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/u_format.h"
27
28#include <nvif/class.h>
29
30static void
31nv98_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   MAYBE_UNUSED 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
62static const struct nouveau_mclass
63nv98_decoder_msvld[] = {
64   { G98_MSVLD, -1 },
65   { IGT21A_MSVLD, -1 },
66   { GT212_MSVLD, -1 },
67   {}
68};
69
70static const struct nouveau_mclass
71nv98_decoder_mspdec[] = {
72   { G98_MSPDEC, -1 },
73   { GT212_MSPDEC, -1 },
74   {}
75};
76
77static const struct nouveau_mclass
78nv98_decoder_msppp[] = {
79   { G98_MSPPP, -1 },
80   { GT212_MSPPP, -1 },
81   {}
82};
83
84struct pipe_video_codec *
85nv98_create_decoder(struct pipe_context *context,
86                    const struct pipe_video_codec *templ)
87{
88   struct nouveau_screen *screen = &((struct nv50_context *)context)->screen->base;
89   struct nouveau_vp3_decoder *dec;
90   struct nouveau_pushbuf **push;
91   struct nv04_fifo nv04_data = {.vram = 0xbeef0201, .gart = 0xbeef0202};
92
93   int ret, i;
94   uint32_t codec = 1, ppp_codec = 3;
95   uint32_t timeout;
96   u32 tmp_size = 0;
97
98   if (getenv("XVMC_VL"))
99       return vl_create_decoder(context, templ);
100
101   if (templ->entrypoint != PIPE_VIDEO_ENTRYPOINT_BITSTREAM) {
102      debug_printf("%x\n", templ->entrypoint);
103      return NULL;
104   }
105
106   dec = CALLOC_STRUCT(nouveau_vp3_decoder);
107   if (!dec)
108      return NULL;
109   dec->client = screen->client;
110   dec->base = *templ;
111   nouveau_vp3_decoder_init_common(&dec->base);
112
113   dec->bsp_idx = 5;
114   dec->vp_idx = 6;
115   dec->ppp_idx = 7;
116
117   ret = nouveau_object_new(&screen->device->object, 0,
118                            NOUVEAU_FIFO_CHANNEL_CLASS,
119                            &nv04_data, sizeof(nv04_data), &dec->channel[0]);
120
121   if (!ret)
122      ret = nouveau_pushbuf_new(screen->client, dec->channel[0], 4,
123                                32 * 1024, true, &dec->pushbuf[0]);
124
125   for (i = 1; i < 3; ++i) {
126      dec->channel[i] = dec->channel[0];
127      dec->pushbuf[i] = dec->pushbuf[0];
128   }
129   push = dec->pushbuf;
130
131   if (!ret) {
132      ret = nouveau_object_mclass(dec->channel[0], nv98_decoder_msvld);
133      if (ret >= 0) {
134         ret = nouveau_object_new(dec->channel[0], 0xbeef85b1,
135                                  nv98_decoder_msvld[ret].oclass, NULL, 0,
136                                  &dec->bsp);
137      }
138   }
139
140   if (!ret) {
141      ret = nouveau_object_mclass(dec->channel[1], nv98_decoder_mspdec);
142      if (ret >= 0) {
143         ret = nouveau_object_new(dec->channel[1], 0xbeef85b2,
144                                  nv98_decoder_mspdec[ret].oclass, NULL, 0,
145                                  &dec->vp);
146      }
147   }
148
149   if (!ret) {
150      ret = nouveau_object_mclass(dec->channel[2], nv98_decoder_msppp);
151      if (ret >= 0) {
152         ret = nouveau_object_new(dec->channel[2], 0xbeef85b3,
153                                  nv98_decoder_msppp[ret].oclass, NULL, 0,
154                                  &dec->ppp);
155      }
156   }
157
158   if (ret)
159      goto fail;
160
161   BEGIN_NV04(push[0], SUBC_BSP(NV01_SUBCHAN_OBJECT), 1);
162   PUSH_DATA (push[0], dec->bsp->handle);
163
164   BEGIN_NV04(push[0], SUBC_BSP(0x180), 5);
165   for (i = 0; i < 5; i++)
166      PUSH_DATA (push[0], nv04_data.vram);
167
168   BEGIN_NV04(push[1], SUBC_VP(NV01_SUBCHAN_OBJECT), 1);
169   PUSH_DATA (push[1], dec->vp->handle);
170
171   BEGIN_NV04(push[1], SUBC_VP(0x180), 6);
172   for (i = 0; i < 6; i++)
173      PUSH_DATA (push[1], nv04_data.vram);
174
175   BEGIN_NV04(push[2], SUBC_PPP(NV01_SUBCHAN_OBJECT), 1);
176   PUSH_DATA (push[2], dec->ppp->handle);
177
178   BEGIN_NV04(push[2], SUBC_PPP(0x180), 5);
179   for (i = 0; i < 5; i++)
180      PUSH_DATA (push[2], nv04_data.vram);
181
182   dec->base.context = context;
183   dec->base.decode_bitstream = nv98_decoder_decode_bitstream;
184
185   for (i = 0; i < NOUVEAU_VP3_VIDEO_QDEPTH && !ret; ++i)
186      ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
187                           0, 1 << 20, NULL, &dec->bsp_bo[i]);
188   if (!ret)
189      ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
190                           0x100, 4 << 20, NULL, &dec->inter_bo[0]);
191   if (!ret)
192      nouveau_bo_ref(dec->inter_bo[0], &dec->inter_bo[1]);
193   if (ret)
194      goto fail;
195
196   switch (u_reduce_video_profile(templ->profile)) {
197   case PIPE_VIDEO_FORMAT_MPEG12: {
198      codec = 1;
199      assert(templ->max_references <= 2);
200      break;
201   }
202   case PIPE_VIDEO_FORMAT_MPEG4: {
203      codec = 4;
204      tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
205      assert(templ->max_references <= 2);
206      break;
207   }
208   case PIPE_VIDEO_FORMAT_VC1: {
209      ppp_codec = codec = 2;
210      tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
211      assert(templ->max_references <= 2);
212      break;
213   }
214   case PIPE_VIDEO_FORMAT_MPEG4_AVC: {
215      codec = 3;
216      dec->tmp_stride = 16 * mb_half(templ->width) * nouveau_vp3_video_align(templ->height) * 3 / 2;
217      tmp_size = dec->tmp_stride * (templ->max_references + 1);
218      assert(templ->max_references <= 16);
219      break;
220   }
221   default:
222      fprintf(stderr, "invalid codec\n");
223      goto fail;
224   }
225
226   ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
227                           0x4000, NULL, &dec->fw_bo);
228   if (ret)
229      goto fail;
230
231   ret = nouveau_vp3_load_firmware(dec, templ->profile, screen->device->chipset);
232   if (ret)
233      goto fw_fail;
234
235   if (codec != 3) {
236      ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
237                           0x400, NULL, &dec->bitplane_bo);
238      if (ret)
239         goto fail;
240   }
241
242   dec->ref_stride = mb(templ->width)*16 * (mb_half(templ->height)*32 + nouveau_vp3_video_align(templ->height)/2);
243   ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
244                        dec->ref_stride * (templ->max_references+2) + tmp_size,
245                        NULL, &dec->ref_bo);
246   if (ret)
247      goto fail;
248
249   timeout = 0;
250
251   BEGIN_NV04(push[0], SUBC_BSP(0x200), 2);
252   PUSH_DATA (push[0], codec);
253   PUSH_DATA (push[0], timeout);
254
255   BEGIN_NV04(push[1], SUBC_VP(0x200), 2);
256   PUSH_DATA (push[1], codec);
257   PUSH_DATA (push[1], timeout);
258
259   BEGIN_NV04(push[2], SUBC_PPP(0x200), 2);
260   PUSH_DATA (push[2], ppp_codec);
261   PUSH_DATA (push[2], timeout);
262
263   ++dec->fence_seq;
264
265#if NOUVEAU_VP3_DEBUG_FENCE
266   ret = nouveau_bo_new(screen->device, NOUVEAU_BO_GART|NOUVEAU_BO_MAP,
267                        0, 0x1000, NULL, &dec->fence_bo);
268   if (ret)
269      goto fail;
270
271   nouveau_bo_map(dec->fence_bo, NOUVEAU_BO_RDWR, screen->client);
272   dec->fence_map = dec->fence_bo->map;
273   dec->fence_map[0] = dec->fence_map[4] = dec->fence_map[8] = 0;
274   dec->comm = (struct comm *)(dec->fence_map + (COMM_OFFSET/sizeof(*dec->fence_map)));
275
276   /* So lets test if the fence is working? */
277   nouveau_pushbuf_space(push[0], 16, 1, 0);
278   PUSH_REFN (push[0], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
279   BEGIN_NV04(push[0], SUBC_BSP(0x240), 3);
280   PUSH_DATAh(push[0], dec->fence_bo->offset);
281   PUSH_DATA (push[0], dec->fence_bo->offset);
282   PUSH_DATA (push[0], dec->fence_seq);
283
284   BEGIN_NV04(push[0], SUBC_BSP(0x304), 1);
285   PUSH_DATA (push[0], 0);
286   PUSH_KICK (push[0]);
287
288   nouveau_pushbuf_space(push[1], 16, 1, 0);
289   PUSH_REFN (push[1], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
290   BEGIN_NV04(push[1], SUBC_VP(0x240), 3);
291   PUSH_DATAh(push[1], (dec->fence_bo->offset + 0x10));
292   PUSH_DATA (push[1], (dec->fence_bo->offset + 0x10));
293   PUSH_DATA (push[1], dec->fence_seq);
294
295   BEGIN_NV04(push[1], SUBC_VP(0x304), 1);
296   PUSH_DATA (push[1], 0);
297   PUSH_KICK (push[1]);
298
299   nouveau_pushbuf_space(push[2], 16, 1, 0);
300   PUSH_REFN (push[2], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
301   BEGIN_NV04(push[2], SUBC_PPP(0x240), 3);
302   PUSH_DATAh(push[2], (dec->fence_bo->offset + 0x20));
303   PUSH_DATA (push[2], (dec->fence_bo->offset + 0x20));
304   PUSH_DATA (push[2], dec->fence_seq);
305
306   BEGIN_NV04(push[2], SUBC_PPP(0x304), 1);
307   PUSH_DATA (push[2], 0);
308   PUSH_KICK (push[2]);
309
310   usleep(100);
311   while (dec->fence_seq > dec->fence_map[0] ||
312          dec->fence_seq > dec->fence_map[4] ||
313          dec->fence_seq > dec->fence_map[8]) {
314      debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
315      usleep(100);
316   }
317   debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
318#endif
319
320   return &dec->base;
321
322fw_fail:
323   debug_printf("Cannot create decoder without firmware..\n");
324   dec->base.destroy(&dec->base);
325   return NULL;
326
327fail:
328   debug_printf("Creation failed: %s (%i)\n", strerror(-ret), ret);
329   dec->base.destroy(&dec->base);
330   return NULL;
331}
332
333struct pipe_video_buffer *
334nv98_video_buffer_create(struct pipe_context *pipe,
335                         const struct pipe_video_buffer *templat)
336{
337   return nouveau_vp3_video_buffer_create(
338         pipe, templat, NV50_RESOURCE_FLAG_VIDEO);
339}
340