]> git.sesse.net Git - ffmpeg/blob - libavcodec/armv4l/dsputil_arm.c
Fix regression introduced by r12929.
[ffmpeg] / libavcodec / armv4l / dsputil_arm.c
1 /*
2  * ARMv4L optimized DSP utils
3  * Copyright (c) 2001 Lionel Ulmer.
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21
22 #include "dsputil.h"
23 #ifdef HAVE_IPP
24 #include "ipp.h"
25 #endif
26
27 extern void dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx);
28
29 extern void j_rev_dct_ARM(DCTELEM *data);
30 extern void simple_idct_ARM(DCTELEM *data);
31
32 extern void simple_idct_armv5te(DCTELEM *data);
33 extern void simple_idct_put_armv5te(uint8_t *dest, int line_size,
34                                     DCTELEM *data);
35 extern void simple_idct_add_armv5te(uint8_t *dest, int line_size,
36                                     DCTELEM *data);
37
38 extern void ff_simple_idct_armv6(DCTELEM *data);
39 extern void ff_simple_idct_put_armv6(uint8_t *dest, int line_size,
40                                      DCTELEM *data);
41 extern void ff_simple_idct_add_armv6(uint8_t *dest, int line_size,
42                                      DCTELEM *data);
43
44 /* XXX: local hack */
45 static void (*ff_put_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size);
46 static void (*ff_add_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size);
47
48 void put_pixels8_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
49 void put_pixels8_x2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
50 void put_pixels8_y2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
51 void put_pixels8_xy2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
52
53 void put_no_rnd_pixels8_x2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
54 void put_no_rnd_pixels8_y2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
55 void put_no_rnd_pixels8_xy2_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
56
57 void put_pixels16_arm(uint8_t *block, const uint8_t *pixels, int line_size, int h);
58
59 CALL_2X_PIXELS(put_pixels16_x2_arm , put_pixels8_x2_arm , 8)
60 CALL_2X_PIXELS(put_pixels16_y2_arm , put_pixels8_y2_arm , 8)
61 CALL_2X_PIXELS(put_pixels16_xy2_arm, put_pixels8_xy2_arm, 8)
62 CALL_2X_PIXELS(put_no_rnd_pixels16_x2_arm , put_no_rnd_pixels8_x2_arm , 8)
63 CALL_2X_PIXELS(put_no_rnd_pixels16_y2_arm , put_no_rnd_pixels8_y2_arm , 8)
64 CALL_2X_PIXELS(put_no_rnd_pixels16_xy2_arm, put_no_rnd_pixels8_xy2_arm, 8)
65
66 static void add_pixels_clamped_ARM(short *block, unsigned char *dest, int line_size)
67 {
68     asm volatile (
69                   "mov r10, #8 \n\t"
70
71                   "1: \n\t"
72
73                   /* load dest */
74                   "ldr r4, [%1] \n\t"
75                   /* block[0] and block[1]*/
76                   "ldrsh r5, [%0] \n\t"
77                   "ldrsh r7, [%0, #2] \n\t"
78                   "and r6, r4, #0xFF \n\t"
79                   "and r8, r4, #0xFF00 \n\t"
80                   "add r6, r5, r6 \n\t"
81                   "add r8, r7, r8, lsr #8 \n\t"
82                   "mvn r5, r5 \n\t"
83                   "mvn r7, r7 \n\t"
84                   "tst r6, #0x100 \n\t"
85                   "movne r6, r5, lsr #24 \n\t"
86                   "tst r8, #0x100 \n\t"
87                   "movne r8, r7, lsr #24 \n\t"
88                   "mov r9, r6 \n\t"
89                   "ldrsh r5, [%0, #4] \n\t" /* moved form [A] */
90                   "orr r9, r9, r8, lsl #8 \n\t"
91                   /* block[2] and block[3] */
92                   /* [A] */
93                   "ldrsh r7, [%0, #6] \n\t"
94                   "and r6, r4, #0xFF0000 \n\t"
95                   "and r8, r4, #0xFF000000 \n\t"
96                   "add r6, r5, r6, lsr #16 \n\t"
97                   "add r8, r7, r8, lsr #24 \n\t"
98                   "mvn r5, r5 \n\t"
99                   "mvn r7, r7 \n\t"
100                   "tst r6, #0x100 \n\t"
101                   "movne r6, r5, lsr #24 \n\t"
102                   "tst r8, #0x100 \n\t"
103                   "movne r8, r7, lsr #24 \n\t"
104                   "orr r9, r9, r6, lsl #16 \n\t"
105                   "ldr r4, [%1, #4] \n\t"       /* moved form [B] */
106                   "orr r9, r9, r8, lsl #24 \n\t"
107                   /* store dest */
108                   "ldrsh r5, [%0, #8] \n\t" /* moved form [C] */
109                   "str r9, [%1] \n\t"
110
111                   /* load dest */
112                   /* [B] */
113                   /* block[4] and block[5] */
114                   /* [C] */
115                   "ldrsh r7, [%0, #10] \n\t"
116                   "and r6, r4, #0xFF \n\t"
117                   "and r8, r4, #0xFF00 \n\t"
118                   "add r6, r5, r6 \n\t"
119                   "add r8, r7, r8, lsr #8 \n\t"
120                   "mvn r5, r5 \n\t"
121                   "mvn r7, r7 \n\t"
122                   "tst r6, #0x100 \n\t"
123                   "movne r6, r5, lsr #24 \n\t"
124                   "tst r8, #0x100 \n\t"
125                   "movne r8, r7, lsr #24 \n\t"
126                   "mov r9, r6 \n\t"
127                   "ldrsh r5, [%0, #12] \n\t" /* moved from [D] */
128                   "orr r9, r9, r8, lsl #8 \n\t"
129                   /* block[6] and block[7] */
130                   /* [D] */
131                   "ldrsh r7, [%0, #14] \n\t"
132                   "and r6, r4, #0xFF0000 \n\t"
133                   "and r8, r4, #0xFF000000 \n\t"
134                   "add r6, r5, r6, lsr #16 \n\t"
135                   "add r8, r7, r8, lsr #24 \n\t"
136                   "mvn r5, r5 \n\t"
137                   "mvn r7, r7 \n\t"
138                   "tst r6, #0x100 \n\t"
139                   "movne r6, r5, lsr #24 \n\t"
140                   "tst r8, #0x100 \n\t"
141                   "movne r8, r7, lsr #24 \n\t"
142                   "orr r9, r9, r6, lsl #16 \n\t"
143                   "add %0, %0, #16 \n\t" /* moved from [E] */
144                   "orr r9, r9, r8, lsl #24 \n\t"
145                   "subs r10, r10, #1 \n\t" /* moved from [F] */
146                   /* store dest */
147                   "str r9, [%1, #4] \n\t"
148
149                   /* [E] */
150                   /* [F] */
151                   "add %1, %1, %2 \n\t"
152                   "bne 1b \n\t"
153                   : "+r"(block),
154                     "+r"(dest)
155                   : "r"(line_size)
156                   : "r4", "r5", "r6", "r7", "r8", "r9", "r10", "cc", "memory" );
157 }
158
159 /* XXX: those functions should be suppressed ASAP when all IDCTs are
160    converted */
161 static void j_rev_dct_ARM_put(uint8_t *dest, int line_size, DCTELEM *block)
162 {
163     j_rev_dct_ARM (block);
164     ff_put_pixels_clamped(block, dest, line_size);
165 }
166 static void j_rev_dct_ARM_add(uint8_t *dest, int line_size, DCTELEM *block)
167 {
168     j_rev_dct_ARM (block);
169     ff_add_pixels_clamped(block, dest, line_size);
170 }
171 static void simple_idct_ARM_put(uint8_t *dest, int line_size, DCTELEM *block)
172 {
173     simple_idct_ARM (block);
174     ff_put_pixels_clamped(block, dest, line_size);
175 }
176 static void simple_idct_ARM_add(uint8_t *dest, int line_size, DCTELEM *block)
177 {
178     simple_idct_ARM (block);
179     ff_add_pixels_clamped(block, dest, line_size);
180 }
181
182 #ifdef HAVE_IPP
183 static void simple_idct_ipp(DCTELEM *block)
184 {
185     ippiDCT8x8Inv_Video_16s_C1I(block);
186 }
187 static void simple_idct_ipp_put(uint8_t *dest, int line_size, DCTELEM *block)
188 {
189     ippiDCT8x8Inv_Video_16s8u_C1R(block, dest, line_size);
190 }
191
192 void add_pixels_clamped_iwmmxt(const DCTELEM *block, uint8_t *pixels, int line_size);
193
194 static void simple_idct_ipp_add(uint8_t *dest, int line_size, DCTELEM *block)
195 {
196     ippiDCT8x8Inv_Video_16s_C1I(block);
197 #ifdef HAVE_IWMMXT
198     add_pixels_clamped_iwmmxt(block, dest, line_size);
199 #else
200     add_pixels_clamped_ARM(block, dest, line_size);
201 #endif
202 }
203 #endif
204
205 void dsputil_init_armv4l(DSPContext* c, AVCodecContext *avctx)
206 {
207     int idct_algo= avctx->idct_algo;
208
209     ff_put_pixels_clamped = c->put_pixels_clamped;
210     ff_add_pixels_clamped = c->add_pixels_clamped;
211
212     if (avctx->lowres == 0) {
213         if(idct_algo == FF_IDCT_AUTO){
214 #if defined(HAVE_IPP)
215             idct_algo = FF_IDCT_IPP;
216 #elif defined(HAVE_ARMV6)
217             idct_algo = FF_IDCT_SIMPLEARMV6;
218 #elif defined(HAVE_ARMV5TE)
219             idct_algo = FF_IDCT_SIMPLEARMV5TE;
220 #else
221             idct_algo = FF_IDCT_ARM;
222 #endif
223         }
224
225         if(idct_algo==FF_IDCT_ARM){
226             c->idct_put= j_rev_dct_ARM_put;
227             c->idct_add= j_rev_dct_ARM_add;
228             c->idct    = j_rev_dct_ARM;
229             c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;/* FF_NO_IDCT_PERM */
230         } else if (idct_algo==FF_IDCT_SIMPLEARM){
231             c->idct_put= simple_idct_ARM_put;
232             c->idct_add= simple_idct_ARM_add;
233             c->idct    = simple_idct_ARM;
234             c->idct_permutation_type= FF_NO_IDCT_PERM;
235 #ifdef HAVE_ARMV6
236         } else if (idct_algo==FF_IDCT_SIMPLEARMV6){
237             c->idct_put= ff_simple_idct_put_armv6;
238             c->idct_add= ff_simple_idct_add_armv6;
239             c->idct    = ff_simple_idct_armv6;
240             c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
241 #endif
242 #ifdef HAVE_ARMV5TE
243         } else if (idct_algo==FF_IDCT_SIMPLEARMV5TE){
244             c->idct_put= simple_idct_put_armv5te;
245             c->idct_add= simple_idct_add_armv5te;
246             c->idct    = simple_idct_armv5te;
247             c->idct_permutation_type = FF_NO_IDCT_PERM;
248 #endif
249 #ifdef HAVE_IPP
250         } else if (idct_algo==FF_IDCT_IPP){
251             c->idct_put= simple_idct_ipp_put;
252             c->idct_add= simple_idct_ipp_add;
253             c->idct    = simple_idct_ipp;
254             c->idct_permutation_type= FF_NO_IDCT_PERM;
255 #endif
256         }
257     }
258
259     c->put_pixels_tab[0][0] = put_pixels16_arm;
260     c->put_pixels_tab[0][1] = put_pixels16_x2_arm; //OK!
261     c->put_pixels_tab[0][2] = put_pixels16_y2_arm; //OK!
262     c->put_pixels_tab[0][3] = put_pixels16_xy2_arm;
263     c->put_no_rnd_pixels_tab[0][0] = put_pixels16_arm;
264     c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_arm; // OK
265     c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_arm; //OK
266     c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_arm;
267     c->put_pixels_tab[1][0] = put_pixels8_arm; //OK
268     c->put_pixels_tab[1][1] = put_pixels8_x2_arm; //OK
269     c->put_pixels_tab[1][2] = put_pixels8_y2_arm;
270     c->put_pixels_tab[1][3] = put_pixels8_xy2_arm;
271     c->put_no_rnd_pixels_tab[1][0] = put_pixels8_arm;//OK
272     c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_arm; //OK
273     c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_arm; //OK
274     c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_arm;
275
276 #ifdef HAVE_IWMMXT
277     dsputil_init_iwmmxt(c, avctx);
278 #endif
279 }