2 * Copyright © 2010 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the
6 * "Software"), to deal in the Software without restriction, including
7 * without limitation the rights to use, copy, modify, merge, publish,
8 * distribute, sub license, and/or sell copies of the Software, and to
9 * permit persons to whom the Software is furnished to do so, subject to
10 * the following conditions:
12 * The above copyright notice and this permission notice (including the
13 * next paragraph) shall be included in all copies or substantial portions
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
19 * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
20 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
21 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
22 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 * Xiang Haihao <haihao.xiang@intel.com>
34 #include "intel_batchbuffer.h"
35 #include "intel_driver.h"
36 #include "i965_defines.h"
37 #include "i965_structs.h"
38 #include "i965_drv_video.h"
39 #include "i965_post_processing.h"
40 #include "i965_render.h"
42 #define HAS_PP(ctx) (IS_IRONLAKE((ctx)->intel.device_id) || \
43 IS_GEN6((ctx)->intel.device_id) || \
44 IS_GEN7((ctx)->intel.device_id))
46 #define SURFACE_STATE_PADDED_SIZE_0_I965 ALIGN(sizeof(struct i965_surface_state), 32)
47 #define SURFACE_STATE_PADDED_SIZE_1_I965 ALIGN(sizeof(struct i965_surface_state2), 32)
48 #define SURFACE_STATE_PADDED_SIZE_I965 MAX(SURFACE_STATE_PADDED_SIZE_0_I965, SURFACE_STATE_PADDED_SIZE_1_I965)
50 #define SURFACE_STATE_PADDED_SIZE_0_GEN7 ALIGN(sizeof(struct gen7_surface_state), 32)
51 #define SURFACE_STATE_PADDED_SIZE_1_GEN7 ALIGN(sizeof(struct gen7_surface_state2), 32)
52 #define SURFACE_STATE_PADDED_SIZE_GEN7 MAX(SURFACE_STATE_PADDED_SIZE_0_GEN7, SURFACE_STATE_PADDED_SIZE_1_GEN7)
54 #define SURFACE_STATE_PADDED_SIZE MAX(SURFACE_STATE_PADDED_SIZE_I965, SURFACE_STATE_PADDED_SIZE_GEN7)
55 #define SURFACE_STATE_OFFSET(index) (SURFACE_STATE_PADDED_SIZE * index)
56 #define BINDING_TABLE_OFFSET SURFACE_STATE_OFFSET(MAX_PP_SURFACES)
58 static const uint32_t pp_null_gen5[][4] = {
59 #include "shaders/post_processing/gen5_6/null.g4b.gen5"
62 static const uint32_t pp_nv12_load_save_nv12_gen5[][4] = {
63 #include "shaders/post_processing/gen5_6/nv12_load_save_nv12.g4b.gen5"
66 static const uint32_t pp_nv12_load_save_pl3_gen5[][4] = {
67 #include "shaders/post_processing/gen5_6/nv12_load_save_pl3.g4b.gen5"
70 static const uint32_t pp_pl3_load_save_nv12_gen5[][4] = {
71 #include "shaders/post_processing/gen5_6/pl3_load_save_nv12.g4b.gen5"
74 static const uint32_t pp_pl3_load_save_pl3_gen5[][4] = {
75 #include "shaders/post_processing/gen5_6/pl3_load_save_pl3.g4b.gen5"
78 static const uint32_t pp_nv12_scaling_gen5[][4] = {
79 #include "shaders/post_processing/gen5_6/nv12_scaling_nv12.g4b.gen5"
82 static const uint32_t pp_nv12_avs_gen5[][4] = {
83 #include "shaders/post_processing/gen5_6/nv12_avs_nv12.g4b.gen5"
86 static const uint32_t pp_nv12_dndi_gen5[][4] = {
87 #include "shaders/post_processing/gen5_6/nv12_dndi_nv12.g4b.gen5"
90 static const uint32_t pp_nv12_dn_gen5[][4] = {
91 #include "shaders/post_processing/gen5_6/nv12_dn_nv12.g4b.gen5"
94 static const uint32_t pp_nv12_load_save_pa_gen5[][4] = {
95 #include "shaders/post_processing/gen5_6/nv12_load_save_pa.g4b.gen5"
98 static const uint32_t pp_pl3_load_save_pa_gen5[][4] = {
99 #include "shaders/post_processing/gen5_6/pl3_load_save_pa.g4b.gen5"
102 static const uint32_t pp_pa_load_save_nv12_gen5[][4] = {
103 #include "shaders/post_processing/gen5_6/pa_load_save_nv12.g4b.gen5"
106 static const uint32_t pp_pa_load_save_pl3_gen5[][4] = {
107 #include "shaders/post_processing/gen5_6/pa_load_save_pl3.g4b.gen5"
110 static const uint32_t pp_rgbx_load_save_nv12_gen5[][4] = {
111 #include "shaders/post_processing/gen5_6/rgbx_load_save_nv12.g4b.gen5"
114 static const uint32_t pp_nv12_load_save_rgbx_gen5[][4] = {
115 #include "shaders/post_processing/gen5_6/nv12_load_save_rgbx.g4b.gen5"
118 static VAStatus pp_null_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
119 const struct i965_surface *src_surface,
120 const VARectangle *src_rect,
121 struct i965_surface *dst_surface,
122 const VARectangle *dst_rect,
124 static VAStatus pp_nv12_avs_initialize_nlas(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
125 const struct i965_surface *src_surface,
126 const VARectangle *src_rect,
127 struct i965_surface *dst_surface,
128 const VARectangle *dst_rect,
130 static VAStatus pp_nv12_scaling_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
131 const struct i965_surface *src_surface,
132 const VARectangle *src_rect,
133 struct i965_surface *dst_surface,
134 const VARectangle *dst_rect,
136 static VAStatus gen6_nv12_scaling_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
137 const struct i965_surface *src_surface,
138 const VARectangle *src_rect,
139 struct i965_surface *dst_surface,
140 const VARectangle *dst_rect,
142 static VAStatus pp_plx_load_save_plx_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
143 const struct i965_surface *src_surface,
144 const VARectangle *src_rect,
145 struct i965_surface *dst_surface,
146 const VARectangle *dst_rect,
148 static VAStatus pp_nv12_dndi_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
149 const struct i965_surface *src_surface,
150 const VARectangle *src_rect,
151 struct i965_surface *dst_surface,
152 const VARectangle *dst_rect,
154 static VAStatus pp_nv12_dn_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
155 const struct i965_surface *src_surface,
156 const VARectangle *src_rect,
157 struct i965_surface *dst_surface,
158 const VARectangle *dst_rect,
161 static struct pp_module pp_modules_gen5[] = {
164 "NULL module (for testing)",
167 sizeof(pp_null_gen5),
177 PP_NV12_LOAD_SAVE_N12,
178 pp_nv12_load_save_nv12_gen5,
179 sizeof(pp_nv12_load_save_nv12_gen5),
183 pp_plx_load_save_plx_initialize,
189 PP_NV12_LOAD_SAVE_PL3,
190 pp_nv12_load_save_pl3_gen5,
191 sizeof(pp_nv12_load_save_pl3_gen5),
195 pp_plx_load_save_plx_initialize,
201 PP_PL3_LOAD_SAVE_N12,
202 pp_pl3_load_save_nv12_gen5,
203 sizeof(pp_pl3_load_save_nv12_gen5),
207 pp_plx_load_save_plx_initialize,
213 PP_PL3_LOAD_SAVE_N12,
214 pp_pl3_load_save_pl3_gen5,
215 sizeof(pp_pl3_load_save_pl3_gen5),
219 pp_plx_load_save_plx_initialize
224 "NV12 Scaling module",
226 pp_nv12_scaling_gen5,
227 sizeof(pp_nv12_scaling_gen5),
231 pp_nv12_scaling_initialize,
239 sizeof(pp_nv12_avs_gen5),
243 pp_nv12_avs_initialize_nlas,
251 sizeof(pp_nv12_dndi_gen5),
255 pp_nv12_dndi_initialize,
263 sizeof(pp_nv12_dn_gen5),
267 pp_nv12_dn_initialize,
273 PP_NV12_LOAD_SAVE_PA,
274 pp_nv12_load_save_pa_gen5,
275 sizeof(pp_nv12_load_save_pa_gen5),
279 pp_plx_load_save_plx_initialize,
286 pp_pl3_load_save_pa_gen5,
287 sizeof(pp_pl3_load_save_pa_gen5),
291 pp_plx_load_save_plx_initialize,
297 PP_PA_LOAD_SAVE_NV12,
298 pp_pa_load_save_nv12_gen5,
299 sizeof(pp_pa_load_save_nv12_gen5),
303 pp_plx_load_save_plx_initialize,
310 pp_pa_load_save_pl3_gen5,
311 sizeof(pp_pa_load_save_pl3_gen5),
315 pp_plx_load_save_plx_initialize,
321 PP_RGBX_LOAD_SAVE_NV12,
322 pp_rgbx_load_save_nv12_gen5,
323 sizeof(pp_rgbx_load_save_nv12_gen5),
327 pp_plx_load_save_plx_initialize,
333 PP_NV12_LOAD_SAVE_RGBX,
334 pp_nv12_load_save_rgbx_gen5,
335 sizeof(pp_nv12_load_save_rgbx_gen5),
339 pp_plx_load_save_plx_initialize,
344 static const uint32_t pp_null_gen6[][4] = {
345 #include "shaders/post_processing/gen5_6/null.g6b"
348 static const uint32_t pp_nv12_load_save_nv12_gen6[][4] = {
349 #include "shaders/post_processing/gen5_6/nv12_load_save_nv12.g6b"
352 static const uint32_t pp_nv12_load_save_pl3_gen6[][4] = {
353 #include "shaders/post_processing/gen5_6/nv12_load_save_pl3.g6b"
356 static const uint32_t pp_pl3_load_save_nv12_gen6[][4] = {
357 #include "shaders/post_processing/gen5_6/pl3_load_save_nv12.g6b"
360 static const uint32_t pp_pl3_load_save_pl3_gen6[][4] = {
361 #include "shaders/post_processing/gen5_6/pl3_load_save_pl3.g6b"
364 static const uint32_t pp_nv12_scaling_gen6[][4] = {
365 #include "shaders/post_processing/gen5_6/nv12_avs_nv12.g6b"
368 static const uint32_t pp_nv12_avs_gen6[][4] = {
369 #include "shaders/post_processing/gen5_6/nv12_avs_nv12.g6b"
372 static const uint32_t pp_nv12_dndi_gen6[][4] = {
373 #include "shaders/post_processing/gen5_6/nv12_dndi_nv12.g6b"
376 static const uint32_t pp_nv12_dn_gen6[][4] = {
377 #include "shaders/post_processing/gen5_6/nv12_dn_nv12.g6b"
380 static const uint32_t pp_nv12_load_save_pa_gen6[][4] = {
381 #include "shaders/post_processing/gen5_6/nv12_load_save_pa.g6b"
384 static const uint32_t pp_pl3_load_save_pa_gen6[][4] = {
385 #include "shaders/post_processing/gen5_6/pl3_load_save_pa.g6b"
388 static const uint32_t pp_pa_load_save_nv12_gen6[][4] = {
389 #include "shaders/post_processing/gen5_6/pa_load_save_nv12.g6b"
392 static const uint32_t pp_pa_load_save_pl3_gen6[][4] = {
393 #include "shaders/post_processing/gen5_6/pa_load_save_pl3.g6b"
396 static const uint32_t pp_rgbx_load_save_nv12_gen6[][4] = {
397 #include "shaders/post_processing/gen5_6/rgbx_load_save_nv12.g6b"
400 static const uint32_t pp_nv12_load_save_rgbx_gen6[][4] = {
401 #include "shaders/post_processing/gen5_6/nv12_load_save_rgbx.g6b"
404 static struct pp_module pp_modules_gen6[] = {
407 "NULL module (for testing)",
410 sizeof(pp_null_gen6),
420 PP_NV12_LOAD_SAVE_N12,
421 pp_nv12_load_save_nv12_gen6,
422 sizeof(pp_nv12_load_save_nv12_gen6),
426 pp_plx_load_save_plx_initialize,
432 PP_NV12_LOAD_SAVE_PL3,
433 pp_nv12_load_save_pl3_gen6,
434 sizeof(pp_nv12_load_save_pl3_gen6),
438 pp_plx_load_save_plx_initialize,
444 PP_PL3_LOAD_SAVE_N12,
445 pp_pl3_load_save_nv12_gen6,
446 sizeof(pp_pl3_load_save_nv12_gen6),
450 pp_plx_load_save_plx_initialize,
456 PP_PL3_LOAD_SAVE_N12,
457 pp_pl3_load_save_pl3_gen6,
458 sizeof(pp_pl3_load_save_pl3_gen6),
462 pp_plx_load_save_plx_initialize,
467 "NV12 Scaling module",
469 pp_nv12_scaling_gen6,
470 sizeof(pp_nv12_scaling_gen6),
474 gen6_nv12_scaling_initialize,
482 sizeof(pp_nv12_avs_gen6),
486 pp_nv12_avs_initialize_nlas,
494 sizeof(pp_nv12_dndi_gen6),
498 pp_nv12_dndi_initialize,
506 sizeof(pp_nv12_dn_gen6),
510 pp_nv12_dn_initialize,
515 PP_NV12_LOAD_SAVE_PA,
516 pp_nv12_load_save_pa_gen6,
517 sizeof(pp_nv12_load_save_pa_gen6),
521 pp_plx_load_save_plx_initialize,
528 pp_pl3_load_save_pa_gen6,
529 sizeof(pp_pl3_load_save_pa_gen6),
533 pp_plx_load_save_plx_initialize,
539 PP_PA_LOAD_SAVE_NV12,
540 pp_pa_load_save_nv12_gen6,
541 sizeof(pp_pa_load_save_nv12_gen6),
545 pp_plx_load_save_plx_initialize,
552 pp_pa_load_save_pl3_gen6,
553 sizeof(pp_pa_load_save_pl3_gen6),
557 pp_plx_load_save_plx_initialize,
563 PP_RGBX_LOAD_SAVE_NV12,
564 pp_rgbx_load_save_nv12_gen6,
565 sizeof(pp_rgbx_load_save_nv12_gen6),
569 pp_plx_load_save_plx_initialize,
575 PP_NV12_LOAD_SAVE_RGBX,
576 pp_nv12_load_save_rgbx_gen6,
577 sizeof(pp_nv12_load_save_rgbx_gen6),
581 pp_plx_load_save_plx_initialize,
585 static const uint32_t pp_null_gen7[][4] = {
588 static const uint32_t pp_nv12_load_save_nv12_gen7[][4] = {
589 #include "shaders/post_processing/gen7/pl2_to_pl2.g7b"
592 static const uint32_t pp_nv12_load_save_pl3_gen7[][4] = {
593 #include "shaders/post_processing/gen7/pl2_to_pl3.g7b"
596 static const uint32_t pp_pl3_load_save_nv12_gen7[][4] = {
597 #include "shaders/post_processing/gen7/pl3_to_pl2.g7b"
600 static const uint32_t pp_pl3_load_save_pl3_gen7[][4] = {
601 #include "shaders/post_processing/gen7/pl3_to_pl3.g7b"
604 static const uint32_t pp_nv12_scaling_gen7[][4] = {
605 #include "shaders/post_processing/gen7/avs.g7b"
608 static const uint32_t pp_nv12_avs_gen7[][4] = {
609 #include "shaders/post_processing/gen7/avs.g7b"
612 static const uint32_t pp_nv12_dndi_gen7[][4] = {
613 #include "shaders/post_processing/gen7/dndi.g7b"
616 static const uint32_t pp_nv12_dn_gen7[][4] = {
617 #include "shaders/post_processing/gen7/nv12_dn_nv12.g7b"
619 static const uint32_t pp_nv12_load_save_pa_gen7[][4] = {
620 #include "shaders/post_processing/gen7/pl2_to_pa.g7b"
622 static const uint32_t pp_pl3_load_save_pa_gen7[][4] = {
623 #include "shaders/post_processing/gen7/pl3_to_pa.g7b"
625 static const uint32_t pp_pa_load_save_nv12_gen7[][4] = {
626 #include "shaders/post_processing/gen7/pa_to_pl2.g7b"
628 static const uint32_t pp_pa_load_save_pl3_gen7[][4] = {
629 #include "shaders/post_processing/gen7/pa_to_pl3.g7b"
631 static const uint32_t pp_rgbx_load_save_nv12_gen7[][4] = {
633 static const uint32_t pp_nv12_load_save_rgbx_gen7[][4] = {
636 static VAStatus gen7_pp_plx_avs_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
637 const struct i965_surface *src_surface,
638 const VARectangle *src_rect,
639 struct i965_surface *dst_surface,
640 const VARectangle *dst_rect,
642 static VAStatus gen7_pp_nv12_dndi_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
643 const struct i965_surface *src_surface,
644 const VARectangle *src_rect,
645 struct i965_surface *dst_surface,
646 const VARectangle *dst_rect,
648 static VAStatus gen7_pp_nv12_dn_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
649 const struct i965_surface *src_surface,
650 const VARectangle *src_rect,
651 struct i965_surface *dst_surface,
652 const VARectangle *dst_rect,
655 static struct pp_module pp_modules_gen7[] = {
658 "NULL module (for testing)",
661 sizeof(pp_null_gen7),
671 PP_NV12_LOAD_SAVE_N12,
672 pp_nv12_load_save_nv12_gen7,
673 sizeof(pp_nv12_load_save_nv12_gen7),
677 gen7_pp_plx_avs_initialize,
683 PP_NV12_LOAD_SAVE_PL3,
684 pp_nv12_load_save_pl3_gen7,
685 sizeof(pp_nv12_load_save_pl3_gen7),
689 gen7_pp_plx_avs_initialize,
695 PP_PL3_LOAD_SAVE_N12,
696 pp_pl3_load_save_nv12_gen7,
697 sizeof(pp_pl3_load_save_nv12_gen7),
701 gen7_pp_plx_avs_initialize,
707 PP_PL3_LOAD_SAVE_N12,
708 pp_pl3_load_save_pl3_gen7,
709 sizeof(pp_pl3_load_save_pl3_gen7),
713 gen7_pp_plx_avs_initialize,
718 "NV12 Scaling module",
720 pp_nv12_scaling_gen7,
721 sizeof(pp_nv12_scaling_gen7),
725 gen7_pp_plx_avs_initialize,
733 sizeof(pp_nv12_avs_gen7),
737 gen7_pp_plx_avs_initialize,
745 sizeof(pp_nv12_dndi_gen7),
749 gen7_pp_nv12_dndi_initialize,
757 sizeof(pp_nv12_dn_gen7),
761 gen7_pp_nv12_dn_initialize,
766 PP_NV12_LOAD_SAVE_PA,
767 pp_nv12_load_save_pa_gen7,
768 sizeof(pp_nv12_load_save_pa_gen7),
772 gen7_pp_plx_avs_initialize,
779 pp_pl3_load_save_pa_gen7,
780 sizeof(pp_pl3_load_save_pa_gen7),
784 gen7_pp_plx_avs_initialize,
790 PP_PA_LOAD_SAVE_NV12,
791 pp_pa_load_save_nv12_gen7,
792 sizeof(pp_pa_load_save_nv12_gen7),
796 gen7_pp_plx_avs_initialize,
803 pp_pa_load_save_pl3_gen7,
804 sizeof(pp_pa_load_save_pl3_gen7),
808 gen7_pp_plx_avs_initialize,
814 PP_RGBX_LOAD_SAVE_NV12,
815 pp_rgbx_load_save_nv12_gen7,
816 sizeof(pp_rgbx_load_save_nv12_gen7),
820 pp_plx_load_save_plx_initialize,
826 PP_NV12_LOAD_SAVE_RGBX,
827 pp_nv12_load_save_rgbx_gen7,
828 sizeof(pp_nv12_load_save_rgbx_gen7),
832 pp_plx_load_save_plx_initialize,
838 pp_get_surface_fourcc(VADriverContextP ctx, const struct i965_surface *surface)
840 struct i965_driver_data *i965 = i965_driver_data(ctx);
843 if (surface->type == I965_SURFACE_TYPE_IMAGE) {
844 struct object_image *obj_image = IMAGE(surface->id);
845 fourcc = obj_image->image.format.fourcc;
847 struct object_surface *obj_surface = SURFACE(surface->id);
848 fourcc = obj_surface->fourcc;
855 pp_set_surface_tiling(struct i965_surface_state *ss, unsigned int tiling)
858 case I915_TILING_NONE:
859 ss->ss3.tiled_surface = 0;
860 ss->ss3.tile_walk = 0;
863 ss->ss3.tiled_surface = 1;
864 ss->ss3.tile_walk = I965_TILEWALK_XMAJOR;
867 ss->ss3.tiled_surface = 1;
868 ss->ss3.tile_walk = I965_TILEWALK_YMAJOR;
874 pp_set_surface2_tiling(struct i965_surface_state2 *ss, unsigned int tiling)
877 case I915_TILING_NONE:
878 ss->ss2.tiled_surface = 0;
879 ss->ss2.tile_walk = 0;
882 ss->ss2.tiled_surface = 1;
883 ss->ss2.tile_walk = I965_TILEWALK_XMAJOR;
886 ss->ss2.tiled_surface = 1;
887 ss->ss2.tile_walk = I965_TILEWALK_YMAJOR;
893 gen7_pp_set_surface_tiling(struct gen7_surface_state *ss, unsigned int tiling)
896 case I915_TILING_NONE:
897 ss->ss0.tiled_surface = 0;
898 ss->ss0.tile_walk = 0;
901 ss->ss0.tiled_surface = 1;
902 ss->ss0.tile_walk = I965_TILEWALK_XMAJOR;
905 ss->ss0.tiled_surface = 1;
906 ss->ss0.tile_walk = I965_TILEWALK_YMAJOR;
912 gen7_pp_set_surface2_tiling(struct gen7_surface_state2 *ss, unsigned int tiling)
915 case I915_TILING_NONE:
916 ss->ss2.tiled_surface = 0;
917 ss->ss2.tile_walk = 0;
920 ss->ss2.tiled_surface = 1;
921 ss->ss2.tile_walk = I965_TILEWALK_XMAJOR;
924 ss->ss2.tiled_surface = 1;
925 ss->ss2.tile_walk = I965_TILEWALK_YMAJOR;
931 ironlake_pp_interface_descriptor_table(struct i965_post_processing_context *pp_context)
933 struct i965_interface_descriptor *desc;
935 int pp_index = pp_context->current_pp;
937 bo = pp_context->idrt.bo;
941 memset(desc, 0, sizeof(*desc));
942 desc->desc0.grf_reg_blocks = 10;
943 desc->desc0.kernel_start_pointer = pp_context->pp_modules[pp_index].kernel.bo->offset >> 6; /* reloc */
944 desc->desc1.const_urb_entry_read_offset = 0;
945 desc->desc1.const_urb_entry_read_len = 4; /* grf 1-4 */
946 desc->desc2.sampler_state_pointer = pp_context->sampler_state_table.bo->offset >> 5;
947 desc->desc2.sampler_count = 0;
948 desc->desc3.binding_table_entry_count = 0;
949 desc->desc3.binding_table_pointer = (BINDING_TABLE_OFFSET >> 5);
951 dri_bo_emit_reloc(bo,
952 I915_GEM_DOMAIN_INSTRUCTION, 0,
953 desc->desc0.grf_reg_blocks,
954 offsetof(struct i965_interface_descriptor, desc0),
955 pp_context->pp_modules[pp_index].kernel.bo);
957 dri_bo_emit_reloc(bo,
958 I915_GEM_DOMAIN_INSTRUCTION, 0,
959 desc->desc2.sampler_count << 2,
960 offsetof(struct i965_interface_descriptor, desc2),
961 pp_context->sampler_state_table.bo);
964 pp_context->idrt.num_interface_descriptors++;
968 ironlake_pp_vfe_state(struct i965_post_processing_context *pp_context)
970 struct i965_vfe_state *vfe_state;
973 bo = pp_context->vfe_state.bo;
976 vfe_state = bo->virtual;
977 memset(vfe_state, 0, sizeof(*vfe_state));
978 vfe_state->vfe1.max_threads = pp_context->urb.num_vfe_entries - 1;
979 vfe_state->vfe1.urb_entry_alloc_size = pp_context->urb.size_vfe_entry - 1;
980 vfe_state->vfe1.num_urb_entries = pp_context->urb.num_vfe_entries;
981 vfe_state->vfe1.vfe_mode = VFE_GENERIC_MODE;
982 vfe_state->vfe1.children_present = 0;
983 vfe_state->vfe2.interface_descriptor_base =
984 pp_context->idrt.bo->offset >> 4; /* reloc */
985 dri_bo_emit_reloc(bo,
986 I915_GEM_DOMAIN_INSTRUCTION, 0,
988 offsetof(struct i965_vfe_state, vfe2),
989 pp_context->idrt.bo);
994 ironlake_pp_upload_constants(struct i965_post_processing_context *pp_context)
996 unsigned char *constant_buffer;
997 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
999 assert(sizeof(*pp_static_parameter) == 128);
1000 dri_bo_map(pp_context->curbe.bo, 1);
1001 assert(pp_context->curbe.bo->virtual);
1002 constant_buffer = pp_context->curbe.bo->virtual;
1003 memcpy(constant_buffer, pp_static_parameter, sizeof(*pp_static_parameter));
1004 dri_bo_unmap(pp_context->curbe.bo);
1008 ironlake_pp_states_setup(VADriverContextP ctx,
1009 struct i965_post_processing_context *pp_context)
1011 ironlake_pp_interface_descriptor_table(pp_context);
1012 ironlake_pp_vfe_state(pp_context);
1013 ironlake_pp_upload_constants(pp_context);
1017 ironlake_pp_pipeline_select(VADriverContextP ctx,
1018 struct i965_post_processing_context *pp_context)
1020 struct intel_batchbuffer *batch = pp_context->batch;
1022 BEGIN_BATCH(batch, 1);
1023 OUT_BATCH(batch, CMD_PIPELINE_SELECT | PIPELINE_SELECT_MEDIA);
1024 ADVANCE_BATCH(batch);
1028 ironlake_pp_urb_layout(VADriverContextP ctx,
1029 struct i965_post_processing_context *pp_context)
1031 struct intel_batchbuffer *batch = pp_context->batch;
1032 unsigned int vfe_fence, cs_fence;
1034 vfe_fence = pp_context->urb.cs_start;
1035 cs_fence = pp_context->urb.size;
1037 BEGIN_BATCH(batch, 3);
1038 OUT_BATCH(batch, CMD_URB_FENCE | UF0_VFE_REALLOC | UF0_CS_REALLOC | 1);
1039 OUT_BATCH(batch, 0);
1041 (vfe_fence << UF2_VFE_FENCE_SHIFT) | /* VFE_SIZE */
1042 (cs_fence << UF2_CS_FENCE_SHIFT)); /* CS_SIZE */
1043 ADVANCE_BATCH(batch);
1047 ironlake_pp_state_base_address(VADriverContextP ctx,
1048 struct i965_post_processing_context *pp_context)
1050 struct intel_batchbuffer *batch = pp_context->batch;
1052 BEGIN_BATCH(batch, 8);
1053 OUT_BATCH(batch, CMD_STATE_BASE_ADDRESS | 6);
1054 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1055 OUT_RELOC(batch, pp_context->surface_state_binding_table.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, BASE_ADDRESS_MODIFY);
1056 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1057 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1058 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1059 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1060 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
1061 ADVANCE_BATCH(batch);
1065 ironlake_pp_state_pointers(VADriverContextP ctx,
1066 struct i965_post_processing_context *pp_context)
1068 struct intel_batchbuffer *batch = pp_context->batch;
1070 BEGIN_BATCH(batch, 3);
1071 OUT_BATCH(batch, CMD_MEDIA_STATE_POINTERS | 1);
1072 OUT_BATCH(batch, 0);
1073 OUT_RELOC(batch, pp_context->vfe_state.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
1074 ADVANCE_BATCH(batch);
1078 ironlake_pp_cs_urb_layout(VADriverContextP ctx,
1079 struct i965_post_processing_context *pp_context)
1081 struct intel_batchbuffer *batch = pp_context->batch;
1083 BEGIN_BATCH(batch, 2);
1084 OUT_BATCH(batch, CMD_CS_URB_STATE | 0);
1086 ((pp_context->urb.size_cs_entry - 1) << 4) | /* URB Entry Allocation Size */
1087 (pp_context->urb.num_cs_entries << 0)); /* Number of URB Entries */
1088 ADVANCE_BATCH(batch);
1092 ironlake_pp_constant_buffer(VADriverContextP ctx,
1093 struct i965_post_processing_context *pp_context)
1095 struct intel_batchbuffer *batch = pp_context->batch;
1097 BEGIN_BATCH(batch, 2);
1098 OUT_BATCH(batch, CMD_CONSTANT_BUFFER | (1 << 8) | (2 - 2));
1099 OUT_RELOC(batch, pp_context->curbe.bo,
1100 I915_GEM_DOMAIN_INSTRUCTION, 0,
1101 pp_context->urb.size_cs_entry - 1);
1102 ADVANCE_BATCH(batch);
1106 ironlake_pp_object_walker(VADriverContextP ctx,
1107 struct i965_post_processing_context *pp_context)
1109 struct intel_batchbuffer *batch = pp_context->batch;
1110 int x, x_steps, y, y_steps;
1111 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1113 x_steps = pp_context->pp_x_steps(&pp_context->private_context);
1114 y_steps = pp_context->pp_y_steps(&pp_context->private_context);
1116 for (y = 0; y < y_steps; y++) {
1117 for (x = 0; x < x_steps; x++) {
1118 if (!pp_context->pp_set_block_parameter(pp_context, x, y)) {
1119 BEGIN_BATCH(batch, 20);
1120 OUT_BATCH(batch, CMD_MEDIA_OBJECT | 18);
1121 OUT_BATCH(batch, 0);
1122 OUT_BATCH(batch, 0); /* no indirect data */
1123 OUT_BATCH(batch, 0);
1125 /* inline data grf 5-6 */
1126 assert(sizeof(*pp_inline_parameter) == 64);
1127 intel_batchbuffer_data(batch, pp_inline_parameter, sizeof(*pp_inline_parameter));
1129 ADVANCE_BATCH(batch);
1136 ironlake_pp_pipeline_setup(VADriverContextP ctx,
1137 struct i965_post_processing_context *pp_context)
1139 struct intel_batchbuffer *batch = pp_context->batch;
1141 intel_batchbuffer_start_atomic(batch, 0x1000);
1142 intel_batchbuffer_emit_mi_flush(batch);
1143 ironlake_pp_pipeline_select(ctx, pp_context);
1144 ironlake_pp_state_base_address(ctx, pp_context);
1145 ironlake_pp_state_pointers(ctx, pp_context);
1146 ironlake_pp_urb_layout(ctx, pp_context);
1147 ironlake_pp_cs_urb_layout(ctx, pp_context);
1148 ironlake_pp_constant_buffer(ctx, pp_context);
1149 ironlake_pp_object_walker(ctx, pp_context);
1150 intel_batchbuffer_end_atomic(batch);
1153 // update u/v offset when the surface format are packed yuv
1154 static void i965_update_src_surface_static_parameter(
1155 VADriverContextP ctx,
1156 struct i965_post_processing_context *pp_context,
1157 const struct i965_surface *surface)
1159 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1160 int fourcc = pp_get_surface_fourcc(ctx, surface);
1163 case VA_FOURCC('Y', 'U', 'Y', '2'):
1164 pp_static_parameter->grf1.source_packed_u_offset = 1;
1165 pp_static_parameter->grf1.source_packed_v_offset = 3;
1167 case VA_FOURCC('U', 'Y', 'V', 'Y'):
1168 pp_static_parameter->grf1.source_packed_y_offset = 1;
1169 pp_static_parameter->grf1.source_packed_v_offset = 2;
1171 case VA_FOURCC('B', 'G', 'R', 'X'):
1172 case VA_FOURCC('B', 'G', 'R', 'A'):
1173 pp_static_parameter->grf1.source_rgb_layout = 0;
1175 case VA_FOURCC('R', 'G', 'B', 'X'):
1176 case VA_FOURCC('R', 'G', 'B', 'A'):
1177 pp_static_parameter->grf1.source_rgb_layout = 1;
1185 static void i965_update_dst_surface_static_parameter(
1186 VADriverContextP ctx,
1187 struct i965_post_processing_context *pp_context,
1188 const struct i965_surface *surface)
1190 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1191 int fourcc = pp_get_surface_fourcc(ctx, surface);
1194 case VA_FOURCC('Y', 'U', 'Y', '2'):
1195 pp_static_parameter->grf1.r1_2.load_and_save.destination_packed_u_offset = 1;
1196 pp_static_parameter->grf1.r1_2.load_and_save.destination_packed_v_offset = 3;
1198 case VA_FOURCC('U', 'Y', 'V', 'Y'):
1199 pp_static_parameter->grf1.r1_2.load_and_save.destination_packed_y_offset = 1;
1200 pp_static_parameter->grf1.r1_2.load_and_save.destination_packed_v_offset = 2;
1202 case VA_FOURCC('B', 'G', 'R', 'X'):
1203 case VA_FOURCC('B', 'G', 'R', 'A'):
1204 pp_static_parameter->grf1.r1_2.csc.destination_rgb_layout = 0;
1206 case VA_FOURCC('R', 'G', 'B', 'X'):
1207 case VA_FOURCC('R', 'G', 'B', 'A'):
1208 pp_static_parameter->grf1.r1_2.csc.destination_rgb_layout = 1;
1217 i965_pp_set_surface_state(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1218 dri_bo *surf_bo, unsigned long surf_bo_offset,
1219 int width, int height, int pitch, int format,
1220 int index, int is_target)
1222 struct i965_surface_state *ss;
1224 unsigned int tiling;
1225 unsigned int swizzle;
1227 dri_bo_get_tiling(surf_bo, &tiling, &swizzle);
1228 ss_bo = pp_context->surface_state_binding_table.bo;
1231 dri_bo_map(ss_bo, True);
1232 assert(ss_bo->virtual);
1233 ss = (struct i965_surface_state *)((char *)ss_bo->virtual + SURFACE_STATE_OFFSET(index));
1234 memset(ss, 0, sizeof(*ss));
1235 ss->ss0.surface_type = I965_SURFACE_2D;
1236 ss->ss0.surface_format = format;
1237 ss->ss1.base_addr = surf_bo->offset + surf_bo_offset;
1238 ss->ss2.width = width - 1;
1239 ss->ss2.height = height - 1;
1240 ss->ss3.pitch = pitch - 1;
1241 pp_set_surface_tiling(ss, tiling);
1242 dri_bo_emit_reloc(ss_bo,
1243 I915_GEM_DOMAIN_RENDER, is_target ? I915_GEM_DOMAIN_RENDER : 0,
1245 SURFACE_STATE_OFFSET(index) + offsetof(struct i965_surface_state, ss1),
1247 ((unsigned int *)((char *)ss_bo->virtual + BINDING_TABLE_OFFSET))[index] = SURFACE_STATE_OFFSET(index);
1248 dri_bo_unmap(ss_bo);
1252 i965_pp_set_surface2_state(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1253 dri_bo *surf_bo, unsigned long surf_bo_offset,
1254 int width, int height, int wpitch,
1255 int xoffset, int yoffset,
1256 int format, int interleave_chroma,
1259 struct i965_surface_state2 *ss2;
1261 unsigned int tiling;
1262 unsigned int swizzle;
1264 dri_bo_get_tiling(surf_bo, &tiling, &swizzle);
1265 ss2_bo = pp_context->surface_state_binding_table.bo;
1268 dri_bo_map(ss2_bo, True);
1269 assert(ss2_bo->virtual);
1270 ss2 = (struct i965_surface_state2 *)((char *)ss2_bo->virtual + SURFACE_STATE_OFFSET(index));
1271 memset(ss2, 0, sizeof(*ss2));
1272 ss2->ss0.surface_base_address = surf_bo->offset + surf_bo_offset;
1273 ss2->ss1.cbcr_pixel_offset_v_direction = 0;
1274 ss2->ss1.width = width - 1;
1275 ss2->ss1.height = height - 1;
1276 ss2->ss2.pitch = wpitch - 1;
1277 ss2->ss2.interleave_chroma = interleave_chroma;
1278 ss2->ss2.surface_format = format;
1279 ss2->ss3.x_offset_for_cb = xoffset;
1280 ss2->ss3.y_offset_for_cb = yoffset;
1281 pp_set_surface2_tiling(ss2, tiling);
1282 dri_bo_emit_reloc(ss2_bo,
1283 I915_GEM_DOMAIN_RENDER, 0,
1285 SURFACE_STATE_OFFSET(index) + offsetof(struct i965_surface_state2, ss0),
1287 ((unsigned int *)((char *)ss2_bo->virtual + BINDING_TABLE_OFFSET))[index] = SURFACE_STATE_OFFSET(index);
1288 dri_bo_unmap(ss2_bo);
1292 gen7_pp_set_surface_state(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1293 dri_bo *surf_bo, unsigned long surf_bo_offset,
1294 int width, int height, int pitch, int format,
1295 int index, int is_target)
1297 struct gen7_surface_state *ss;
1299 unsigned int tiling;
1300 unsigned int swizzle;
1302 dri_bo_get_tiling(surf_bo, &tiling, &swizzle);
1303 ss_bo = pp_context->surface_state_binding_table.bo;
1306 dri_bo_map(ss_bo, True);
1307 assert(ss_bo->virtual);
1308 ss = (struct gen7_surface_state *)((char *)ss_bo->virtual + SURFACE_STATE_OFFSET(index));
1309 memset(ss, 0, sizeof(*ss));
1310 ss->ss0.surface_type = I965_SURFACE_2D;
1311 ss->ss0.surface_format = format;
1312 ss->ss1.base_addr = surf_bo->offset + surf_bo_offset;
1313 ss->ss2.width = width - 1;
1314 ss->ss2.height = height - 1;
1315 ss->ss3.pitch = pitch - 1;
1316 gen7_pp_set_surface_tiling(ss, tiling);
1317 dri_bo_emit_reloc(ss_bo,
1318 I915_GEM_DOMAIN_RENDER, is_target ? I915_GEM_DOMAIN_RENDER : 0,
1320 SURFACE_STATE_OFFSET(index) + offsetof(struct gen7_surface_state, ss1),
1322 ((unsigned int *)((char *)ss_bo->virtual + BINDING_TABLE_OFFSET))[index] = SURFACE_STATE_OFFSET(index);
1323 dri_bo_unmap(ss_bo);
1327 gen7_pp_set_surface2_state(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1328 dri_bo *surf_bo, unsigned long surf_bo_offset,
1329 int width, int height, int wpitch,
1330 int xoffset, int yoffset,
1331 int format, int interleave_chroma,
1334 struct gen7_surface_state2 *ss2;
1336 unsigned int tiling;
1337 unsigned int swizzle;
1339 dri_bo_get_tiling(surf_bo, &tiling, &swizzle);
1340 ss2_bo = pp_context->surface_state_binding_table.bo;
1343 dri_bo_map(ss2_bo, True);
1344 assert(ss2_bo->virtual);
1345 ss2 = (struct gen7_surface_state2 *)((char *)ss2_bo->virtual + SURFACE_STATE_OFFSET(index));
1346 memset(ss2, 0, sizeof(*ss2));
1347 ss2->ss0.surface_base_address = surf_bo->offset + surf_bo_offset;
1348 ss2->ss1.cbcr_pixel_offset_v_direction = 0;
1349 ss2->ss1.width = width - 1;
1350 ss2->ss1.height = height - 1;
1351 ss2->ss2.pitch = wpitch - 1;
1352 ss2->ss2.interleave_chroma = interleave_chroma;
1353 ss2->ss2.surface_format = format;
1354 ss2->ss3.x_offset_for_cb = xoffset;
1355 ss2->ss3.y_offset_for_cb = yoffset;
1356 gen7_pp_set_surface2_tiling(ss2, tiling);
1357 dri_bo_emit_reloc(ss2_bo,
1358 I915_GEM_DOMAIN_RENDER, 0,
1360 SURFACE_STATE_OFFSET(index) + offsetof(struct gen7_surface_state2, ss0),
1362 ((unsigned int *)((char *)ss2_bo->virtual + BINDING_TABLE_OFFSET))[index] = SURFACE_STATE_OFFSET(index);
1363 dri_bo_unmap(ss2_bo);
1367 pp_set_media_rw_message_surface(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1368 const struct i965_surface *surface,
1369 int base_index, int is_target,
1370 int *width, int *height, int *pitch, int *offset)
1372 struct i965_driver_data *i965 = i965_driver_data(ctx);
1373 struct object_surface *obj_surface;
1374 struct object_image *obj_image;
1376 int fourcc = pp_get_surface_fourcc(ctx, surface);
1378 const int U = fourcc == VA_FOURCC('Y', 'V', '1', '2') ? 2 : 1;
1379 const int V = fourcc == VA_FOURCC('Y', 'V', '1', '2') ? 1 : 2;
1381 int interleaved_uv = fourcc == VA_FOURCC('N', 'V', '1', '2');
1382 int packed_yuv = (fourcc == VA_FOURCC('Y', 'U', 'Y', '2') || fourcc == VA_FOURCC('U', 'Y', 'V', 'Y'));
1383 int full_packed_format = (fourcc == VA_FOURCC('R', 'G', 'B', 'A') ||
1384 fourcc == VA_FOURCC('R', 'G', 'B', 'X') ||
1385 fourcc == VA_FOURCC('B', 'G', 'R', 'A') ||
1386 fourcc == VA_FOURCC('B', 'G', 'R', 'X'));
1387 int scale_factor_of_1st_plane_width_in_byte = 1;
1389 if (surface->type == I965_SURFACE_TYPE_SURFACE) {
1390 obj_surface = SURFACE(surface->id);
1391 bo = obj_surface->bo;
1392 width[0] = obj_surface->orig_width;
1393 height[0] = obj_surface->orig_height;
1394 pitch[0] = obj_surface->width;
1397 if (full_packed_format) {
1398 scale_factor_of_1st_plane_width_in_byte = 4;
1399 pitch[0] = obj_surface->width * 4;
1401 else if (packed_yuv ) {
1402 scale_factor_of_1st_plane_width_in_byte = 2;
1403 pitch[0] = obj_surface->width * 2;
1405 else if (interleaved_uv) {
1406 width[1] = obj_surface->orig_width;
1407 height[1] = obj_surface->orig_height / 2;
1408 pitch[1] = obj_surface->width;
1409 offset[1] = offset[0] + obj_surface->width * obj_surface->height;
1411 width[1] = obj_surface->orig_width / 2;
1412 height[1] = obj_surface->orig_height / 2;
1413 pitch[1] = obj_surface->width / 2;
1414 offset[1] = offset[0] + obj_surface->width * obj_surface->height;
1415 width[2] = obj_surface->orig_width / 2;
1416 height[2] = obj_surface->orig_height / 2;
1417 pitch[2] = obj_surface->width / 2;
1418 offset[2] = offset[1] + (obj_surface->width / 2) * (obj_surface->height / 2);
1421 obj_image = IMAGE(surface->id);
1423 width[0] = obj_image->image.width;
1424 height[0] = obj_image->image.height;
1425 pitch[0] = obj_image->image.pitches[0];
1426 offset[0] = obj_image->image.offsets[0];
1428 if (full_packed_format) {
1429 scale_factor_of_1st_plane_width_in_byte = 4;
1431 else if (packed_yuv ) {
1432 scale_factor_of_1st_plane_width_in_byte = 2;
1434 else if (interleaved_uv) {
1435 width[1] = obj_image->image.width;
1436 height[1] = obj_image->image.height / 2;
1437 pitch[1] = obj_image->image.pitches[1];
1438 offset[1] = obj_image->image.offsets[1];
1440 width[1] = obj_image->image.width / 2;
1441 height[1] = obj_image->image.height / 2;
1442 pitch[1] = obj_image->image.pitches[1];
1443 offset[1] = obj_image->image.offsets[1];
1444 width[2] = obj_image->image.width / 2;
1445 height[2] = obj_image->image.height / 2;
1446 pitch[2] = obj_image->image.pitches[2];
1447 offset[2] = obj_image->image.offsets[2];
1452 i965_pp_set_surface_state(ctx, pp_context,
1454 width[Y] *scale_factor_of_1st_plane_width_in_byte / 4, height[Y], pitch[Y], I965_SURFACEFORMAT_R8_UNORM,
1455 base_index, is_target);
1457 if (!packed_yuv && !full_packed_format) {
1458 if (interleaved_uv) {
1459 i965_pp_set_surface_state(ctx, pp_context,
1461 width[UV] / 4, height[UV], pitch[UV], I965_SURFACEFORMAT_R8_UNORM,
1462 base_index + 1, is_target);
1465 i965_pp_set_surface_state(ctx, pp_context,
1467 width[U] / 4, height[U], pitch[U], I965_SURFACEFORMAT_R8_UNORM,
1468 base_index + 1, is_target);
1471 i965_pp_set_surface_state(ctx, pp_context,
1473 width[V] / 4, height[V], pitch[V], I965_SURFACEFORMAT_R8_UNORM,
1474 base_index + 2, is_target);
1481 gen7_pp_set_media_rw_message_surface(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1482 const struct i965_surface *surface,
1483 int base_index, int is_target,
1484 int *width, int *height, int *pitch, int *offset)
1486 struct i965_driver_data *i965 = i965_driver_data(ctx);
1487 struct object_surface *obj_surface;
1488 struct object_image *obj_image;
1490 int fourcc = pp_get_surface_fourcc(ctx, surface);
1491 const int U = (fourcc == VA_FOURCC('Y', 'V', '1', '2') ||
1492 fourcc == VA_FOURCC('I', 'M', 'C', '1')) ? 2 : 1;
1493 const int V = (fourcc == VA_FOURCC('Y', 'V', '1', '2') ||
1494 fourcc == VA_FOURCC('I', 'M', 'C', '1')) ? 1 : 2;
1495 int interleaved_uv = fourcc == VA_FOURCC('N', 'V', '1', '2');
1496 int packed_yuv = (fourcc == VA_FOURCC('Y', 'U', 'Y', '2') || fourcc == VA_FOURCC('U', 'Y', 'V', 'Y'));
1498 if (surface->type == I965_SURFACE_TYPE_SURFACE) {
1499 obj_surface = SURFACE(surface->id);
1500 bo = obj_surface->bo;
1501 width[0] = obj_surface->orig_width;
1502 height[0] = obj_surface->orig_height;
1503 pitch[0] = obj_surface->width;
1508 width[0] = obj_surface->orig_width * 2; /* surface format is R8, so double the width */
1510 width[0] = obj_surface->orig_width; /* surface foramt is YCBCR, width is specified in units of pixels */
1512 pitch[0] = obj_surface->width * 2;
1515 width[1] = obj_surface->cb_cr_width;
1516 height[1] = obj_surface->cb_cr_height;
1517 pitch[1] = obj_surface->cb_cr_pitch;
1518 offset[1] = obj_surface->y_cb_offset * obj_surface->width;
1520 width[2] = obj_surface->cb_cr_width;
1521 height[2] = obj_surface->cb_cr_height;
1522 pitch[2] = obj_surface->cb_cr_pitch;
1523 offset[2] = obj_surface->y_cr_offset * obj_surface->width;
1525 obj_image = IMAGE(surface->id);
1527 width[0] = obj_image->image.width;
1528 height[0] = obj_image->image.height;
1529 pitch[0] = obj_image->image.pitches[0];
1530 offset[0] = obj_image->image.offsets[0];
1534 width[0] = obj_image->image.width * 2; /* surface format is R8, so double the width */
1536 width[0] = obj_image->image.width; /* surface foramt is YCBCR, width is specified in units of pixels */
1537 } else if (interleaved_uv) {
1538 width[1] = obj_image->image.width / 2;
1539 height[1] = obj_image->image.height / 2;
1540 pitch[1] = obj_image->image.pitches[1];
1541 offset[1] = obj_image->image.offsets[1];
1543 width[1] = obj_image->image.width / 2;
1544 height[1] = obj_image->image.height / 2;
1545 pitch[1] = obj_image->image.pitches[U];
1546 offset[1] = obj_image->image.offsets[U];
1547 width[2] = obj_image->image.width / 2;
1548 height[2] = obj_image->image.height / 2;
1549 pitch[2] = obj_image->image.pitches[V];
1550 offset[2] = obj_image->image.offsets[V];
1555 gen7_pp_set_surface_state(ctx, pp_context,
1557 width[0] / 4, height[0], pitch[0],
1558 I965_SURFACEFORMAT_R8_SINT,
1562 if (interleaved_uv) {
1563 gen7_pp_set_surface_state(ctx, pp_context,
1565 width[1] / 2, height[1], pitch[1],
1566 I965_SURFACEFORMAT_R8G8_SINT,
1569 gen7_pp_set_surface_state(ctx, pp_context,
1571 width[1] / 4, height[1], pitch[1],
1572 I965_SURFACEFORMAT_R8_SINT,
1574 gen7_pp_set_surface_state(ctx, pp_context,
1576 width[2] / 4, height[2], pitch[2],
1577 I965_SURFACEFORMAT_R8_SINT,
1582 int format0 = SURFACE_FORMAT_Y8_UNORM;
1585 case VA_FOURCC('Y', 'U', 'Y', '2'):
1586 format0 = SURFACE_FORMAT_YCRCB_NORMAL;
1589 case VA_FOURCC('U', 'Y', 'V', 'Y'):
1590 format0 = SURFACE_FORMAT_YCRCB_SWAPY;
1597 gen7_pp_set_surface2_state(ctx, pp_context,
1599 width[0], height[0], pitch[0],
1605 if (interleaved_uv) {
1606 gen7_pp_set_surface2_state(ctx, pp_context,
1608 width[1], height[1], pitch[1],
1610 SURFACE_FORMAT_R8B8_UNORM, 0,
1613 gen7_pp_set_surface2_state(ctx, pp_context,
1615 width[1], height[1], pitch[1],
1617 SURFACE_FORMAT_R8_UNORM, 0,
1619 gen7_pp_set_surface2_state(ctx, pp_context,
1621 width[2], height[2], pitch[2],
1623 SURFACE_FORMAT_R8_UNORM, 0,
1631 pp_null_x_steps(void *private_context)
1637 pp_null_y_steps(void *private_context)
1643 pp_null_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
1649 pp_null_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1650 const struct i965_surface *src_surface,
1651 const VARectangle *src_rect,
1652 struct i965_surface *dst_surface,
1653 const VARectangle *dst_rect,
1656 /* private function & data */
1657 pp_context->pp_x_steps = pp_null_x_steps;
1658 pp_context->pp_y_steps = pp_null_y_steps;
1659 pp_context->pp_set_block_parameter = pp_null_set_block_parameter;
1661 dst_surface->flags = src_surface->flags;
1663 return VA_STATUS_SUCCESS;
1667 pp_load_save_x_steps(void *private_context)
1673 pp_load_save_y_steps(void *private_context)
1675 struct pp_load_save_context *pp_load_save_context = private_context;
1677 return pp_load_save_context->dest_h / 8;
1681 pp_load_save_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
1683 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1685 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16;
1686 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8;
1691 static void calculate_boundary_block_mask(struct i965_post_processing_context *pp_context, int width, int height)
1693 #define BLOCK_WIDTH 16
1694 #define BLOCK_HEIGHT 8
1696 if (width%BLOCK_WIDTH){
1697 pp_context->block_horizontal_mask = (1 << (width%BLOCK_WIDTH)) - 1;
1700 pp_context->block_horizontal_mask = 0xffff;
1703 if (height%BLOCK_HEIGHT){
1704 pp_context->block_vertical_mask = (1 << (height%BLOCK_HEIGHT)) - 1;
1707 pp_context->block_vertical_mask = 0xff;
1712 pp_plx_load_save_plx_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1713 const struct i965_surface *src_surface,
1714 const VARectangle *src_rect,
1715 struct i965_surface *dst_surface,
1716 const VARectangle *dst_rect,
1719 struct pp_load_save_context *pp_load_save_context = (struct pp_load_save_context *)&pp_context->private_context;
1720 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1721 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1722 int width[3], height[3], pitch[3], offset[3];
1725 /* source surface */
1726 pp_set_media_rw_message_surface(ctx, pp_context, src_surface, 1, 0,
1727 width, height, pitch, offset);
1729 /* destination surface */
1730 pp_set_media_rw_message_surface(ctx, pp_context, dst_surface, 7, 1,
1731 width, height, pitch, offset);
1733 /* private function & data */
1734 pp_context->pp_x_steps = pp_load_save_x_steps;
1735 pp_context->pp_y_steps = pp_load_save_y_steps;
1736 pp_context->pp_set_block_parameter = pp_load_save_set_block_parameter;
1737 pp_load_save_context->dest_h = ALIGN(height[Y], 8);
1738 pp_load_save_context->dest_w = ALIGN(width[Y], 16);
1740 pp_inline_parameter->grf5.block_count_x = ALIGN(width[Y], 16) / 16; /* 1 x N */
1741 pp_inline_parameter->grf5.number_blocks = ALIGN(width[Y], 16) / 16;
1743 pp_static_parameter->grf3.horizontal_origin_offset = src_rect->x;
1744 pp_static_parameter->grf3.vertical_origin_offset = src_rect->y;
1746 // update u/v offset for packed yuv
1747 i965_update_src_surface_static_parameter (ctx, pp_context, src_surface);
1748 i965_update_dst_surface_static_parameter (ctx, pp_context, dst_surface);
1750 dst_surface->flags = src_surface->flags;
1752 return VA_STATUS_SUCCESS;
1756 pp_scaling_x_steps(void *private_context)
1762 pp_scaling_y_steps(void *private_context)
1764 struct pp_scaling_context *pp_scaling_context = private_context;
1766 return pp_scaling_context->dest_h / 8;
1770 pp_scaling_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
1772 struct pp_scaling_context *pp_scaling_context = (struct pp_scaling_context *)&pp_context->private_context;
1773 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1774 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1775 float src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1776 float src_y_steping = pp_static_parameter->grf1.r1_6.normalized_video_y_scaling_step;
1778 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin = src_x_steping * x * 16 + pp_scaling_context->src_normalized_x;
1779 pp_inline_parameter->grf5.source_surface_block_normalized_vertical_origin = src_y_steping * y * 8 + pp_scaling_context->src_normalized_y;
1780 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16 + pp_scaling_context->dest_x;
1781 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8 + pp_scaling_context->dest_y;
1787 pp_nv12_scaling_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
1788 const struct i965_surface *src_surface,
1789 const VARectangle *src_rect,
1790 struct i965_surface *dst_surface,
1791 const VARectangle *dst_rect,
1794 struct i965_driver_data *i965 = i965_driver_data(ctx);
1795 struct pp_scaling_context *pp_scaling_context = (struct pp_scaling_context *)&pp_context->private_context;
1796 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1797 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1798 struct object_surface *obj_surface;
1799 struct i965_sampler_state *sampler_state;
1800 int in_w, in_h, in_wpitch, in_hpitch;
1801 int out_w, out_h, out_wpitch, out_hpitch;
1803 /* source surface */
1804 obj_surface = SURFACE(src_surface->id);
1805 in_w = obj_surface->orig_width;
1806 in_h = obj_surface->orig_height;
1807 in_wpitch = obj_surface->width;
1808 in_hpitch = obj_surface->height;
1810 /* source Y surface index 1 */
1811 i965_pp_set_surface_state(ctx, pp_context,
1813 in_w, in_h, in_wpitch, I965_SURFACEFORMAT_R8_UNORM,
1816 /* source UV surface index 2 */
1817 i965_pp_set_surface_state(ctx, pp_context,
1818 obj_surface->bo, in_wpitch * in_hpitch,
1819 in_w / 2, in_h / 2, in_wpitch, I965_SURFACEFORMAT_R8G8_UNORM,
1822 /* destination surface */
1823 obj_surface = SURFACE(dst_surface->id);
1824 out_w = obj_surface->orig_width;
1825 out_h = obj_surface->orig_height;
1826 out_wpitch = obj_surface->width;
1827 out_hpitch = obj_surface->height;
1829 /* destination Y surface index 7 */
1830 i965_pp_set_surface_state(ctx, pp_context,
1832 out_w / 4, out_h, out_wpitch, I965_SURFACEFORMAT_R8_UNORM,
1835 /* destination UV surface index 8 */
1836 i965_pp_set_surface_state(ctx, pp_context,
1837 obj_surface->bo, out_wpitch * out_hpitch,
1838 out_w / 4, out_h / 2, out_wpitch, I965_SURFACEFORMAT_R8G8_UNORM,
1842 dri_bo_map(pp_context->sampler_state_table.bo, True);
1843 assert(pp_context->sampler_state_table.bo->virtual);
1844 sampler_state = pp_context->sampler_state_table.bo->virtual;
1846 /* SIMD16 Y index 1 */
1847 sampler_state[1].ss0.min_filter = I965_MAPFILTER_LINEAR;
1848 sampler_state[1].ss0.mag_filter = I965_MAPFILTER_LINEAR;
1849 sampler_state[1].ss1.r_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1850 sampler_state[1].ss1.s_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1851 sampler_state[1].ss1.t_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1853 /* SIMD16 UV index 2 */
1854 sampler_state[2].ss0.min_filter = I965_MAPFILTER_LINEAR;
1855 sampler_state[2].ss0.mag_filter = I965_MAPFILTER_LINEAR;
1856 sampler_state[2].ss1.r_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1857 sampler_state[2].ss1.s_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1858 sampler_state[2].ss1.t_wrap_mode = I965_TEXCOORDMODE_CLAMP;
1860 dri_bo_unmap(pp_context->sampler_state_table.bo);
1862 /* private function & data */
1863 pp_context->pp_x_steps = pp_scaling_x_steps;
1864 pp_context->pp_y_steps = pp_scaling_y_steps;
1865 pp_context->pp_set_block_parameter = pp_scaling_set_block_parameter;
1867 pp_scaling_context->dest_x = dst_rect->x;
1868 pp_scaling_context->dest_y = dst_rect->y;
1869 pp_scaling_context->dest_w = ALIGN(dst_rect->width, 16);
1870 pp_scaling_context->dest_h = ALIGN(dst_rect->height, 16);
1871 pp_scaling_context->src_normalized_x = (float)src_rect->x / in_w;
1872 pp_scaling_context->src_normalized_y = (float)src_rect->y / in_h;
1874 pp_static_parameter->grf1.r1_6.normalized_video_y_scaling_step = (float) src_rect->height / in_h / dst_rect->height;
1876 pp_inline_parameter->grf5.normalized_video_x_scaling_step = (float) src_rect->width / in_w / dst_rect->width;
1877 pp_inline_parameter->grf5.block_count_x = pp_scaling_context->dest_w / 16; /* 1 x N */
1878 pp_inline_parameter->grf5.number_blocks = pp_scaling_context->dest_w / 16;
1880 dst_surface->flags = src_surface->flags;
1882 return VA_STATUS_SUCCESS;
1886 pp_avs_x_steps(void *private_context)
1888 struct pp_avs_context *pp_avs_context = private_context;
1890 return pp_avs_context->dest_w / 16;
1894 pp_avs_y_steps(void *private_context)
1900 pp_avs_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
1902 struct pp_avs_context *pp_avs_context = (struct pp_avs_context *)&pp_context->private_context;
1903 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
1904 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
1905 float src_x_steping, src_y_steping, video_step_delta;
1906 int tmp_w = ALIGN(pp_avs_context->dest_h * pp_avs_context->src_w / pp_avs_context->src_h, 16);
1908 if (pp_static_parameter->grf4.r4_2.avs.nlas == 0) {
1909 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1910 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin = src_x_steping * x * 16 + pp_avs_context->src_normalized_x;
1911 } else if (tmp_w >= pp_avs_context->dest_w) {
1912 pp_inline_parameter->grf5.normalized_video_x_scaling_step = 1.0 / tmp_w;
1913 pp_inline_parameter->grf6.video_step_delta = 0;
1916 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin = (float)(tmp_w - pp_avs_context->dest_w) / tmp_w / 2 +
1917 pp_avs_context->src_normalized_x;
1919 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1920 video_step_delta = pp_inline_parameter->grf6.video_step_delta;
1921 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin += src_x_steping * 16 +
1922 16 * 15 * video_step_delta / 2;
1925 int n0, n1, n2, nls_left, nls_right;
1926 int factor_a = 5, factor_b = 4;
1929 n0 = (pp_avs_context->dest_w - tmp_w) / (16 * 2);
1930 n1 = (pp_avs_context->dest_w - tmp_w) / 16 - n0;
1931 n2 = tmp_w / (16 * factor_a);
1933 nls_right = n1 + n2;
1934 f = (float) n2 * 16 / tmp_w;
1937 pp_inline_parameter->grf6.video_step_delta = 0.0;
1940 pp_inline_parameter->grf5.normalized_video_x_scaling_step = 1.0 / pp_avs_context->dest_w;
1941 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin = pp_avs_context->src_normalized_x;
1943 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1944 video_step_delta = pp_inline_parameter->grf6.video_step_delta;
1945 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin += src_x_steping * 16 +
1946 16 * 15 * video_step_delta / 2;
1950 /* f = a * nls_left * 16 + b * nls_left * 16 * (nls_left * 16 - 1) / 2 */
1951 float a = f / (nls_left * 16 * factor_b);
1952 float b = (f - nls_left * 16 * a) * 2 / (nls_left * 16 * (nls_left * 16 - 1));
1954 pp_inline_parameter->grf6.video_step_delta = b;
1957 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin = pp_avs_context->src_normalized_x;
1958 pp_inline_parameter->grf5.normalized_video_x_scaling_step = a;
1960 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1961 video_step_delta = pp_inline_parameter->grf6.video_step_delta;
1962 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin += src_x_steping * 16 +
1963 16 * 15 * video_step_delta / 2;
1964 pp_inline_parameter->grf5.normalized_video_x_scaling_step += 16 * b;
1966 } else if (x < (pp_avs_context->dest_w / 16 - nls_right)) {
1967 /* scale the center linearly */
1968 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1969 video_step_delta = pp_inline_parameter->grf6.video_step_delta;
1970 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin += src_x_steping * 16 +
1971 16 * 15 * video_step_delta / 2;
1972 pp_inline_parameter->grf6.video_step_delta = 0.0;
1973 pp_inline_parameter->grf5.normalized_video_x_scaling_step = 1.0 / tmp_w;
1975 float a = f / (nls_right * 16 * factor_b);
1976 float b = (f - nls_right * 16 * a) * 2 / (nls_right * 16 * (nls_right * 16 - 1));
1978 src_x_steping = pp_inline_parameter->grf5.normalized_video_x_scaling_step;
1979 video_step_delta = pp_inline_parameter->grf6.video_step_delta;
1980 pp_inline_parameter->grf5.r5_1.source_surface_block_normalized_horizontal_origin += src_x_steping * 16 +
1981 16 * 15 * video_step_delta / 2;
1982 pp_inline_parameter->grf6.video_step_delta = -b;
1984 if (x == (pp_avs_context->dest_w / 16 - nls_right))
1985 pp_inline_parameter->grf5.normalized_video_x_scaling_step = a + (nls_right * 16 - 1) * b;
1987 pp_inline_parameter->grf5.normalized_video_x_scaling_step -= b * 16;
1992 src_y_steping = pp_static_parameter->grf1.r1_6.normalized_video_y_scaling_step;
1993 pp_inline_parameter->grf5.source_surface_block_normalized_vertical_origin = src_y_steping * y * 8 + pp_avs_context->src_normalized_y;
1994 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16 + pp_avs_context->dest_x;
1995 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8 + pp_avs_context->dest_y;
2001 pp_nv12_avs_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2002 const struct i965_surface *src_surface,
2003 const VARectangle *src_rect,
2004 struct i965_surface *dst_surface,
2005 const VARectangle *dst_rect,
2009 struct i965_driver_data *i965 = i965_driver_data(ctx);
2010 struct pp_avs_context *pp_avs_context = (struct pp_avs_context *)&pp_context->private_context;
2011 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2012 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2013 struct object_surface *obj_surface;
2014 struct i965_sampler_8x8 *sampler_8x8;
2015 struct i965_sampler_8x8_state *sampler_8x8_state;
2017 int in_w, in_h, in_wpitch, in_hpitch;
2018 int out_w, out_h, out_wpitch, out_hpitch;
2022 obj_surface = SURFACE(src_surface->id);
2023 in_w = obj_surface->orig_width;
2024 in_h = obj_surface->orig_height;
2025 in_wpitch = obj_surface->width;
2026 in_hpitch = obj_surface->height;
2028 /* source Y surface index 1 */
2029 i965_pp_set_surface2_state(ctx, pp_context,
2031 in_w, in_h, in_wpitch,
2033 SURFACE_FORMAT_Y8_UNORM, 0,
2036 /* source UV surface index 2 */
2037 i965_pp_set_surface2_state(ctx, pp_context,
2038 obj_surface->bo, in_wpitch * in_hpitch,
2039 in_w / 2, in_h / 2, in_wpitch,
2041 SURFACE_FORMAT_R8B8_UNORM, 0,
2044 /* destination surface */
2045 obj_surface = SURFACE(dst_surface->id);
2046 out_w = obj_surface->orig_width;
2047 out_h = obj_surface->orig_height;
2048 out_wpitch = obj_surface->width;
2049 out_hpitch = obj_surface->height;
2050 assert(out_w <= out_wpitch && out_h <= out_hpitch);
2052 /* destination Y surface index 7 */
2053 i965_pp_set_surface_state(ctx, pp_context,
2055 out_w / 4, out_h, out_wpitch, I965_SURFACEFORMAT_R8_UNORM,
2058 /* destination UV surface index 8 */
2059 i965_pp_set_surface_state(ctx, pp_context,
2060 obj_surface->bo, out_wpitch * out_hpitch,
2061 out_w / 4, out_h / 2, out_wpitch, I965_SURFACEFORMAT_R8G8_UNORM,
2064 /* sampler 8x8 state */
2065 dri_bo_map(pp_context->sampler_state_table.bo_8x8, True);
2066 assert(pp_context->sampler_state_table.bo_8x8->virtual);
2067 assert(sizeof(*sampler_8x8_state) == sizeof(int) * 138);
2068 sampler_8x8_state = pp_context->sampler_state_table.bo_8x8->virtual;
2069 memset(sampler_8x8_state, 0, sizeof(*sampler_8x8_state));
2071 for (i = 0; i < 17; i++) {
2072 /* for Y channel, currently ignore */
2073 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c0 = 0x00;
2074 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c1 = 0x00;
2075 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c2 = 0x08;
2076 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c3 = 0x18;
2077 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c4 = 0x18;
2078 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c5 = 0x08;
2079 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c6 = 0x00;
2080 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c7 = 0x00;
2081 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c0 = 0x00;
2082 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c1 = 0x00;
2083 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c2 = 0x10;
2084 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c3 = 0x10;
2085 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c4 = 0x10;
2086 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c5 = 0x10;
2087 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c6 = 0x00;
2088 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c7 = 0x00;
2089 /* for U/V channel, 0.25 */
2090 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c0 = 0x0;
2091 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c1 = 0x0;
2092 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c2 = 0x10;
2093 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c3 = 0x10;
2094 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c4 = 0x10;
2095 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c5 = 0x10;
2096 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c6 = 0x0;
2097 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c7 = 0x0;
2098 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c0 = 0x0;
2099 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c1 = 0x0;
2100 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c2 = 0x10;
2101 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c3 = 0x10;
2102 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c4 = 0x10;
2103 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c5 = 0x10;
2104 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c6 = 0x0;
2105 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c7 = 0x0;
2108 sampler_8x8_state->dw136.default_sharpness_level = 0;
2109 sampler_8x8_state->dw137.adaptive_filter_for_all_channel = 1;
2110 sampler_8x8_state->dw137.bypass_y_adaptive_filtering = 1;
2111 sampler_8x8_state->dw137.bypass_x_adaptive_filtering = 1;
2112 dri_bo_unmap(pp_context->sampler_state_table.bo_8x8);
2115 dri_bo_map(pp_context->sampler_state_table.bo, True);
2116 assert(pp_context->sampler_state_table.bo->virtual);
2117 assert(sizeof(*sampler_8x8) == sizeof(int) * 16);
2118 sampler_8x8 = pp_context->sampler_state_table.bo->virtual;
2120 /* sample_8x8 Y index 1 */
2122 memset(&sampler_8x8[index], 0, sizeof(*sampler_8x8));
2123 sampler_8x8[index].dw0.avs_filter_type = AVS_FILTER_ADAPTIVE_8_TAP;
2124 sampler_8x8[index].dw0.ief_bypass = 1;
2125 sampler_8x8[index].dw0.ief_filter_type = IEF_FILTER_DETAIL;
2126 sampler_8x8[index].dw0.ief_filter_size = IEF_FILTER_SIZE_5X5;
2127 sampler_8x8[index].dw1.sampler_8x8_state_pointer = pp_context->sampler_state_table.bo_8x8->offset >> 5;
2128 sampler_8x8[index].dw2.global_noise_estimation = 22;
2129 sampler_8x8[index].dw2.strong_edge_threshold = 8;
2130 sampler_8x8[index].dw2.weak_edge_threshold = 1;
2131 sampler_8x8[index].dw3.strong_edge_weight = 7;
2132 sampler_8x8[index].dw3.regular_weight = 2;
2133 sampler_8x8[index].dw3.non_edge_weight = 0;
2134 sampler_8x8[index].dw3.gain_factor = 40;
2135 sampler_8x8[index].dw4.steepness_boost = 0;
2136 sampler_8x8[index].dw4.steepness_threshold = 0;
2137 sampler_8x8[index].dw4.mr_boost = 0;
2138 sampler_8x8[index].dw4.mr_threshold = 5;
2139 sampler_8x8[index].dw5.pwl1_point_1 = 4;
2140 sampler_8x8[index].dw5.pwl1_point_2 = 12;
2141 sampler_8x8[index].dw5.pwl1_point_3 = 16;
2142 sampler_8x8[index].dw5.pwl1_point_4 = 26;
2143 sampler_8x8[index].dw6.pwl1_point_5 = 40;
2144 sampler_8x8[index].dw6.pwl1_point_6 = 160;
2145 sampler_8x8[index].dw6.pwl1_r3_bias_0 = 127;
2146 sampler_8x8[index].dw6.pwl1_r3_bias_1 = 98;
2147 sampler_8x8[index].dw7.pwl1_r3_bias_2 = 88;
2148 sampler_8x8[index].dw7.pwl1_r3_bias_3 = 64;
2149 sampler_8x8[index].dw7.pwl1_r3_bias_4 = 44;
2150 sampler_8x8[index].dw7.pwl1_r3_bias_5 = 0;
2151 sampler_8x8[index].dw8.pwl1_r3_bias_6 = 0;
2152 sampler_8x8[index].dw8.pwl1_r5_bias_0 = 3;
2153 sampler_8x8[index].dw8.pwl1_r5_bias_1 = 32;
2154 sampler_8x8[index].dw8.pwl1_r5_bias_2 = 32;
2155 sampler_8x8[index].dw9.pwl1_r5_bias_3 = 58;
2156 sampler_8x8[index].dw9.pwl1_r5_bias_4 = 100;
2157 sampler_8x8[index].dw9.pwl1_r5_bias_5 = 108;
2158 sampler_8x8[index].dw9.pwl1_r5_bias_6 = 88;
2159 sampler_8x8[index].dw10.pwl1_r3_slope_0 = -116;
2160 sampler_8x8[index].dw10.pwl1_r3_slope_1 = -20;
2161 sampler_8x8[index].dw10.pwl1_r3_slope_2 = -96;
2162 sampler_8x8[index].dw10.pwl1_r3_slope_3 = -32;
2163 sampler_8x8[index].dw11.pwl1_r3_slope_4 = -50;
2164 sampler_8x8[index].dw11.pwl1_r3_slope_5 = 0;
2165 sampler_8x8[index].dw11.pwl1_r3_slope_6 = 0;
2166 sampler_8x8[index].dw11.pwl1_r5_slope_0 = 116;
2167 sampler_8x8[index].dw12.pwl1_r5_slope_1 = 0;
2168 sampler_8x8[index].dw12.pwl1_r5_slope_2 = 114;
2169 sampler_8x8[index].dw12.pwl1_r5_slope_3 = 67;
2170 sampler_8x8[index].dw12.pwl1_r5_slope_4 = 9;
2171 sampler_8x8[index].dw13.pwl1_r5_slope_5 = -3;
2172 sampler_8x8[index].dw13.pwl1_r5_slope_6 = -15;
2173 sampler_8x8[index].dw13.limiter_boost = 0;
2174 sampler_8x8[index].dw13.minimum_limiter = 10;
2175 sampler_8x8[index].dw13.maximum_limiter = 11;
2176 sampler_8x8[index].dw14.clip_limiter = 130;
2177 dri_bo_emit_reloc(pp_context->sampler_state_table.bo,
2178 I915_GEM_DOMAIN_RENDER,
2181 sizeof(*sampler_8x8) * index + offsetof(struct i965_sampler_8x8, dw1),
2182 pp_context->sampler_state_table.bo_8x8);
2184 /* sample_8x8 UV index 2 */
2186 memset(&sampler_8x8[index], 0, sizeof(*sampler_8x8));
2187 sampler_8x8[index].dw0.avs_filter_type = AVS_FILTER_ADAPTIVE_8_TAP;
2188 sampler_8x8[index].dw0.ief_bypass = 1;
2189 sampler_8x8[index].dw0.ief_filter_type = IEF_FILTER_DETAIL;
2190 sampler_8x8[index].dw0.ief_filter_size = IEF_FILTER_SIZE_5X5;
2191 sampler_8x8[index].dw1.sampler_8x8_state_pointer = pp_context->sampler_state_table.bo_8x8->offset >> 5;
2192 sampler_8x8[index].dw2.global_noise_estimation = 22;
2193 sampler_8x8[index].dw2.strong_edge_threshold = 8;
2194 sampler_8x8[index].dw2.weak_edge_threshold = 1;
2195 sampler_8x8[index].dw3.strong_edge_weight = 7;
2196 sampler_8x8[index].dw3.regular_weight = 2;
2197 sampler_8x8[index].dw3.non_edge_weight = 0;
2198 sampler_8x8[index].dw3.gain_factor = 40;
2199 sampler_8x8[index].dw4.steepness_boost = 0;
2200 sampler_8x8[index].dw4.steepness_threshold = 0;
2201 sampler_8x8[index].dw4.mr_boost = 0;
2202 sampler_8x8[index].dw4.mr_threshold = 5;
2203 sampler_8x8[index].dw5.pwl1_point_1 = 4;
2204 sampler_8x8[index].dw5.pwl1_point_2 = 12;
2205 sampler_8x8[index].dw5.pwl1_point_3 = 16;
2206 sampler_8x8[index].dw5.pwl1_point_4 = 26;
2207 sampler_8x8[index].dw6.pwl1_point_5 = 40;
2208 sampler_8x8[index].dw6.pwl1_point_6 = 160;
2209 sampler_8x8[index].dw6.pwl1_r3_bias_0 = 127;
2210 sampler_8x8[index].dw6.pwl1_r3_bias_1 = 98;
2211 sampler_8x8[index].dw7.pwl1_r3_bias_2 = 88;
2212 sampler_8x8[index].dw7.pwl1_r3_bias_3 = 64;
2213 sampler_8x8[index].dw7.pwl1_r3_bias_4 = 44;
2214 sampler_8x8[index].dw7.pwl1_r3_bias_5 = 0;
2215 sampler_8x8[index].dw8.pwl1_r3_bias_6 = 0;
2216 sampler_8x8[index].dw8.pwl1_r5_bias_0 = 3;
2217 sampler_8x8[index].dw8.pwl1_r5_bias_1 = 32;
2218 sampler_8x8[index].dw8.pwl1_r5_bias_2 = 32;
2219 sampler_8x8[index].dw9.pwl1_r5_bias_3 = 58;
2220 sampler_8x8[index].dw9.pwl1_r5_bias_4 = 100;
2221 sampler_8x8[index].dw9.pwl1_r5_bias_5 = 108;
2222 sampler_8x8[index].dw9.pwl1_r5_bias_6 = 88;
2223 sampler_8x8[index].dw10.pwl1_r3_slope_0 = -116;
2224 sampler_8x8[index].dw10.pwl1_r3_slope_1 = -20;
2225 sampler_8x8[index].dw10.pwl1_r3_slope_2 = -96;
2226 sampler_8x8[index].dw10.pwl1_r3_slope_3 = -32;
2227 sampler_8x8[index].dw11.pwl1_r3_slope_4 = -50;
2228 sampler_8x8[index].dw11.pwl1_r3_slope_5 = 0;
2229 sampler_8x8[index].dw11.pwl1_r3_slope_6 = 0;
2230 sampler_8x8[index].dw11.pwl1_r5_slope_0 = 116;
2231 sampler_8x8[index].dw12.pwl1_r5_slope_1 = 0;
2232 sampler_8x8[index].dw12.pwl1_r5_slope_2 = 114;
2233 sampler_8x8[index].dw12.pwl1_r5_slope_3 = 67;
2234 sampler_8x8[index].dw12.pwl1_r5_slope_4 = 9;
2235 sampler_8x8[index].dw13.pwl1_r5_slope_5 = -3;
2236 sampler_8x8[index].dw13.pwl1_r5_slope_6 = -15;
2237 sampler_8x8[index].dw13.limiter_boost = 0;
2238 sampler_8x8[index].dw13.minimum_limiter = 10;
2239 sampler_8x8[index].dw13.maximum_limiter = 11;
2240 sampler_8x8[index].dw14.clip_limiter = 130;
2241 dri_bo_emit_reloc(pp_context->sampler_state_table.bo,
2242 I915_GEM_DOMAIN_RENDER,
2245 sizeof(*sampler_8x8) * index + offsetof(struct i965_sampler_8x8, dw1),
2246 pp_context->sampler_state_table.bo_8x8);
2248 dri_bo_unmap(pp_context->sampler_state_table.bo);
2250 /* private function & data */
2251 pp_context->pp_x_steps = pp_avs_x_steps;
2252 pp_context->pp_y_steps = pp_avs_y_steps;
2253 pp_context->pp_set_block_parameter = pp_avs_set_block_parameter;
2255 pp_avs_context->dest_x = dst_rect->x;
2256 pp_avs_context->dest_y = dst_rect->y;
2257 pp_avs_context->dest_w = ALIGN(dst_rect->width, 16);
2258 pp_avs_context->dest_h = ALIGN(dst_rect->height, 8);
2259 pp_avs_context->src_normalized_x = (float)src_rect->x / in_w;
2260 pp_avs_context->src_normalized_y = (float)src_rect->y / in_h;
2261 pp_avs_context->src_w = src_rect->width;
2262 pp_avs_context->src_h = src_rect->height;
2264 pp_static_parameter->grf4.r4_2.avs.nlas = nlas;
2265 pp_static_parameter->grf1.r1_6.normalized_video_y_scaling_step = (float) src_rect->height / in_h / dst_rect->height;
2267 pp_inline_parameter->grf5.normalized_video_x_scaling_step = (float) src_rect->width / in_w / dst_rect->width;
2268 pp_inline_parameter->grf5.block_count_x = 1; /* M x 1 */
2269 pp_inline_parameter->grf5.number_blocks = pp_avs_context->dest_h / 8;
2270 pp_inline_parameter->grf6.video_step_delta = 0.0;
2272 dst_surface->flags = src_surface->flags;
2274 return VA_STATUS_SUCCESS;
2278 pp_nv12_avs_initialize_nlas(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2279 const struct i965_surface *src_surface,
2280 const VARectangle *src_rect,
2281 struct i965_surface *dst_surface,
2282 const VARectangle *dst_rect,
2285 return pp_nv12_avs_initialize(ctx, pp_context,
2295 gen6_nv12_scaling_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2296 const struct i965_surface *src_surface,
2297 const VARectangle *src_rect,
2298 struct i965_surface *dst_surface,
2299 const VARectangle *dst_rect,
2302 return pp_nv12_avs_initialize(ctx, pp_context,
2312 gen7_pp_avs_x_steps(void *private_context)
2314 struct pp_avs_context *pp_avs_context = private_context;
2316 return pp_avs_context->dest_w / 16;
2320 gen7_pp_avs_y_steps(void *private_context)
2322 struct pp_avs_context *pp_avs_context = private_context;
2324 return pp_avs_context->dest_h / 16;
2328 gen7_pp_avs_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
2330 struct pp_avs_context *pp_avs_context = (struct pp_avs_context *)&pp_context->private_context;
2331 struct gen7_pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2333 pp_inline_parameter->grf7.destination_block_horizontal_origin = x * 16 + pp_avs_context->dest_x;
2334 pp_inline_parameter->grf7.destination_block_vertical_origin = y * 16 + pp_avs_context->dest_y;
2335 pp_inline_parameter->grf7.constant_0 = 0xffffffff;
2336 pp_inline_parameter->grf7.sampler_load_main_video_x_scaling_step = 1.0 / pp_avs_context->src_w;
2341 static void gen7_update_src_surface_uv_offset(VADriverContextP ctx,
2342 struct i965_post_processing_context *pp_context,
2343 const struct i965_surface *surface)
2345 struct gen7_pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2346 int fourcc = pp_get_surface_fourcc(ctx, surface);
2348 if (fourcc == VA_FOURCC('Y', 'U', 'Y', '2')) {
2349 pp_static_parameter->grf2.di_destination_packed_y_component_offset = 0;
2350 pp_static_parameter->grf2.di_destination_packed_u_component_offset = 1;
2351 pp_static_parameter->grf2.di_destination_packed_v_component_offset = 3;
2352 } else if (fourcc == VA_FOURCC('U', 'Y', 'V', 'Y')) {
2353 pp_static_parameter->grf2.di_destination_packed_y_component_offset = 1;
2354 pp_static_parameter->grf2.di_destination_packed_u_component_offset = 0;
2355 pp_static_parameter->grf2.di_destination_packed_v_component_offset = 2;
2360 gen7_pp_plx_avs_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2361 const struct i965_surface *src_surface,
2362 const VARectangle *src_rect,
2363 struct i965_surface *dst_surface,
2364 const VARectangle *dst_rect,
2367 struct pp_avs_context *pp_avs_context = (struct pp_avs_context *)&pp_context->private_context;
2368 struct gen7_pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2369 struct gen7_sampler_8x8 *sampler_8x8;
2370 struct i965_sampler_8x8_state *sampler_8x8_state;
2372 int width[3], height[3], pitch[3], offset[3];
2373 int src_width, src_height;
2375 /* source surface */
2376 gen7_pp_set_media_rw_message_surface(ctx, pp_context, src_surface, 0, 0,
2377 width, height, pitch, offset);
2378 src_width = width[0];
2379 src_height = height[0];
2381 /* destination surface */
2382 gen7_pp_set_media_rw_message_surface(ctx, pp_context, dst_surface, 24, 1,
2383 width, height, pitch, offset);
2385 /* sampler 8x8 state */
2386 dri_bo_map(pp_context->sampler_state_table.bo_8x8, True);
2387 assert(pp_context->sampler_state_table.bo_8x8->virtual);
2388 assert(sizeof(*sampler_8x8_state) == sizeof(int) * 138);
2389 sampler_8x8_state = pp_context->sampler_state_table.bo_8x8->virtual;
2390 memset(sampler_8x8_state, 0, sizeof(*sampler_8x8_state));
2392 for (i = 0; i < 17; i++) {
2393 /* for Y channel, currently ignore */
2394 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c0 = 0x0;
2395 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c1 = 0x0;
2396 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c2 = 0x0;
2397 sampler_8x8_state->coefficients[i].dw0.table_0x_filter_c3 = 0x0;
2398 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c4 = 0x0;
2399 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c5 = 0x0;
2400 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c6 = 0x0;
2401 sampler_8x8_state->coefficients[i].dw1.table_0x_filter_c7 = 0x0;
2402 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c0 = 0x0;
2403 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c1 = 0x0;
2404 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c2 = 0x0;
2405 sampler_8x8_state->coefficients[i].dw2.table_0y_filter_c3 = 0x0;
2406 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c4 = 0x0;
2407 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c5 = 0x0;
2408 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c6 = 0x0;
2409 sampler_8x8_state->coefficients[i].dw3.table_0y_filter_c7 = 0x0;
2410 /* for U/V channel, 0.25 */
2411 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c0 = 0x0;
2412 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c1 = 0x0;
2413 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c2 = 0x10;
2414 sampler_8x8_state->coefficients[i].dw4.table_1x_filter_c3 = 0x10;
2415 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c4 = 0x10;
2416 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c5 = 0x10;
2417 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c6 = 0x0;
2418 sampler_8x8_state->coefficients[i].dw5.table_1x_filter_c7 = 0x0;
2419 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c0 = 0x0;
2420 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c1 = 0x0;
2421 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c2 = 0x10;
2422 sampler_8x8_state->coefficients[i].dw6.table_1y_filter_c3 = 0x10;
2423 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c4 = 0x10;
2424 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c5 = 0x10;
2425 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c6 = 0x0;
2426 sampler_8x8_state->coefficients[i].dw7.table_1y_filter_c7 = 0x0;
2429 sampler_8x8_state->dw136.default_sharpness_level = 0;
2430 sampler_8x8_state->dw137.adaptive_filter_for_all_channel = 1;
2431 sampler_8x8_state->dw137.bypass_y_adaptive_filtering = 1;
2432 sampler_8x8_state->dw137.bypass_x_adaptive_filtering = 1;
2433 dri_bo_unmap(pp_context->sampler_state_table.bo_8x8);
2436 dri_bo_map(pp_context->sampler_state_table.bo, True);
2437 assert(pp_context->sampler_state_table.bo->virtual);
2438 assert(sizeof(*sampler_8x8) == sizeof(int) * 4);
2439 sampler_8x8 = pp_context->sampler_state_table.bo->virtual;
2441 /* sample_8x8 Y index 4 */
2443 memset(&sampler_8x8[index], 0, sizeof(*sampler_8x8));
2444 sampler_8x8[index].dw0.global_noise_estimation = 255;
2445 sampler_8x8[index].dw0.ief_bypass = 1;
2447 sampler_8x8[index].dw1.sampler_8x8_state_pointer = pp_context->sampler_state_table.bo_8x8->offset >> 5;
2449 sampler_8x8[index].dw2.weak_edge_threshold = 1;
2450 sampler_8x8[index].dw2.strong_edge_threshold = 8;
2451 sampler_8x8[index].dw2.r5x_coefficient = 9;
2452 sampler_8x8[index].dw2.r5cx_coefficient = 8;
2453 sampler_8x8[index].dw2.r5c_coefficient = 3;
2455 sampler_8x8[index].dw3.r3x_coefficient = 27;
2456 sampler_8x8[index].dw3.r3c_coefficient = 5;
2457 sampler_8x8[index].dw3.gain_factor = 40;
2458 sampler_8x8[index].dw3.non_edge_weight = 1;
2459 sampler_8x8[index].dw3.regular_weight = 2;
2460 sampler_8x8[index].dw3.strong_edge_weight = 7;
2461 sampler_8x8[index].dw3.ief4_smooth_enable = 0;
2463 dri_bo_emit_reloc(pp_context->sampler_state_table.bo,
2464 I915_GEM_DOMAIN_RENDER,
2467 sizeof(*sampler_8x8) * index + offsetof(struct i965_sampler_8x8, dw1),
2468 pp_context->sampler_state_table.bo_8x8);
2470 /* sample_8x8 UV index 8 */
2472 memset(&sampler_8x8[index], 0, sizeof(*sampler_8x8));
2473 sampler_8x8[index].dw0.disable_8x8_filter = 0;
2474 sampler_8x8[index].dw0.global_noise_estimation = 255;
2475 sampler_8x8[index].dw0.ief_bypass = 1;
2476 sampler_8x8[index].dw1.sampler_8x8_state_pointer = pp_context->sampler_state_table.bo_8x8->offset >> 5;
2477 sampler_8x8[index].dw2.weak_edge_threshold = 1;
2478 sampler_8x8[index].dw2.strong_edge_threshold = 8;
2479 sampler_8x8[index].dw2.r5x_coefficient = 9;
2480 sampler_8x8[index].dw2.r5cx_coefficient = 8;
2481 sampler_8x8[index].dw2.r5c_coefficient = 3;
2482 sampler_8x8[index].dw3.r3x_coefficient = 27;
2483 sampler_8x8[index].dw3.r3c_coefficient = 5;
2484 sampler_8x8[index].dw3.gain_factor = 40;
2485 sampler_8x8[index].dw3.non_edge_weight = 1;
2486 sampler_8x8[index].dw3.regular_weight = 2;
2487 sampler_8x8[index].dw3.strong_edge_weight = 7;
2488 sampler_8x8[index].dw3.ief4_smooth_enable = 0;
2490 dri_bo_emit_reloc(pp_context->sampler_state_table.bo,
2491 I915_GEM_DOMAIN_RENDER,
2494 sizeof(*sampler_8x8) * index + offsetof(struct i965_sampler_8x8, dw1),
2495 pp_context->sampler_state_table.bo_8x8);
2497 /* sampler_8x8 V, index 12 */
2499 memset(&sampler_8x8[index], 0, sizeof(*sampler_8x8));
2500 sampler_8x8[index].dw0.disable_8x8_filter = 0;
2501 sampler_8x8[index].dw0.global_noise_estimation = 255;
2502 sampler_8x8[index].dw0.ief_bypass = 1;
2503 sampler_8x8[index].dw1.sampler_8x8_state_pointer = pp_context->sampler_state_table.bo_8x8->offset >> 5;
2504 sampler_8x8[index].dw2.weak_edge_threshold = 1;
2505 sampler_8x8[index].dw2.strong_edge_threshold = 8;
2506 sampler_8x8[index].dw2.r5x_coefficient = 9;
2507 sampler_8x8[index].dw2.r5cx_coefficient = 8;
2508 sampler_8x8[index].dw2.r5c_coefficient = 3;
2509 sampler_8x8[index].dw3.r3x_coefficient = 27;
2510 sampler_8x8[index].dw3.r3c_coefficient = 5;
2511 sampler_8x8[index].dw3.gain_factor = 40;
2512 sampler_8x8[index].dw3.non_edge_weight = 1;
2513 sampler_8x8[index].dw3.regular_weight = 2;
2514 sampler_8x8[index].dw3.strong_edge_weight = 7;
2515 sampler_8x8[index].dw3.ief4_smooth_enable = 0;
2517 dri_bo_emit_reloc(pp_context->sampler_state_table.bo,
2518 I915_GEM_DOMAIN_RENDER,
2521 sizeof(*sampler_8x8) * index + offsetof(struct i965_sampler_8x8, dw1),
2522 pp_context->sampler_state_table.bo_8x8);
2524 dri_bo_unmap(pp_context->sampler_state_table.bo);
2526 /* private function & data */
2527 pp_context->pp_x_steps = gen7_pp_avs_x_steps;
2528 pp_context->pp_y_steps = gen7_pp_avs_y_steps;
2529 pp_context->pp_set_block_parameter = gen7_pp_avs_set_block_parameter;
2531 pp_avs_context->dest_x = dst_rect->x;
2532 pp_avs_context->dest_y = dst_rect->y;
2533 pp_avs_context->dest_w = ALIGN(dst_rect->width, 16);
2534 pp_avs_context->dest_h = ALIGN(dst_rect->height, 16);
2535 pp_avs_context->src_w = src_rect->width;
2536 pp_avs_context->src_h = src_rect->height;
2538 int dw = (pp_avs_context->src_w - 1) / 16 + 1;
2539 dw = MAX(dw, pp_avs_context->dest_w);
2541 pp_static_parameter->grf1.pointer_to_inline_parameter = 7;
2542 pp_static_parameter->grf2.avs_wa_enable = 1; /* must be set for GEN7 */
2543 pp_static_parameter->grf2.avs_wa_width = dw;
2544 pp_static_parameter->grf2.avs_wa_one_div_256_width = (float) 1.0 / (256 * dw);
2545 pp_static_parameter->grf2.avs_wa_five_div_256_width = (float) 5.0 / (256 * dw);
2547 pp_static_parameter->grf3.sampler_load_horizontal_scaling_step_ratio = (float) pp_avs_context->src_w / dw;
2548 pp_static_parameter->grf4.sampler_load_vertical_scaling_step = (float) src_rect->height / src_height / pp_avs_context->dest_h;
2549 pp_static_parameter->grf5.sampler_load_vertical_frame_origin = -(float)pp_avs_context->dest_y / pp_avs_context->dest_h;
2550 pp_static_parameter->grf6.sampler_load_horizontal_frame_origin = -(float)pp_avs_context->dest_x / dw;
2552 gen7_update_src_surface_uv_offset(ctx, pp_context, dst_surface);
2554 dst_surface->flags = src_surface->flags;
2556 return VA_STATUS_SUCCESS;
2560 pp_dndi_x_steps(void *private_context)
2566 pp_dndi_y_steps(void *private_context)
2568 struct pp_dndi_context *pp_dndi_context = private_context;
2570 return pp_dndi_context->dest_h / 4;
2574 pp_dndi_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
2576 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2578 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16;
2579 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 4;
2585 pp_nv12_dndi_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2586 const struct i965_surface *src_surface,
2587 const VARectangle *src_rect,
2588 struct i965_surface *dst_surface,
2589 const VARectangle *dst_rect,
2592 struct i965_driver_data *i965 = i965_driver_data(ctx);
2593 struct pp_dndi_context *pp_dndi_context = (struct pp_dndi_context *)&pp_context->private_context;
2594 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2595 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2596 struct object_surface *obj_surface;
2597 struct i965_sampler_dndi *sampler_dndi;
2601 int dndi_top_first = 1;
2603 if (src_surface->flags == I965_SURFACE_FLAG_FRAME)
2604 return VA_STATUS_ERROR_FLAG_NOT_SUPPORTED;
2606 if (src_surface->flags == I965_SURFACE_FLAG_TOP_FIELD_FIRST)
2612 obj_surface = SURFACE(src_surface->id);
2613 orig_w = obj_surface->orig_width;
2614 orig_h = obj_surface->orig_height;
2615 w = obj_surface->width;
2616 h = obj_surface->height;
2618 if (pp_context->stmm.bo == NULL) {
2619 pp_context->stmm.bo = dri_bo_alloc(i965->intel.bufmgr,
2623 assert(pp_context->stmm.bo);
2626 /* source UV surface index 2 */
2627 i965_pp_set_surface_state(ctx, pp_context,
2628 obj_surface->bo, w * h,
2629 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
2632 /* source YUV surface index 4 */
2633 i965_pp_set_surface2_state(ctx, pp_context,
2637 SURFACE_FORMAT_PLANAR_420_8, 1,
2640 /* source STMM surface index 20 */
2641 i965_pp_set_surface_state(ctx, pp_context,
2642 pp_context->stmm.bo, 0,
2643 orig_w, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
2646 /* destination surface */
2647 obj_surface = SURFACE(dst_surface->id);
2648 orig_w = obj_surface->orig_width;
2649 orig_h = obj_surface->orig_height;
2650 w = obj_surface->width;
2651 h = obj_surface->height;
2653 /* destination Y surface index 7 */
2654 i965_pp_set_surface_state(ctx, pp_context,
2656 orig_w / 4, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
2659 /* destination UV surface index 8 */
2660 i965_pp_set_surface_state(ctx, pp_context,
2661 obj_surface->bo, w * h,
2662 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
2665 dri_bo_map(pp_context->sampler_state_table.bo, True);
2666 assert(pp_context->sampler_state_table.bo->virtual);
2667 assert(sizeof(*sampler_dndi) == sizeof(int) * 8);
2668 sampler_dndi = pp_context->sampler_state_table.bo->virtual;
2670 /* sample dndi index 1 */
2672 sampler_dndi[index].dw0.denoise_asd_threshold = 0;
2673 sampler_dndi[index].dw0.denoise_history_delta = 8; // 0-15, default is 8
2674 sampler_dndi[index].dw0.denoise_maximum_history = 128; // 128-240
2675 sampler_dndi[index].dw0.denoise_stad_threshold = 0;
2677 sampler_dndi[index].dw1.denoise_threshold_for_sum_of_complexity_measure = 64;
2678 sampler_dndi[index].dw1.denoise_moving_pixel_threshold = 4;
2679 sampler_dndi[index].dw1.stmm_c2 = 1;
2680 sampler_dndi[index].dw1.low_temporal_difference_threshold = 8;
2681 sampler_dndi[index].dw1.temporal_difference_threshold = 16;
2683 sampler_dndi[index].dw2.block_noise_estimate_noise_threshold = 15; // 0-31
2684 sampler_dndi[index].dw2.block_noise_estimate_edge_threshold = 7; // 0-15
2685 sampler_dndi[index].dw2.denoise_edge_threshold = 7; // 0-15
2686 sampler_dndi[index].dw2.good_neighbor_threshold = 4; // 0-63
2688 sampler_dndi[index].dw3.maximum_stmm = 128;
2689 sampler_dndi[index].dw3.multipler_for_vecm = 2;
2690 sampler_dndi[index].dw3.blending_constant_across_time_for_small_values_of_stmm = 0;
2691 sampler_dndi[index].dw3.blending_constant_across_time_for_large_values_of_stmm = 64;
2692 sampler_dndi[index].dw3.stmm_blending_constant_select = 0;
2694 sampler_dndi[index].dw4.sdi_delta = 8;
2695 sampler_dndi[index].dw4.sdi_threshold = 128;
2696 sampler_dndi[index].dw4.stmm_output_shift = 7; // stmm_max - stmm_min = 2 ^ stmm_output_shift
2697 sampler_dndi[index].dw4.stmm_shift_up = 0;
2698 sampler_dndi[index].dw4.stmm_shift_down = 0;
2699 sampler_dndi[index].dw4.minimum_stmm = 0;
2701 sampler_dndi[index].dw5.fmd_temporal_difference_threshold = 8;
2702 sampler_dndi[index].dw5.sdi_fallback_mode_2_constant = 32;
2703 sampler_dndi[index].dw5.sdi_fallback_mode_1_t2_constant = 64;
2704 sampler_dndi[index].dw5.sdi_fallback_mode_1_t1_constant = 32;
2706 sampler_dndi[index].dw6.dn_enable = 1;
2707 sampler_dndi[index].dw6.di_enable = 1;
2708 sampler_dndi[index].dw6.di_partial = 0;
2709 sampler_dndi[index].dw6.dndi_top_first = dndi_top_first;
2710 sampler_dndi[index].dw6.dndi_stream_id = 0;
2711 sampler_dndi[index].dw6.dndi_first_frame = 1;
2712 sampler_dndi[index].dw6.progressive_dn = 0;
2713 sampler_dndi[index].dw6.fmd_tear_threshold = 63;
2714 sampler_dndi[index].dw6.fmd2_vertical_difference_threshold = 32;
2715 sampler_dndi[index].dw6.fmd1_vertical_difference_threshold = 32;
2717 sampler_dndi[index].dw7.fmd_for_1st_field_of_current_frame = 0;
2718 sampler_dndi[index].dw7.fmd_for_2nd_field_of_previous_frame = 0;
2719 sampler_dndi[index].dw7.vdi_walker_enable = 0;
2720 sampler_dndi[index].dw7.column_width_minus1 = 0;
2722 dri_bo_unmap(pp_context->sampler_state_table.bo);
2724 /* private function & data */
2725 pp_context->pp_x_steps = pp_dndi_x_steps;
2726 pp_context->pp_y_steps = pp_dndi_y_steps;
2727 pp_context->pp_set_block_parameter = pp_dndi_set_block_parameter;
2729 pp_static_parameter->grf1.statistics_surface_picth = w / 2;
2730 pp_static_parameter->grf1.r1_6.di.top_field_first = dndi_top_first;
2731 pp_static_parameter->grf4.r4_2.di.motion_history_coefficient_m2 = 0;
2732 pp_static_parameter->grf4.r4_2.di.motion_history_coefficient_m1 = 0;
2734 pp_inline_parameter->grf5.block_count_x = w / 16; /* 1 x N */
2735 pp_inline_parameter->grf5.number_blocks = w / 16;
2736 pp_inline_parameter->grf5.block_vertical_mask = 0xff;
2737 pp_inline_parameter->grf5.block_horizontal_mask = 0xffff;
2739 pp_dndi_context->dest_w = w;
2740 pp_dndi_context->dest_h = h;
2742 dst_surface->flags = I965_SURFACE_FLAG_FRAME;
2744 return VA_STATUS_SUCCESS;
2748 pp_dn_x_steps(void *private_context)
2754 pp_dn_y_steps(void *private_context)
2756 struct pp_dn_context *pp_dn_context = private_context;
2758 return pp_dn_context->dest_h / 8;
2762 pp_dn_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
2764 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2766 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16;
2767 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8;
2773 pp_nv12_dn_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2774 const struct i965_surface *src_surface,
2775 const VARectangle *src_rect,
2776 struct i965_surface *dst_surface,
2777 const VARectangle *dst_rect,
2780 struct i965_driver_data *i965 = i965_driver_data(ctx);
2781 struct pp_dn_context *pp_dn_context = (struct pp_dn_context *)&pp_context->private_context;
2782 struct object_surface *obj_surface;
2783 struct i965_sampler_dndi *sampler_dndi;
2784 struct pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2785 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2786 VAProcFilterParameterBuffer *dn_filter_param = filter_param; /* FIXME: parameter */
2790 int dn_strength = 15;
2791 int dndi_top_first = 1;
2792 int dn_progressive = 0;
2794 if (src_surface->flags == I965_SURFACE_FLAG_FRAME) {
2797 } else if (src_surface->flags == I965_SURFACE_FLAG_TOP_FIELD_FIRST) {
2805 if (dn_filter_param) {
2806 float value = dn_filter_param->value;
2814 dn_strength = (int)(value * 31.0F);
2818 obj_surface = SURFACE(src_surface->id);
2819 orig_w = obj_surface->orig_width;
2820 orig_h = obj_surface->orig_height;
2821 w = obj_surface->width;
2822 h = obj_surface->height;
2824 if (pp_context->stmm.bo == NULL) {
2825 pp_context->stmm.bo = dri_bo_alloc(i965->intel.bufmgr,
2829 assert(pp_context->stmm.bo);
2832 /* source UV surface index 2 */
2833 i965_pp_set_surface_state(ctx, pp_context,
2834 obj_surface->bo, w * h,
2835 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
2838 /* source YUV surface index 4 */
2839 i965_pp_set_surface2_state(ctx, pp_context,
2843 SURFACE_FORMAT_PLANAR_420_8, 1,
2846 /* source STMM surface index 20 */
2847 i965_pp_set_surface_state(ctx, pp_context,
2848 pp_context->stmm.bo, 0,
2849 orig_w, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
2852 /* destination surface */
2853 obj_surface = SURFACE(dst_surface->id);
2854 orig_w = obj_surface->orig_width;
2855 orig_h = obj_surface->orig_height;
2856 w = obj_surface->width;
2857 h = obj_surface->height;
2859 /* destination Y surface index 7 */
2860 i965_pp_set_surface_state(ctx, pp_context,
2862 orig_w / 4, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
2865 /* destination UV surface index 8 */
2866 i965_pp_set_surface_state(ctx, pp_context,
2867 obj_surface->bo, w * h,
2868 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
2871 dri_bo_map(pp_context->sampler_state_table.bo, True);
2872 assert(pp_context->sampler_state_table.bo->virtual);
2873 assert(sizeof(*sampler_dndi) == sizeof(int) * 8);
2874 sampler_dndi = pp_context->sampler_state_table.bo->virtual;
2876 /* sample dndi index 1 */
2878 sampler_dndi[index].dw0.denoise_asd_threshold = 0;
2879 sampler_dndi[index].dw0.denoise_history_delta = 8; // 0-15, default is 8
2880 sampler_dndi[index].dw0.denoise_maximum_history = 128; // 128-240
2881 sampler_dndi[index].dw0.denoise_stad_threshold = 0;
2883 sampler_dndi[index].dw1.denoise_threshold_for_sum_of_complexity_measure = 64;
2884 sampler_dndi[index].dw1.denoise_moving_pixel_threshold = 0;
2885 sampler_dndi[index].dw1.stmm_c2 = 0;
2886 sampler_dndi[index].dw1.low_temporal_difference_threshold = 8;
2887 sampler_dndi[index].dw1.temporal_difference_threshold = 16;
2889 sampler_dndi[index].dw2.block_noise_estimate_noise_threshold = dn_strength; // 0-31
2890 sampler_dndi[index].dw2.block_noise_estimate_edge_threshold = 7; // 0-15
2891 sampler_dndi[index].dw2.denoise_edge_threshold = 7; // 0-15
2892 sampler_dndi[index].dw2.good_neighbor_threshold = 7; // 0-63
2894 sampler_dndi[index].dw3.maximum_stmm = 128;
2895 sampler_dndi[index].dw3.multipler_for_vecm = 2;
2896 sampler_dndi[index].dw3.blending_constant_across_time_for_small_values_of_stmm = 0;
2897 sampler_dndi[index].dw3.blending_constant_across_time_for_large_values_of_stmm = 64;
2898 sampler_dndi[index].dw3.stmm_blending_constant_select = 0;
2900 sampler_dndi[index].dw4.sdi_delta = 8;
2901 sampler_dndi[index].dw4.sdi_threshold = 128;
2902 sampler_dndi[index].dw4.stmm_output_shift = 7; // stmm_max - stmm_min = 2 ^ stmm_output_shift
2903 sampler_dndi[index].dw4.stmm_shift_up = 0;
2904 sampler_dndi[index].dw4.stmm_shift_down = 0;
2905 sampler_dndi[index].dw4.minimum_stmm = 0;
2907 sampler_dndi[index].dw5.fmd_temporal_difference_threshold = 0;
2908 sampler_dndi[index].dw5.sdi_fallback_mode_2_constant = 0;
2909 sampler_dndi[index].dw5.sdi_fallback_mode_1_t2_constant = 0;
2910 sampler_dndi[index].dw5.sdi_fallback_mode_1_t1_constant = 0;
2912 sampler_dndi[index].dw6.dn_enable = 1;
2913 sampler_dndi[index].dw6.di_enable = 0;
2914 sampler_dndi[index].dw6.di_partial = 0;
2915 sampler_dndi[index].dw6.dndi_top_first = dndi_top_first;
2916 sampler_dndi[index].dw6.dndi_stream_id = 1;
2917 sampler_dndi[index].dw6.dndi_first_frame = 1;
2918 sampler_dndi[index].dw6.progressive_dn = dn_progressive;
2919 sampler_dndi[index].dw6.fmd_tear_threshold = 32;
2920 sampler_dndi[index].dw6.fmd2_vertical_difference_threshold = 32;
2921 sampler_dndi[index].dw6.fmd1_vertical_difference_threshold = 32;
2923 sampler_dndi[index].dw7.fmd_for_1st_field_of_current_frame = 2;
2924 sampler_dndi[index].dw7.fmd_for_2nd_field_of_previous_frame = 1;
2925 sampler_dndi[index].dw7.vdi_walker_enable = 0;
2926 sampler_dndi[index].dw7.column_width_minus1 = w / 16;
2928 dri_bo_unmap(pp_context->sampler_state_table.bo);
2930 /* private function & data */
2931 pp_context->pp_x_steps = pp_dn_x_steps;
2932 pp_context->pp_y_steps = pp_dn_y_steps;
2933 pp_context->pp_set_block_parameter = pp_dn_set_block_parameter;
2935 pp_static_parameter->grf1.statistics_surface_picth = w / 2;
2936 pp_static_parameter->grf1.r1_6.di.top_field_first = 0;
2937 pp_static_parameter->grf4.r4_2.di.motion_history_coefficient_m2 = 64;
2938 pp_static_parameter->grf4.r4_2.di.motion_history_coefficient_m1 = 192;
2940 pp_inline_parameter->grf5.block_count_x = w / 16; /* 1 x N */
2941 pp_inline_parameter->grf5.number_blocks = w / 16;
2942 pp_inline_parameter->grf5.block_vertical_mask = 0xff;
2943 pp_inline_parameter->grf5.block_horizontal_mask = 0xffff;
2945 pp_dn_context->dest_w = w;
2946 pp_dn_context->dest_h = h;
2948 dst_surface->flags = src_surface->flags;
2950 return VA_STATUS_SUCCESS;
2954 gen7_pp_dndi_x_steps(void *private_context)
2956 struct pp_dndi_context *pp_dndi_context = private_context;
2958 return pp_dndi_context->dest_w / 16;
2962 gen7_pp_dndi_y_steps(void *private_context)
2964 struct pp_dndi_context *pp_dndi_context = private_context;
2966 return pp_dndi_context->dest_h / 4;
2970 gen7_pp_dndi_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
2972 struct gen7_pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
2974 pp_inline_parameter->grf7.destination_block_horizontal_origin = x * 16;
2975 pp_inline_parameter->grf7.destination_block_vertical_origin = y * 4;
2981 gen7_pp_nv12_dndi_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
2982 const struct i965_surface *src_surface,
2983 const VARectangle *src_rect,
2984 struct i965_surface *dst_surface,
2985 const VARectangle *dst_rect,
2988 struct i965_driver_data *i965 = i965_driver_data(ctx);
2989 struct pp_dndi_context *pp_dndi_context = (struct pp_dndi_context *)&pp_context->private_context;
2990 struct gen7_pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
2991 struct object_surface *obj_surface;
2992 struct gen7_sampler_dndi *sampler_dndi;
2996 int dndi_top_first = 1;
2998 if (src_surface->flags == I965_SURFACE_FLAG_FRAME)
2999 return VA_STATUS_ERROR_FLAG_NOT_SUPPORTED;
3001 if (src_surface->flags == I965_SURFACE_FLAG_TOP_FIELD_FIRST)
3007 obj_surface = SURFACE(src_surface->id);
3008 orig_w = obj_surface->orig_width;
3009 orig_h = obj_surface->orig_height;
3010 w = obj_surface->width;
3011 h = obj_surface->height;
3013 if (pp_context->stmm.bo == NULL) {
3014 pp_context->stmm.bo = dri_bo_alloc(i965->intel.bufmgr,
3018 assert(pp_context->stmm.bo);
3021 /* source UV surface index 1 */
3022 gen7_pp_set_surface_state(ctx, pp_context,
3023 obj_surface->bo, w * h,
3024 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
3027 /* source YUV surface index 3 */
3028 gen7_pp_set_surface2_state(ctx, pp_context,
3032 SURFACE_FORMAT_PLANAR_420_8, 1,
3035 /* source (temporal reference) YUV surface index 4 */
3036 gen7_pp_set_surface2_state(ctx, pp_context,
3040 SURFACE_FORMAT_PLANAR_420_8, 1,
3043 /* STMM / History Statistics input surface, index 5 */
3044 gen7_pp_set_surface_state(ctx, pp_context,
3045 pp_context->stmm.bo, 0,
3046 orig_w, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3049 /* destination surface */
3050 obj_surface = SURFACE(dst_surface->id);
3051 orig_w = obj_surface->orig_width;
3052 orig_h = obj_surface->orig_height;
3053 w = obj_surface->width;
3054 h = obj_surface->height;
3056 /* destination(Previous frame) Y surface index 27 */
3057 gen7_pp_set_surface_state(ctx, pp_context,
3059 orig_w / 4, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3062 /* destination(Previous frame) UV surface index 28 */
3063 gen7_pp_set_surface_state(ctx, pp_context,
3064 obj_surface->bo, w * h,
3065 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
3068 /* destination(Current frame) Y surface index 30 */
3069 gen7_pp_set_surface_state(ctx, pp_context,
3071 orig_w / 4, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3074 /* destination(Current frame) UV surface index 31 */
3075 gen7_pp_set_surface_state(ctx, pp_context,
3076 obj_surface->bo, w * h,
3077 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
3080 /* STMM output surface, index 33 */
3081 gen7_pp_set_surface_state(ctx, pp_context,
3082 pp_context->stmm.bo, 0,
3083 orig_w, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3088 dri_bo_map(pp_context->sampler_state_table.bo, True);
3089 assert(pp_context->sampler_state_table.bo->virtual);
3090 assert(sizeof(*sampler_dndi) == sizeof(int) * 8);
3091 sampler_dndi = pp_context->sampler_state_table.bo->virtual;
3093 /* sample dndi index 0 */
3095 sampler_dndi[index].dw0.denoise_asd_threshold = 0;
3096 sampler_dndi[index].dw0.dnmh_delt = 8;
3097 sampler_dndi[index].dw0.vdi_walker_y_stride = 0;
3098 sampler_dndi[index].dw0.vdi_walker_frame_sharing_enable = 0;
3099 sampler_dndi[index].dw0.denoise_maximum_history = 128; // 128-240
3100 sampler_dndi[index].dw0.denoise_stad_threshold = 0;
3102 sampler_dndi[index].dw1.denoise_threshold_for_sum_of_complexity_measure = 64;
3103 sampler_dndi[index].dw1.denoise_moving_pixel_threshold = 0;
3104 sampler_dndi[index].dw1.stmm_c2 = 0;
3105 sampler_dndi[index].dw1.low_temporal_difference_threshold = 8;
3106 sampler_dndi[index].dw1.temporal_difference_threshold = 16;
3108 sampler_dndi[index].dw2.block_noise_estimate_noise_threshold = 15; // 0-31
3109 sampler_dndi[index].dw2.bne_edge_th = 1;
3110 sampler_dndi[index].dw2.smooth_mv_th = 0;
3111 sampler_dndi[index].dw2.sad_tight_th = 5;
3112 sampler_dndi[index].dw2.cat_slope_minus1 = 9;
3113 sampler_dndi[index].dw2.good_neighbor_th = 4;
3115 sampler_dndi[index].dw3.maximum_stmm = 128;
3116 sampler_dndi[index].dw3.multipler_for_vecm = 2;
3117 sampler_dndi[index].dw3.blending_constant_across_time_for_small_values_of_stmm = 0;
3118 sampler_dndi[index].dw3.blending_constant_across_time_for_large_values_of_stmm = 64;
3119 sampler_dndi[index].dw3.stmm_blending_constant_select = 0;
3121 sampler_dndi[index].dw4.sdi_delta = 8;
3122 sampler_dndi[index].dw4.sdi_threshold = 128;
3123 sampler_dndi[index].dw4.stmm_output_shift = 7; // stmm_max - stmm_min = 2 ^ stmm_output_shift
3124 sampler_dndi[index].dw4.stmm_shift_up = 0;
3125 sampler_dndi[index].dw4.stmm_shift_down = 0;
3126 sampler_dndi[index].dw4.minimum_stmm = 0;
3128 sampler_dndi[index].dw5.fmd_temporal_difference_threshold = 0;
3129 sampler_dndi[index].dw5.sdi_fallback_mode_2_constant = 0;
3130 sampler_dndi[index].dw5.sdi_fallback_mode_1_t2_constant = 0;
3131 sampler_dndi[index].dw5.sdi_fallback_mode_1_t1_constant = 0;
3133 sampler_dndi[index].dw6.dn_enable = 0;
3134 sampler_dndi[index].dw6.di_enable = 1;
3135 sampler_dndi[index].dw6.di_partial = 0;
3136 sampler_dndi[index].dw6.dndi_top_first = dndi_top_first;
3137 sampler_dndi[index].dw6.dndi_stream_id = 1;
3138 sampler_dndi[index].dw6.dndi_first_frame = 1;
3139 sampler_dndi[index].dw6.progressive_dn = 0;
3140 sampler_dndi[index].dw6.mcdi_enable = 0;
3141 sampler_dndi[index].dw6.fmd_tear_threshold = 32;
3142 sampler_dndi[index].dw6.cat_th1 = 0;
3143 sampler_dndi[index].dw6.fmd2_vertical_difference_threshold = 32;
3144 sampler_dndi[index].dw6.fmd1_vertical_difference_threshold = 32;
3146 sampler_dndi[index].dw7.sad_tha = 5;
3147 sampler_dndi[index].dw7.sad_thb = 10;
3148 sampler_dndi[index].dw7.fmd_for_1st_field_of_current_frame = 0;
3149 sampler_dndi[index].dw7.mc_pixel_consistency_th = 25;
3150 sampler_dndi[index].dw7.fmd_for_2nd_field_of_previous_frame = 0;
3151 sampler_dndi[index].dw7.vdi_walker_enable = 0;
3152 sampler_dndi[index].dw7.neighborpixel_th = 10;
3153 sampler_dndi[index].dw7.column_width_minus1 = w / 16;
3155 dri_bo_unmap(pp_context->sampler_state_table.bo);
3157 /* private function & data */
3158 pp_context->pp_x_steps = gen7_pp_dndi_x_steps;
3159 pp_context->pp_y_steps = gen7_pp_dndi_y_steps;
3160 pp_context->pp_set_block_parameter = gen7_pp_dndi_set_block_parameter;
3162 pp_static_parameter->grf1.di_statistics_surface_pitch_div2 = w / 2;
3163 pp_static_parameter->grf1.di_statistics_surface_height_div4 = h / 4;
3164 pp_static_parameter->grf1.di_top_field_first = 0;
3165 pp_static_parameter->grf1.pointer_to_inline_parameter = 7;
3167 pp_static_parameter->grf2.di_destination_packed_y_component_offset = 0;
3168 pp_static_parameter->grf2.di_destination_packed_u_component_offset = 1;
3169 pp_static_parameter->grf2.di_destination_packed_v_component_offset = 3;
3171 pp_static_parameter->grf4.di_hoffset_svf_from_dvf = 0;
3172 pp_static_parameter->grf4.di_voffset_svf_from_dvf = 0;
3174 pp_dndi_context->dest_w = w;
3175 pp_dndi_context->dest_h = h;
3177 dst_surface->flags = I965_SURFACE_FLAG_FRAME;
3179 return VA_STATUS_SUCCESS;
3183 gen7_pp_dn_x_steps(void *private_context)
3185 struct pp_dn_context *pp_dn_context = private_context;
3187 return pp_dn_context->dest_w / 16;
3191 gen7_pp_dn_y_steps(void *private_context)
3193 struct pp_dn_context *pp_dn_context = private_context;
3195 return pp_dn_context->dest_h / 4;
3199 gen7_pp_dn_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
3201 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
3203 pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16;
3204 pp_inline_parameter->grf5.destination_block_vertical_origin = y * 4;
3210 gen7_pp_nv12_dn_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
3211 const struct i965_surface *src_surface,
3212 const VARectangle *src_rect,
3213 struct i965_surface *dst_surface,
3214 const VARectangle *dst_rect,
3217 struct i965_driver_data *i965 = i965_driver_data(ctx);
3218 struct pp_dn_context *pp_dn_context = (struct pp_dn_context *)&pp_context->private_context;
3219 struct gen7_pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
3220 struct object_surface *obj_surface;
3221 struct gen7_sampler_dndi *sampler_dn;
3222 VAProcFilterParameterBuffer *dn_filter_param = filter_param; /* FIXME: parameter */
3226 int dn_strength = 15;
3227 int dndi_top_first = 1;
3228 int dn_progressive = 0;
3230 if (src_surface->flags == I965_SURFACE_FLAG_FRAME) {
3233 } else if (src_surface->flags == I965_SURFACE_FLAG_TOP_FIELD_FIRST) {
3241 if (dn_filter_param) {
3242 float value = dn_filter_param->value;
3250 dn_strength = (int)(value * 31.0F);
3254 obj_surface = SURFACE(src_surface->id);
3255 orig_w = obj_surface->orig_width;
3256 orig_h = obj_surface->orig_height;
3257 w = obj_surface->width;
3258 h = obj_surface->height;
3260 if (pp_context->stmm.bo == NULL) {
3261 pp_context->stmm.bo = dri_bo_alloc(i965->intel.bufmgr,
3265 assert(pp_context->stmm.bo);
3268 /* source UV surface index 1 */
3269 gen7_pp_set_surface_state(ctx, pp_context,
3270 obj_surface->bo, w * h,
3271 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
3274 /* source YUV surface index 3 */
3275 gen7_pp_set_surface2_state(ctx, pp_context,
3279 SURFACE_FORMAT_PLANAR_420_8, 1,
3282 /* source (temporal reference) YUV surface index 4 */
3283 gen7_pp_set_surface2_state(ctx, pp_context,
3287 SURFACE_FORMAT_PLANAR_420_8, 1,
3290 /* STMM / History Statistics input surface, index 5 */
3291 gen7_pp_set_surface_state(ctx, pp_context,
3292 pp_context->stmm.bo, 0,
3293 orig_w, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3296 /* destination surface */
3297 obj_surface = SURFACE(dst_surface->id);
3298 orig_w = obj_surface->orig_width;
3299 orig_h = obj_surface->orig_height;
3300 w = obj_surface->width;
3301 h = obj_surface->height;
3303 /* destination Y surface index 24 */
3304 gen7_pp_set_surface_state(ctx, pp_context,
3306 orig_w / 4, orig_h, w, I965_SURFACEFORMAT_R8_UNORM,
3309 /* destination UV surface index 25 */
3310 gen7_pp_set_surface_state(ctx, pp_context,
3311 obj_surface->bo, w * h,
3312 orig_w / 4, orig_h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM,
3316 dri_bo_map(pp_context->sampler_state_table.bo, True);
3317 assert(pp_context->sampler_state_table.bo->virtual);
3318 assert(sizeof(*sampler_dn) == sizeof(int) * 8);
3319 sampler_dn = pp_context->sampler_state_table.bo->virtual;
3321 /* sample dn index 1 */
3323 sampler_dn[index].dw0.denoise_asd_threshold = 0;
3324 sampler_dn[index].dw0.dnmh_delt = 8;
3325 sampler_dn[index].dw0.vdi_walker_y_stride = 0;
3326 sampler_dn[index].dw0.vdi_walker_frame_sharing_enable = 0;
3327 sampler_dn[index].dw0.denoise_maximum_history = 128; // 128-240
3328 sampler_dn[index].dw0.denoise_stad_threshold = 0;
3330 sampler_dn[index].dw1.denoise_threshold_for_sum_of_complexity_measure = 64;
3331 sampler_dn[index].dw1.denoise_moving_pixel_threshold = 0;
3332 sampler_dn[index].dw1.stmm_c2 = 0;
3333 sampler_dn[index].dw1.low_temporal_difference_threshold = 8;
3334 sampler_dn[index].dw1.temporal_difference_threshold = 16;
3336 sampler_dn[index].dw2.block_noise_estimate_noise_threshold = dn_strength; // 0-31
3337 sampler_dn[index].dw2.bne_edge_th = 1;
3338 sampler_dn[index].dw2.smooth_mv_th = 0;
3339 sampler_dn[index].dw2.sad_tight_th = 5;
3340 sampler_dn[index].dw2.cat_slope_minus1 = 9;
3341 sampler_dn[index].dw2.good_neighbor_th = 4;
3343 sampler_dn[index].dw3.maximum_stmm = 128;
3344 sampler_dn[index].dw3.multipler_for_vecm = 2;
3345 sampler_dn[index].dw3.blending_constant_across_time_for_small_values_of_stmm = 0;
3346 sampler_dn[index].dw3.blending_constant_across_time_for_large_values_of_stmm = 64;
3347 sampler_dn[index].dw3.stmm_blending_constant_select = 0;
3349 sampler_dn[index].dw4.sdi_delta = 8;
3350 sampler_dn[index].dw4.sdi_threshold = 128;
3351 sampler_dn[index].dw4.stmm_output_shift = 7; // stmm_max - stmm_min = 2 ^ stmm_output_shift
3352 sampler_dn[index].dw4.stmm_shift_up = 0;
3353 sampler_dn[index].dw4.stmm_shift_down = 0;
3354 sampler_dn[index].dw4.minimum_stmm = 0;
3356 sampler_dn[index].dw5.fmd_temporal_difference_threshold = 0;
3357 sampler_dn[index].dw5.sdi_fallback_mode_2_constant = 0;
3358 sampler_dn[index].dw5.sdi_fallback_mode_1_t2_constant = 0;
3359 sampler_dn[index].dw5.sdi_fallback_mode_1_t1_constant = 0;
3361 sampler_dn[index].dw6.dn_enable = 1;
3362 sampler_dn[index].dw6.di_enable = 0;
3363 sampler_dn[index].dw6.di_partial = 0;
3364 sampler_dn[index].dw6.dndi_top_first = dndi_top_first;
3365 sampler_dn[index].dw6.dndi_stream_id = 1;
3366 sampler_dn[index].dw6.dndi_first_frame = 1;
3367 sampler_dn[index].dw6.progressive_dn = dn_progressive;
3368 sampler_dn[index].dw6.mcdi_enable = 0;
3369 sampler_dn[index].dw6.fmd_tear_threshold = 32;
3370 sampler_dn[index].dw6.cat_th1 = 0;
3371 sampler_dn[index].dw6.fmd2_vertical_difference_threshold = 32;
3372 sampler_dn[index].dw6.fmd1_vertical_difference_threshold = 32;
3374 sampler_dn[index].dw7.sad_tha = 5;
3375 sampler_dn[index].dw7.sad_thb = 10;
3376 sampler_dn[index].dw7.fmd_for_1st_field_of_current_frame = 2;
3377 sampler_dn[index].dw7.mc_pixel_consistency_th = 25;
3378 sampler_dn[index].dw7.fmd_for_2nd_field_of_previous_frame = 1;
3379 sampler_dn[index].dw7.vdi_walker_enable = 0;
3380 sampler_dn[index].dw7.neighborpixel_th = 10;
3381 sampler_dn[index].dw7.column_width_minus1 = w / 16;
3383 dri_bo_unmap(pp_context->sampler_state_table.bo);
3385 /* private function & data */
3386 pp_context->pp_x_steps = gen7_pp_dn_x_steps;
3387 pp_context->pp_y_steps = gen7_pp_dn_y_steps;
3388 pp_context->pp_set_block_parameter = gen7_pp_dn_set_block_parameter;
3390 pp_static_parameter->grf1.di_statistics_surface_pitch_div2 = w / 2;
3391 pp_static_parameter->grf1.di_statistics_surface_height_div4 = h / 4;
3392 pp_static_parameter->grf1.di_top_field_first = 0;
3393 pp_static_parameter->grf1.pointer_to_inline_parameter = 7;
3395 pp_static_parameter->grf2.di_destination_packed_y_component_offset = 0;
3396 pp_static_parameter->grf2.di_destination_packed_u_component_offset = 1;
3397 pp_static_parameter->grf2.di_destination_packed_v_component_offset = 3;
3399 pp_static_parameter->grf4.di_hoffset_svf_from_dvf = 0;
3400 pp_static_parameter->grf4.di_voffset_svf_from_dvf = 0;
3402 pp_dn_context->dest_w = w;
3403 pp_dn_context->dest_h = h;
3405 dst_surface->flags = src_surface->flags;
3407 return VA_STATUS_SUCCESS;
3411 ironlake_pp_initialize(
3412 VADriverContextP ctx,
3413 struct i965_post_processing_context *pp_context,
3414 const struct i965_surface *src_surface,
3415 const VARectangle *src_rect,
3416 struct i965_surface *dst_surface,
3417 const VARectangle *dst_rect,
3423 struct i965_driver_data *i965 = i965_driver_data(ctx);
3424 struct pp_module *pp_module;
3426 int static_param_size, inline_param_size;
3428 dri_bo_unreference(pp_context->surface_state_binding_table.bo);
3429 bo = dri_bo_alloc(i965->intel.bufmgr,
3430 "surface state & binding table",
3431 (SURFACE_STATE_PADDED_SIZE + sizeof(unsigned int)) * MAX_PP_SURFACES,
3434 pp_context->surface_state_binding_table.bo = bo;
3436 dri_bo_unreference(pp_context->curbe.bo);
3437 bo = dri_bo_alloc(i965->intel.bufmgr,
3442 pp_context->curbe.bo = bo;
3444 dri_bo_unreference(pp_context->idrt.bo);
3445 bo = dri_bo_alloc(i965->intel.bufmgr,
3446 "interface discriptor",
3447 sizeof(struct i965_interface_descriptor),
3450 pp_context->idrt.bo = bo;
3451 pp_context->idrt.num_interface_descriptors = 0;
3453 dri_bo_unreference(pp_context->sampler_state_table.bo);
3454 bo = dri_bo_alloc(i965->intel.bufmgr,
3455 "sampler state table",
3459 dri_bo_map(bo, True);
3460 memset(bo->virtual, 0, bo->size);
3462 pp_context->sampler_state_table.bo = bo;
3464 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8);
3465 bo = dri_bo_alloc(i965->intel.bufmgr,
3466 "sampler 8x8 state ",
3470 pp_context->sampler_state_table.bo_8x8 = bo;
3472 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8_uv);
3473 bo = dri_bo_alloc(i965->intel.bufmgr,
3474 "sampler 8x8 state ",
3478 pp_context->sampler_state_table.bo_8x8_uv = bo;
3480 dri_bo_unreference(pp_context->vfe_state.bo);
3481 bo = dri_bo_alloc(i965->intel.bufmgr,
3483 sizeof(struct i965_vfe_state),
3486 pp_context->vfe_state.bo = bo;
3488 static_param_size = sizeof(struct pp_static_parameter);
3489 inline_param_size = sizeof(struct pp_inline_parameter);
3491 memset(pp_context->pp_static_parameter, 0, static_param_size);
3492 memset(pp_context->pp_inline_parameter, 0, inline_param_size);
3494 assert(pp_index >= PP_NULL && pp_index < NUM_PP_MODULES);
3495 pp_context->current_pp = pp_index;
3496 pp_module = &pp_context->pp_modules[pp_index];
3498 if (pp_module->initialize)
3499 va_status = pp_module->initialize(ctx, pp_context,
3506 va_status = VA_STATUS_ERROR_UNIMPLEMENTED;
3508 calculate_boundary_block_mask(pp_context, dst_rect->width, dst_rect->height);
3514 ironlake_post_processing(
3515 VADriverContextP ctx,
3516 struct i965_post_processing_context *pp_context,
3517 const struct i965_surface *src_surface,
3518 const VARectangle *src_rect,
3519 struct i965_surface *dst_surface,
3520 const VARectangle *dst_rect,
3527 va_status = ironlake_pp_initialize(ctx, pp_context,
3535 if (va_status == VA_STATUS_SUCCESS) {
3536 ironlake_pp_states_setup(ctx, pp_context);
3537 ironlake_pp_pipeline_setup(ctx, pp_context);
3545 VADriverContextP ctx,
3546 struct i965_post_processing_context *pp_context,
3547 const struct i965_surface *src_surface,
3548 const VARectangle *src_rect,
3549 struct i965_surface *dst_surface,
3550 const VARectangle *dst_rect,
3556 struct i965_driver_data *i965 = i965_driver_data(ctx);
3557 struct pp_module *pp_module;
3559 int static_param_size, inline_param_size;
3561 dri_bo_unreference(pp_context->surface_state_binding_table.bo);
3562 bo = dri_bo_alloc(i965->intel.bufmgr,
3563 "surface state & binding table",
3564 (SURFACE_STATE_PADDED_SIZE + sizeof(unsigned int)) * MAX_PP_SURFACES,
3567 pp_context->surface_state_binding_table.bo = bo;
3569 dri_bo_unreference(pp_context->curbe.bo);
3570 bo = dri_bo_alloc(i965->intel.bufmgr,
3575 pp_context->curbe.bo = bo;
3577 dri_bo_unreference(pp_context->idrt.bo);
3578 bo = dri_bo_alloc(i965->intel.bufmgr,
3579 "interface discriptor",
3580 sizeof(struct gen6_interface_descriptor_data),
3583 pp_context->idrt.bo = bo;
3584 pp_context->idrt.num_interface_descriptors = 0;
3586 dri_bo_unreference(pp_context->sampler_state_table.bo);
3587 bo = dri_bo_alloc(i965->intel.bufmgr,
3588 "sampler state table",
3592 dri_bo_map(bo, True);
3593 memset(bo->virtual, 0, bo->size);
3595 pp_context->sampler_state_table.bo = bo;
3597 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8);
3598 bo = dri_bo_alloc(i965->intel.bufmgr,
3599 "sampler 8x8 state ",
3603 pp_context->sampler_state_table.bo_8x8 = bo;
3605 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8_uv);
3606 bo = dri_bo_alloc(i965->intel.bufmgr,
3607 "sampler 8x8 state ",
3611 pp_context->sampler_state_table.bo_8x8_uv = bo;
3613 dri_bo_unreference(pp_context->vfe_state.bo);
3614 bo = dri_bo_alloc(i965->intel.bufmgr,
3616 sizeof(struct i965_vfe_state),
3619 pp_context->vfe_state.bo = bo;
3621 if (IS_GEN7(i965->intel.device_id)) {
3622 static_param_size = sizeof(struct gen7_pp_static_parameter);
3623 inline_param_size = sizeof(struct gen7_pp_inline_parameter);
3625 static_param_size = sizeof(struct pp_static_parameter);
3626 inline_param_size = sizeof(struct pp_inline_parameter);
3629 memset(pp_context->pp_static_parameter, 0, static_param_size);
3630 memset(pp_context->pp_inline_parameter, 0, inline_param_size);
3632 assert(pp_index >= PP_NULL && pp_index < NUM_PP_MODULES);
3633 pp_context->current_pp = pp_index;
3634 pp_module = &pp_context->pp_modules[pp_index];
3636 if (pp_module->initialize)
3637 va_status = pp_module->initialize(ctx, pp_context,
3644 va_status = VA_STATUS_ERROR_UNIMPLEMENTED;
3650 gen6_pp_interface_descriptor_table(VADriverContextP ctx,
3651 struct i965_post_processing_context *pp_context)
3653 struct i965_driver_data *i965 = i965_driver_data(ctx);
3654 struct gen6_interface_descriptor_data *desc;
3656 int pp_index = pp_context->current_pp;
3658 bo = pp_context->idrt.bo;
3659 dri_bo_map(bo, True);
3660 assert(bo->virtual);
3662 memset(desc, 0, sizeof(*desc));
3663 desc->desc0.kernel_start_pointer =
3664 pp_context->pp_modules[pp_index].kernel.bo->offset >> 6; /* reloc */
3665 desc->desc1.single_program_flow = 1;
3666 desc->desc1.floating_point_mode = FLOATING_POINT_IEEE_754;
3667 desc->desc2.sampler_count = 1; /* 1 - 4 samplers used */
3668 desc->desc2.sampler_state_pointer =
3669 pp_context->sampler_state_table.bo->offset >> 5;
3670 desc->desc3.binding_table_entry_count = 0;
3671 desc->desc3.binding_table_pointer = (BINDING_TABLE_OFFSET >> 5);
3672 desc->desc4.constant_urb_entry_read_offset = 0;
3674 if (IS_GEN7(i965->intel.device_id))
3675 desc->desc4.constant_urb_entry_read_length = 6; /* grf 1-6 */
3677 desc->desc4.constant_urb_entry_read_length = 4; /* grf 1-4 */
3679 dri_bo_emit_reloc(bo,
3680 I915_GEM_DOMAIN_INSTRUCTION, 0,
3682 offsetof(struct gen6_interface_descriptor_data, desc0),
3683 pp_context->pp_modules[pp_index].kernel.bo);
3685 dri_bo_emit_reloc(bo,
3686 I915_GEM_DOMAIN_INSTRUCTION, 0,
3687 desc->desc2.sampler_count << 2,
3688 offsetof(struct gen6_interface_descriptor_data, desc2),
3689 pp_context->sampler_state_table.bo);
3692 pp_context->idrt.num_interface_descriptors++;
3696 gen6_pp_upload_constants(VADriverContextP ctx,
3697 struct i965_post_processing_context *pp_context)
3699 struct i965_driver_data *i965 = i965_driver_data(ctx);
3700 unsigned char *constant_buffer;
3703 assert(sizeof(struct pp_static_parameter) == 128);
3704 assert(sizeof(struct gen7_pp_static_parameter) == 192);
3706 if (IS_GEN7(i965->intel.device_id))
3707 param_size = sizeof(struct gen7_pp_static_parameter);
3709 param_size = sizeof(struct pp_static_parameter);
3711 dri_bo_map(pp_context->curbe.bo, 1);
3712 assert(pp_context->curbe.bo->virtual);
3713 constant_buffer = pp_context->curbe.bo->virtual;
3714 memcpy(constant_buffer, pp_context->pp_static_parameter, param_size);
3715 dri_bo_unmap(pp_context->curbe.bo);
3719 gen6_pp_states_setup(VADriverContextP ctx,
3720 struct i965_post_processing_context *pp_context)
3722 gen6_pp_interface_descriptor_table(ctx, pp_context);
3723 gen6_pp_upload_constants(ctx, pp_context);
3727 gen6_pp_pipeline_select(VADriverContextP ctx,
3728 struct i965_post_processing_context *pp_context)
3730 struct intel_batchbuffer *batch = pp_context->batch;
3732 BEGIN_BATCH(batch, 1);
3733 OUT_BATCH(batch, CMD_PIPELINE_SELECT | PIPELINE_SELECT_MEDIA);
3734 ADVANCE_BATCH(batch);
3738 gen6_pp_state_base_address(VADriverContextP ctx,
3739 struct i965_post_processing_context *pp_context)
3741 struct intel_batchbuffer *batch = pp_context->batch;
3743 BEGIN_BATCH(batch, 10);
3744 OUT_BATCH(batch, CMD_STATE_BASE_ADDRESS | (10 - 2));
3745 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3746 OUT_RELOC(batch, pp_context->surface_state_binding_table.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, BASE_ADDRESS_MODIFY); /* Surface state base address */
3747 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3748 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3749 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3750 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3751 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3752 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3753 OUT_BATCH(batch, 0 | BASE_ADDRESS_MODIFY);
3754 ADVANCE_BATCH(batch);
3758 gen6_pp_vfe_state(VADriverContextP ctx,
3759 struct i965_post_processing_context *pp_context)
3761 struct intel_batchbuffer *batch = pp_context->batch;
3763 BEGIN_BATCH(batch, 8);
3764 OUT_BATCH(batch, CMD_MEDIA_VFE_STATE | (8 - 2));
3765 OUT_BATCH(batch, 0);
3767 (pp_context->urb.num_vfe_entries - 1) << 16 |
3768 pp_context->urb.num_vfe_entries << 8);
3769 OUT_BATCH(batch, 0);
3771 (pp_context->urb.size_vfe_entry * 2) << 16 | /* URB Entry Allocation Size, in 256 bits unit */
3772 (pp_context->urb.size_cs_entry * pp_context->urb.num_cs_entries * 2)); /* CURBE Allocation Size, in 256 bits unit */
3773 OUT_BATCH(batch, 0);
3774 OUT_BATCH(batch, 0);
3775 OUT_BATCH(batch, 0);
3776 ADVANCE_BATCH(batch);
3780 gen6_pp_curbe_load(VADriverContextP ctx,
3781 struct i965_post_processing_context *pp_context)
3783 struct intel_batchbuffer *batch = pp_context->batch;
3785 assert(pp_context->urb.size_cs_entry * pp_context->urb.num_cs_entries * 2 * 32 <= pp_context->curbe.bo->size);
3787 BEGIN_BATCH(batch, 4);
3788 OUT_BATCH(batch, CMD_MEDIA_CURBE_LOAD | (4 - 2));
3789 OUT_BATCH(batch, 0);
3791 pp_context->urb.size_cs_entry * pp_context->urb.num_cs_entries * 2 * 32);
3793 pp_context->curbe.bo,
3794 I915_GEM_DOMAIN_INSTRUCTION, 0,
3796 ADVANCE_BATCH(batch);
3800 gen6_interface_descriptor_load(VADriverContextP ctx,
3801 struct i965_post_processing_context *pp_context)
3803 struct intel_batchbuffer *batch = pp_context->batch;
3805 BEGIN_BATCH(batch, 4);
3806 OUT_BATCH(batch, CMD_MEDIA_INTERFACE_DESCRIPTOR_LOAD | (4 - 2));
3807 OUT_BATCH(batch, 0);
3809 pp_context->idrt.num_interface_descriptors * sizeof(struct gen6_interface_descriptor_data));
3811 pp_context->idrt.bo,
3812 I915_GEM_DOMAIN_INSTRUCTION, 0,
3814 ADVANCE_BATCH(batch);
3817 static void update_block_mask_parameter(struct i965_post_processing_context *pp_context, int x, int y, int x_steps, int y_steps)
3819 struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
3821 pp_inline_parameter->grf5.block_vertical_mask = 0xff;
3822 pp_inline_parameter->grf5.block_horizontal_mask = 0xffff;
3823 pp_inline_parameter->grf6.block_vertical_mask = pp_context->block_vertical_mask;
3824 pp_inline_parameter->grf6.block_horizontal_mask = pp_context->block_horizontal_mask;
3828 if (y == y_steps-1) {
3829 pp_inline_parameter->grf5.block_vertical_mask = pp_context->block_vertical_mask;
3832 pp_inline_parameter->grf6.block_vertical_mask = 0xff;
3838 if (x == x_steps-1) {
3839 pp_inline_parameter->grf5.block_horizontal_mask = pp_context->block_horizontal_mask;
3842 pp_inline_parameter->grf6.block_horizontal_mask = 0xffff;
3849 gen6_pp_object_walker(VADriverContextP ctx,
3850 struct i965_post_processing_context *pp_context)
3852 struct i965_driver_data *i965 = i965_driver_data(ctx);
3853 struct intel_batchbuffer *batch = pp_context->batch;
3854 int x, x_steps, y, y_steps;
3855 int param_size, command_length_in_dws;
3856 dri_bo *command_buffer;
3857 unsigned int *command_ptr;
3859 if (IS_GEN7(i965->intel.device_id))
3860 param_size = sizeof(struct gen7_pp_inline_parameter);
3862 param_size = sizeof(struct pp_inline_parameter);
3864 x_steps = pp_context->pp_x_steps(&pp_context->private_context);
3865 y_steps = pp_context->pp_y_steps(&pp_context->private_context);
3866 command_length_in_dws = 6 + (param_size >> 2);
3867 command_buffer = dri_bo_alloc(i965->intel.bufmgr,
3868 "command objects buffer",
3869 command_length_in_dws * 4 * x_steps * y_steps + 8,
3872 dri_bo_map(command_buffer, 1);
3873 command_ptr = command_buffer->virtual;
3875 for (y = 0; y < y_steps; y++) {
3876 for (x = 0; x < x_steps; x++) {
3877 if (!pp_context->pp_set_block_parameter(pp_context, x, y)) {
3878 // some common block parameter update goes here, apply to all pp functions
3879 update_block_mask_parameter (pp_context, x, y, x_steps, y_steps);
3881 *command_ptr++ = (CMD_MEDIA_OBJECT | (command_length_in_dws - 2));
3887 memcpy(command_ptr, pp_context->pp_inline_parameter, param_size);
3888 command_ptr += (param_size >> 2);
3893 if (command_length_in_dws * x_steps * y_steps % 2 == 0)
3896 *command_ptr = MI_BATCH_BUFFER_END;
3898 dri_bo_unmap(command_buffer);
3900 BEGIN_BATCH(batch, 2);
3901 OUT_BATCH(batch, MI_BATCH_BUFFER_START | (2 << 6));
3902 OUT_RELOC(batch, command_buffer,
3903 I915_GEM_DOMAIN_COMMAND, 0,
3905 ADVANCE_BATCH(batch);
3907 dri_bo_unreference(command_buffer);
3909 /* Have to execute the batch buffer here becuase MI_BATCH_BUFFER_END
3910 * will cause control to pass back to ring buffer
3912 intel_batchbuffer_end_atomic(batch);
3913 intel_batchbuffer_flush(batch);
3914 intel_batchbuffer_start_atomic(batch, 0x1000);
3918 gen6_pp_pipeline_setup(VADriverContextP ctx,
3919 struct i965_post_processing_context *pp_context)
3921 struct intel_batchbuffer *batch = pp_context->batch;
3923 intel_batchbuffer_start_atomic(batch, 0x1000);
3924 intel_batchbuffer_emit_mi_flush(batch);
3925 gen6_pp_pipeline_select(ctx, pp_context);
3926 gen6_pp_state_base_address(ctx, pp_context);
3927 gen6_pp_vfe_state(ctx, pp_context);
3928 gen6_pp_curbe_load(ctx, pp_context);
3929 gen6_interface_descriptor_load(ctx, pp_context);
3930 gen6_pp_object_walker(ctx, pp_context);
3931 intel_batchbuffer_end_atomic(batch);
3935 gen6_post_processing(
3936 VADriverContextP ctx,
3937 struct i965_post_processing_context *pp_context,
3938 const struct i965_surface *src_surface,
3939 const VARectangle *src_rect,
3940 struct i965_surface *dst_surface,
3941 const VARectangle *dst_rect,
3948 va_status = gen6_pp_initialize(ctx, pp_context,
3956 if (va_status == VA_STATUS_SUCCESS) {
3957 gen6_pp_states_setup(ctx, pp_context);
3958 gen6_pp_pipeline_setup(ctx, pp_context);
3965 i965_post_processing_internal(
3966 VADriverContextP ctx,
3967 struct i965_post_processing_context *pp_context,
3968 const struct i965_surface *src_surface,
3969 const VARectangle *src_rect,
3970 struct i965_surface *dst_surface,
3971 const VARectangle *dst_rect,
3976 struct i965_driver_data *i965 = i965_driver_data(ctx);
3979 if (IS_GEN6(i965->intel.device_id) ||
3980 IS_GEN7(i965->intel.device_id))
3981 va_status = gen6_post_processing(ctx, pp_context, src_surface, src_rect, dst_surface, dst_rect, pp_index, filter_param);
3983 va_status = ironlake_post_processing(ctx, pp_context, src_surface, src_rect, dst_surface, dst_rect, pp_index, filter_param);
3989 i965_DestroySurfaces(VADriverContextP ctx,
3990 VASurfaceID *surface_list,
3993 i965_CreateSurfaces(VADriverContextP ctx,
3998 VASurfaceID *surfaces);
4001 rgb_to_yuv(unsigned int argb,
4007 int r = ((argb >> 16) & 0xff);
4008 int g = ((argb >> 8) & 0xff);
4009 int b = ((argb >> 0) & 0xff);
4011 *y = (257 * r + 504 * g + 98 * b) / 1000 + 16;
4012 *v = (439 * r - 368 * g - 71 * b) / 1000 + 128;
4013 *u = (-148 * r - 291 * g + 439 * b) / 1000 + 128;
4014 *a = ((argb >> 24) & 0xff);
4018 i965_vpp_clear_surface(VADriverContextP ctx,
4019 struct i965_post_processing_context *pp_context,
4020 VASurfaceID surface,
4023 struct i965_driver_data *i965 = i965_driver_data(ctx);
4024 struct intel_batchbuffer *batch = pp_context->batch;
4025 struct object_surface *obj_surface = SURFACE(surface);
4026 unsigned int blt_cmd, br13;
4027 unsigned int tiling = 0, swizzle = 0;
4029 unsigned char y, u, v, a = 0;
4031 /* Currently only support NV12 surface */
4032 if (!obj_surface || obj_surface->fourcc != VA_FOURCC('N', 'V', '1', '2'))
4035 rgb_to_yuv(color, &y, &u, &v, &a);
4040 dri_bo_get_tiling(obj_surface->bo, &tiling, &swizzle);
4041 blt_cmd = XY_COLOR_BLT_CMD;
4042 pitch = obj_surface->width;
4044 if (tiling != I915_TILING_NONE) {
4045 blt_cmd |= XY_COLOR_BLT_DST_TILED;
4053 if (IS_GEN6(i965->intel.device_id) ||
4054 IS_GEN7(i965->intel.device_id)) {
4055 intel_batchbuffer_start_atomic_blt(batch, 48);
4056 BEGIN_BLT_BATCH(batch, 12);
4058 intel_batchbuffer_start_atomic(batch, 48);
4059 BEGIN_BATCH(batch, 12);
4062 OUT_BATCH(batch, blt_cmd);
4063 OUT_BATCH(batch, br13);
4068 obj_surface->height << 16 |
4069 obj_surface->width);
4070 OUT_RELOC(batch, obj_surface->bo,
4071 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
4073 OUT_BATCH(batch, y);
4079 OUT_BATCH(batch, blt_cmd);
4080 OUT_BATCH(batch, br13);
4085 obj_surface->height / 2 << 16 |
4086 obj_surface->width / 2);
4087 OUT_RELOC(batch, obj_surface->bo,
4088 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
4089 obj_surface->width * obj_surface->y_cb_offset);
4090 OUT_BATCH(batch, v << 8 | u);
4092 ADVANCE_BATCH(batch);
4093 intel_batchbuffer_end_atomic(batch);
4097 i965_post_processing(
4098 VADriverContextP ctx,
4099 VASurfaceID surface,
4100 const VARectangle *src_rect,
4101 const VARectangle *dst_rect,
4103 int *has_done_scaling
4106 struct i965_driver_data *i965 = i965_driver_data(ctx);
4107 VASurfaceID in_surface_id = surface;
4108 VASurfaceID out_surface_id = VA_INVALID_ID;
4110 *has_done_scaling = 0;
4113 struct object_surface *obj_surface;
4115 struct i965_surface src_surface;
4116 struct i965_surface dst_surface;
4118 obj_surface = SURFACE(in_surface_id);
4120 /* Currently only support post processing for NV12 surface */
4121 if (obj_surface->fourcc != VA_FOURCC('N', 'V', '1', '2'))
4122 return out_surface_id;
4124 _i965LockMutex(&i965->pp_mutex);
4126 if (flags & I965_PP_FLAG_MCDI) {
4127 status = i965_CreateSurfaces(ctx,
4128 obj_surface->orig_width,
4129 obj_surface->orig_height,
4130 VA_RT_FORMAT_YUV420,
4133 assert(status == VA_STATUS_SUCCESS);
4134 obj_surface = SURFACE(out_surface_id);
4135 i965_check_alloc_surface_bo(ctx, obj_surface, 0, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420);
4136 i965_vpp_clear_surface(ctx, i965->pp_context, out_surface_id, 0);
4137 src_surface.id = in_surface_id;
4138 src_surface.type = I965_SURFACE_TYPE_SURFACE;
4139 src_surface.flags = (flags & I965_PP_FLAG_TOP_FIELD) ?
4140 I965_SURFACE_FLAG_TOP_FIELD_FIRST : I965_SURFACE_FLAG_BOTTOME_FIELD_FIRST;
4141 dst_surface.id = out_surface_id;
4142 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4143 dst_surface.flags = I965_SURFACE_FLAG_FRAME;
4145 i965_post_processing_internal(ctx, i965->pp_context,
4154 if (flags & I965_PP_FLAG_AVS) {
4155 struct i965_render_state *render_state = &i965->render_state;
4156 struct intel_region *dest_region = render_state->draw_region;
4158 if (out_surface_id != VA_INVALID_ID)
4159 in_surface_id = out_surface_id;
4161 status = i965_CreateSurfaces(ctx,
4163 dest_region->height,
4164 VA_RT_FORMAT_YUV420,
4167 assert(status == VA_STATUS_SUCCESS);
4168 obj_surface = SURFACE(out_surface_id);
4169 i965_check_alloc_surface_bo(ctx, obj_surface, 0, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420);
4170 i965_vpp_clear_surface(ctx, i965->pp_context, out_surface_id, 0);
4171 src_surface.id = in_surface_id;
4172 src_surface.type = I965_SURFACE_TYPE_SURFACE;
4173 src_surface.flags = I965_SURFACE_FLAG_FRAME;
4174 dst_surface.id = out_surface_id;
4175 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4176 dst_surface.flags = I965_SURFACE_FLAG_FRAME;
4178 i965_post_processing_internal(ctx, i965->pp_context,
4186 if (in_surface_id != surface)
4187 i965_DestroySurfaces(ctx, &in_surface_id, 1);
4189 *has_done_scaling = 1;
4192 _i965UnlockMutex(&i965->pp_mutex);
4195 return out_surface_id;
4199 i965_image_pl1_rgbx_processing(VADriverContextP ctx,
4200 const struct i965_surface *src_surface,
4201 const VARectangle *src_rect,
4202 struct i965_surface *dst_surface,
4203 const VARectangle *dst_rect)
4205 struct i965_driver_data *i965 = i965_driver_data(ctx);
4206 struct i965_post_processing_context *pp_context = i965->pp_context;
4207 int fourcc = pp_get_surface_fourcc(ctx, dst_surface);
4209 if (fourcc == VA_FOURCC('N', 'V', '1', '2')) {
4210 i965_post_processing_internal(ctx, i965->pp_context,
4215 PP_RGBX_LOAD_SAVE_NV12,
4219 return VA_STATUS_ERROR_UNKNOWN;
4222 intel_batchbuffer_flush(pp_context->batch);
4224 return VA_STATUS_SUCCESS;
4228 i965_image_pl3_processing(VADriverContextP ctx,
4229 const struct i965_surface *src_surface,
4230 const VARectangle *src_rect,
4231 struct i965_surface *dst_surface,
4232 const VARectangle *dst_rect)
4234 struct i965_driver_data *i965 = i965_driver_data(ctx);
4235 struct i965_post_processing_context *pp_context = i965->pp_context;
4236 int fourcc = pp_get_surface_fourcc(ctx, dst_surface);
4237 VAStatus vaStatus = VA_STATUS_ERROR_UNIMPLEMENTED;
4239 if (fourcc == VA_FOURCC('N', 'V', '1', '2')) {
4240 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4245 PP_PL3_LOAD_SAVE_N12,
4247 } else if (fourcc == VA_FOURCC('I', 'M', 'C', '1') ||
4248 fourcc == VA_FOURCC('I', 'M', 'C', '3') ||
4249 fourcc == VA_FOURCC('Y', 'V', '1', '2') ||
4250 fourcc == VA_FOURCC('I', '4', '2', '0')) {
4251 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4256 PP_PL3_LOAD_SAVE_PL3,
4258 } else if (fourcc == VA_FOURCC('Y', 'U', 'Y', '2') ||
4259 fourcc == VA_FOURCC('U', 'Y', 'V', 'Y')) {
4260 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4265 PP_PL3_LOAD_SAVE_PA,
4272 intel_batchbuffer_flush(pp_context->batch);
4278 i965_image_pl2_processing(VADriverContextP ctx,
4279 const struct i965_surface *src_surface,
4280 const VARectangle *src_rect,
4281 struct i965_surface *dst_surface,
4282 const VARectangle *dst_rect)
4284 struct i965_driver_data *i965 = i965_driver_data(ctx);
4285 struct i965_post_processing_context *pp_context = i965->pp_context;
4286 int fourcc = pp_get_surface_fourcc(ctx, dst_surface);
4287 VAStatus vaStatus = VA_STATUS_ERROR_UNIMPLEMENTED;
4289 if (fourcc == VA_FOURCC('N', 'V', '1', '2')) {
4290 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4295 PP_NV12_LOAD_SAVE_N12,
4297 } else if (fourcc == VA_FOURCC('I', 'M', 'C', '1') ||
4298 fourcc == VA_FOURCC('I', 'M', 'C', '3') ||
4299 fourcc == VA_FOURCC('Y', 'V', '1', '2') ||
4300 fourcc == VA_FOURCC('I', '4', '2', '0') ) {
4301 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4306 PP_NV12_LOAD_SAVE_PL3,
4308 } else if (fourcc == VA_FOURCC('Y', 'U', 'Y', '2') ||
4309 fourcc == VA_FOURCC('U', 'Y', 'V', 'Y')) {
4310 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4315 PP_NV12_LOAD_SAVE_PA,
4317 } else if (fourcc == VA_FOURCC('B', 'G', 'R', 'X') ||
4318 fourcc == VA_FOURCC('B', 'G', 'R', 'A') ||
4319 fourcc == VA_FOURCC('R', 'G', 'B', 'X') ||
4320 fourcc == VA_FOURCC('R', 'G', 'B', 'A') ) {
4321 vaStatus = i965_post_processing_internal(ctx, i965->pp_context,
4326 PP_NV12_LOAD_SAVE_RGBX,
4330 return VA_STATUS_ERROR_UNKNOWN;
4333 intel_batchbuffer_flush(pp_context->batch);
4339 i965_image_pl1_processing(VADriverContextP ctx,
4340 const struct i965_surface *src_surface,
4341 const VARectangle *src_rect,
4342 struct i965_surface *dst_surface,
4343 const VARectangle *dst_rect)
4345 struct i965_driver_data *i965 = i965_driver_data(ctx);
4346 struct i965_post_processing_context *pp_context = i965->pp_context;
4347 int fourcc = pp_get_surface_fourcc(ctx, dst_surface);
4349 if (fourcc == VA_FOURCC('N', 'V', '1', '2')) {
4350 i965_post_processing_internal(ctx, i965->pp_context,
4355 PP_PA_LOAD_SAVE_NV12,
4358 else if (fourcc == VA_FOURCC_YV12) {
4359 i965_post_processing_internal(ctx, i965->pp_context,
4364 PP_PA_LOAD_SAVE_PL3,
4369 return VA_STATUS_ERROR_UNKNOWN;
4372 intel_batchbuffer_flush(pp_context->batch);
4374 return VA_STATUS_SUCCESS;
4378 i965_image_processing(VADriverContextP ctx,
4379 const struct i965_surface *src_surface,
4380 const VARectangle *src_rect,
4381 struct i965_surface *dst_surface,
4382 const VARectangle *dst_rect)
4384 struct i965_driver_data *i965 = i965_driver_data(ctx);
4385 VAStatus status = VA_STATUS_ERROR_UNIMPLEMENTED;
4388 int fourcc = pp_get_surface_fourcc(ctx, src_surface);
4390 _i965LockMutex(&i965->pp_mutex);
4393 case VA_FOURCC('Y', 'V', '1', '2'):
4394 case VA_FOURCC('I', '4', '2', '0'):
4395 case VA_FOURCC('I', 'M', 'C', '1'):
4396 case VA_FOURCC('I', 'M', 'C', '3'):
4397 status = i965_image_pl3_processing(ctx,
4404 case VA_FOURCC('N', 'V', '1', '2'):
4405 status = i965_image_pl2_processing(ctx,
4411 case VA_FOURCC('Y', 'U', 'Y', '2'):
4412 case VA_FOURCC('U', 'Y', 'V', 'Y'):
4413 status = i965_image_pl1_processing(ctx,
4419 case VA_FOURCC('B', 'G', 'R', 'A'):
4420 case VA_FOURCC('B', 'G', 'R', 'X'):
4421 case VA_FOURCC('R', 'G', 'B', 'A'):
4422 case VA_FOURCC('R', 'G', 'B', 'X'):
4423 status = i965_image_pl1_rgbx_processing(ctx,
4430 status = VA_STATUS_ERROR_UNIMPLEMENTED;
4434 _i965UnlockMutex(&i965->pp_mutex);
4441 i965_post_processing_context_finalize(struct i965_post_processing_context *pp_context)
4445 dri_bo_unreference(pp_context->surface_state_binding_table.bo);
4446 pp_context->surface_state_binding_table.bo = NULL;
4448 dri_bo_unreference(pp_context->curbe.bo);
4449 pp_context->curbe.bo = NULL;
4451 dri_bo_unreference(pp_context->sampler_state_table.bo);
4452 pp_context->sampler_state_table.bo = NULL;
4454 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8);
4455 pp_context->sampler_state_table.bo_8x8 = NULL;
4457 dri_bo_unreference(pp_context->sampler_state_table.bo_8x8_uv);
4458 pp_context->sampler_state_table.bo_8x8_uv = NULL;
4460 dri_bo_unreference(pp_context->idrt.bo);
4461 pp_context->idrt.bo = NULL;
4462 pp_context->idrt.num_interface_descriptors = 0;
4464 dri_bo_unreference(pp_context->vfe_state.bo);
4465 pp_context->vfe_state.bo = NULL;
4467 dri_bo_unreference(pp_context->stmm.bo);
4468 pp_context->stmm.bo = NULL;
4470 for (i = 0; i < NUM_PP_MODULES; i++) {
4471 struct pp_module *pp_module = &pp_context->pp_modules[i];
4473 dri_bo_unreference(pp_module->kernel.bo);
4474 pp_module->kernel.bo = NULL;
4477 free(pp_context->pp_static_parameter);
4478 free(pp_context->pp_inline_parameter);
4479 pp_context->pp_static_parameter = NULL;
4480 pp_context->pp_inline_parameter = NULL;
4484 i965_post_processing_terminate(VADriverContextP ctx)
4486 struct i965_driver_data *i965 = i965_driver_data(ctx);
4487 struct i965_post_processing_context *pp_context = i965->pp_context;
4490 i965_post_processing_context_finalize(pp_context);
4494 i965->pp_context = NULL;
4500 i965_post_processing_context_init(VADriverContextP ctx,
4501 struct i965_post_processing_context *pp_context,
4502 struct intel_batchbuffer *batch)
4504 struct i965_driver_data *i965 = i965_driver_data(ctx);
4507 pp_context->urb.size = URB_SIZE((&i965->intel));
4508 pp_context->urb.num_vfe_entries = 32;
4509 pp_context->urb.size_vfe_entry = 1; /* in 512 bits unit */
4510 pp_context->urb.num_cs_entries = 1;
4512 if (IS_GEN7(i965->intel.device_id))
4513 pp_context->urb.size_cs_entry = 4; /* in 512 bits unit */
4515 pp_context->urb.size_cs_entry = 2;
4517 pp_context->urb.vfe_start = 0;
4518 pp_context->urb.cs_start = pp_context->urb.vfe_start +
4519 pp_context->urb.num_vfe_entries * pp_context->urb.size_vfe_entry;
4520 assert(pp_context->urb.cs_start +
4521 pp_context->urb.num_cs_entries * pp_context->urb.size_cs_entry <= URB_SIZE((&i965->intel)));
4523 assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen5));
4524 assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen6));
4525 assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen7));
4527 if (IS_GEN7(i965->intel.device_id))
4528 memcpy(pp_context->pp_modules, pp_modules_gen7, sizeof(pp_context->pp_modules));
4529 else if (IS_GEN6(i965->intel.device_id))
4530 memcpy(pp_context->pp_modules, pp_modules_gen6, sizeof(pp_context->pp_modules));
4531 else if (IS_IRONLAKE(i965->intel.device_id))
4532 memcpy(pp_context->pp_modules, pp_modules_gen5, sizeof(pp_context->pp_modules));
4534 for (i = 0; i < NUM_PP_MODULES; i++) {
4535 struct pp_module *pp_module = &pp_context->pp_modules[i];
4536 dri_bo_unreference(pp_module->kernel.bo);
4537 if (pp_module->kernel.bin && pp_module->kernel.size) {
4538 pp_module->kernel.bo = dri_bo_alloc(i965->intel.bufmgr,
4539 pp_module->kernel.name,
4540 pp_module->kernel.size,
4542 assert(pp_module->kernel.bo);
4543 dri_bo_subdata(pp_module->kernel.bo, 0, pp_module->kernel.size, pp_module->kernel.bin);
4545 pp_module->kernel.bo = NULL;
4549 /* static & inline parameters */
4550 if (IS_GEN7(i965->intel.device_id)) {
4551 pp_context->pp_static_parameter = calloc(sizeof(struct gen7_pp_static_parameter), 1);
4552 pp_context->pp_inline_parameter = calloc(sizeof(struct gen7_pp_inline_parameter), 1);
4554 pp_context->pp_static_parameter = calloc(sizeof(struct pp_static_parameter), 1);
4555 pp_context->pp_inline_parameter = calloc(sizeof(struct pp_inline_parameter), 1);
4558 pp_context->batch = batch;
4562 i965_post_processing_init(VADriverContextP ctx)
4564 struct i965_driver_data *i965 = i965_driver_data(ctx);
4565 struct i965_post_processing_context *pp_context = i965->pp_context;
4568 if (pp_context == NULL) {
4569 pp_context = calloc(1, sizeof(*pp_context));
4570 i965_post_processing_context_init(ctx, pp_context, i965->batch);
4571 i965->pp_context = pp_context;
4578 static const int procfilter_to_pp_flag[VAProcFilterCount] = {
4579 PP_NULL, /* VAProcFilterNone */
4580 PP_NV12_DN, /* VAProcFilterNoiseReduction */
4581 PP_NULL, /* VAProcFilterDeblocking */
4582 PP_NV12_DNDI, /* VAProcFilterDeinterlacing */
4583 PP_NULL, /* VAProcFilterSharpening */
4584 PP_NULL, /* VAProcFilterColorBalance */
4585 PP_NULL, /* VAProcFilterColorStandard */
4586 PP_NULL, /* VAProcFilterFrameRateConversion */
4589 static const int proc_frame_to_pp_frame[3] = {
4590 I965_SURFACE_FLAG_FRAME,
4591 I965_SURFACE_FLAG_TOP_FIELD_FIRST,
4592 I965_SURFACE_FLAG_BOTTOME_FIELD_FIRST
4596 i965_proc_picture(VADriverContextP ctx,
4598 union codec_state *codec_state,
4599 struct hw_context *hw_context)
4601 struct i965_driver_data *i965 = i965_driver_data(ctx);
4602 struct i965_proc_context *proc_context = (struct i965_proc_context *)hw_context;
4603 struct proc_state *proc_state = &codec_state->proc;
4604 VAProcPipelineParameterBuffer *pipeline_param = (VAProcPipelineParameterBuffer *)proc_state->pipeline_param->buffer;
4605 struct object_surface *obj_surface;
4606 struct i965_surface src_surface, dst_surface;
4607 VARectangle src_rect, dst_rect;
4610 VASurfaceID tmp_surfaces[VAProcFilterCount + 4];
4611 int num_tmp_surfaces = 0;
4612 unsigned int tiling = 0, swizzle = 0;
4613 int in_width, in_height;
4615 assert(pipeline_param->surface != VA_INVALID_ID);
4616 assert(proc_state->current_render_target != VA_INVALID_ID);
4618 obj_surface = SURFACE(pipeline_param->surface);
4619 in_width = obj_surface->orig_width;
4620 in_height = obj_surface->orig_height;
4621 dri_bo_get_tiling(obj_surface->bo, &tiling, &swizzle);
4623 src_surface.id = pipeline_param->surface;
4624 src_surface.type = I965_SURFACE_TYPE_SURFACE;
4625 src_surface.flags = proc_frame_to_pp_frame[pipeline_param->filter_flags & 0x3];
4627 VASurfaceID out_surface_id = VA_INVALID_ID;
4628 if (obj_surface->fourcc != VA_FOURCC('N', 'V', '1', '2')) {
4629 src_surface.id = pipeline_param->surface;
4630 src_surface.type = I965_SURFACE_TYPE_SURFACE;
4631 src_surface.flags = I965_SURFACE_FLAG_FRAME;
4634 src_rect.width = in_width;
4635 src_rect.height = in_height;
4637 status = i965_CreateSurfaces(ctx,
4640 VA_RT_FORMAT_YUV420,
4643 assert(status == VA_STATUS_SUCCESS);
4644 tmp_surfaces[num_tmp_surfaces++] = out_surface_id;
4645 obj_surface = SURFACE(out_surface_id);
4646 i965_check_alloc_surface_bo(ctx, obj_surface, !!tiling, VA_FOURCC('N', 'V', '1', '2'), SUBSAMPLE_YUV420);
4648 dst_surface.id = out_surface_id;
4649 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4650 dst_surface.flags = I965_SURFACE_FLAG_FRAME;
4653 dst_rect.width = in_width;
4654 dst_rect.height = in_height;
4656 status = i965_image_processing(ctx,
4661 assert(status == VA_STATUS_SUCCESS);
4663 src_surface.id = out_surface_id;
4664 src_surface.type = I965_SURFACE_TYPE_SURFACE;
4665 src_surface.flags = proc_frame_to_pp_frame[pipeline_param->filter_flags & 0x3];
4668 if (pipeline_param->surface_region) {
4669 src_rect.x = pipeline_param->surface_region->x;
4670 src_rect.y = pipeline_param->surface_region->y;
4671 src_rect.width = pipeline_param->surface_region->width;
4672 src_rect.height = pipeline_param->surface_region->height;
4676 src_rect.width = in_width;
4677 src_rect.height = in_height;
4680 if (pipeline_param->output_region) {
4681 dst_rect.x = pipeline_param->output_region->x;
4682 dst_rect.y = pipeline_param->output_region->y;
4683 dst_rect.width = pipeline_param->output_region->width;
4684 dst_rect.height = pipeline_param->output_region->height;
4688 dst_rect.width = in_width;
4689 dst_rect.height = in_height;
4692 for (i = 0; i < pipeline_param->num_filters; i++) {
4693 struct object_buffer *obj_buffer = BUFFER(pipeline_param->filters[i]);
4694 VAProcFilterParameterBufferBase *filter_param = (VAProcFilterParameterBufferBase *)obj_buffer->buffer_store->buffer;
4695 VAProcFilterType filter_type = filter_param->type;
4696 out_surface_id = VA_INVALID_ID;
4697 int kernel_index = procfilter_to_pp_flag[filter_type];
4699 if (kernel_index != PP_NULL &&
4700 proc_context->pp_context.pp_modules[kernel_index].kernel.bo != NULL) {
4701 status = i965_CreateSurfaces(ctx,
4704 VA_RT_FORMAT_YUV420,
4707 assert(status == VA_STATUS_SUCCESS);
4708 tmp_surfaces[num_tmp_surfaces++] = out_surface_id;
4709 obj_surface = SURFACE(out_surface_id);
4710 i965_check_alloc_surface_bo(ctx, obj_surface, !!tiling, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420);
4711 dst_surface.id = out_surface_id;
4712 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4713 status = i965_post_processing_internal(ctx, &proc_context->pp_context,
4721 if (status == VA_STATUS_SUCCESS) {
4722 src_surface.id = dst_surface.id;
4723 src_surface.type = dst_surface.type;
4724 src_surface.flags = dst_surface.flags;
4729 obj_surface = SURFACE(proc_state->current_render_target);
4731 if (obj_surface->fourcc && obj_surface->fourcc != VA_FOURCC('N','V','1','2')){
4733 out_surface_id = VA_INVALID_ID;
4734 status = i965_CreateSurfaces(ctx,
4735 obj_surface->orig_width,
4736 obj_surface->orig_height,
4737 VA_RT_FORMAT_YUV420,
4740 assert(status == VA_STATUS_SUCCESS);
4741 tmp_surfaces[num_tmp_surfaces++] = out_surface_id;
4742 struct object_surface *csc_surface = SURFACE(out_surface_id);
4743 i965_check_alloc_surface_bo(ctx, csc_surface, !!tiling, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420);
4744 dst_surface.id = out_surface_id;
4746 i965_check_alloc_surface_bo(ctx, obj_surface, !!tiling, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420);
4747 dst_surface.id = proc_state->current_render_target;
4750 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4751 i965_vpp_clear_surface(ctx, &proc_context->pp_context, proc_state->current_render_target, pipeline_param->output_background_color);
4752 if (src_rect.width == dst_rect.width &&
4753 src_rect.height == dst_rect.height) {
4754 i965_post_processing_internal(ctx, &proc_context->pp_context,
4759 PP_NV12_LOAD_SAVE_N12,
4763 i965_post_processing_internal(ctx, &proc_context->pp_context,
4768 (pipeline_param->filter_flags & VA_FILTER_SCALING_MASK) == VA_FILTER_SCALING_NL_ANAMORPHIC ?
4769 PP_NV12_AVS : PP_NV12_SCALING,
4774 src_surface.id = dst_surface.id;
4775 src_surface.type = dst_surface.type;
4776 src_surface.flags = dst_surface.flags;
4777 dst_surface.id = proc_state->current_render_target;
4778 dst_surface.type = I965_SURFACE_TYPE_SURFACE;
4779 i965_image_processing(ctx, &src_surface, &dst_rect, &dst_surface, &dst_rect);
4782 if (num_tmp_surfaces)
4783 i965_DestroySurfaces(ctx,
4787 intel_batchbuffer_flush(hw_context->batch);
4791 i965_proc_context_destroy(void *hw_context)
4793 struct i965_proc_context *proc_context = (struct i965_proc_context *)hw_context;
4795 i965_post_processing_context_finalize(&proc_context->pp_context);
4796 intel_batchbuffer_free(proc_context->base.batch);
4801 i965_proc_context_init(VADriverContextP ctx, struct object_config *obj_config)
4803 struct intel_driver_data *intel = intel_driver_data(ctx);
4804 struct i965_proc_context *proc_context = calloc(1, sizeof(struct i965_proc_context));
4806 proc_context->base.destroy = i965_proc_context_destroy;
4807 proc_context->base.run = i965_proc_picture;
4808 proc_context->base.batch = intel_batchbuffer_new(intel, I915_EXEC_RENDER);
4809 i965_post_processing_context_init(ctx, &proc_context->pp_context, proc_context->base.batch);
4811 return (struct hw_context *)proc_context;