## ffmpeg / libavcodec / bfin / pixels_bfin.S @ 2912e87a

History | View | Annotate | Download (23.7 KB)

1 |
/* |
---|---|

2 |
* Blackfin Pixel Operations |

3 |
* Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com> |

4 |
* |

5 |
* This file is part of Libav. |

6 |
* |

7 |
* Libav is free software; you can redistribute it and/or |

8 |
* modify it under the terms of the GNU Lesser General Public |

9 |
* License as published by the Free Software Foundation; either |

10 |
* version 2.1 of the License, or (at your option) any later version. |

11 |
* |

12 |
* Libav is distributed in the hope that it will be useful, |

13 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |

14 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |

15 |
* Lesser General Public License for more details. |

16 |
* |

17 |
* You should have received a copy of the GNU Lesser General Public |

18 |
* License along with Libav; if not, write to the Free Software |

19 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |

20 |
*/ |

21 |
#include "config_bfin.h" |

22 | |

23 |
DEFUN(put_pixels_clamped,mL1, |

24 |
(DCTELEM *block, uint8_t *dest, int line_size)): |

25 |
[--SP] = (R7:4); |

26 |
R4 = 0; |

27 |
R5.l = 0x00ff; |

28 |
R5.h = 0x00ff; |

29 |
I0 = R0; // block |

30 |
I1 = R1; // dest |

31 |
R2 += -4; // line_size |

32 |
M1 = R2; |

33 |
P0 = 8; |

34 |
R0 = [I0++]; |

35 |
R1 = [I0++]; |

36 |
R2 = MAX(R0, R4) (V); |

37 |
LSETUP (ppc$0,ppc$1) LC0=P0; |

38 |
ppc$0: R2 = MIN(R2, R5) (V); |

39 |
R3 = MAX(R1, R4) (V); |

40 |
R3 = MIN(R3, R5) (V) || R0 = [I0++]; |

41 |
R6 = BYTEPACK (R2,R3) || R1 = [I0++]; |

42 |
R2 = MAX(R0, R4) (V) || [I1++] = R6; |

43 |
R2 = MIN(R2, R5) (V); |

44 |
R3 = MAX(R1, R4) (V); |

45 |
R3 = MIN(R3, R5) (V) || R0 = [I0++]; |

46 |
R6 = BYTEPACK (R2,R3) || R1 = [I0++]; |

47 |
ppc$1: R2 = Max(R0, R4) (V) || [I1++M1] = R6; |

48 | |

49 |
(R7:4) = [SP++]; |

50 |
RTS; |

51 |
DEFUN_END(put_pixels_clamped) |

52 | |

53 |
DEFUN(add_pixels_clamped,mL1, |

54 |
(DCTELEM *block, uint8_t *dest, int line_size)): |

55 |
[-- SP] = (R7:4); |

56 |
R4 = 0; |

57 |
I0 = 0; |

58 |
R2 += -4; // line_size |

59 |
M0 = R2; |

60 |
I1 = R1; // dest |

61 |
I3 = R0; // block |

62 |
I2 = R1; // dest |

63 |
P0 = 8; |

64 |
M3 = 2; |

65 |
R0 = [I3++] || R2 = [I1]; |

66 |
R2 = R2 << 8 || R0.H = W[I3--] || R3 = [I1++]; |

67 |
R3 = R3 >> 8 || R1.L = W[I3] || I3 += 4; |

68 |
R6 = BYTEOP3P(R1:0, R3:2) (LO) || R1.H = W[I3++] || R2 = [I1]; |

69 | |

70 |
LSETUP(apc$2,apc$3) LC1 = P0; |

71 |
apc$2: R7 = BYTEOP3P(R1:0, R3:2) (HI, R) || R0 = [I3++] || R3 = [I1++M0]; |

72 |
R2 = R2 << 8 || R0.H = W[I3--]; |

73 |
R3 = R3 >> 8 || R1.L = W[I3] || I3 += 4; |

74 |
R6 = R6 + R7 (S) || R1.H = W[I3]; |

75 |
R6 = BYTEOP3P(R1:0, R3:2) (LO) || I3+=M3 || [I2++]=R6; |

76 |
R7 = BYTEOP3P(R1:0, R3:2) (HI, R) || R0 = [I3++] || R2 = [I1]; |

77 |
R2 = R2 << 8 || R0.H = W[I3--] || R3 = [I1++]; |

78 |
R3 = R3 >> 8 || R1.L = W[I3] || I3 += 4; |

79 |
R6 = R6 + R7 (S) || R1.H = W[I3++]; |

80 |
apc$3: R6 = BYTEOP3P(R1:0, R3:2) (LO) || [I2++M0] = R6 || R2 = [I1]; |

81 | |

82 |
(R7:4) = [SP++]; |

83 |
RTS; |

84 |
DEFUN_END(add_pixels_clamped) |

85 | |

86 | |

87 |
/* |

88 |
motion compensation |

89 |
primitives |

90 | |

91 |
* Halfpel motion compensation with rounding (a+b+1)>>1. |

92 |
* This is an array[4][4] of motion compensation funcions for 4 |

93 |
* horizontal blocksizes (8,16) and the 4 halfpel positions<br> |

94 |
* *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ] |

95 |
* @param block destination where the result is stored |

96 |
* @param pixels source |

97 |
* @param line_size number of bytes in a horizontal line of block |

98 |
* @param h height |

99 | |

100 |
*/ |

101 | |

102 |
DEFUN(put_pixels8uc,mL1, |

103 |
(uint8_t *block, const uint8_t *s0, const uint8_t *s1, |

104 |
int dest_size, int line_size, int h)): |

105 |
i3=r0; // dest |

106 |
i0=r1; // src0 |

107 |
i1=r2; // src1 |

108 |
r0=[sp+12]; // dest_size |

109 |
r2=[sp+16]; // line_size |

110 |
p0=[sp+20]; // h |

111 |
[--sp] = (r7:6); |

112 |
r0+=-4; |

113 |
m3=r0; |

114 |
r2+=-8; |

115 |
m0=r2; |

116 |
LSETUP(pp8$0,pp8$1) LC0=P0; |

117 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

118 | |

119 |
pp8$0: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

120 |
R6 = BYTEOP1P(R1:0,R3:2) || R0 = [I0++M0]|| R2 =[I1++M0]; |

121 |
R7 = BYTEOP1P(R1:0,R3:2)(R) || R0 = [I0++] || [I3++] = R6 ; |

122 |
pp8$1: DISALGNEXCPT || R2 = [I1++] || [I3++M3] = R7; |

123 | |

124 |
(r7:6) = [sp++]; |

125 |
RTS; |

126 |
DEFUN_END(put_pixels8uc) |

127 | |

128 |
DEFUN(put_pixels16uc,mL1, |

129 |
(uint8_t *block, const uint8_t *s0, const uint8_t *s1, |

130 |
int dest_size, int line_size, int h)): |

131 |
link 0; |

132 |
[--sp] = (r7:6); |

133 |
i3=r0; // dest |

134 |
i0=r1; // src0 |

135 |
i1=r2; // src1 |

136 |
r0=[fp+20]; // dest_size |

137 |
r2=[fp+24]; // line_size |

138 |
p0=[fp+28]; // h |

139 | |

140 | |

141 |
r0+=-12; |

142 |
m3=r0; // line_size |

143 |
r2+=-16; |

144 |
m0=r2; |

145 | |

146 |
LSETUP(pp16$0,pp16$1) LC0=P0; |

147 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

148 | |

149 |
pp16$0: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

150 |
R6 = BYTEOP1P(R1:0,R3:2) || R0 = [I0++] || R2 =[I1++]; |

151 |
R7 = BYTEOP1P(R1:0,R3:2)(R) || R1 = [I0++] || R3 =[I1++]; |

152 |
[I3++] = R6; |

153 |
R6 = BYTEOP1P(R1:0,R3:2) || R0 = [I0++M0] || R2 =[I1++M0]; |

154 |
R7 = BYTEOP1P(R1:0,R3:2)(R) || R0 = [I0++] || [I3++] = R7 ; |

155 |
[I3++] = R6; |

156 |
pp16$1: DISALGNEXCPT || R2 = [I1++] || [I3++M3] = R7; |

157 | |

158 |
(r7:6) = [sp++]; |

159 |
unlink; |

160 |
RTS; |

161 |
DEFUN_END(put_pixels16uc) |

162 | |

163 | |

164 | |

165 | |

166 | |

167 | |

168 |
DEFUN(put_pixels8uc_nornd,mL1, |

169 |
(uint8_t *block, const uint8_t *s0, const uint8_t *s1, |

170 |
int line_size, int h)): |

171 |
i3=r0; // dest |

172 |
i0=r1; // src0 |

173 |
i1=r2; // src1 |

174 |
r2=[sp+12]; // line_size |

175 |
p0=[sp+16]; // h |

176 |
[--sp] = (r7:6); |

177 |
r2+=-4; |

178 |
m3=r2; |

179 |
r2+=-4; |

180 |
m0=r2; |

181 |
LSETUP(pp8$2,pp8$3) LC0=P0; |

182 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

183 | |

184 |
pp8$2: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

185 |
R6 = BYTEOP1P(R1:0,R3:2)(T) || R0 = [I0++M0]|| R2 =[I1++M0]; |

186 |
R7 = BYTEOP1P(R1:0,R3:2)(T,R) || R0 = [I0++] || [I3++] = R6 ; |

187 |
pp8$3: DISALGNEXCPT || R2 = [I1++] || [I3++M3] = R7; |

188 | |

189 |
(r7:6) = [sp++]; |

190 |
RTS; |

191 |
DEFUN_END(put_pixels8uc_nornd) |

192 | |

193 |
DEFUN(put_pixels16uc_nornd,mL1, |

194 |
(uint8_t *block, const uint8_t *s0, const uint8_t *s1, |

195 |
int line_size, int h)): |

196 |
i3=r0; // dest |

197 |
i0=r1; // src0 |

198 |
i1=r2; // src1 |

199 |
r2=[sp+12]; // line_size |

200 |
p0=[sp+16]; // h |

201 | |

202 |
[--sp] = (r7:6); |

203 |
r2+=-12; |

204 |
m3=r2; // line_size |

205 |
r2+=-4; |

206 |
m0=r2; |

207 | |

208 |
LSETUP(pp16$2,pp16$3) LC0=P0; |

209 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

210 | |

211 |
pp16$2: |

212 |
DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

213 |
R6 = BYTEOP1P(R1:0,R3:2)(T) || R0 = [I0++] || R2 =[I1++]; |

214 |
R7 = BYTEOP1P(R1:0,R3:2)(T,R) || R1 = [I0++] || R3 =[I1++]; |

215 |
[I3++] = R6; |

216 | |

217 |
R6 = BYTEOP1P(R1:0,R3:2)(T) || R0 = [I0++M0] || R2 =[I1++M0]; |

218 |
R7 = BYTEOP1P(R1:0,R3:2)(T,R) || R0 = [I0++] || [I3++] = R7 ; |

219 |
[I3++] = R6; |

220 |
pp16$3: DISALGNEXCPT || R2 = [I1++] || [I3++M3] = R7; |

221 | |

222 |
(r7:6) = [sp++]; |

223 | |

224 |
RTS; |

225 |
DEFUN_END(put_pixels16uc_nornd) |

226 | |

227 |
DEFUN(z_put_pixels16_xy2,mL1, |

228 |
(uint8_t *block, const uint8_t *s0, |

229 |
int dest_size, int line_size, int h)): |

230 |
link 0; |

231 |
[--sp] = (r7:4); |

232 |
i3=r0; // dest |

233 |
i0=r1; // src0--> pixels |

234 |
i1=r1; // src1--> pixels + line_size |

235 |
r2+=-12; |

236 |
m2=r2; // m2=dest_width-4 |

237 |
r2=[fp+20]; |

238 |
m3=r2; // line_size |

239 |
p0=[fp+24]; // h |

240 |
r2+=-16; |

241 |
i1+=m3; /* src1 + line_size */ |

242 |
m0=r2; /* line-size - 20 */ |

243 | |

244 |
B0 = I0; |

245 |
B1 = I1; |

246 |
B3 = I3; |

247 | |

248 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

249 | |

250 |
LSETUP(LS$16E,LE$16E) LC0=P0; |

251 |
LS$16E: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

252 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDL) || R0 = [I0++] || R2 =[I1++]; |

253 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDL,R) || R1 = [I0++] || [I3++] = R4 ; |

254 |
DISALGNEXCPT || R3 = [I1++] || [I3++] = R5; |

255 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDL) || R0 = [I0++M0]|| R2 = [I1++M0]; |

256 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDL,R) || R0 = [I0++] || [I3++] = R4 ; |

257 |
LE$16E: DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

258 | |

259 |
M1 = 1; |

260 |
I3 = B3; |

261 |
I1 = B1; |

262 |
I0 = B0; |

263 | |

264 |
I0 += M1; |

265 |
I1 += M1; |

266 | |

267 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

268 |
LSETUP(LS$16O,LE$16O) LC0=P0; |

269 |
LS$16O: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

270 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDH) || R0 = [I0++] || R2 =[I1++]; |

271 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDH,R) || R1 = [I0++] || R6 =[I3++]; |

272 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

273 |
R5 = R5 +|+ R7 || [I3++] = R4; |

274 |
DISALGNEXCPT || R3 =[I1++] || [I3++] = R5; |

275 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDH) || R0 = [I0++M0]|| R2 = [I1++M0]; |

276 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDH,R) || R0 = [I0++] || R6 = [I3++]; |

277 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

278 |
R5 = R5 +|+ R7 || [I3++] = R4; |

279 |
LE$16O: DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

280 | |

281 |
(r7:4) = [sp++]; |

282 |
unlink; |

283 |
rts; |

284 |
DEFUN_END(z_put_pixels16_xy2) |

285 | |

286 |
DEFUN(put_pixels16_xy2_nornd,mL1, |

287 |
(uint8_t *block, const uint8_t *s0, |

288 |
int line_size, int h)): |

289 |
link 0; |

290 |
[--sp] = (r7:4); |

291 |
i3=r0; // dest |

292 |
i0=r1; // src0--> pixels |

293 |
i1=r1; // src1--> pixels + line_size |

294 |
m3=r2; |

295 |
r2+=-12; |

296 |
m2=r2; |

297 |
r2+=-4; |

298 |
i1+=m3; /* src1 + line_size */ |

299 |
m0=r2; /* line-size - 20 */ |

300 |
p0=[fp+20]; // h |

301 | |

302 |
B0=I0; |

303 |
B1=I1; |

304 |
B3=I3; |

305 | |

306 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

307 | |

308 |
LSETUP(LS$16ET,LE$16ET) LC0=P0; |

309 |
LS$16ET:DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

310 |
R4 = BYTEOP2P (R3:2,R1:0) (TL) || R0 = [I0++] || R2 =[I1++]; |

311 |
R5 = BYTEOP2P (R3:2,R1:0) (TL,R) || R1 = [I0++] || [I3++] = R4 ; |

312 |
DISALGNEXCPT || R3 = [I1++] || [I3++] = R5; |

313 |
R4 = BYTEOP2P (R3:2,R1:0) (TL) || R0 = [I0++M0]|| R2 = [I1++M0]; |

314 |
R5 = BYTEOP2P (R3:2,R1:0) (TL,R) || R0 = [I0++] || [I3++] = R4 ; |

315 |
LE$16ET:DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

316 | |

317 |
M1 = 1; |

318 |
I3=B3; |

319 |
I1=B1; |

320 |
I0=B0; |

321 | |

322 |
I0 += M1; |

323 |
I1 += M1; |

324 | |

325 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

326 |
LSETUP(LS$16OT,LE$16OT) LC0=P0; |

327 |
LS$16OT:DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

328 |
R4 = BYTEOP2P (R3:2,R1:0) (TH) || R0 = [I0++] || R2 =[I1++]; |

329 |
R5 = BYTEOP2P (R3:2,R1:0) (TH,R) || R1 = [I0++] || R6 =[I3++]; |

330 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

331 |
R5 = R5 +|+ R7 || [I3++] = R4; |

332 |
DISALGNEXCPT || R3 =[I1++] || [I3++] = R5; |

333 |
R4 = BYTEOP2P (R3:2,R1:0) (TH) || R0 = [I0++M0]|| R2 = [I1++M0]; |

334 |
R5 = BYTEOP2P (R3:2,R1:0) (TH,R) || R0 = [I0++] || R6 = [I3++]; |

335 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

336 |
R5 = R5 +|+ R7 || [I3++] = R4; |

337 |
LE$16OT:DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

338 | |

339 |
(r7:4) = [sp++]; |

340 |
unlink; |

341 |
rts; |

342 |
DEFUN_END(put_pixels16_xy2_nornd) |

343 | |

344 |
DEFUN(z_put_pixels8_xy2,mL1, |

345 |
(uint8_t *block, const uint8_t *s0, |

346 |
int dest_size, int line_size, int h)): |

347 |
link 0; |

348 |
[--sp] = (r7:4); |

349 |
i3=r0; // dest |

350 |
i0=r1; // src0--> pixels |

351 |
i1=r1; // src1--> pixels + line_size |

352 |
r2+=-4; |

353 |
m2=r2; // m2=dest_width-4 |

354 |
r2=[fp+20]; |

355 |
m3=r2; // line_size |

356 |
p0=[fp+24]; // h |

357 |
r2+=-8; |

358 |
i1+=m3; /* src1 + line_size */ |

359 |
m0=r2; /* line-size - 20 */ |

360 | |

361 |
b0 = I0; |

362 |
b1 = I1; |

363 |
b3 = I3; |

364 | |

365 |
LSETUP(LS$8E,LE$8E) LC0=P0; |

366 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

367 |
LS$8E: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

368 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDL) || R0 = [I0++M0] || R2 =[I1++M0]; |

369 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDL,R) || R0 = [I0++] || [I3++] = R4 ; |

370 |
LE$8E: DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

371 | |

372 |
M1 = 1; |

373 |
I3 = b3; |

374 |
I1 = b1; |

375 |
I0 = b0; |

376 | |

377 |
I0 += M1; |

378 |
I1 += M1; |

379 | |

380 |
LSETUP(LS$8O,LE$8O) LC0=P0; |

381 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

382 |
LS$8O: DISALGNEXCPT || R1 = [I0++] || R3 =[I1++]; |

383 |
R4 = BYTEOP2P (R3:2,R1:0) (RNDH) || R0 = [I0++M0] || R2 =[I1++M0]; |

384 |
R5 = BYTEOP2P (R3:2,R1:0) (RNDH,R) || R0 = [I0++] || R6 =[I3++]; |

385 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

386 |
R5 = R5 +|+ R7 || [I3++] = R4; |

387 |
LE$8O: DISALGNEXCPT || R2 =[I1++] || [I3++M2] = R5; |

388 | |

389 |
(r7:4) = [sp++]; |

390 |
unlink; |

391 |
rts; |

392 |
DEFUN_END(z_put_pixels8_xy2) |

393 | |

394 |
DEFUN(put_pixels8_xy2_nornd,mL1, |

395 |
(uint8_t *block, const uint8_t *s0, int line_size, int h)): |

396 |
link 0; |

397 |
[--sp] = (r7:4); |

398 |
i3=r0; // dest |

399 |
i0=r1; // src0--> pixels |

400 |
i1=r1; // src1--> pixels + line_size |

401 |
m3=r2; |

402 |
r2+=-4; |

403 |
m2=r2; |

404 |
r2+=-4; |

405 |
i1+=m3; /* src1 + line_size */ |

406 |
m0=r2; /* line-size - 20 */ |

407 |
p0=[fp+20]; // h |

408 | |

409 | |

410 |
b0 = I0; |

411 |
b1 = I1; |

412 |
b3 = I3; |

413 | |

414 |
LSETUP(LS$8ET,LE$8ET) LC0=P0; |

415 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

416 | |

417 |
LS$8ET: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

418 |
R4 = BYTEOP2P (R3:2,R1:0) (TL) || R0 = [I0++M0] || R2 = [I1++M0]; |

419 |
R5 = BYTEOP2P (R3:2,R1:0) (TL,R) || R0 = [I0++] || [I3++] = R4 ; |

420 |
LE$8ET: DISALGNEXCPT || R2 = [I1++] || [I3++M2] = R5; |

421 | |

422 |
M1 = 1; |

423 |
I3 = b3; |

424 |
I1 = b1; |

425 |
I0 = b0; |

426 | |

427 |
I0 += M1; |

428 |
I1 += M1; |

429 | |

430 |
LSETUP(LS$8OT,LE$8OT) LC0=P0; |

431 |
DISALGNEXCPT || R0 = [I0++] || R2 = [I1++]; |

432 | |

433 |
LS$8OT: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

434 |
R4 = BYTEOP2P (R3:2,R1:0) (TH) || R0 = [I0++M0] || R2 = [I1++M0]; |

435 |
R5 = BYTEOP2P (R3:2,R1:0) (TH,R) || R0 = [I0++] || R6 = [I3++]; |

436 |
R4 = R4 +|+ R6 || R7 = [I3--]; |

437 |
R5 = R5 +|+ R7 || [I3++] = R4; |

438 |
LE$8OT: DISALGNEXCPT || R2 =[I1++] || [I3++M2] = R5; |

439 | |

440 |
(r7:4) = [sp++]; |

441 |
unlink; |

442 |
rts; |

443 | |

444 |
DEFUN(diff_pixels,mL1, |

445 |
(DCTELEM *block, uint8_t *s1, uint8_t *s2, int stride)): |

446 |
link 0; |

447 |
[--sp] = (r7:4); |

448 |
p0=8; |

449 |
i3=r0; // block |

450 |
i0=r1; // s1 |

451 |
i1=r2; // s2 |

452 |
r2=[fp+20]; // stride |

453 |
r2+=-8; |

454 |
m0=r2; |

455 | |

456 | |

457 |
LSETUP(.LS0,.LE0) LC0=P0; |

458 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

459 | |

460 |
.LS0: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

461 |
(R5,R4) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0]; |

462 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || [I3++] = R4; |

463 |
DISALGNEXCPT || R2 = [I1++] || [I3++] = R5; |

464 |
[i3++]=r6; |

465 |
.LE0: [i3++]=r7; |

466 | |

467 |
(r7:4) = [sp++]; |

468 |
unlink; |

469 |
rts; |

470 |
DEFUN_END(put_pixels8_xy2_nornd) |

471 | |

472 |
/* |

473 |
for (i = 0; i < 16; i++) { |

474 |
for (j = 0; j < 16; j++) { |

475 |
sum += pix[j]; |

476 |
} |

477 |
pix += line_size; |

478 |
} |

479 |
*/ |

480 |
DEFUN(pix_sum,mL1, |

481 |
(uint8_t *p, int stride)): |

482 |
link 0; |

483 |
[--sp] = (r7:4); |

484 |
p0=8; |

485 |
i0=r0; // s1 |

486 |
i1=r0; |

487 |
m1=r1; |

488 |
r1=r1+r1; |

489 |
r1+=-16; // stride |

490 |
m0=r1; |

491 |
i1+=m1; |

492 | |

493 |
r6=0; |

494 | |

495 |
LSETUP(LS$PS,LE$PS) LC0=P0; |

496 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

497 | |

498 |
LS$PS: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

499 |
(R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++] || R2 = [I1++]; |

500 |
r6=r6+|+r5; |

501 |
r6=r6+|+r4; |

502 |
(R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R1 = [I0++] || R3 = [I1++]; |

503 |
r6=r6+|+r5; |

504 |
r6=r6+|+r4; |

505 |
(R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++m0] || R2 = [I1++m0]; |

506 |
r6=r6+|+r5; |

507 |
r6=r6+|+r4; |

508 |
(R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R0 = [I0++] || R2 = [I1++]; |

509 |
r6=r6+|+r5; |

510 |
LE$PS: r6=r6+|+r4; |

511 |
r0.l=r6.l+r6.h; |

512 |
r0.h=0; |

513 | |

514 |
(r7:4) = [sp++]; |

515 |
unlink; |

516 |
rts; |

517 |
DEFUN_END(pix_sum) |

518 | |

519 | |

520 |
DEFUN(get_pixels,mL1, |

521 |
(DCTELEM *restrict block, const uint8_t *pixels, int line_size)): |

522 |
[--sp] = (r7:4); |

523 |
i3=r0; // dest |

524 |
i0=r1; // src0 |

525 |
p0=8; |

526 |
r2+=-8; |

527 |
m0=r2; |

528 |
LSETUP(gp8$0,gp8$1) LC0=P0; |

529 | |

530 |
DISALGNEXCPT || R0 = [I0++]; |

531 |
DISALGNEXCPT || R1 = [I0++]; |

532 | |

533 |
gp8$0: (R7,R6) = byteunpack R1:0 || R0 = [I0++M0]; |

534 |
(R5,R4) = byteunpack R1:0 (R) || R0 = [I0++] || [I3++]=R6; |

535 |
DISALGNEXCPT || R1 = [I0++] || [I3++]=R7; |

536 |
[I3++]=R4; |

537 |
gp8$1: [I3++]=R5 |

538 | |

539 | |

540 |
(r7:4) = [sp++]; |

541 |
RTS; |

542 |
DEFUN_END(get_pixels) |

543 | |

544 | |

545 |
/* sad = sad16x16 (ubyte *mb, ubyte *refwin, srcwidth, refwinwidth, h) */ |

546 |
/* 91 cycles */ |

547 |
DEFUN(z_sad16x16,mL1, |

548 |
(uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)): |

549 |
link 0; |

550 |
I0 = R0; |

551 |
I1 = R1; |

552 | |

553 |
A1 = A0 = 0; |

554 |
R0 = [sp+20]; // rwidth |

555 |
P2 = [sp+24]; // height |

556 |
R3 = 16; |

557 |
R0 = R0 - R3; |

558 |
R3 = R2 - R3; |

559 |
M1 = R0; |

560 |
M0 = R3; |

561 | |

562 |
DISALGNEXCPT || R0 = [I0++] || R2 = [I1++]; |

563 |
LSETUP (s$16, e$16) LC0=P2; |

564 |
s$16: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

565 |
SAA (R1:0,R3:2) || R0 = [I0++] || R2 = [I1++]; |

566 |
SAA (R1:0,R3:2) (R) || R1 = [I0++] || R3 = [I1++]; |

567 |
SAA (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M1]; |

568 |
e$16: SAA (R1:0,R3:2) (R) || R0 = [I0++] || R2 = [I1++]; |

569 | |

570 |
R3=A1.L+A1.H, R2=A0.L+A0.H ; |

571 |
R0 = R2 + R3 ; |

572 |
unlink; |

573 |
RTS; |

574 |
DEFUN_END(z_sad16x16) |

575 | |

576 |
/* sad = sad8x8 (ubyte *mb, ubyte *refwin, int srcwidth, int refwinwidth, int h) */ |

577 |
/* 36 cycles */ |

578 |
DEFUN(z_sad8x8,mL1, |

579 |
(uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)): |

580 |
I0 = R0; |

581 |
I1 = R1; |

582 | |

583 |
A1 = A0 = 0; |

584 |
r0 = [sp+12]; // rwidth |

585 |
P2 = [sp+16]; //height |

586 |
R3 = 8; |

587 |
R0 = R0 - R3; |

588 |
R3 = R2 - R3; |

589 |
M0 = R3; |

590 |
M1 = R0; |

591 | |

592 |
LSETUP (s$8, e$8) LC0=P2; |

593 |
DISALGNEXCPT || R0 = [I0++] || R2 = [I1++]; |

594 |
DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

595 |
s$8: SAA (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M1]; |

596 |
SAA (R1:0,R3:2) (R) || R0 = [I0++] || R2 = [I1++]; |

597 |
e$8: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

598 | |

599 |
R3=A1.L+A1.H, R2=A0.L+A0.H ; |

600 |
R0 = R2 + R3 ; |

601 |
RTS; |

602 |
DEFUN_END(z_sad8x8) |

603 | |

604 |
DEFUN(pix_norm1,mL1, |

605 |
(uint8_t * pix, int line_size)): |

606 |
[--SP]=(R7:4,P5:3); |

607 | |

608 |
// Fetch the input arguments. |

609 |
P1 = R0; // pix |

610 |
P0 = R1; // line_size |

611 |
P5 = 16; // loop ctr. |

612 |
P0 -= P5; |

613 |
M0 = P0; // M0 = line_size-16; |

614 |
// Now for the real work. |

615 |
A1 = A0 = 0; |

616 |
lsetup(_pix_norm1_blkfn_loopStart, _pix_norm1_blkfn_loopEnd) LC1 = P5; |

617 |
I0 = P1; |

618 |
DISALGNEXCPT || r0 = [i0++]; |

619 | |

620 |
_pix_norm1_blkfn_loopStart: |

621 |
// following unpacks pix1[0..15] pix1+line_size[0..15] |

622 |
DISALGNEXCPT || r1 = [i0++]; |

623 | |

624 |
(r5, r4) = byteunpack r1:0 || r0 = [i0++]; |

625 |
a1 += r5.h * r5.h, a0 += r5.l * r5.l (is); |

626 |
a1 += r4.h * r4.h, a0 += r4.l * r4.l (is); |

627 |
(r5, r4) = byteunpack r1:0(r) || r1 = [i0++]; |

628 |
a1 += r5.h * r5.h, a0 += r5.l * r5.l (is); |

629 |
a1 += r4.h * r4.h, a0 += r4.l * r4.l (is); |

630 |
(r5, r4) = byteunpack r1:0 || r0 = [i0++M0]; |

631 |
a1 += r5.h * r5.h, a0 += r5.l * r5.l (is); |

632 |
a1 += r4.h * r4.h, a0 += r4.l * r4.l (is); |

633 |
(r5, r4) = byteunpack r1:0(r) || r0 = [i0++]; |

634 |
a1 += r5.h * r5.h, a0 += r5.l * r5.l (is); |

635 |
_pix_norm1_blkfn_loopEnd: |

636 |
a1 += r4.h * r4.h, a0 += r4.l * r4.l (is); |

637 | |

638 | |

639 |
// Clean up at the end: |

640 |
R2 = A0, R3 = A1; |

641 |
R0 = R2 + R3 (S); |

642 | |

643 |
(R7:4,P5:3)=[SP++]; |

644 | |

645 |
RTS; |

646 |
DEFUN_END(pix_norm1) |

647 | |

648 |
DEFUN(sse4,mL1, |

649 |
(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)): |

650 |
link 0; |

651 |
[--sp] = (r7:6); |

652 |
p0=[fp+24]; // h |

653 |
i0=r1; // pix1 |

654 |
i1=r2; // pix2 |

655 |
r2=[fp+20]; // line_size |

656 |
r2+=-4; |

657 |
m0=r2; |

658 | |

659 |
a0=a1=0; |

660 |
LSETUP(.S40,.E40) LC0=P0; |

661 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

662 | |

663 |
.S40: DISALGNEXCPT || R1 = [I0++M0] || R3 = [I1++M0]; |

664 |
(R7,R6) = BYTEOP16M (R1:0,R3:2); |

665 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

666 |
.E40: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

667 |
a0 += a1; |

668 |
r0 = a0; |

669 | |

670 |
(r7:6) = [sp++]; |

671 |
unlink; |

672 |
rts; |

673 |
DEFUN_END(sse4) |

674 | |

675 |
DEFUN(sse8,mL1, |

676 |
(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)): |

677 |
link 0; |

678 |
[--sp] = (r7:6); |

679 |
p0=[fp+24]; // h |

680 |
i0=r1; // pix1 |

681 |
i1=r2; // pix2 |

682 |
r2=[fp+20]; // line_size |

683 |
r2+=-8; |

684 |
m0=r2; |

685 | |

686 |
a0=a1=0; |

687 |
LSETUP(.S80,.E80) LC0=P0; |

688 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

689 | |

690 |
.S80: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

691 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0]; |

692 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

693 |
a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

694 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || R2 = [I1++]; |

695 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

696 |
.E80: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

697 |
a0 += a1; |

698 |
r0 = a0; |

699 | |

700 |
(r7:6) = [sp++]; |

701 |
unlink; |

702 |
rts; |

703 |
DEFUN_END(sse8) |

704 | |

705 |
DEFUN(sse16,mL1, |

706 |
(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)): |

707 |
link 0; |

708 |
[--sp] = (r7:6); |

709 |
p0=[fp+24]; // h |

710 |
i0=r1; // pix1 |

711 |
i1=r2; // pix2 |

712 |
r2=[fp+20]; // line_size |

713 |
r2+=-16; |

714 |
m0=r2; |

715 | |

716 |
a0=a1=0; |

717 |
DISALGNEXCPT || R0 = [I0++] || R2 =[I1++]; |

718 |
LSETUP(.S160,.E160) LC0=P0; |

719 | |

720 |
.S160: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++]; |

721 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++] || R2 = [I1++]; |

722 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

723 |
a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

724 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R1 = [I0++] || R3 = [I1++]; |

725 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

726 |
a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

727 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0]; |

728 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

729 |
a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

730 |
(R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || R2 = [I1++]; |

731 |
a0 += r7.l * r7.l, a1 += r7.h * r7.h (is); |

732 |
.E160: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is); |

733 |
a0 += a1; |

734 |
r0 = a0; |

735 | |

736 |
(r7:6) = [sp++]; |

737 |
unlink; |

738 |
rts; |

739 |
DEFUN_END(sse16) |

740 | |

741 |