2 * ARM NEON optimised DSP functions
3 * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
5 * This file is part of FFmpeg.
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/arm/asm.S"
24 function ff_clear_block_neon, export=1
27 vst1.16 {q0}, [r0,:128]!
32 function ff_clear_blocks_neon, export=1
35 vst1.16 {q0}, [r0,:128]!
40 .macro pixels16 rnd=1, avg=0
44 1: vld1.8 {q0}, [r1], r2
53 vld1.8 {q8}, [r12,:128], r2
55 vld1.8 {q9}, [r12,:128], r2
57 vld1.8 {q10}, [r12,:128], r2
59 vld1.8 {q11}, [r12,:128], r2
63 vst1.64 {q0}, [r0,:128], r2
64 vst1.64 {q1}, [r0,:128], r2
65 vst1.64 {q2}, [r0,:128], r2
66 vst1.64 {q3}, [r0,:128], r2
71 .macro pixels16_x2 rnd=1, avg=0
72 1: vld1.8 {d0-d2}, [r1], r2
73 vld1.8 {d4-d6}, [r1], r2
82 vld1.8 {q1}, [r0,:128], r2
83 vld1.8 {q3}, [r0,:128]
88 vst1.8 {q0}, [r0,:128], r2
89 vst1.8 {q2}, [r0,:128], r2
94 .macro pixels16_y2 rnd=1, avg=0
100 vld1.8 {q0}, [r1], r2
102 vld1.8 {q1}, [r1], r2
106 vld1.8 {q8}, [r0,:128], r2
107 vld1.8 {q9}, [r0,:128]
112 vst1.8 {q2}, [r0,:128], r2
113 vst1.8 {q3}, [r0,:128], r2
117 vld1.8 {q0}, [r1], r2
120 vld1.8 {q8}, [r0,:128], r2
121 vld1.8 {q9}, [r0,:128]
126 vst1.8 {q2}, [r0,:128], r2
127 vst1.8 {q3}, [r0,:128], r2
132 .macro pixels16_xy2 rnd=1, avg=0
134 vld1.8 {d0-d2}, [r1], r2
135 vld1.8 {d4-d6}, [r1], r2
136 NRND vmov.i16 q13, #1
139 vext.8 q1, q0, q1, #1
140 vext.8 q3, q2, q3, #1
146 vld1.8 {d0-d2}, [r1], r2
149 NRND vadd.u16 q12, q12, q13
150 vext.8 q15, q0, q1, #1
151 vadd.u16 q1 , q10, q11
153 NRND vadd.u16 q1, q1, q13
156 vld1.8 {q8}, [r0,:128]
157 vrhadd.u8 q14, q14, q8
160 vld1.8 {d2-d4}, [r1], r2
161 vaddl.u8 q10, d1, d31
162 vst1.8 {q14}, [r0,:128], r2
165 NRND vadd.u16 q12, q12, q13
166 vext.8 q2, q1, q2, #1
167 vadd.u16 q0, q10, q11
169 NRND vadd.u16 q0, q0, q13
172 vld1.8 {q9}, [r0,:128]
173 vrhadd.u8 q15, q15, q9
177 vst1.8 {q15}, [r0,:128], r2
180 vld1.8 {d0-d2}, [r1], r2
182 NRND vadd.u16 q12, q12, q13
183 vext.8 q15, q0, q1, #1
184 vadd.u16 q1 , q10, q11
186 NRND vadd.u16 q1, q1, q13
189 vld1.8 {q8}, [r0,:128]
190 vrhadd.u8 q14, q14, q8
193 vaddl.u8 q10, d1, d31
194 vst1.8 {q14}, [r0,:128], r2
196 NRND vadd.u16 q12, q12, q13
197 vadd.u16 q0, q10, q11
199 NRND vadd.u16 q0, q0, q13
202 vld1.8 {q9}, [r0,:128]
203 vrhadd.u8 q15, q15, q9
205 vst1.8 {q15}, [r0,:128], r2
210 .macro pixels8 rnd=1, avg=0
211 1: vld1.8 {d0}, [r1], r2
212 vld1.8 {d1}, [r1], r2
213 vld1.8 {d2}, [r1], r2
215 vld1.8 {d3}, [r1], r2
220 vld1.8 {d4}, [r0,:64], r2
222 vld1.8 {d5}, [r0,:64], r2
224 vld1.8 {d6}, [r0,:64], r2
226 vld1.8 {d7}, [r0,:64], r2
228 sub r0, r0, r2, lsl #2
231 vst1.8 {d0}, [r0,:64], r2
232 vst1.8 {d1}, [r0,:64], r2
233 vst1.8 {d2}, [r0,:64], r2
234 vst1.8 {d3}, [r0,:64], r2
239 .macro pixels8_x2 rnd=1, avg=0
240 1: vld1.8 {q0}, [r1], r2
241 vext.8 d1, d0, d1, #1
242 vld1.8 {q1}, [r1], r2
243 vext.8 d3, d2, d3, #1
250 vld1.8 {d4}, [r0,:64], r2
251 vld1.8 {d5}, [r0,:64]
255 vst1.8 {d0}, [r0,:64], r2
256 vst1.8 {d1}, [r0,:64], r2
261 .macro pixels8_y2 rnd=1, avg=0
263 vld1.8 {d0}, [r1], r2
264 vld1.8 {d1}, [r1], r2
267 vld1.8 {d0}, [r1], r2
269 vld1.8 {d1}, [r1], r2
273 vld1.8 {d2}, [r0,:64], r2
274 vld1.8 {d3}, [r0,:64]
278 vst1.8 {d4}, [r0,:64], r2
279 vst1.8 {d5}, [r0,:64], r2
283 vld1.8 {d0}, [r1], r2
286 vld1.8 {d2}, [r0,:64], r2
287 vld1.8 {d3}, [r0,:64]
291 vst1.8 {d4}, [r0,:64], r2
292 vst1.8 {d5}, [r0,:64], r2
297 .macro pixels8_xy2 rnd=1, avg=0
299 vld1.8 {q0}, [r1], r2
300 vld1.8 {q1}, [r1], r2
301 NRND vmov.i16 q11, #1
304 vext.8 d4, d0, d1, #1
305 vext.8 d6, d2, d3, #1
309 vld1.8 {q0}, [r1], r2
312 vext.8 d4, d0, d1, #1
313 NRND vadd.u16 q10, q10, q11
316 vld1.8 {q1}, [r1], r2
320 vld1.8 {d7}, [r0,:64]
323 NRND vadd.u16 q10, q10, q11
324 vst1.8 {d5}, [r0,:64], r2
327 vld1.8 {d5}, [r0,:64]
330 vext.8 d6, d2, d3, #1
332 vst1.8 {d7}, [r0,:64], r2
335 vld1.8 {q0}, [r1], r2
337 vext.8 d4, d0, d1, #1
338 NRND vadd.u16 q10, q10, q11
343 vld1.8 {d7}, [r0,:64]
346 NRND vadd.u16 q10, q10, q11
347 vst1.8 {d5}, [r0,:64], r2
350 vld1.8 {d5}, [r0,:64]
353 vst1.8 {d7}, [r0,:64], r2
358 .macro pixfunc pfx, name, suf, rnd=1, avg=0
360 .macro avg rd, rn, rm
361 vrhadd.u8 \rd, \rn, \rm
363 .macro shrn rd, rn, rm
364 vrshrn.u16 \rd, \rn, \rm
366 .macro NRND insn:vararg
369 .macro avg rd, rn, rm
370 vhadd.u8 \rd, \rn, \rm
372 .macro shrn rd, rn, rm
373 vshrn.u16 \rd, \rn, \rm
375 .macro NRND insn:vararg
379 function ff_\pfx\name\suf\()_neon, export=1
387 .macro pixfunc2 pfx, name, avg=0
388 pixfunc \pfx, \name, rnd=1, avg=\avg
389 pixfunc \pfx, \name, _no_rnd, rnd=0, avg=\avg
392 function ff_put_h264_qpel16_mc00_neon, export=1
396 pixfunc put_, pixels16, avg=0
397 pixfunc2 put_, pixels16_x2, avg=0
398 pixfunc2 put_, pixels16_y2, avg=0
399 pixfunc2 put_, pixels16_xy2, avg=0
401 function ff_avg_h264_qpel16_mc00_neon, export=1
405 pixfunc avg_, pixels16, avg=1
406 pixfunc2 avg_, pixels16_x2, avg=1
407 pixfunc2 avg_, pixels16_y2, avg=1
408 pixfunc2 avg_, pixels16_xy2, avg=1
410 function ff_put_h264_qpel8_mc00_neon, export=1
414 pixfunc put_, pixels8, avg=0
415 pixfunc2 put_, pixels8_x2, avg=0
416 pixfunc2 put_, pixels8_y2, avg=0
417 pixfunc2 put_, pixels8_xy2, avg=0
419 function ff_avg_h264_qpel8_mc00_neon, export=1
423 pixfunc avg_, pixels8, avg=1
424 pixfunc2 avg_, pixels8_x2, avg=1
425 pixfunc2 avg_, pixels8_y2, avg=1
426 pixfunc2 avg_, pixels8_xy2, avg=1
428 function ff_put_pixels_clamped_neon, export=1
429 vld1.16 {d16-d19}, [r0,:128]!
431 vld1.16 {d20-d23}, [r0,:128]!
433 vld1.16 {d24-d27}, [r0,:128]!
435 vld1.16 {d28-d31}, [r0,:128]!
437 vst1.8 {d0}, [r1,:64], r2
439 vst1.8 {d1}, [r1,:64], r2
441 vst1.8 {d2}, [r1,:64], r2
443 vst1.8 {d3}, [r1,:64], r2
445 vst1.8 {d4}, [r1,:64], r2
446 vst1.8 {d5}, [r1,:64], r2
447 vst1.8 {d6}, [r1,:64], r2
448 vst1.8 {d7}, [r1,:64], r2
452 function ff_put_signed_pixels_clamped_neon, export=1
454 vld1.16 {d16-d17}, [r0,:128]!
456 vld1.16 {d18-d19}, [r0,:128]!
458 vld1.16 {d16-d17}, [r0,:128]!
460 vld1.16 {d18-d19}, [r0,:128]!
462 vld1.16 {d20-d21}, [r0,:128]!
464 vld1.16 {d22-d23}, [r0,:128]!
466 vst1.8 {d0}, [r1,:64], r2
468 vst1.8 {d1}, [r1,:64], r2
470 vst1.8 {d2}, [r1,:64], r2
472 vld1.16 {d24-d25}, [r0,:128]!
474 vld1.16 {d26-d27}, [r0,:128]!
477 vst1.8 {d3}, [r1,:64], r2
479 vst1.8 {d4}, [r1,:64], r2
481 vst1.8 {d5}, [r1,:64], r2
484 vst1.8 {d6}, [r1,:64], r2
485 vst1.8 {d7}, [r1,:64], r2
489 function ff_add_pixels_clamped_neon, export=1
491 vld1.8 {d16}, [r1,:64], r2
492 vld1.16 {d0-d1}, [r0,:128]!
494 vld1.8 {d17}, [r1,:64], r2
495 vld1.16 {d2-d3}, [r0,:128]!
497 vld1.8 {d18}, [r1,:64], r2
499 vld1.16 {d4-d5}, [r0,:128]!
501 vst1.8 {d0}, [r3,:64], r2
503 vld1.8 {d19}, [r1,:64], r2
504 vld1.16 {d6-d7}, [r0,:128]!
507 vst1.8 {d2}, [r3,:64], r2
508 vld1.8 {d16}, [r1,:64], r2
510 vld1.16 {d0-d1}, [r0,:128]!
512 vst1.8 {d4}, [r3,:64], r2
513 vld1.8 {d17}, [r1,:64], r2
514 vld1.16 {d2-d3}, [r0,:128]!
516 vst1.8 {d6}, [r3,:64], r2
518 vld1.8 {d18}, [r1,:64], r2
519 vld1.16 {d4-d5}, [r0,:128]!
521 vst1.8 {d0}, [r3,:64], r2
523 vld1.8 {d19}, [r1,:64], r2
525 vld1.16 {d6-d7}, [r0,:128]!
527 vst1.8 {d2}, [r3,:64], r2
529 vst1.8 {d4}, [r3,:64], r2
530 vst1.8 {d6}, [r3,:64], r2
534 function ff_butterflies_float_neon, export=1
535 1: vld1.32 {q0},[r0,:128]
536 vld1.32 {q1},[r1,:128]
539 vst1.32 {q2},[r1,:128]!
540 vst1.32 {q1},[r0,:128]!
546 function ff_scalarproduct_float_neon, export=1
548 1: vld1.32 {q0},[r0,:128]!
549 vld1.32 {q1},[r1,:128]!
555 NOVFP vmov.32 r0, d0[0]
559 function ff_vector_fmul_reverse_neon, export=1
560 add r2, r2, r3, lsl #2
563 vld1.32 {q0-q1}, [r1,:128]!
564 vld1.32 {q2-q3}, [r2,:128], r12
575 vld1.32 {q0-q1}, [r1,:128]!
576 vld1.32 {q2-q3}, [r2,:128], r12
577 vst1.32 {q8-q9}, [r0,:128]!
579 2: vst1.32 {q8-q9}, [r0,:128]!
583 function ff_vector_fmul_add_neon, export=1
585 vld1.32 {q0-q1}, [r1,:128]!
586 vld1.32 {q8-q9}, [r2,:128]!
587 vld1.32 {q2-q3}, [r3,:128]!
590 1: vadd.f32 q12, q2, q10
591 vadd.f32 q13, q3, q11
597 vld1.32 {q0}, [r1,:128]!
598 vld1.32 {q8}, [r2,:128]!
600 vld1.32 {q1}, [r1,:128]!
601 vld1.32 {q9}, [r2,:128]!
603 vld1.32 {q2-q3}, [r3,:128]!
604 vst1.32 {q12-q13},[r0,:128]!
606 2: vst1.32 {q12-q13},[r0,:128]!
610 function ff_vector_clipf_neon, export=1
611 VFP vdup.32 q1, d0[1]
612 VFP vdup.32 q0, d0[0]
616 vld1.f32 {q2},[r1,:128]!
618 vld1.f32 {q3},[r1,:128]!
620 1: vmax.f32 q8, q10, q0
624 vld1.f32 {q2},[r1,:128]!
626 vld1.f32 {q3},[r1,:128]!
628 vst1.f32 {q8},[r0,:128]!
629 vst1.f32 {q9},[r0,:128]!
631 2: vst1.f32 {q8},[r0,:128]!
632 vst1.f32 {q9},[r0,:128]!
636 function ff_apply_window_int16_neon, export=1
638 add r4, r1, r3, lsl #1
639 add lr, r0, r3, lsl #1
644 vld1.16 {q0}, [r1,:128]!
645 vld1.16 {q2}, [r2,:128]!
646 vld1.16 {q1}, [r4,:128], r12
648 vqrdmulh.s16 q0, q0, q2
649 vqrdmulh.s16 d2, d2, d7
650 vqrdmulh.s16 d3, d3, d6
651 vst1.16 {q0}, [r0,:128]!
652 vst1.16 {q1}, [lr,:128], r12
659 function ff_vector_clip_int32_neon, export=1
664 vld1.32 {q2-q3}, [r1,:128]!
669 vst1.32 {q2-q3}, [r0,:128]!