1 /*
2 * Copyright (c) 2012 The LibYuv project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #include "libyuv/row.h"
12
13 #ifdef __cplusplus
14 namespace libyuv {
15 extern "C" {
16 #endif
17
18 // The following are available on Mips platforms:
19 #if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips__) && \
20 (_MIPS_SIM == _MIPS_SIM_ABI32)
21
22 #ifdef HAS_COPYROW_MIPS
CopyRow_MIPS(const uint8 * src,uint8 * dst,int count)23 void CopyRow_MIPS(const uint8* src, uint8* dst, int count) {
24 __asm__ __volatile__ (
25 ".set noreorder \n"
26 ".set noat \n"
27 "slti $at, %[count], 8 \n"
28 "bne $at ,$zero, $last8 \n"
29 "xor $t8, %[src], %[dst] \n"
30 "andi $t8, $t8, 0x3 \n"
31
32 "bne $t8, $zero, unaligned \n"
33 "negu $a3, %[dst] \n"
34 // make dst/src aligned
35 "andi $a3, $a3, 0x3 \n"
36 "beq $a3, $zero, $chk16w \n"
37 // word-aligned now count is the remining bytes count
38 "subu %[count], %[count], $a3 \n"
39
40 "lwr $t8, 0(%[src]) \n"
41 "addu %[src], %[src], $a3 \n"
42 "swr $t8, 0(%[dst]) \n"
43 "addu %[dst], %[dst], $a3 \n"
44
45 // Now the dst/src are mutually word-aligned with word-aligned addresses
46 "$chk16w: \n"
47 "andi $t8, %[count], 0x3f \n" // whole 64-B chunks?
48 // t8 is the byte count after 64-byte chunks
49 "beq %[count], $t8, chk8w \n"
50 // There will be at most 1 32-byte chunk after it
51 "subu $a3, %[count], $t8 \n" // the reminder
52 // Here a3 counts bytes in 16w chunks
53 "addu $a3, %[dst], $a3 \n"
54 // Now a3 is the final dst after 64-byte chunks
55 "addu $t0, %[dst], %[count] \n"
56 // t0 is the "past the end" address
57
58 // When in the loop we exercise "pref 30,x(a1)", the a1+x should not be past
59 // the "t0-32" address
60 // This means: for x=128 the last "safe" a1 address is "t0-160"
61 // Alternatively, for x=64 the last "safe" a1 address is "t0-96"
62 // we will use "pref 30,128(a1)", so "t0-160" is the limit
63 "subu $t9, $t0, 160 \n"
64 // t9 is the "last safe pref 30,128(a1)" address
65 "pref 0, 0(%[src]) \n" // first line of src
66 "pref 0, 32(%[src]) \n" // second line of src
67 "pref 0, 64(%[src]) \n"
68 "pref 30, 32(%[dst]) \n"
69 // In case the a1 > t9 don't use "pref 30" at all
70 "sgtu $v1, %[dst], $t9 \n"
71 "bgtz $v1, $loop16w \n"
72 "nop \n"
73 // otherwise, start with using pref30
74 "pref 30, 64(%[dst]) \n"
75 "$loop16w: \n"
76 "pref 0, 96(%[src]) \n"
77 "lw $t0, 0(%[src]) \n"
78 "bgtz $v1, $skip_pref30_96 \n" // skip
79 "lw $t1, 4(%[src]) \n"
80 "pref 30, 96(%[dst]) \n" // continue
81 "$skip_pref30_96: \n"
82 "lw $t2, 8(%[src]) \n"
83 "lw $t3, 12(%[src]) \n"
84 "lw $t4, 16(%[src]) \n"
85 "lw $t5, 20(%[src]) \n"
86 "lw $t6, 24(%[src]) \n"
87 "lw $t7, 28(%[src]) \n"
88 "pref 0, 128(%[src]) \n"
89 // bring the next lines of src, addr 128
90 "sw $t0, 0(%[dst]) \n"
91 "sw $t1, 4(%[dst]) \n"
92 "sw $t2, 8(%[dst]) \n"
93 "sw $t3, 12(%[dst]) \n"
94 "sw $t4, 16(%[dst]) \n"
95 "sw $t5, 20(%[dst]) \n"
96 "sw $t6, 24(%[dst]) \n"
97 "sw $t7, 28(%[dst]) \n"
98 "lw $t0, 32(%[src]) \n"
99 "bgtz $v1, $skip_pref30_128 \n" // skip pref 30,128(a1)
100 "lw $t1, 36(%[src]) \n"
101 "pref 30, 128(%[dst]) \n" // set dest, addr 128
102 "$skip_pref30_128: \n"
103 "lw $t2, 40(%[src]) \n"
104 "lw $t3, 44(%[src]) \n"
105 "lw $t4, 48(%[src]) \n"
106 "lw $t5, 52(%[src]) \n"
107 "lw $t6, 56(%[src]) \n"
108 "lw $t7, 60(%[src]) \n"
109 "pref 0, 160(%[src]) \n"
110 // bring the next lines of src, addr 160
111 "sw $t0, 32(%[dst]) \n"
112 "sw $t1, 36(%[dst]) \n"
113 "sw $t2, 40(%[dst]) \n"
114 "sw $t3, 44(%[dst]) \n"
115 "sw $t4, 48(%[dst]) \n"
116 "sw $t5, 52(%[dst]) \n"
117 "sw $t6, 56(%[dst]) \n"
118 "sw $t7, 60(%[dst]) \n"
119
120 "addiu %[dst], %[dst], 64 \n" // adding 64 to dest
121 "sgtu $v1, %[dst], $t9 \n"
122 "bne %[dst], $a3, $loop16w \n"
123 " addiu %[src], %[src], 64 \n" // adding 64 to src
124 "move %[count], $t8 \n"
125
126 // Here we have src and dest word-aligned but less than 64-bytes to go
127
128 "chk8w: \n"
129 "pref 0, 0x0(%[src]) \n"
130 "andi $t8, %[count], 0x1f \n" // 32-byte chunk?
131 // the t8 is the reminder count past 32-bytes
132 "beq %[count], $t8, chk1w \n"
133 // count=t8,no 32-byte chunk
134 " nop \n"
135
136 "lw $t0, 0(%[src]) \n"
137 "lw $t1, 4(%[src]) \n"
138 "lw $t2, 8(%[src]) \n"
139 "lw $t3, 12(%[src]) \n"
140 "lw $t4, 16(%[src]) \n"
141 "lw $t5, 20(%[src]) \n"
142 "lw $t6, 24(%[src]) \n"
143 "lw $t7, 28(%[src]) \n"
144 "addiu %[src], %[src], 32 \n"
145
146 "sw $t0, 0(%[dst]) \n"
147 "sw $t1, 4(%[dst]) \n"
148 "sw $t2, 8(%[dst]) \n"
149 "sw $t3, 12(%[dst]) \n"
150 "sw $t4, 16(%[dst]) \n"
151 "sw $t5, 20(%[dst]) \n"
152 "sw $t6, 24(%[dst]) \n"
153 "sw $t7, 28(%[dst]) \n"
154 "addiu %[dst], %[dst], 32 \n"
155
156 "chk1w: \n"
157 "andi %[count], $t8, 0x3 \n"
158 // now count is the reminder past 1w chunks
159 "beq %[count], $t8, $last8 \n"
160 " subu $a3, $t8, %[count] \n"
161 // a3 is count of bytes in 1w chunks
162 "addu $a3, %[dst], $a3 \n"
163 // now a3 is the dst address past the 1w chunks
164 // copying in words (4-byte chunks)
165 "$wordCopy_loop: \n"
166 "lw $t3, 0(%[src]) \n"
167 // the first t3 may be equal t0 ... optimize?
168 "addiu %[src], %[src],4 \n"
169 "addiu %[dst], %[dst],4 \n"
170 "bne %[dst], $a3,$wordCopy_loop \n"
171 " sw $t3, -4(%[dst]) \n"
172
173 // For the last (<8) bytes
174 "$last8: \n"
175 "blez %[count], leave \n"
176 " addu $a3, %[dst], %[count] \n" // a3 -last dst address
177 "$last8loop: \n"
178 "lb $v1, 0(%[src]) \n"
179 "addiu %[src], %[src], 1 \n"
180 "addiu %[dst], %[dst], 1 \n"
181 "bne %[dst], $a3, $last8loop \n"
182 " sb $v1, -1(%[dst]) \n"
183
184 "leave: \n"
185 " j $ra \n"
186 " nop \n"
187
188 //
189 // UNALIGNED case
190 //
191
192 "unaligned: \n"
193 // got here with a3="negu a1"
194 "andi $a3, $a3, 0x3 \n" // a1 is word aligned?
195 "beqz $a3, $ua_chk16w \n"
196 " subu %[count], %[count], $a3 \n"
197 // bytes left after initial a3 bytes
198 "lwr $v1, 0(%[src]) \n"
199 "lwl $v1, 3(%[src]) \n"
200 "addu %[src], %[src], $a3 \n" // a3 may be 1, 2 or 3
201 "swr $v1, 0(%[dst]) \n"
202 "addu %[dst], %[dst], $a3 \n"
203 // below the dst will be word aligned (NOTE1)
204 "$ua_chk16w: \n"
205 "andi $t8, %[count], 0x3f \n" // whole 64-B chunks?
206 // t8 is the byte count after 64-byte chunks
207 "beq %[count], $t8, ua_chk8w \n"
208 // if a2==t8, no 64-byte chunks
209 // There will be at most 1 32-byte chunk after it
210 "subu $a3, %[count], $t8 \n" // the reminder
211 // Here a3 counts bytes in 16w chunks
212 "addu $a3, %[dst], $a3 \n"
213 // Now a3 is the final dst after 64-byte chunks
214 "addu $t0, %[dst], %[count] \n" // t0 "past the end"
215 "subu $t9, $t0, 160 \n"
216 // t9 is the "last safe pref 30,128(a1)" address
217 "pref 0, 0(%[src]) \n" // first line of src
218 "pref 0, 32(%[src]) \n" // second line addr 32
219 "pref 0, 64(%[src]) \n"
220 "pref 30, 32(%[dst]) \n"
221 // safe, as we have at least 64 bytes ahead
222 // In case the a1 > t9 don't use "pref 30" at all
223 "sgtu $v1, %[dst], $t9 \n"
224 "bgtz $v1, $ua_loop16w \n"
225 // skip "pref 30,64(a1)" for too short arrays
226 " nop \n"
227 // otherwise, start with using pref30
228 "pref 30, 64(%[dst]) \n"
229 "$ua_loop16w: \n"
230 "pref 0, 96(%[src]) \n"
231 "lwr $t0, 0(%[src]) \n"
232 "lwl $t0, 3(%[src]) \n"
233 "lwr $t1, 4(%[src]) \n"
234 "bgtz $v1, $ua_skip_pref30_96 \n"
235 " lwl $t1, 7(%[src]) \n"
236 "pref 30, 96(%[dst]) \n"
237 // continue setting up the dest, addr 96
238 "$ua_skip_pref30_96: \n"
239 "lwr $t2, 8(%[src]) \n"
240 "lwl $t2, 11(%[src]) \n"
241 "lwr $t3, 12(%[src]) \n"
242 "lwl $t3, 15(%[src]) \n"
243 "lwr $t4, 16(%[src]) \n"
244 "lwl $t4, 19(%[src]) \n"
245 "lwr $t5, 20(%[src]) \n"
246 "lwl $t5, 23(%[src]) \n"
247 "lwr $t6, 24(%[src]) \n"
248 "lwl $t6, 27(%[src]) \n"
249 "lwr $t7, 28(%[src]) \n"
250 "lwl $t7, 31(%[src]) \n"
251 "pref 0, 128(%[src]) \n"
252 // bring the next lines of src, addr 128
253 "sw $t0, 0(%[dst]) \n"
254 "sw $t1, 4(%[dst]) \n"
255 "sw $t2, 8(%[dst]) \n"
256 "sw $t3, 12(%[dst]) \n"
257 "sw $t4, 16(%[dst]) \n"
258 "sw $t5, 20(%[dst]) \n"
259 "sw $t6, 24(%[dst]) \n"
260 "sw $t7, 28(%[dst]) \n"
261 "lwr $t0, 32(%[src]) \n"
262 "lwl $t0, 35(%[src]) \n"
263 "lwr $t1, 36(%[src]) \n"
264 "bgtz $v1, ua_skip_pref30_128 \n"
265 " lwl $t1, 39(%[src]) \n"
266 "pref 30, 128(%[dst]) \n"
267 // continue setting up the dest, addr 128
268 "ua_skip_pref30_128: \n"
269
270 "lwr $t2, 40(%[src]) \n"
271 "lwl $t2, 43(%[src]) \n"
272 "lwr $t3, 44(%[src]) \n"
273 "lwl $t3, 47(%[src]) \n"
274 "lwr $t4, 48(%[src]) \n"
275 "lwl $t4, 51(%[src]) \n"
276 "lwr $t5, 52(%[src]) \n"
277 "lwl $t5, 55(%[src]) \n"
278 "lwr $t6, 56(%[src]) \n"
279 "lwl $t6, 59(%[src]) \n"
280 "lwr $t7, 60(%[src]) \n"
281 "lwl $t7, 63(%[src]) \n"
282 "pref 0, 160(%[src]) \n"
283 // bring the next lines of src, addr 160
284 "sw $t0, 32(%[dst]) \n"
285 "sw $t1, 36(%[dst]) \n"
286 "sw $t2, 40(%[dst]) \n"
287 "sw $t3, 44(%[dst]) \n"
288 "sw $t4, 48(%[dst]) \n"
289 "sw $t5, 52(%[dst]) \n"
290 "sw $t6, 56(%[dst]) \n"
291 "sw $t7, 60(%[dst]) \n"
292
293 "addiu %[dst],%[dst],64 \n" // adding 64 to dest
294 "sgtu $v1,%[dst],$t9 \n"
295 "bne %[dst],$a3,$ua_loop16w \n"
296 " addiu %[src],%[src],64 \n" // adding 64 to src
297 "move %[count],$t8 \n"
298
299 // Here we have src and dest word-aligned but less than 64-bytes to go
300
301 "ua_chk8w: \n"
302 "pref 0, 0x0(%[src]) \n"
303 "andi $t8, %[count], 0x1f \n" // 32-byte chunk?
304 // the t8 is the reminder count
305 "beq %[count], $t8, $ua_chk1w \n"
306 // when count==t8, no 32-byte chunk
307
308 "lwr $t0, 0(%[src]) \n"
309 "lwl $t0, 3(%[src]) \n"
310 "lwr $t1, 4(%[src]) \n"
311 "lwl $t1, 7(%[src]) \n"
312 "lwr $t2, 8(%[src]) \n"
313 "lwl $t2, 11(%[src]) \n"
314 "lwr $t3, 12(%[src]) \n"
315 "lwl $t3, 15(%[src]) \n"
316 "lwr $t4, 16(%[src]) \n"
317 "lwl $t4, 19(%[src]) \n"
318 "lwr $t5, 20(%[src]) \n"
319 "lwl $t5, 23(%[src]) \n"
320 "lwr $t6, 24(%[src]) \n"
321 "lwl $t6, 27(%[src]) \n"
322 "lwr $t7, 28(%[src]) \n"
323 "lwl $t7, 31(%[src]) \n"
324 "addiu %[src], %[src], 32 \n"
325
326 "sw $t0, 0(%[dst]) \n"
327 "sw $t1, 4(%[dst]) \n"
328 "sw $t2, 8(%[dst]) \n"
329 "sw $t3, 12(%[dst]) \n"
330 "sw $t4, 16(%[dst]) \n"
331 "sw $t5, 20(%[dst]) \n"
332 "sw $t6, 24(%[dst]) \n"
333 "sw $t7, 28(%[dst]) \n"
334 "addiu %[dst], %[dst], 32 \n"
335
336 "$ua_chk1w: \n"
337 "andi %[count], $t8, 0x3 \n"
338 // now count is the reminder past 1w chunks
339 "beq %[count], $t8, ua_smallCopy \n"
340 "subu $a3, $t8, %[count] \n"
341 // a3 is count of bytes in 1w chunks
342 "addu $a3, %[dst], $a3 \n"
343 // now a3 is the dst address past the 1w chunks
344
345 // copying in words (4-byte chunks)
346 "$ua_wordCopy_loop: \n"
347 "lwr $v1, 0(%[src]) \n"
348 "lwl $v1, 3(%[src]) \n"
349 "addiu %[src], %[src], 4 \n"
350 "addiu %[dst], %[dst], 4 \n"
351 // note: dst=a1 is word aligned here, see NOTE1
352 "bne %[dst], $a3, $ua_wordCopy_loop \n"
353 " sw $v1,-4(%[dst]) \n"
354
355 // Now less than 4 bytes (value in count) left to copy
356 "ua_smallCopy: \n"
357 "beqz %[count], leave \n"
358 " addu $a3, %[dst], %[count] \n" // a3 = last dst address
359 "$ua_smallCopy_loop: \n"
360 "lb $v1, 0(%[src]) \n"
361 "addiu %[src], %[src], 1 \n"
362 "addiu %[dst], %[dst], 1 \n"
363 "bne %[dst],$a3,$ua_smallCopy_loop \n"
364 " sb $v1, -1(%[dst]) \n"
365
366 "j $ra \n"
367 " nop \n"
368 ".set at \n"
369 ".set reorder \n"
370 : [dst] "+r" (dst), [src] "+r" (src)
371 : [count] "r" (count)
372 : "t0", "t1", "t2", "t3", "t4", "t5", "t6", "t7",
373 "t8", "t9", "a3", "v1", "at"
374 );
375 }
376 #endif // HAS_COPYROW_MIPS
377
378 // MIPS DSPR2 functions
379 #if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips_dsp) && \
380 (__mips_dsp_rev >= 2) && \
381 (_MIPS_SIM == _MIPS_SIM_ABI32) && (__mips_isa_rev < 6)
382
SplitUVRow_MIPS_DSPR2(const uint8 * src_uv,uint8 * dst_u,uint8 * dst_v,int width)383 void SplitUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
384 int width) {
385 __asm__ __volatile__ (
386 ".set push \n"
387 ".set noreorder \n"
388 "srl $t4, %[width], 4 \n" // multiplies of 16
389 "blez $t4, 2f \n"
390 " andi %[width], %[width], 0xf \n" // residual
391
392 ".p2align 2 \n"
393 "1: \n"
394 "addiu $t4, $t4, -1 \n"
395 "lw $t0, 0(%[src_uv]) \n" // V1 | U1 | V0 | U0
396 "lw $t1, 4(%[src_uv]) \n" // V3 | U3 | V2 | U2
397 "lw $t2, 8(%[src_uv]) \n" // V5 | U5 | V4 | U4
398 "lw $t3, 12(%[src_uv]) \n" // V7 | U7 | V6 | U6
399 "lw $t5, 16(%[src_uv]) \n" // V9 | U9 | V8 | U8
400 "lw $t6, 20(%[src_uv]) \n" // V11 | U11 | V10 | U10
401 "lw $t7, 24(%[src_uv]) \n" // V13 | U13 | V12 | U12
402 "lw $t8, 28(%[src_uv]) \n" // V15 | U15 | V14 | U14
403 "addiu %[src_uv], %[src_uv], 32 \n"
404 "precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0
405 "precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0
406 "precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4
407 "precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4
408 "precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8
409 "precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8
410 "precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12
411 "precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12
412 "sw $t9, 0(%[dst_v]) \n"
413 "sw $t0, 0(%[dst_u]) \n"
414 "sw $t1, 4(%[dst_v]) \n"
415 "sw $t2, 4(%[dst_u]) \n"
416 "sw $t3, 8(%[dst_v]) \n"
417 "sw $t5, 8(%[dst_u]) \n"
418 "sw $t6, 12(%[dst_v]) \n"
419 "sw $t7, 12(%[dst_u]) \n"
420 "addiu %[dst_v], %[dst_v], 16 \n"
421 "bgtz $t4, 1b \n"
422 " addiu %[dst_u], %[dst_u], 16 \n"
423
424 "beqz %[width], 3f \n"
425 " nop \n"
426
427 "2: \n"
428 "lbu $t0, 0(%[src_uv]) \n"
429 "lbu $t1, 1(%[src_uv]) \n"
430 "addiu %[src_uv], %[src_uv], 2 \n"
431 "addiu %[width], %[width], -1 \n"
432 "sb $t0, 0(%[dst_u]) \n"
433 "sb $t1, 0(%[dst_v]) \n"
434 "addiu %[dst_u], %[dst_u], 1 \n"
435 "bgtz %[width], 2b \n"
436 " addiu %[dst_v], %[dst_v], 1 \n"
437
438 "3: \n"
439 ".set pop \n"
440 : [src_uv] "+r" (src_uv),
441 [width] "+r" (width),
442 [dst_u] "+r" (dst_u),
443 [dst_v] "+r" (dst_v)
444 :
445 : "t0", "t1", "t2", "t3",
446 "t4", "t5", "t6", "t7", "t8", "t9"
447 );
448 }
449
MirrorRow_MIPS_DSPR2(const uint8 * src,uint8 * dst,int width)450 void MirrorRow_MIPS_DSPR2(const uint8* src, uint8* dst, int width) {
451 __asm__ __volatile__ (
452 ".set push \n"
453 ".set noreorder \n"
454
455 "srl $t4, %[width], 4 \n" // multiplies of 16
456 "andi $t5, %[width], 0xf \n"
457 "blez $t4, 2f \n"
458 " addu %[src], %[src], %[width] \n" // src += width
459
460 ".p2align 2 \n"
461 "1: \n"
462 "lw $t0, -16(%[src]) \n" // |3|2|1|0|
463 "lw $t1, -12(%[src]) \n" // |7|6|5|4|
464 "lw $t2, -8(%[src]) \n" // |11|10|9|8|
465 "lw $t3, -4(%[src]) \n" // |15|14|13|12|
466 "wsbh $t0, $t0 \n" // |2|3|0|1|
467 "wsbh $t1, $t1 \n" // |6|7|4|5|
468 "wsbh $t2, $t2 \n" // |10|11|8|9|
469 "wsbh $t3, $t3 \n" // |14|15|12|13|
470 "rotr $t0, $t0, 16 \n" // |0|1|2|3|
471 "rotr $t1, $t1, 16 \n" // |4|5|6|7|
472 "rotr $t2, $t2, 16 \n" // |8|9|10|11|
473 "rotr $t3, $t3, 16 \n" // |12|13|14|15|
474 "addiu %[src], %[src], -16 \n"
475 "addiu $t4, $t4, -1 \n"
476 "sw $t3, 0(%[dst]) \n" // |15|14|13|12|
477 "sw $t2, 4(%[dst]) \n" // |11|10|9|8|
478 "sw $t1, 8(%[dst]) \n" // |7|6|5|4|
479 "sw $t0, 12(%[dst]) \n" // |3|2|1|0|
480 "bgtz $t4, 1b \n"
481 " addiu %[dst], %[dst], 16 \n"
482 "beqz $t5, 3f \n"
483 " nop \n"
484
485 "2: \n"
486 "lbu $t0, -1(%[src]) \n"
487 "addiu $t5, $t5, -1 \n"
488 "addiu %[src], %[src], -1 \n"
489 "sb $t0, 0(%[dst]) \n"
490 "bgez $t5, 2b \n"
491 " addiu %[dst], %[dst], 1 \n"
492
493 "3: \n"
494 ".set pop \n"
495 : [src] "+r" (src), [dst] "+r" (dst)
496 : [width] "r" (width)
497 : "t0", "t1", "t2", "t3", "t4", "t5"
498 );
499 }
500
MirrorUVRow_MIPS_DSPR2(const uint8 * src_uv,uint8 * dst_u,uint8 * dst_v,int width)501 void MirrorUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
502 int width) {
503 int x = 0;
504 int y = 0;
505 __asm__ __volatile__ (
506 ".set push \n"
507 ".set noreorder \n"
508
509 "addu $t4, %[width], %[width] \n"
510 "srl %[x], %[width], 4 \n"
511 "andi %[y], %[width], 0xf \n"
512 "blez %[x], 2f \n"
513 " addu %[src_uv], %[src_uv], $t4 \n"
514
515 ".p2align 2 \n"
516 "1: \n"
517 "lw $t0, -32(%[src_uv]) \n" // |3|2|1|0|
518 "lw $t1, -28(%[src_uv]) \n" // |7|6|5|4|
519 "lw $t2, -24(%[src_uv]) \n" // |11|10|9|8|
520 "lw $t3, -20(%[src_uv]) \n" // |15|14|13|12|
521 "lw $t4, -16(%[src_uv]) \n" // |19|18|17|16|
522 "lw $t6, -12(%[src_uv]) \n" // |23|22|21|20|
523 "lw $t7, -8(%[src_uv]) \n" // |27|26|25|24|
524 "lw $t8, -4(%[src_uv]) \n" // |31|30|29|28|
525
526 "rotr $t0, $t0, 16 \n" // |1|0|3|2|
527 "rotr $t1, $t1, 16 \n" // |5|4|7|6|
528 "rotr $t2, $t2, 16 \n" // |9|8|11|10|
529 "rotr $t3, $t3, 16 \n" // |13|12|15|14|
530 "rotr $t4, $t4, 16 \n" // |17|16|19|18|
531 "rotr $t6, $t6, 16 \n" // |21|20|23|22|
532 "rotr $t7, $t7, 16 \n" // |25|24|27|26|
533 "rotr $t8, $t8, 16 \n" // |29|28|31|30|
534 "precr.qb.ph $t9, $t0, $t1 \n" // |0|2|4|6|
535 "precrq.qb.ph $t5, $t0, $t1 \n" // |1|3|5|7|
536 "precr.qb.ph $t0, $t2, $t3 \n" // |8|10|12|14|
537 "precrq.qb.ph $t1, $t2, $t3 \n" // |9|11|13|15|
538 "precr.qb.ph $t2, $t4, $t6 \n" // |16|18|20|22|
539 "precrq.qb.ph $t3, $t4, $t6 \n" // |17|19|21|23|
540 "precr.qb.ph $t4, $t7, $t8 \n" // |24|26|28|30|
541 "precrq.qb.ph $t6, $t7, $t8 \n" // |25|27|29|31|
542 "addiu %[src_uv], %[src_uv], -32 \n"
543 "addiu %[x], %[x], -1 \n"
544 "swr $t4, 0(%[dst_u]) \n"
545 "swl $t4, 3(%[dst_u]) \n" // |30|28|26|24|
546 "swr $t6, 0(%[dst_v]) \n"
547 "swl $t6, 3(%[dst_v]) \n" // |31|29|27|25|
548 "swr $t2, 4(%[dst_u]) \n"
549 "swl $t2, 7(%[dst_u]) \n" // |22|20|18|16|
550 "swr $t3, 4(%[dst_v]) \n"
551 "swl $t3, 7(%[dst_v]) \n" // |23|21|19|17|
552 "swr $t0, 8(%[dst_u]) \n"
553 "swl $t0, 11(%[dst_u]) \n" // |14|12|10|8|
554 "swr $t1, 8(%[dst_v]) \n"
555 "swl $t1, 11(%[dst_v]) \n" // |15|13|11|9|
556 "swr $t9, 12(%[dst_u]) \n"
557 "swl $t9, 15(%[dst_u]) \n" // |6|4|2|0|
558 "swr $t5, 12(%[dst_v]) \n"
559 "swl $t5, 15(%[dst_v]) \n" // |7|5|3|1|
560 "addiu %[dst_v], %[dst_v], 16 \n"
561 "bgtz %[x], 1b \n"
562 " addiu %[dst_u], %[dst_u], 16 \n"
563 "beqz %[y], 3f \n"
564 " nop \n"
565 "b 2f \n"
566 " nop \n"
567
568 "2: \n"
569 "lbu $t0, -2(%[src_uv]) \n"
570 "lbu $t1, -1(%[src_uv]) \n"
571 "addiu %[src_uv], %[src_uv], -2 \n"
572 "addiu %[y], %[y], -1 \n"
573 "sb $t0, 0(%[dst_u]) \n"
574 "sb $t1, 0(%[dst_v]) \n"
575 "addiu %[dst_u], %[dst_u], 1 \n"
576 "bgtz %[y], 2b \n"
577 " addiu %[dst_v], %[dst_v], 1 \n"
578
579 "3: \n"
580 ".set pop \n"
581 : [src_uv] "+r" (src_uv),
582 [dst_u] "+r" (dst_u),
583 [dst_v] "+r" (dst_v),
584 [x] "=&r" (x),
585 [y] "+r" (y)
586 : [width] "r" (width)
587 : "t0", "t1", "t2", "t3", "t4",
588 "t5", "t7", "t8", "t9"
589 );
590 }
591
592 // Convert (4 Y and 2 VU) I422 and arrange RGB values into
593 // t5 = | 0 | B0 | 0 | b0 |
594 // t4 = | 0 | B1 | 0 | b1 |
595 // t9 = | 0 | G0 | 0 | g0 |
596 // t8 = | 0 | G1 | 0 | g1 |
597 // t2 = | 0 | R0 | 0 | r0 |
598 // t1 = | 0 | R1 | 0 | r1 |
599 #define I422ToTransientMipsRGB \
600 "lw $t0, 0(%[y_buf]) \n" \
601 "lhu $t1, 0(%[u_buf]) \n" \
602 "lhu $t2, 0(%[v_buf]) \n" \
603 "preceu.ph.qbr $t1, $t1 \n" \
604 "preceu.ph.qbr $t2, $t2 \n" \
605 "preceu.ph.qbra $t3, $t0 \n" \
606 "preceu.ph.qbla $t0, $t0 \n" \
607 "subu.ph $t1, $t1, $s5 \n" \
608 "subu.ph $t2, $t2, $s5 \n" \
609 "subu.ph $t3, $t3, $s4 \n" \
610 "subu.ph $t0, $t0, $s4 \n" \
611 "mul.ph $t3, $t3, $s0 \n" \
612 "mul.ph $t0, $t0, $s0 \n" \
613 "shll.ph $t4, $t1, 0x7 \n" \
614 "subu.ph $t4, $t4, $t1 \n" \
615 "mul.ph $t6, $t1, $s1 \n" \
616 "mul.ph $t1, $t2, $s2 \n" \
617 "addq_s.ph $t5, $t4, $t3 \n" \
618 "addq_s.ph $t4, $t4, $t0 \n" \
619 "shra.ph $t5, $t5, 6 \n" \
620 "shra.ph $t4, $t4, 6 \n" \
621 "addiu %[u_buf], 2 \n" \
622 "addiu %[v_buf], 2 \n" \
623 "addu.ph $t6, $t6, $t1 \n" \
624 "mul.ph $t1, $t2, $s3 \n" \
625 "addu.ph $t9, $t6, $t3 \n" \
626 "addu.ph $t8, $t6, $t0 \n" \
627 "shra.ph $t9, $t9, 6 \n" \
628 "shra.ph $t8, $t8, 6 \n" \
629 "addu.ph $t2, $t1, $t3 \n" \
630 "addu.ph $t1, $t1, $t0 \n" \
631 "shra.ph $t2, $t2, 6 \n" \
632 "shra.ph $t1, $t1, 6 \n" \
633 "subu.ph $t5, $t5, $s5 \n" \
634 "subu.ph $t4, $t4, $s5 \n" \
635 "subu.ph $t9, $t9, $s5 \n" \
636 "subu.ph $t8, $t8, $s5 \n" \
637 "subu.ph $t2, $t2, $s5 \n" \
638 "subu.ph $t1, $t1, $s5 \n" \
639 "shll_s.ph $t5, $t5, 8 \n" \
640 "shll_s.ph $t4, $t4, 8 \n" \
641 "shll_s.ph $t9, $t9, 8 \n" \
642 "shll_s.ph $t8, $t8, 8 \n" \
643 "shll_s.ph $t2, $t2, 8 \n" \
644 "shll_s.ph $t1, $t1, 8 \n" \
645 "shra.ph $t5, $t5, 8 \n" \
646 "shra.ph $t4, $t4, 8 \n" \
647 "shra.ph $t9, $t9, 8 \n" \
648 "shra.ph $t8, $t8, 8 \n" \
649 "shra.ph $t2, $t2, 8 \n" \
650 "shra.ph $t1, $t1, 8 \n" \
651 "addu.ph $t5, $t5, $s5 \n" \
652 "addu.ph $t4, $t4, $s5 \n" \
653 "addu.ph $t9, $t9, $s5 \n" \
654 "addu.ph $t8, $t8, $s5 \n" \
655 "addu.ph $t2, $t2, $s5 \n" \
656 "addu.ph $t1, $t1, $s5 \n"
657
I422ToARGBRow_MIPS_DSPR2(const uint8 * y_buf,const uint8 * u_buf,const uint8 * v_buf,uint8 * rgb_buf,int width)658 void I422ToARGBRow_MIPS_DSPR2(const uint8* y_buf,
659 const uint8* u_buf,
660 const uint8* v_buf,
661 uint8* rgb_buf,
662 int width) {
663 __asm__ __volatile__ (
664 ".set push \n"
665 ".set noreorder \n"
666 "beqz %[width], 2f \n"
667 " repl.ph $s0, 74 \n" // |YG|YG| = |74|74|
668 "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
669 "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
670 "repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
671 "repl.ph $s4, 16 \n" // |0|16|0|16|
672 "repl.ph $s5, 128 \n" // |128|128| // clipping
673 "lui $s6, 0xff00 \n"
674 "ori $s6, 0xff00 \n" // |ff|00|ff|00|ff|
675
676 ".p2align 2 \n"
677 "1: \n"
678 I422ToTransientMipsRGB
679 // Arranging into argb format
680 "precr.qb.ph $t4, $t8, $t4 \n" // |G1|g1|B1|b1|
681 "precr.qb.ph $t5, $t9, $t5 \n" // |G0|g0|B0|b0|
682 "addiu %[width], -4 \n"
683 "precrq.qb.ph $t8, $t4, $t5 \n" // |G1|B1|G0|B0|
684 "precr.qb.ph $t9, $t4, $t5 \n" // |g1|b1|g0|b0|
685 "precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0|
686
687 "addiu %[y_buf], 4 \n"
688 "preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0|
689 "preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0|
690 "or $t1, $t1, $s6 \n" // |ff|R1|ff|R0|
691 "or $t2, $t2, $s6 \n" // |ff|r1|ff|r0|
692 "precrq.ph.w $t0, $t2, $t9 \n" // |ff|r1|g1|b1|
693 "precrq.ph.w $t3, $t1, $t8 \n" // |ff|R1|G1|B1|
694 "sll $t9, $t9, 16 \n"
695 "sll $t8, $t8, 16 \n"
696 "packrl.ph $t2, $t2, $t9 \n" // |ff|r0|g0|b0|
697 "packrl.ph $t1, $t1, $t8 \n" // |ff|R0|G0|B0|
698 // Store results.
699 "sw $t2, 0(%[rgb_buf]) \n"
700 "sw $t0, 4(%[rgb_buf]) \n"
701 "sw $t1, 8(%[rgb_buf]) \n"
702 "sw $t3, 12(%[rgb_buf]) \n"
703 "bnez %[width], 1b \n"
704 " addiu %[rgb_buf], 16 \n"
705 "2: \n"
706 ".set pop \n"
707 :[y_buf] "+r" (y_buf),
708 [u_buf] "+r" (u_buf),
709 [v_buf] "+r" (v_buf),
710 [width] "+r" (width),
711 [rgb_buf] "+r" (rgb_buf)
712 :
713 : "t0", "t1", "t2", "t3", "t4", "t5",
714 "t6", "t7", "t8", "t9",
715 "s0", "s1", "s2", "s3",
716 "s4", "s5", "s6"
717 );
718 }
719
I422ToABGRRow_MIPS_DSPR2(const uint8 * y_buf,const uint8 * u_buf,const uint8 * v_buf,uint8 * rgb_buf,int width)720 void I422ToABGRRow_MIPS_DSPR2(const uint8* y_buf,
721 const uint8* u_buf,
722 const uint8* v_buf,
723 uint8* rgb_buf,
724 int width) {
725 __asm__ __volatile__ (
726 ".set push \n"
727 ".set noreorder \n"
728 "beqz %[width], 2f \n"
729 " repl.ph $s0, 74 \n" // |YG|YG| = |74|74|
730 "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
731 "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
732 "repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
733 "repl.ph $s4, 16 \n" // |0|16|0|16|
734 "repl.ph $s5, 128 \n" // |128|128|
735 "lui $s6, 0xff00 \n"
736 "ori $s6, 0xff00 \n" // |ff|00|ff|00|
737
738 ".p2align 2 \n"
739 "1: \n"
740 I422ToTransientMipsRGB
741 // Arranging into abgr format
742 "precr.qb.ph $t0, $t8, $t1 \n" // |G1|g1|R1|r1|
743 "precr.qb.ph $t3, $t9, $t2 \n" // |G0|g0|R0|r0|
744 "precrq.qb.ph $t8, $t0, $t3 \n" // |G1|R1|G0|R0|
745 "precr.qb.ph $t9, $t0, $t3 \n" // |g1|r1|g0|r0|
746
747 "precr.qb.ph $t2, $t4, $t5 \n" // |B1|b1|B0|b0|
748 "addiu %[width], -4 \n"
749 "addiu %[y_buf], 4 \n"
750 "preceu.ph.qbla $t1, $t2 \n" // |0 |B1|0 |B0|
751 "preceu.ph.qbra $t2, $t2 \n" // |0 |b1|0 |b0|
752 "or $t1, $t1, $s6 \n" // |ff|B1|ff|B0|
753 "or $t2, $t2, $s6 \n" // |ff|b1|ff|b0|
754 "precrq.ph.w $t0, $t2, $t9 \n" // |ff|b1|g1|r1|
755 "precrq.ph.w $t3, $t1, $t8 \n" // |ff|B1|G1|R1|
756 "sll $t9, $t9, 16 \n"
757 "sll $t8, $t8, 16 \n"
758 "packrl.ph $t2, $t2, $t9 \n" // |ff|b0|g0|r0|
759 "packrl.ph $t1, $t1, $t8 \n" // |ff|B0|G0|R0|
760 // Store results.
761 "sw $t2, 0(%[rgb_buf]) \n"
762 "sw $t0, 4(%[rgb_buf]) \n"
763 "sw $t1, 8(%[rgb_buf]) \n"
764 "sw $t3, 12(%[rgb_buf]) \n"
765 "bnez %[width], 1b \n"
766 " addiu %[rgb_buf], 16 \n"
767 "2: \n"
768 ".set pop \n"
769 :[y_buf] "+r" (y_buf),
770 [u_buf] "+r" (u_buf),
771 [v_buf] "+r" (v_buf),
772 [width] "+r" (width),
773 [rgb_buf] "+r" (rgb_buf)
774 :
775 : "t0", "t1", "t2", "t3", "t4", "t5",
776 "t6", "t7", "t8", "t9",
777 "s0", "s1", "s2", "s3",
778 "s4", "s5", "s6"
779 );
780 }
781
I422ToBGRARow_MIPS_DSPR2(const uint8 * y_buf,const uint8 * u_buf,const uint8 * v_buf,uint8 * rgb_buf,int width)782 void I422ToBGRARow_MIPS_DSPR2(const uint8* y_buf,
783 const uint8* u_buf,
784 const uint8* v_buf,
785 uint8* rgb_buf,
786 int width) {
787 __asm__ __volatile__ (
788 ".set push \n"
789 ".set noreorder \n"
790 "beqz %[width], 2f \n"
791 " repl.ph $s0, 74 \n" // |YG|YG| = |74 |74 |
792 "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25|
793 "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52|
794 "repl.ph $s3, 102 \n" // |VR|VR| = |102|102|
795 "repl.ph $s4, 16 \n" // |0|16|0|16|
796 "repl.ph $s5, 128 \n" // |128|128|
797 "lui $s6, 0xff \n"
798 "ori $s6, 0xff \n" // |00|ff|00|ff|
799
800 ".p2align 2 \n"
801 "1: \n"
802 I422ToTransientMipsRGB
803 // Arranging into bgra format
804 "precr.qb.ph $t4, $t4, $t8 \n" // |B1|b1|G1|g1|
805 "precr.qb.ph $t5, $t5, $t9 \n" // |B0|b0|G0|g0|
806 "precrq.qb.ph $t8, $t4, $t5 \n" // |B1|G1|B0|G0|
807 "precr.qb.ph $t9, $t4, $t5 \n" // |b1|g1|b0|g0|
808
809 "precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0|
810 "addiu %[width], -4 \n"
811 "addiu %[y_buf], 4 \n"
812 "preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0|
813 "preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0|
814 "sll $t1, $t1, 8 \n" // |R1|0 |R0|0 |
815 "sll $t2, $t2, 8 \n" // |r1|0 |r0|0 |
816 "or $t1, $t1, $s6 \n" // |R1|ff|R0|ff|
817 "or $t2, $t2, $s6 \n" // |r1|ff|r0|ff|
818 "precrq.ph.w $t0, $t9, $t2 \n" // |b1|g1|r1|ff|
819 "precrq.ph.w $t3, $t8, $t1 \n" // |B1|G1|R1|ff|
820 "sll $t1, $t1, 16 \n"
821 "sll $t2, $t2, 16 \n"
822 "packrl.ph $t2, $t9, $t2 \n" // |b0|g0|r0|ff|
823 "packrl.ph $t1, $t8, $t1 \n" // |B0|G0|R0|ff|
824 // Store results.
825 "sw $t2, 0(%[rgb_buf]) \n"
826 "sw $t0, 4(%[rgb_buf]) \n"
827 "sw $t1, 8(%[rgb_buf]) \n"
828 "sw $t3, 12(%[rgb_buf]) \n"
829 "bnez %[width], 1b \n"
830 " addiu %[rgb_buf], 16 \n"
831 "2: \n"
832 ".set pop \n"
833 :[y_buf] "+r" (y_buf),
834 [u_buf] "+r" (u_buf),
835 [v_buf] "+r" (v_buf),
836 [width] "+r" (width),
837 [rgb_buf] "+r" (rgb_buf)
838 :
839 : "t0", "t1", "t2", "t3", "t4", "t5",
840 "t6", "t7", "t8", "t9",
841 "s0", "s1", "s2", "s3",
842 "s4", "s5", "s6"
843 );
844 }
845
846 // Bilinear filter 8x2 -> 8x1
InterpolateRow_MIPS_DSPR2(uint8 * dst_ptr,const uint8 * src_ptr,ptrdiff_t src_stride,int dst_width,int source_y_fraction)847 void InterpolateRow_MIPS_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
848 ptrdiff_t src_stride, int dst_width,
849 int source_y_fraction) {
850 int y0_fraction = 256 - source_y_fraction;
851 const uint8* src_ptr1 = src_ptr + src_stride;
852
853 __asm__ __volatile__ (
854 ".set push \n"
855 ".set noreorder \n"
856
857 "replv.ph $t0, %[y0_fraction] \n"
858 "replv.ph $t1, %[source_y_fraction] \n"
859
860 ".p2align 2 \n"
861 "1: \n"
862 "lw $t2, 0(%[src_ptr]) \n"
863 "lw $t3, 0(%[src_ptr1]) \n"
864 "lw $t4, 4(%[src_ptr]) \n"
865 "lw $t5, 4(%[src_ptr1]) \n"
866 "muleu_s.ph.qbl $t6, $t2, $t0 \n"
867 "muleu_s.ph.qbr $t7, $t2, $t0 \n"
868 "muleu_s.ph.qbl $t8, $t3, $t1 \n"
869 "muleu_s.ph.qbr $t9, $t3, $t1 \n"
870 "muleu_s.ph.qbl $t2, $t4, $t0 \n"
871 "muleu_s.ph.qbr $t3, $t4, $t0 \n"
872 "muleu_s.ph.qbl $t4, $t5, $t1 \n"
873 "muleu_s.ph.qbr $t5, $t5, $t1 \n"
874 "addq.ph $t6, $t6, $t8 \n"
875 "addq.ph $t7, $t7, $t9 \n"
876 "addq.ph $t2, $t2, $t4 \n"
877 "addq.ph $t3, $t3, $t5 \n"
878 "shra.ph $t6, $t6, 8 \n"
879 "shra.ph $t7, $t7, 8 \n"
880 "shra.ph $t2, $t2, 8 \n"
881 "shra.ph $t3, $t3, 8 \n"
882 "precr.qb.ph $t6, $t6, $t7 \n"
883 "precr.qb.ph $t2, $t2, $t3 \n"
884 "addiu %[src_ptr], %[src_ptr], 8 \n"
885 "addiu %[src_ptr1], %[src_ptr1], 8 \n"
886 "addiu %[dst_width], %[dst_width], -8 \n"
887 "sw $t6, 0(%[dst_ptr]) \n"
888 "sw $t2, 4(%[dst_ptr]) \n"
889 "bgtz %[dst_width], 1b \n"
890 " addiu %[dst_ptr], %[dst_ptr], 8 \n"
891
892 ".set pop \n"
893 : [dst_ptr] "+r" (dst_ptr),
894 [src_ptr1] "+r" (src_ptr1),
895 [src_ptr] "+r" (src_ptr),
896 [dst_width] "+r" (dst_width)
897 : [source_y_fraction] "r" (source_y_fraction),
898 [y0_fraction] "r" (y0_fraction),
899 [src_stride] "r" (src_stride)
900 : "t0", "t1", "t2", "t3", "t4", "t5",
901 "t6", "t7", "t8", "t9"
902 );
903 }
904 #endif // __mips_dsp_rev >= 2
905
906 #endif // defined(__mips__)
907
908 #ifdef __cplusplus
909 } // extern "C"
910 } // namespace libyuv
911 #endif
912