1 /*
2 * Copyright 2011-2013 Maarten Lankhorst, Ilia Mirkin
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 */
22
23 #include "nv50/nv98_video.h"
24
25 #include "util/u_sampler.h"
26 #include "util/format/u_format.h"
27
28 #include <nvif/class.h>
29
30 static void
nv98_decoder_decode_bitstream(struct pipe_video_codec * decoder,struct pipe_video_buffer * video_target,struct pipe_picture_desc * picture,unsigned num_buffers,const void * const * data,const unsigned * num_bytes)31 nv98_decoder_decode_bitstream(struct pipe_video_codec *decoder,
32 struct pipe_video_buffer *video_target,
33 struct pipe_picture_desc *picture,
34 unsigned num_buffers,
35 const void *const *data,
36 const unsigned *num_bytes)
37 {
38 struct nouveau_vp3_decoder *dec = (struct nouveau_vp3_decoder *)decoder;
39 struct nouveau_vp3_video_buffer *target = (struct nouveau_vp3_video_buffer *)video_target;
40 uint32_t comm_seq = ++dec->fence_seq;
41 union pipe_desc desc;
42
43 unsigned vp_caps, is_ref;
44 ASSERTED unsigned ret; /* used in debug checks */
45 struct nouveau_vp3_video_buffer *refs[16] = {};
46
47 desc.base = picture;
48
49 assert(target->base.buffer_format == PIPE_FORMAT_NV12);
50
51 ret = nv98_decoder_bsp(dec, desc, target, comm_seq,
52 num_buffers, data, num_bytes,
53 &vp_caps, &is_ref, refs);
54
55 /* did we decode bitstream correctly? */
56 assert(ret == 2);
57
58 nv98_decoder_vp(dec, desc, target, comm_seq, vp_caps, is_ref, refs);
59 nv98_decoder_ppp(dec, desc, target, comm_seq);
60 }
61
62 static const struct nouveau_mclass
63 nv98_decoder_msvld[] = {
64 { G98_MSVLD, -1 },
65 { IGT21A_MSVLD, -1 },
66 { GT212_MSVLD, -1 },
67 {}
68 };
69
70 static const struct nouveau_mclass
71 nv98_decoder_mspdec[] = {
72 { G98_MSPDEC, -1 },
73 { GT212_MSPDEC, -1 },
74 {}
75 };
76
77 static const struct nouveau_mclass
78 nv98_decoder_msppp[] = {
79 { G98_MSPPP, -1 },
80 { GT212_MSPPP, -1 },
81 {}
82 };
83
84 struct pipe_video_codec *
nv98_create_decoder(struct pipe_context * context,const struct pipe_video_codec * templ)85 nv98_create_decoder(struct pipe_context *context,
86 const struct pipe_video_codec *templ)
87 {
88 struct nv50_context *nv50 = nv50_context(context);
89 struct nouveau_screen *screen = &nv50->screen->base;
90 struct nouveau_vp3_decoder *dec;
91 struct nouveau_pushbuf **push;
92 struct nv04_fifo nv04_data = {.vram = 0xbeef0201, .gart = 0xbeef0202};
93
94 int ret, i;
95 uint32_t codec = 1, ppp_codec = 3;
96 uint32_t timeout;
97 u32 tmp_size = 0;
98
99 if (templ->entrypoint != PIPE_VIDEO_ENTRYPOINT_BITSTREAM) {
100 debug_printf("%x\n", templ->entrypoint);
101 return NULL;
102 }
103
104 dec = CALLOC_STRUCT(nouveau_vp3_decoder);
105 if (!dec)
106 return NULL;
107 dec->client = nv50->base.client;
108 dec->base = *templ;
109 nouveau_vp3_decoder_init_common(&dec->base);
110
111 dec->bsp_idx = 5;
112 dec->vp_idx = 6;
113 dec->ppp_idx = 7;
114
115 ret = nouveau_object_new(&screen->device->object, 0,
116 NOUVEAU_FIFO_CHANNEL_CLASS,
117 &nv04_data, sizeof(nv04_data), &dec->channel[0]);
118
119 if (!ret)
120 ret = nouveau_pushbuf_create(screen, &nv50->base, nv50->base.client, dec->channel[0],
121 4, 32 * 1024, &dec->pushbuf[0]);
122
123 for (i = 1; i < 3; ++i) {
124 dec->channel[i] = dec->channel[0];
125 dec->pushbuf[i] = dec->pushbuf[0];
126 }
127 push = dec->pushbuf;
128
129 if (!ret) {
130 ret = nouveau_object_mclass(dec->channel[0], nv98_decoder_msvld);
131 if (ret >= 0) {
132 ret = nouveau_object_new(dec->channel[0], 0xbeef85b1,
133 nv98_decoder_msvld[ret].oclass, NULL, 0,
134 &dec->bsp);
135 }
136 }
137
138 if (!ret) {
139 ret = nouveau_object_mclass(dec->channel[1], nv98_decoder_mspdec);
140 if (ret >= 0) {
141 ret = nouveau_object_new(dec->channel[1], 0xbeef85b2,
142 nv98_decoder_mspdec[ret].oclass, NULL, 0,
143 &dec->vp);
144 }
145 }
146
147 if (!ret) {
148 ret = nouveau_object_mclass(dec->channel[2], nv98_decoder_msppp);
149 if (ret >= 0) {
150 ret = nouveau_object_new(dec->channel[2], 0xbeef85b3,
151 nv98_decoder_msppp[ret].oclass, NULL, 0,
152 &dec->ppp);
153 }
154 }
155
156 if (ret)
157 goto fail;
158
159 BEGIN_NV04(push[0], SUBC_BSP(NV01_SUBCHAN_OBJECT), 1);
160 PUSH_DATA (push[0], dec->bsp->handle);
161
162 BEGIN_NV04(push[0], SUBC_BSP(0x180), 5);
163 for (i = 0; i < 5; i++)
164 PUSH_DATA (push[0], nv04_data.vram);
165
166 BEGIN_NV04(push[1], SUBC_VP(NV01_SUBCHAN_OBJECT), 1);
167 PUSH_DATA (push[1], dec->vp->handle);
168
169 BEGIN_NV04(push[1], SUBC_VP(0x180), 6);
170 for (i = 0; i < 6; i++)
171 PUSH_DATA (push[1], nv04_data.vram);
172
173 BEGIN_NV04(push[2], SUBC_PPP(NV01_SUBCHAN_OBJECT), 1);
174 PUSH_DATA (push[2], dec->ppp->handle);
175
176 BEGIN_NV04(push[2], SUBC_PPP(0x180), 5);
177 for (i = 0; i < 5; i++)
178 PUSH_DATA (push[2], nv04_data.vram);
179
180 dec->base.context = context;
181 dec->base.decode_bitstream = nv98_decoder_decode_bitstream;
182
183 for (i = 0; i < NOUVEAU_VP3_VIDEO_QDEPTH && !ret; ++i)
184 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
185 0, 1 << 20, NULL, &dec->bsp_bo[i]);
186 if (!ret)
187 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
188 0x100, 4 << 20, NULL, &dec->inter_bo[0]);
189 if (!ret)
190 nouveau_bo_ref(dec->inter_bo[0], &dec->inter_bo[1]);
191 if (ret)
192 goto fail;
193
194 switch (u_reduce_video_profile(templ->profile)) {
195 case PIPE_VIDEO_FORMAT_MPEG12: {
196 codec = 1;
197 assert(templ->max_references <= 2);
198 break;
199 }
200 case PIPE_VIDEO_FORMAT_MPEG4: {
201 codec = 4;
202 tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
203 assert(templ->max_references <= 2);
204 break;
205 }
206 case PIPE_VIDEO_FORMAT_VC1: {
207 ppp_codec = codec = 2;
208 tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
209 assert(templ->max_references <= 2);
210 break;
211 }
212 case PIPE_VIDEO_FORMAT_MPEG4_AVC: {
213 codec = 3;
214 dec->tmp_stride = 16 * mb_half(templ->width) * nouveau_vp3_video_align(templ->height) * 3 / 2;
215 tmp_size = dec->tmp_stride * (templ->max_references + 1);
216 assert(templ->max_references <= 16);
217 break;
218 }
219 default:
220 fprintf(stderr, "invalid codec\n");
221 goto fail;
222 }
223
224 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
225 0x4000, NULL, &dec->fw_bo);
226 if (ret)
227 goto fail;
228
229 ret = nouveau_vp3_load_firmware(dec, templ->profile, screen->device->chipset);
230 if (ret)
231 goto fw_fail;
232
233 if (codec != 3) {
234 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
235 0x400, NULL, &dec->bitplane_bo);
236 if (ret)
237 goto fail;
238 }
239
240 dec->ref_stride = mb(templ->width)*16 * (mb_half(templ->height)*32 + nouveau_vp3_video_align(templ->height)/2);
241 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
242 dec->ref_stride * (templ->max_references+2) + tmp_size,
243 NULL, &dec->ref_bo);
244 if (ret)
245 goto fail;
246
247 timeout = 0;
248
249 BEGIN_NV04(push[0], SUBC_BSP(0x200), 2);
250 PUSH_DATA (push[0], codec);
251 PUSH_DATA (push[0], timeout);
252
253 BEGIN_NV04(push[1], SUBC_VP(0x200), 2);
254 PUSH_DATA (push[1], codec);
255 PUSH_DATA (push[1], timeout);
256
257 BEGIN_NV04(push[2], SUBC_PPP(0x200), 2);
258 PUSH_DATA (push[2], ppp_codec);
259 PUSH_DATA (push[2], timeout);
260
261 ++dec->fence_seq;
262
263 #if NOUVEAU_VP3_DEBUG_FENCE
264 ret = nouveau_bo_new(screen->device, NOUVEAU_BO_GART|NOUVEAU_BO_MAP,
265 0, 0x1000, NULL, &dec->fence_bo);
266 if (ret)
267 goto fail;
268
269 BO_MAP(screen, dec->fence_bo, NOUVEAU_BO_RDWR, screen->client);
270 dec->fence_map = dec->fence_bo->map;
271 dec->fence_map[0] = dec->fence_map[4] = dec->fence_map[8] = 0;
272 dec->comm = (struct comm *)(dec->fence_map + (COMM_OFFSET/sizeof(*dec->fence_map)));
273
274 /* So lets test if the fence is working? */
275 PUSH_SPACE_EX(push[0], 16, 1, 0);
276 PUSH_REF1 (push[0], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
277 BEGIN_NV04(push[0], SUBC_BSP(0x240), 3);
278 PUSH_DATAh(push[0], dec->fence_bo->offset);
279 PUSH_DATA (push[0], dec->fence_bo->offset);
280 PUSH_DATA (push[0], dec->fence_seq);
281
282 BEGIN_NV04(push[0], SUBC_BSP(0x304), 1);
283 PUSH_DATA (push[0], 0);
284 PUSH_KICK (push[0]);
285
286 PUSH_SPACE_EX(push[1], 16, 1, 0);
287 PUSH_REF1 (push[1], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
288 BEGIN_NV04(push[1], SUBC_VP(0x240), 3);
289 PUSH_DATAh(push[1], (dec->fence_bo->offset + 0x10));
290 PUSH_DATA (push[1], (dec->fence_bo->offset + 0x10));
291 PUSH_DATA (push[1], dec->fence_seq);
292
293 BEGIN_NV04(push[1], SUBC_VP(0x304), 1);
294 PUSH_DATA (push[1], 0);
295 PUSH_KICK (push[1]);
296
297 PUSH_SPACE_EX(push[2], 16, 1, 0);
298 PUSH_REF1 (push[2], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
299 BEGIN_NV04(push[2], SUBC_PPP(0x240), 3);
300 PUSH_DATAh(push[2], (dec->fence_bo->offset + 0x20));
301 PUSH_DATA (push[2], (dec->fence_bo->offset + 0x20));
302 PUSH_DATA (push[2], dec->fence_seq);
303
304 BEGIN_NV04(push[2], SUBC_PPP(0x304), 1);
305 PUSH_DATA (push[2], 0);
306 PUSH_KICK (push[2]);
307
308 usleep(100);
309 while (dec->fence_seq > dec->fence_map[0] ||
310 dec->fence_seq > dec->fence_map[4] ||
311 dec->fence_seq > dec->fence_map[8]) {
312 debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
313 usleep(100);
314 }
315 debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
316 #endif
317
318 return &dec->base;
319
320 fw_fail:
321 debug_printf("Cannot create decoder without firmware..\n");
322 dec->base.destroy(&dec->base);
323 return NULL;
324
325 fail:
326 debug_printf("Creation failed: %s (%i)\n", strerror(-ret), ret);
327 dec->base.destroy(&dec->base);
328 return NULL;
329 }
330
331 struct pipe_video_buffer *
nv98_video_buffer_create(struct pipe_context * pipe,const struct pipe_video_buffer * templat)332 nv98_video_buffer_create(struct pipe_context *pipe,
333 const struct pipe_video_buffer *templat)
334 {
335 return nouveau_vp3_video_buffer_create(
336 pipe, templat, NV50_RESOURCE_FLAG_VIDEO);
337 }
338