2 * Blackfin Pixel Operations
3 * Copyright (C) 2007 Marc Hoffman <marc.hoffman@analog.com>
5 * This file is part of Libav.
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.
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.
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
21 #include "config_bfin.h"
23 DEFUN(put_pixels_clamped,mL1,
24 (int16_t *block, uint8_t *dest, int line_size)):
31 R2 += -4; // line_size
37 LSETUP (ppc$0,ppc$1) LC0=P0;
38 ppc$0: R2 = MIN(R2, R5) (V);
40 R3 = MIN(R3, R5) (V) || R0 = [I0++];
41 R6 = BYTEPACK (R2,R3) || R1 = [I0++];
42 R2 = MAX(R0, R4) (V) || [I1++] = R6;
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;
51 DEFUN_END(put_pixels_clamped)
53 DEFUN(add_pixels_clamped,mL1,
54 (int16_t *block, uint8_t *dest, int line_size)):
58 R2 += -4; // line_size
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];
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$3: R6 = BYTEOP3P(R1:0, R3:2) (LO) || [I2++M0] = R6 || R2 = [I1];
84 DEFUN_END(add_pixels_clamped)
86 DEFUN(diff_pixels,mL1,
87 (int16_t *block, uint8_t *s1, uint8_t *s2, int stride)):
99 LSETUP(.LS0,.LE0) LC0=P0;
100 DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
102 .LS0: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
103 (R5,R4) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0];
104 (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || [I3++] = R4;
105 DISALGNEXCPT || R2 = [I1++] || [I3++] = R5;
112 DEFUN_END(diff_pixels)
115 for (i = 0; i < 16; i++) {
116 for (j = 0; j < 16; j++) {
123 (uint8_t *p, int stride)):
137 LSETUP(LS$PS,LE$PS) LC0=P0;
138 DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
140 LS$PS: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
141 (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++] || R2 = [I1++];
144 (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R1 = [I0++] || R3 = [I1++];
147 (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++m0] || R2 = [I1++m0];
150 (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R0 = [I0++] || R2 = [I1++];
162 DEFUN(get_pixels,mL1,
163 (int16_t *restrict block, const uint8_t *pixels, int line_size)):
170 LSETUP(gp8$0,gp8$1) LC0=P0;
172 DISALGNEXCPT || R0 = [I0++];
173 DISALGNEXCPT || R1 = [I0++];
175 gp8$0: (R7,R6) = byteunpack R1:0 || R0 = [I0++M0];
176 (R5,R4) = byteunpack R1:0 (R) || R0 = [I0++] || [I3++]=R6;
177 DISALGNEXCPT || R1 = [I0++] || [I3++]=R7;
184 DEFUN_END(get_pixels)
187 /* sad = sad16x16 (ubyte *mb, ubyte *refwin, srcwidth, refwinwidth, h) */
189 DEFUN(z_sad16x16,mL1,
190 (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)):
196 R0 = [sp+20]; // rwidth
197 P2 = [sp+24]; // height
204 DISALGNEXCPT || R0 = [I0++] || R2 = [I1++];
205 LSETUP (s$16, e$16) LC0=P2;
206 s$16: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
207 SAA (R1:0,R3:2) || R0 = [I0++] || R2 = [I1++];
208 SAA (R1:0,R3:2) (R) || R1 = [I0++] || R3 = [I1++];
209 SAA (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M1];
210 e$16: SAA (R1:0,R3:2) (R) || R0 = [I0++] || R2 = [I1++];
212 R3=A1.L+A1.H, R2=A0.L+A0.H ;
216 DEFUN_END(z_sad16x16)
218 /* sad = sad8x8 (ubyte *mb, ubyte *refwin, int srcwidth, int refwinwidth, int h) */
221 (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h)):
226 r0 = [sp+12]; // rwidth
227 P2 = [sp+16]; //height
234 LSETUP (s$8, e$8) LC0=P2;
235 DISALGNEXCPT || R0 = [I0++] || R2 = [I1++];
236 DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
237 s$8: SAA (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M1];
238 SAA (R1:0,R3:2) (R) || R0 = [I0++] || R2 = [I1++];
239 e$8: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
241 R3=A1.L+A1.H, R2=A0.L+A0.H ;
247 (uint8_t * pix, int line_size)):
250 // Fetch the input arguments.
252 P0 = R1; // line_size
253 P5 = 16; // loop ctr.
255 M0 = P0; // M0 = line_size-16;
256 // Now for the real work.
258 lsetup(_pix_norm1_blkfn_loopStart, _pix_norm1_blkfn_loopEnd) LC1 = P5;
260 DISALGNEXCPT || r0 = [i0++];
262 _pix_norm1_blkfn_loopStart:
263 // following unpacks pix1[0..15] pix1+line_size[0..15]
264 DISALGNEXCPT || r1 = [i0++];
266 (r5, r4) = byteunpack r1:0 || r0 = [i0++];
267 a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
268 a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
269 (r5, r4) = byteunpack r1:0(r) || r1 = [i0++];
270 a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
271 a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
272 (r5, r4) = byteunpack r1:0 || r0 = [i0++M0];
273 a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
274 a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
275 (r5, r4) = byteunpack r1:0(r) || r0 = [i0++];
276 a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
277 _pix_norm1_blkfn_loopEnd:
278 a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
281 // Clean up at the end:
291 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
297 r2=[fp+20]; // line_size
302 LSETUP(.S40,.E40) LC0=P0;
303 DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
305 .S40: DISALGNEXCPT || R1 = [I0++M0] || R3 = [I1++M0];
306 (R7,R6) = BYTEOP16M (R1:0,R3:2);
307 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
308 .E40: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
318 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
324 r2=[fp+20]; // line_size
329 LSETUP(.S80,.E80) LC0=P0;
330 DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
332 .S80: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
333 (R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0];
334 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
335 a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
336 (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || R2 = [I1++];
337 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
338 .E80: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
348 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
354 r2=[fp+20]; // line_size
359 DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
360 LSETUP(.S160,.E160) LC0=P0;
362 .S160: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
363 (R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++] || R2 = [I1++];
364 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
365 a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
366 (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R1 = [I0++] || R3 = [I1++];
367 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
368 a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
369 (R7,R6) = BYTEOP16M (R1:0,R3:2) || R0 = [I0++M0] || R2 = [I1++M0];
370 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
371 a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);
372 (R7,R6) = BYTEOP16M (R1:0,R3:2) (R)|| R0 = [I0++] || R2 = [I1++];
373 a0 += r7.l * r7.l, a1 += r7.h * r7.h (is);
374 .E160: a0 += r6.l * r6.l, a1 += r6.h * r6.h (is);