vc4: Try to schedule QIR instructions between writing to and reading math.
[mesa.git] / src / gallium / drivers / radeon / radeon_video.c
1 /**************************************************************************
2 *
3 * Copyright 2013 Advanced Micro Devices, Inc.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 /*
29 * Authors:
30 * Christian König <christian.koenig@amd.com>
31 *
32 */
33
34 #include <unistd.h>
35
36 #include "util/u_memory.h"
37 #include "util/u_video.h"
38
39 #include "vl/vl_defines.h"
40 #include "vl/vl_video_buffer.h"
41
42 #include "r600_pipe_common.h"
43 #include "radeon_video.h"
44 #include "radeon_vce.h"
45
46 #define UVD_FW_1_66_16 ((1 << 24) | (66 << 16) | (16 << 8))
47
48 /* generate an stream handle */
49 unsigned rvid_alloc_stream_handle()
50 {
51 static unsigned counter = 0;
52 unsigned stream_handle = 0;
53 unsigned pid = getpid();
54 int i;
55
56 for (i = 0; i < 32; ++i)
57 stream_handle |= ((pid >> i) & 1) << (31 - i);
58
59 stream_handle ^= ++counter;
60 return stream_handle;
61 }
62
63 /* create a buffer in the winsys */
64 bool rvid_create_buffer(struct pipe_screen *screen, struct rvid_buffer *buffer,
65 unsigned size, unsigned usage)
66 {
67 memset(buffer, 0, sizeof(*buffer));
68 buffer->usage = usage;
69
70 /* Hardware buffer placement restrictions require the kernel to be
71 * able to move buffers around individually, so request a
72 * non-sub-allocated buffer.
73 */
74 buffer->res = (struct r600_resource *)
75 pipe_buffer_create(screen, PIPE_BIND_SHARED,
76 usage, size);
77
78 return buffer->res != NULL;
79 }
80
81 /* destroy a buffer */
82 void rvid_destroy_buffer(struct rvid_buffer *buffer)
83 {
84 r600_resource_reference(&buffer->res, NULL);
85 }
86
87 /* reallocate a buffer, preserving its content */
88 bool rvid_resize_buffer(struct pipe_screen *screen, struct radeon_winsys_cs *cs,
89 struct rvid_buffer *new_buf, unsigned new_size)
90 {
91 struct r600_common_screen *rscreen = (struct r600_common_screen *)screen;
92 struct radeon_winsys* ws = rscreen->ws;
93 unsigned bytes = MIN2(new_buf->res->buf->size, new_size);
94 struct rvid_buffer old_buf = *new_buf;
95 void *src = NULL, *dst = NULL;
96
97 if (!rvid_create_buffer(screen, new_buf, new_size, new_buf->usage))
98 goto error;
99
100 src = ws->buffer_map(old_buf.res->buf, cs, PIPE_TRANSFER_READ);
101 if (!src)
102 goto error;
103
104 dst = ws->buffer_map(new_buf->res->buf, cs, PIPE_TRANSFER_WRITE);
105 if (!dst)
106 goto error;
107
108 memcpy(dst, src, bytes);
109 if (new_size > bytes) {
110 new_size -= bytes;
111 dst += bytes;
112 memset(dst, 0, new_size);
113 }
114 ws->buffer_unmap(new_buf->res->buf);
115 ws->buffer_unmap(old_buf.res->buf);
116 rvid_destroy_buffer(&old_buf);
117 return true;
118
119 error:
120 if (src)
121 ws->buffer_unmap(old_buf.res->buf);
122 rvid_destroy_buffer(new_buf);
123 *new_buf = old_buf;
124 return false;
125 }
126
127 /* clear the buffer with zeros */
128 void rvid_clear_buffer(struct pipe_context *context, struct rvid_buffer* buffer)
129 {
130 struct r600_common_context *rctx = (struct r600_common_context*)context;
131
132 rctx->clear_buffer(context, &buffer->res->b.b, 0, buffer->res->buf->size,
133 0, R600_COHERENCY_NONE);
134 context->flush(context, NULL, 0);
135 }
136
137 /**
138 * join surfaces into the same buffer with identical tiling params
139 * sumup their sizes and replace the backend buffers with a single bo
140 */
141 void rvid_join_surfaces(struct radeon_winsys* ws,
142 struct pb_buffer** buffers[VL_NUM_COMPONENTS],
143 struct radeon_surf *surfaces[VL_NUM_COMPONENTS])
144 {
145 unsigned best_tiling, best_wh, off;
146 unsigned size, alignment;
147 struct pb_buffer *pb;
148 unsigned i, j;
149
150 for (i = 0, best_tiling = 0, best_wh = ~0; i < VL_NUM_COMPONENTS; ++i) {
151 unsigned wh;
152
153 if (!surfaces[i])
154 continue;
155
156 /* choose the smallest bank w/h for now */
157 wh = surfaces[i]->bankw * surfaces[i]->bankh;
158 if (wh < best_wh) {
159 best_wh = wh;
160 best_tiling = i;
161 }
162 }
163
164 for (i = 0, off = 0; i < VL_NUM_COMPONENTS; ++i) {
165 if (!surfaces[i])
166 continue;
167
168 /* copy the tiling parameters */
169 surfaces[i]->bankw = surfaces[best_tiling]->bankw;
170 surfaces[i]->bankh = surfaces[best_tiling]->bankh;
171 surfaces[i]->mtilea = surfaces[best_tiling]->mtilea;
172 surfaces[i]->tile_split = surfaces[best_tiling]->tile_split;
173
174 /* adjust the texture layer offsets */
175 off = align(off, surfaces[i]->surf_alignment);
176 for (j = 0; j < ARRAY_SIZE(surfaces[i]->level); ++j)
177 surfaces[i]->level[j].offset += off;
178 off += surfaces[i]->surf_size;
179 }
180
181 for (i = 0, size = 0, alignment = 0; i < VL_NUM_COMPONENTS; ++i) {
182 if (!buffers[i] || !*buffers[i])
183 continue;
184
185 size = align(size, (*buffers[i])->alignment);
186 size += (*buffers[i])->size;
187 alignment = MAX2(alignment, (*buffers[i])->alignment * 1);
188 }
189
190 if (!size)
191 return;
192
193 /* TODO: 2D tiling workaround */
194 alignment *= 2;
195
196 pb = ws->buffer_create(ws, size, alignment, RADEON_DOMAIN_VRAM, 0);
197 if (!pb)
198 return;
199
200 for (i = 0; i < VL_NUM_COMPONENTS; ++i) {
201 if (!buffers[i] || !*buffers[i])
202 continue;
203
204 pb_reference(buffers[i], pb);
205 }
206
207 pb_reference(&pb, NULL);
208 }
209
210 int rvid_get_video_param(struct pipe_screen *screen,
211 enum pipe_video_profile profile,
212 enum pipe_video_entrypoint entrypoint,
213 enum pipe_video_cap param)
214 {
215 struct r600_common_screen *rscreen = (struct r600_common_screen *)screen;
216 enum pipe_video_format codec = u_reduce_video_profile(profile);
217 struct radeon_info info;
218
219 rscreen->ws->query_info(rscreen->ws, &info);
220
221 if (entrypoint == PIPE_VIDEO_ENTRYPOINT_ENCODE) {
222 switch (param) {
223 case PIPE_VIDEO_CAP_SUPPORTED:
224 return codec == PIPE_VIDEO_FORMAT_MPEG4_AVC &&
225 rvce_is_fw_version_supported(rscreen);
226 case PIPE_VIDEO_CAP_NPOT_TEXTURES:
227 return 1;
228 case PIPE_VIDEO_CAP_MAX_WIDTH:
229 return (rscreen->family < CHIP_TONGA) ? 2048 : 4096;
230 case PIPE_VIDEO_CAP_MAX_HEIGHT:
231 return (rscreen->family < CHIP_TONGA) ? 1152 : 2304;
232 case PIPE_VIDEO_CAP_PREFERED_FORMAT:
233 return PIPE_FORMAT_NV12;
234 case PIPE_VIDEO_CAP_PREFERS_INTERLACED:
235 return false;
236 case PIPE_VIDEO_CAP_SUPPORTS_INTERLACED:
237 return false;
238 case PIPE_VIDEO_CAP_SUPPORTS_PROGRESSIVE:
239 return true;
240 case PIPE_VIDEO_CAP_STACKED_FRAMES:
241 return (rscreen->family < CHIP_TONGA) ? 1 : 2;
242 default:
243 return 0;
244 }
245 }
246
247 switch (param) {
248 case PIPE_VIDEO_CAP_SUPPORTED:
249 switch (codec) {
250 case PIPE_VIDEO_FORMAT_MPEG12:
251 return profile != PIPE_VIDEO_PROFILE_MPEG1;
252 case PIPE_VIDEO_FORMAT_MPEG4:
253 /* no support for MPEG4 on older hw */
254 return rscreen->family >= CHIP_PALM;
255 case PIPE_VIDEO_FORMAT_MPEG4_AVC:
256 if ((rscreen->family == CHIP_POLARIS10 ||
257 rscreen->family == CHIP_POLARIS11) &&
258 info.uvd_fw_version < UVD_FW_1_66_16 ) {
259 RVID_ERR("POLARIS10/11 firmware version need to be updated.\n");
260 return false;
261 }
262 return true;
263 case PIPE_VIDEO_FORMAT_VC1:
264 return true;
265 case PIPE_VIDEO_FORMAT_HEVC:
266 /* Carrizo only supports HEVC Main */
267 if (rscreen->family >= CHIP_STONEY)
268 return (profile == PIPE_VIDEO_PROFILE_HEVC_MAIN ||
269 profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10);
270 else if (rscreen->family >= CHIP_CARRIZO)
271 return profile == PIPE_VIDEO_PROFILE_HEVC_MAIN;
272 default:
273 return false;
274 }
275 case PIPE_VIDEO_CAP_NPOT_TEXTURES:
276 return 1;
277 case PIPE_VIDEO_CAP_MAX_WIDTH:
278 return (rscreen->family < CHIP_TONGA) ? 2048 : 4096;
279 case PIPE_VIDEO_CAP_MAX_HEIGHT:
280 return (rscreen->family < CHIP_TONGA) ? 1152 : 4096;
281 case PIPE_VIDEO_CAP_PREFERED_FORMAT:
282 return PIPE_FORMAT_NV12;
283 case PIPE_VIDEO_CAP_PREFERS_INTERLACED:
284 case PIPE_VIDEO_CAP_SUPPORTS_INTERLACED:
285 if (rscreen->family < CHIP_PALM) {
286 /* MPEG2 only with shaders and no support for
287 interlacing on R6xx style UVD */
288 return codec != PIPE_VIDEO_FORMAT_MPEG12 &&
289 rscreen->family > CHIP_RV770;
290 } else {
291 if (u_reduce_video_profile(profile) == PIPE_VIDEO_FORMAT_HEVC)
292 return false; //The firmware doesn't support interlaced HEVC.
293 return true;
294 }
295 case PIPE_VIDEO_CAP_SUPPORTS_PROGRESSIVE:
296 return true;
297 case PIPE_VIDEO_CAP_MAX_LEVEL:
298 switch (profile) {
299 case PIPE_VIDEO_PROFILE_MPEG1:
300 return 0;
301 case PIPE_VIDEO_PROFILE_MPEG2_SIMPLE:
302 case PIPE_VIDEO_PROFILE_MPEG2_MAIN:
303 return 3;
304 case PIPE_VIDEO_PROFILE_MPEG4_SIMPLE:
305 return 3;
306 case PIPE_VIDEO_PROFILE_MPEG4_ADVANCED_SIMPLE:
307 return 5;
308 case PIPE_VIDEO_PROFILE_VC1_SIMPLE:
309 return 1;
310 case PIPE_VIDEO_PROFILE_VC1_MAIN:
311 return 2;
312 case PIPE_VIDEO_PROFILE_VC1_ADVANCED:
313 return 4;
314 case PIPE_VIDEO_PROFILE_MPEG4_AVC_BASELINE:
315 case PIPE_VIDEO_PROFILE_MPEG4_AVC_MAIN:
316 case PIPE_VIDEO_PROFILE_MPEG4_AVC_HIGH:
317 return (rscreen->family < CHIP_TONGA) ? 41 : 52;
318 case PIPE_VIDEO_PROFILE_HEVC_MAIN:
319 case PIPE_VIDEO_PROFILE_HEVC_MAIN_10:
320 return 186;
321 default:
322 return 0;
323 }
324 default:
325 return 0;
326 }
327 }
328
329 boolean rvid_is_format_supported(struct pipe_screen *screen,
330 enum pipe_format format,
331 enum pipe_video_profile profile,
332 enum pipe_video_entrypoint entrypoint)
333 {
334 /* we can only handle this one with UVD */
335 if (profile != PIPE_VIDEO_PROFILE_UNKNOWN)
336 return format == PIPE_FORMAT_NV12;
337
338 return vl_video_buffer_is_format_supported(screen, format, profile, entrypoint);
339 }