4 * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "../dsputil.h"
22 #define c1 1.38703984532214752434 /* sqrt(2)*cos(1*pi/16) */
23 #define c2 1.30656296487637657577 /* sqrt(2)*cos(2*pi/16) */
24 #define c3 1.17587560241935884520 /* sqrt(2)*cos(3*pi/16) */
25 #define c4 1.00000000000000000000 /* sqrt(2)*cos(4*pi/16) */
26 #define c5 0.78569495838710234903 /* sqrt(2)*cos(5*pi/16) */
27 #define c6 0.54119610014619712324 /* sqrt(2)*cos(6*pi/16) */
28 #define c7 0.27589937928294311353 /* sqrt(2)*cos(7*pi/16) */
30 static const float even_table[] __attribute__ ((aligned(8))) = {
37 static const float odd_table[] __attribute__ ((aligned(8))) = {
52 #if defined(__SH4_SINGLE__) || defined(__SH4_SINGLE_ONLY__)
54 #define load_matrix(table) \
72 __asm__ volatile("ftrv xmtrx,fv0" \
73 : "=f"(fr0),"=f"(fr1),"=f"(fr2),"=f"(fr3) \
74 : "0"(fr0), "1"(fr1), "2"(fr2), "3"(fr3) );
77 register float fr0 __asm__("fr0"); \
78 register float fr1 __asm__("fr1"); \
79 register float fr2 __asm__("fr2"); \
80 register float fr3 __asm__("fr3")
84 /* generic C code for check */
86 static void ftrv_(const float xf[],float fv[])
93 fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3;
94 fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3;
95 fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3;
96 fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3;
99 static void load_matrix_(float xf[],const float table[])
102 for(i=0;i<16;i++) xf[i]=table[i];
105 #define ftrv() ftrv_(xf,fv)
106 #define load_matrix(table) load_matrix_(xf,table)
119 #define DESCALE(x,n) (x)*(1.0f/(1<<(n)))
121 #define DESCALE(x,n) (((int)(x)+(1<<(n-1)))>>(n))
124 /* this code work worse on gcc cvs. 3.2.3 work fine */
130 void idct_sh4(DCTELEM *block)
135 float tblock[8*8],*fblock;
139 #error "FIXME!! change to single float"
145 load_matrix(even_table);
165 load_matrix(odd_table);
169 // ofs1 = sizeof(float)*1;
170 // ofs2 = sizeof(float)*2;
171 // ofs3 = sizeof(float)*3;
186 *--fblock = t0 - fr0;
187 *--fblock = t1 - fr1;
188 *--fblock = t2 - fr2;
189 *--fblock = t3 - fr3;
190 *--fblock = t3 + fr3;
191 *--fblock = t2 + fr2;
192 *--fblock = t1 + fr1;
193 *--fblock = t0 + fr0;
202 load_matrix(even_table);
204 ofs1 = sizeof(float)*2*8;
205 ofs2 = sizeof(float)*4*8;
206 ofs3 = sizeof(float)*6*8;
210 #define OA(fblock,ofs) *(float*)((char*)fblock + ofs)
214 fr1 = OA(fblock,ofs1);
215 fr2 = OA(fblock,ofs2);
216 fr3 = OA(fblock,ofs3);
219 OA(fblock,ofs1) = fr1;
220 OA(fblock,ofs2) = fr2;
221 OA(fblock,ofs3) = fr3;
226 load_matrix(odd_table);
231 t0 = OA(fblock, 0); /* [8*0] */
232 t1 = OA(fblock,ofs1); /* [8*2] */
233 t2 = OA(fblock,ofs2); /* [8*4] */
234 t3 = OA(fblock,ofs3); /* [8*6] */
236 fr0 = OA(fblock, 0); /* [8*1] */
237 fr1 = OA(fblock,ofs1); /* [8*3] */
238 fr2 = OA(fblock,ofs2); /* [8*5] */
239 fr3 = OA(fblock,ofs3); /* [8*7] */
242 block[8*0] = DESCALE(t0 + fr0,3);
243 block[8*7] = DESCALE(t0 - fr0,3);
244 block[8*1] = DESCALE(t1 + fr1,3);
245 block[8*6] = DESCALE(t1 - fr1,3);
246 block[8*2] = DESCALE(t2 + fr2,3);
247 block[8*5] = DESCALE(t2 - fr2,3);
248 block[8*3] = DESCALE(t3 + fr3,3);
249 block[8*4] = DESCALE(t3 - fr3,3);
254 #error "FIXME!! change to double"
258 void idct_sh4(DCTELEM *block)
263 float tblock[8*8],*fblock;
268 load_matrix(even_table);
288 load_matrix(odd_table);
304 fblock[0] = t0 + fr0;
305 fblock[7] = t0 - fr0;
306 fblock[1] = t1 + fr1;
307 fblock[6] = t1 - fr1;
308 fblock[2] = t2 + fr2;
309 fblock[5] = t2 - fr2;
310 fblock[3] = t3 + fr3;
311 fblock[4] = t3 - fr3;
320 load_matrix(even_table);
338 load_matrix(odd_table);
353 block[8*0] = DESCALE(t0 + fr0,3);
354 block[8*7] = DESCALE(t0 - fr0,3);
355 block[8*1] = DESCALE(t1 + fr1,3);
356 block[8*6] = DESCALE(t1 - fr1,3);
357 block[8*2] = DESCALE(t2 + fr2,3);
358 block[8*5] = DESCALE(t2 - fr2,3);
359 block[8*3] = DESCALE(t3 + fr3,3);
360 block[8*4] = DESCALE(t3 - fr3,3);