4 * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavcodec/dsputil.h"
26 #define c1 1.38703984532214752434 /* sqrt(2)*cos(1*pi/16) */
27 #define c2 1.30656296487637657577 /* sqrt(2)*cos(2*pi/16) */
28 #define c3 1.17587560241935884520 /* sqrt(2)*cos(3*pi/16) */
29 #define c4 1.00000000000000000000 /* sqrt(2)*cos(4*pi/16) */
30 #define c5 0.78569495838710234903 /* sqrt(2)*cos(5*pi/16) */
31 #define c6 0.54119610014619712324 /* sqrt(2)*cos(6*pi/16) */
32 #define c7 0.27589937928294311353 /* sqrt(2)*cos(7*pi/16) */
34 static const float even_table[] __attribute__ ((aligned(8))) = {
41 static const float odd_table[] __attribute__ ((aligned(8))) = {
58 #define load_matrix(table) \
60 const float *t = table; \
77 __asm__ volatile("ftrv xmtrx,fv0" \
78 : "+f"(fr0),"+f"(fr1),"+f"(fr2),"+f"(fr3));
81 register float fr0 __asm__("fr0"); \
82 register float fr1 __asm__("fr1"); \
83 register float fr2 __asm__("fr2"); \
84 register float fr3 __asm__("fr3")
88 /* generic C code for check */
90 static void ftrv_(const float xf[],float fv[])
97 fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3;
98 fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3;
99 fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3;
100 fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3;
103 static void load_matrix_(float xf[],const float table[])
106 for(i=0;i<16;i++) xf[i]=table[i];
109 #define ftrv() ftrv_(xf,fv)
110 #define load_matrix(table) load_matrix_(xf,table)
123 #define DESCALE(x,n) (x)*(1.0f/(1<<(n)))
125 #define DESCALE(x,n) (((int)(x)+(1<<(n-1)))>>(n))
128 /* this code work worse on gcc cvs. 3.2.3 work fine */
134 void idct_sh4(DCTELEM *block)
139 float tblock[8*8],*fblock;
143 fp_single_enter(fpscr);
148 load_matrix(even_table);
168 load_matrix(odd_table);
185 *--fblock = t0 - fr0;
186 *--fblock = t1 - fr1;
187 *--fblock = t2 - fr2;
188 *--fblock = t3 - fr3;
189 *--fblock = t3 + fr3;
190 *--fblock = t2 + fr2;
191 *--fblock = t1 + fr1;
192 *--fblock = t0 + fr0;
201 load_matrix(even_table);
203 ofs1 = sizeof(float)*2*8;
204 ofs2 = sizeof(float)*4*8;
205 ofs3 = sizeof(float)*6*8;
209 #define OA(fblock,ofs) *(float*)((char*)fblock + ofs)
213 fr1 = OA(fblock,ofs1);
214 fr2 = OA(fblock,ofs2);
215 fr3 = OA(fblock,ofs3);
218 OA(fblock,ofs1) = fr1;
219 OA(fblock,ofs2) = fr2;
220 OA(fblock,ofs3) = fr3;
225 load_matrix(odd_table);
230 t0 = OA(fblock, 0); /* [8*0] */
231 t1 = OA(fblock,ofs1); /* [8*2] */
232 t2 = OA(fblock,ofs2); /* [8*4] */
233 t3 = OA(fblock,ofs3); /* [8*6] */
235 fr0 = OA(fblock, 0); /* [8*1] */
236 fr1 = OA(fblock,ofs1); /* [8*3] */
237 fr2 = OA(fblock,ofs2); /* [8*5] */
238 fr3 = OA(fblock,ofs3); /* [8*7] */
241 block[8*0] = DESCALE(t0 + fr0,3);
242 block[8*7] = DESCALE(t0 - fr0,3);
243 block[8*1] = DESCALE(t1 + fr1,3);
244 block[8*6] = DESCALE(t1 - fr1,3);
245 block[8*2] = DESCALE(t2 + fr2,3);
246 block[8*5] = DESCALE(t2 - fr2,3);
247 block[8*3] = DESCALE(t3 + fr3,3);
248 block[8*4] = DESCALE(t3 - fr3,3);
252 fp_single_leave(fpscr);
255 void idct_sh4(DCTELEM *block)
260 float tblock[8*8],*fblock;
265 load_matrix(even_table);
285 load_matrix(odd_table);
301 fblock[0] = t0 + fr0;
302 fblock[7] = t0 - fr0;
303 fblock[1] = t1 + fr1;
304 fblock[6] = t1 - fr1;
305 fblock[2] = t2 + fr2;
306 fblock[5] = t2 - fr2;
307 fblock[3] = t3 + fr3;
308 fblock[4] = t3 - fr3;
317 load_matrix(even_table);
335 load_matrix(odd_table);
350 block[8*0] = DESCALE(t0 + fr0,3);
351 block[8*7] = DESCALE(t0 - fr0,3);
352 block[8*1] = DESCALE(t1 + fr1,3);
353 block[8*6] = DESCALE(t1 - fr1,3);
354 block[8*2] = DESCALE(t2 + fr2,3);
355 block[8*5] = DESCALE(t2 - fr2,3);
356 block[8*3] = DESCALE(t3 + fr3,3);
357 block[8*4] = DESCALE(t3 - fr3,3);