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
273 * A0 = A0 - X2 * cos(pi/8)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
274 * R3 = cos(pi/4)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
275 * R7 = (cos(7pi/8),cos(pi/8))
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
276 * X4 = X4 + X5.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
277 * X5 = X4 - X5.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
278 * X6 = X7 - X6.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
279 * X7 = X7 + X6.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
280 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
281 A1=R1.H*R7.L, A0=R1.L*R7.L || W[P0++P3]=R5.L || R2.L=W[I0];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
282 R2=R2+|+R4, R4=R2-|-R4 || I0+=4 || R3.L=W[I1--];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
283 R6.H=(A1+=R1.L*R7.H),R6.L=(A0 -= R1.H * R7.H) || I0+=4 || R7=[I3++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
284
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
285 /* R2 = (X4, X7) R4 = (X5,X6) R5 = (X1, X0) R6 = (X2,X3). */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
286
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
287 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
288 * A1 = X4 * cos(7pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
289 * A0 = X7 * cos(7pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
290 * A1 = A1 + X7 * cos(pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
291 * A0 = A0 - X4 * cos(pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
292 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
293
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
294 A1=R2.H*R7.L, A0=R2.L*R7.L || W[P0++P3]=R6.H || R0=[I0++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
295 R2.H=(A1+=R2.L*R7.H),R2.L=(A0-=R2.H*R7.H) || W[P0++P3]=R5.H || R7=[I3++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
296
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
297 /*
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
298 * A1 = X5 * cos(3pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
299 * A0 = X6 * cos(3pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
300 * A1 = A1 + X6 * cos(5pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
301 * A0 = A0 - X5 * cos(5pi/16)
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
302 * The output values are written.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
303 */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
304
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
305 A1=R4.H*R7.H, A0=R4.L*R7.H || W[P0++P2]=R6.L || R1.H=W[I0++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
306 R4.H=(A1+=R4.L*R7.L),R4.L=(A0-=R4.H*R7.L) || W[P0++P4]=R2.L || R1.L=W[I0++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
307
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
308
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
309 /* Beginning of next stage, **pipelined** + drain and store the
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
310 rest of the column store. */
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
311
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
312 R0=R0+|+R3,R3=R0-|-R3 || W[P1++P3]=R2.H || R2=[I2++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
313 R1=R1+|+R2,R2=R1-|-R2 (CO) || W[P1++P3]=R4.L || R7=[I3++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
314 .row1: R0=R0+|+R1,R1=R0-|-R1 || W[P1++P5]=R4.H || NOP;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
315
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
316 // Exchange input with output.
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
317 B1 = B0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
318 B0 = B2;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
319 .3: B2 = B1;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
320
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
321 L3=0;
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
322 (r7:4,p5:3) = [sp++];
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
323 RTS;
5001
75bf61c6c385 Blackfin DSP utilities: add DEFUN_END
gpoirier
parents: 4765
diff changeset
324 DEFUN_END(fdct)
4765
85298e8c55c4 bfin dsputils, basic pixel operations sads, diffs, motion compensation
diego
parents:
diff changeset
325