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"
24 #define c1 1.38703984532214752434 /* sqrt(2)*cos(1*pi/16) */
25 #define c2 1.30656296487637657577 /* sqrt(2)*cos(2*pi/16) */
26 #define c3 1.17587560241935884520 /* sqrt(2)*cos(3*pi/16) */
27 #define c4 1.00000000000000000000 /* sqrt(2)*cos(4*pi/16) */
28 #define c5 0.78569495838710234903 /* sqrt(2)*cos(5*pi/16) */
29 #define c6 0.54119610014619712324 /* sqrt(2)*cos(6*pi/16) */
30 #define c7 0.27589937928294311353 /* sqrt(2)*cos(7*pi/16) */
32 static const float even_table[] __attribute__ ((aligned(8))) = {
39 static const float odd_table[] __attribute__ ((aligned(8))) = {
54 #if defined(__SH4_SINGLE__) || defined(__SH4_SINGLE_ONLY__)
56 #define load_matrix(table) \
74 asm volatile("ftrv xmtrx,fv0" \
75 : "=f"(fr0),"=f"(fr1),"=f"(fr2),"=f"(fr3) \
76 : "0"(fr0), "1"(fr1), "2"(fr2), "3"(fr3) );
79 register float fr0 asm("fr0"); \
80 register float fr1 asm("fr1"); \
81 register float fr2 asm("fr2"); \
82 register float fr3 asm("fr3")
86 /* generic C code for check */
88 static void ftrv_(const float xf[],float fv[])
95 fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3;
96 fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3;
97 fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3;
98 fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3;
101 static void load_matrix_(float xf[],const float table[])
104 for(i=0;i<16;i++) xf[i]=table[i];
107 #define ftrv() ftrv_(xf,fv)
108 #define load_matrix(table) load_matrix_(xf,table)
121 #define DESCALE(x,n) (x)*(1.0f/(1<<(n)))
123 #define DESCALE(x,n) (((int)(x)+(1<<(n-1)))>>(n))
126 /* this code work worse on gcc cvs. 3.2.3 work fine */
132 void idct_sh4(DCTELEM *block)
137 float tblock[8*8],*fblock;
141 #error "FIXME!! change to single float"
147 load_matrix(even_table);
167 load_matrix(odd_table);
171 // ofs1 = sizeof(float)*1;
172 // ofs2 = sizeof(float)*2;
173 // ofs3 = sizeof(float)*3;
188 *--fblock = t0 - fr0;
189 *--fblock = t1 - fr1;
190 *--fblock = t2 - fr2;
191 *--fblock = t3 - fr3;
192 *--fblock = t3 + fr3;
193 *--fblock = t2 + fr2;
194 *--fblock = t1 + fr1;
195 *--fblock = t0 + fr0;
204 load_matrix(even_table);
206 ofs1 = sizeof(float)*2*8;
207 ofs2 = sizeof(float)*4*8;
208 ofs3 = sizeof(float)*6*8;
212 #define OA(fblock,ofs) *(float*)((char*)fblock + ofs)
216 fr1 = OA(fblock,ofs1);
217 fr2 = OA(fblock,ofs2);
218 fr3 = OA(fblock,ofs3);
221 OA(fblock,ofs1) = fr1;
222 OA(fblock,ofs2) = fr2;
223 OA(fblock,ofs3) = fr3;
228 load_matrix(odd_table);
233 t0 = OA(fblock, 0); /* [8*0] */
234 t1 = OA(fblock,ofs1); /* [8*2] */
235 t2 = OA(fblock,ofs2); /* [8*4] */
236 t3 = OA(fblock,ofs3); /* [8*6] */
238 fr0 = OA(fblock, 0); /* [8*1] */
239 fr1 = OA(fblock,ofs1); /* [8*3] */
240 fr2 = OA(fblock,ofs2); /* [8*5] */
241 fr3 = OA(fblock,ofs3); /* [8*7] */
244 block[8*0] = DESCALE(t0 + fr0,3);
245 block[8*7] = DESCALE(t0 - fr0,3);
246 block[8*1] = DESCALE(t1 + fr1,3);
247 block[8*6] = DESCALE(t1 - fr1,3);
248 block[8*2] = DESCALE(t2 + fr2,3);
249 block[8*5] = DESCALE(t2 - fr2,3);
250 block[8*3] = DESCALE(t3 + fr3,3);
251 block[8*4] = DESCALE(t3 - fr3,3);
256 #error "FIXME!! change to double"
260 void idct_sh4(DCTELEM *block)
265 float tblock[8*8],*fblock;
270 load_matrix(even_table);
290 load_matrix(odd_table);
306 fblock[0] = t0 + fr0;
307 fblock[7] = t0 - fr0;
308 fblock[1] = t1 + fr1;
309 fblock[6] = t1 - fr1;
310 fblock[2] = t2 + fr2;
311 fblock[5] = t2 - fr2;
312 fblock[3] = t3 + fr3;
313 fblock[4] = t3 - fr3;
322 load_matrix(even_table);
340 load_matrix(odd_table);
355 block[8*0] = DESCALE(t0 + fr0,3);
356 block[8*7] = DESCALE(t0 - fr0,3);
357 block[8*1] = DESCALE(t1 + fr1,3);
358 block[8*6] = DESCALE(t1 - fr1,3);
359 block[8*2] = DESCALE(t2 + fr2,3);
360 block[8*5] = DESCALE(t2 - fr2,3);
361 block[8*3] = DESCALE(t3 + fr3,3);
362 block[8*4] = DESCALE(t3 - fr3,3);