1 /*****************************************************************************
2 * ppccommon.h: h264 encoder
3 *****************************************************************************
4 * Copyright (C) 2003 Laurent Aimar
5 * $Id: ppccommon.h,v 1.1 2004/06/03 19:27:07 fenrir Exp $
7 * Authors: Eric Petit <titer@m0k.org>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
22 *****************************************************************************/
24 /***********************************************************************
25 * For constant vectors, use parentheses on OS X and braces on Linux
26 **********************************************************************/
33 /***********************************************************************
35 **********************************************************************/
36 #define vec_u8_t vector unsigned char
37 #define vec_s8_t vector signed char
38 #define vec_u16_t vector unsigned short
39 #define vec_s16_t vector signed short
40 #define vec_u32_t vector unsigned int
41 #define vec_s32_t vector signed int
43 /***********************************************************************
45 **********************************************************************/
46 #define LOAD_ZERO const vec_u8_t zerov = vec_splat_u8( 0 )
48 #define zero_u8v (vec_u8_t) zerov
49 #define zero_s8v (vec_s8_t) zerov
50 #define zero_u16v (vec_u16_t) zerov
51 #define zero_s16v (vec_s16_t) zerov
52 #define zero_u32v (vec_u32_t) zerov
53 #define zero_s32v (vec_s32_t) zerov
55 /***********************************************************************
56 * 8 <-> 16 bits conversions
57 **********************************************************************/
58 #define vec_u8_to_u16_h(v) (vec_u16_t) vec_mergeh( zero_u8v, (vec_u8_t) v )
59 #define vec_u8_to_u16_l(v) (vec_u16_t) vec_mergel( zero_u8v, (vec_u8_t) v )
60 #define vec_u8_to_s16_h(v) (vec_s16_t) vec_mergeh( zero_u8v, (vec_u8_t) v )
61 #define vec_u8_to_s16_l(v) (vec_s16_t) vec_mergel( zero_u8v, (vec_u8_t) v )
63 #define vec_u8_to_u16(v) vec_u8_to_u16_h(v)
64 #define vec_u8_to_s16(v) vec_u8_to_s16_h(v)
66 #define vec_u16_to_u8(v) vec_pack( v, zero_u16v )
67 #define vec_s16_to_u8(v) vec_pack( v, zero_u16v )
69 /***********************************************************************
70 * PREP_LOAD: declares two vectors required to perform unaligned loads
71 * VEC_LOAD: loads n bytes from u8 * p into vector v of type t
72 **********************************************************************/
76 #define VEC_LOAD( p, v, n, t ) \
77 _hv = vec_ld( 0, p ); \
78 v = (t) vec_lvsl( 0, p ); \
79 _lv = vec_ld( n - 1, p ); \
80 v = (t) vec_perm( _hv, _lv, (vec_u8_t) v )
82 /***********************************************************************
83 * PREP_STORE##n: declares required vectors to store n bytes to a
84 * potentially unaligned address
85 * VEC_STORE##n: stores n bytes from vector v to address p
86 **********************************************************************/
87 #define PREP_STORE16 \
88 vec_u8_t _tmp1v, _tmp2v \
90 #define VEC_STORE16( v, p ) \
91 _hv = vec_ld( 0, p ); \
92 _tmp2v = vec_lvsl( 0, p ); \
93 _lv = vec_ld( 15, p ); \
94 _tmp1v = vec_perm( _lv, _hv, _tmp2v ); \
95 _tmp2v = vec_lvsr( 0, p ); \
96 _lv = vec_perm( (vec_u8_t) v, _tmp1v, _tmp2v ); \
97 vec_st( _lv, 15, (uint8_t *) p ); \
98 _hv = vec_perm( _tmp1v, (vec_u8_t) v, _tmp2v ); \
99 vec_st( _hv, 0, (uint8_t *) p )
101 #define PREP_STORE8 \
103 vec_u8_t _tmp3v, _tmp4v; \
104 const vec_u8_t sel_h = \
105 (vec_u8_t) CV(-1,-1,-1,-1,-1,-1,-1,-1,0,0,0,0,0,0,0,0)
107 #define PREP_STORE8_HL \
109 const vec_u8_t sel_l = \
110 (vec_u8_t) CV(0,0,0,0,0,0,0,0,-1,-1,-1,-1,-1,-1,-1,-1)
112 #define VEC_STORE8 VEC_STORE8_H
114 #define VEC_STORE8_H( v, p ) \
115 _tmp3v = vec_lvsr( 0, (uint8_t *) p ); \
116 _tmp4v = vec_perm( (vec_u8_t) v, (vec_u8_t) v, _tmp3v ); \
117 _lv = vec_ld( 7, (uint8_t *) p ); \
118 _tmp1v = vec_perm( sel_h, zero_u8v, _tmp3v ); \
119 _lv = vec_sel( _lv, _tmp4v, _tmp1v ); \
120 vec_st( _lv, 7, (uint8_t *) p ); \
121 _hv = vec_ld( 0, (uint8_t *) p ); \
122 _tmp2v = vec_perm( zero_u8v, sel_h, _tmp3v ); \
123 _hv = vec_sel( _hv, _tmp4v, _tmp2v ); \
124 vec_st( _hv, 0, (uint8_t *) p )
126 #define VEC_STORE8_L( v, p ) \
127 _tmp3v = vec_lvsr( 8, (uint8_t *) p ); \
128 _tmp4v = vec_perm( (vec_u8_t) v, (vec_u8_t) v, _tmp3v ); \
129 _lv = vec_ld( 7, (uint8_t *) p ); \
130 _tmp1v = vec_perm( sel_l, zero_u8v, _tmp3v ); \
131 _lv = vec_sel( _lv, _tmp4v, _tmp1v ); \
132 vec_st( _lv, 7, (uint8_t *) p ); \
133 _hv = vec_ld( 0, (uint8_t *) p ); \
134 _tmp2v = vec_perm( zero_u8v, sel_l, _tmp3v ); \
135 _hv = vec_sel( _hv, _tmp4v, _tmp2v ); \
136 vec_st( _hv, 0, (uint8_t *) p )
138 #define PREP_STORE4 \
141 const vec_u8_t sel = \
142 (vec_u8_t) CV(-1,-1,-1,-1,0,0,0,0,0,0,0,0,0,0,0,0)
144 #define VEC_STORE4( v, p ) \
145 _tmp3v = vec_lvsr( 0, p ); \
146 v = vec_perm( v, v, _tmp3v ); \
147 _lv = vec_ld( 3, p ); \
148 _tmp1v = vec_perm( sel, zero_u8v, _tmp3v ); \
149 _lv = vec_sel( _lv, v, _tmp1v ); \
150 vec_st( _lv, 3, p ); \
151 _hv = vec_ld( 0, p ); \
152 _tmp2v = vec_perm( zero_u8v, sel, _tmp3v ); \
153 _hv = vec_sel( _hv, v, _tmp2v ); \
156 /***********************************************************************
158 ***********************************************************************
159 * Transposes a 8x8 matrix of s16 vectors
160 **********************************************************************/
161 #define VEC_TRANSPOSE_8(a0,a1,a2,a3,a4,a5,a6,a7,b0,b1,b2,b3,b4,b5,b6,b7) \
162 b0 = vec_mergeh( a0, a4 ); \
163 b1 = vec_mergel( a0, a4 ); \
164 b2 = vec_mergeh( a1, a5 ); \
165 b3 = vec_mergel( a1, a5 ); \
166 b4 = vec_mergeh( a2, a6 ); \
167 b5 = vec_mergel( a2, a6 ); \
168 b6 = vec_mergeh( a3, a7 ); \
169 b7 = vec_mergel( a3, a7 ); \
170 a0 = vec_mergeh( b0, b4 ); \
171 a1 = vec_mergel( b0, b4 ); \
172 a2 = vec_mergeh( b1, b5 ); \
173 a3 = vec_mergel( b1, b5 ); \
174 a4 = vec_mergeh( b2, b6 ); \
175 a5 = vec_mergel( b2, b6 ); \
176 a6 = vec_mergeh( b3, b7 ); \
177 a7 = vec_mergel( b3, b7 ); \
178 b0 = vec_mergeh( a0, a4 ); \
179 b1 = vec_mergel( a0, a4 ); \
180 b2 = vec_mergeh( a1, a5 ); \
181 b3 = vec_mergel( a1, a5 ); \
182 b4 = vec_mergeh( a2, a6 ); \
183 b5 = vec_mergel( a2, a6 ); \
184 b6 = vec_mergeh( a3, a7 ); \
185 b7 = vec_mergel( a3, a7 )
187 /***********************************************************************
189 ***********************************************************************
190 * Transposes a 4x4 matrix of s16 vectors.
191 * Actually source and destination are 8x4. The low elements of the
192 * source are discarded and the low elements of the destination mustn't
194 **********************************************************************/
195 #define VEC_TRANSPOSE_4(a0,a1,a2,a3,b0,b1,b2,b3) \
196 b0 = vec_mergeh( a0, a0 ); \
197 b1 = vec_mergeh( a1, a0 ); \
198 b2 = vec_mergeh( a2, a0 ); \
199 b3 = vec_mergeh( a3, a0 ); \
200 a0 = vec_mergeh( b0, b2 ); \
201 a1 = vec_mergel( b0, b2 ); \
202 a2 = vec_mergeh( b1, b3 ); \
203 a3 = vec_mergel( b1, b3 ); \
204 b0 = vec_mergeh( a0, a2 ); \
205 b1 = vec_mergel( a0, a2 ); \
206 b2 = vec_mergeh( a1, a3 ); \
207 b3 = vec_mergel( a1, a3 )
209 /***********************************************************************
211 ***********************************************************************
216 * Loads n bytes from p1 and p2, do the diff of the high elements into
217 * d, increments p1 and p2 by i1 and i2
218 **********************************************************************/
222 vec_s16_t pix1v, pix2v;
224 #define VEC_DIFF_H(p1,i1,p2,i2,n,d) \
225 VEC_LOAD( p1, pix1v, n, vec_s16_t ); \
226 pix1v = vec_u8_to_s16( pix1v ); \
227 VEC_LOAD( p2, pix2v, n, vec_s16_t ); \
228 pix2v = vec_u8_to_s16( pix2v ); \
229 d = vec_sub( pix1v, pix2v ); \
233 /***********************************************************************
235 ***********************************************************************
240 * Loads 16 bytes from p1 and p2, do the diff of the high elements into
241 * dh, the diff of the low elements into dl, increments p1 and p2 by i1
243 **********************************************************************/
244 #define VEC_DIFF_HL(p1,i1,p2,i2,dh,dl) \
245 VEC_LOAD( p1, pix1v, 16, vec_s16_t ); \
246 temp0v = vec_u8_to_s16_h( pix1v ); \
247 temp1v = vec_u8_to_s16_l( pix1v ); \
248 VEC_LOAD( p2, pix2v, 16, vec_s16_t ); \
249 temp2v = vec_u8_to_s16_h( pix2v ); \
250 temp3v = vec_u8_to_s16_l( pix2v ); \
251 dh = vec_sub( temp0v, temp2v ); \
252 dl = vec_sub( temp1v, temp3v ); \
256 /***********************************************************************
257 * VEC_DIFF_H_8BYTE_ALIGNED
258 ***********************************************************************
263 * Loads n bytes from p1 and p2, do the diff of the high elements into
264 * d, increments p1 and p2 by i1 and i2
265 * Slightly faster when we know we are loading/diffing 8bytes which
266 * are 8 byte aligned. Reduces need for two loads and two vec_lvsl()'s
267 **********************************************************************/
268 #define PREP_DIFF_8BYTEALIGNED \
270 vec_s16_t pix1v, pix2v; \
271 vec_u8_t permPix1, permPix2; \
272 permPix1 = vec_lvsl(0, pix1); \
273 permPix2 = vec_lvsl(0, pix2); \
275 #define VEC_DIFF_H_8BYTE_ALIGNED(p1,i1,p2,i2,n,d) \
276 pix1v = vec_perm(vec_ld(0,p1), zero_u8v, permPix1); \
277 pix2v = vec_perm(vec_ld(0, p2), zero_u8v, permPix2); \
278 pix1v = vec_u8_to_s16( pix1v ); \
279 pix2v = vec_u8_to_s16( pix2v ); \
280 d = vec_sub( pix1v, pix2v); \