1 /**************************************************************************
3 * Copyright 2010 Christian König
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 #include "vl_vertex_buffers.h"
30 #include "util/u_draw.h"
32 #include <pipe/p_context.h>
33 #include <pipe/p_screen.h>
34 #include <util/u_inlines.h>
35 #include <util/u_sampler.h>
36 #include <util/u_format.h>
37 #include <tgsi/tgsi_ureg.h>
41 #define BLOCK_HEIGHT 8
43 #define SCALE_FACTOR_16_TO_9 (32768.0f / 256.0f)
45 #define NR_RENDER_TARGETS 4
56 static const float const_matrix
[8][8] = {
57 { 0.3535530f
, 0.3535530f
, 0.3535530f
, 0.3535530f
, 0.3535530f
, 0.3535530f
, 0.353553f
, 0.3535530f
},
58 { 0.4903930f
, 0.4157350f
, 0.2777850f
, 0.0975451f
, -0.0975452f
, -0.2777850f
, -0.415735f
, -0.4903930f
},
59 { 0.4619400f
, 0.1913420f
, -0.1913420f
, -0.4619400f
, -0.4619400f
, -0.1913420f
, 0.191342f
, 0.4619400f
},
60 { 0.4157350f
, -0.0975452f
, -0.4903930f
, -0.2777850f
, 0.2777850f
, 0.4903930f
, 0.097545f
, -0.4157350f
},
61 { 0.3535530f
, -0.3535530f
, -0.3535530f
, 0.3535540f
, 0.3535530f
, -0.3535540f
, -0.353553f
, 0.3535530f
},
62 { 0.2777850f
, -0.4903930f
, 0.0975452f
, 0.4157350f
, -0.4157350f
, -0.0975451f
, 0.490393f
, -0.2777850f
},
63 { 0.1913420f
, -0.4619400f
, 0.4619400f
, -0.1913420f
, -0.1913410f
, 0.4619400f
, -0.461940f
, 0.1913420f
},
64 { 0.0975451f
, -0.2777850f
, 0.4157350f
, -0.4903930f
, 0.4903930f
, -0.4157350f
, 0.277786f
, -0.0975458f
}
68 calc_addr(struct ureg_program
*shader
, struct ureg_dst addr
[2],
69 struct ureg_src tc
, struct ureg_src start
, bool right_side
,
70 bool transposed
, float size
)
72 unsigned wm_start
= (right_side
== transposed
) ? TGSI_WRITEMASK_X
: TGSI_WRITEMASK_Y
;
73 unsigned sw_start
= right_side
? TGSI_SWIZZLE_Y
: TGSI_SWIZZLE_X
;
75 unsigned wm_tc
= (right_side
== transposed
) ? TGSI_WRITEMASK_Y
: TGSI_WRITEMASK_X
;
76 unsigned sw_tc
= right_side
? TGSI_SWIZZLE_X
: TGSI_SWIZZLE_Y
;
79 * addr[0..1].(start) = right_side ? start.x : tc.x
80 * addr[0..1].(tc) = right_side ? tc.y : start.y
82 * addr[1].(start) += 1.0f / scale
84 ureg_MOV(shader
, ureg_writemask(addr
[0], wm_start
), ureg_scalar(start
, sw_start
));
85 ureg_MOV(shader
, ureg_writemask(addr
[0], wm_tc
), ureg_scalar(tc
, sw_tc
));
86 ureg_MOV(shader
, ureg_writemask(addr
[0], TGSI_WRITEMASK_Z
), tc
);
88 ureg_ADD(shader
, ureg_writemask(addr
[1], wm_start
), ureg_scalar(start
, sw_start
), ureg_imm1f(shader
, 1.0f
/ size
));
89 ureg_MOV(shader
, ureg_writemask(addr
[1], wm_tc
), ureg_scalar(tc
, sw_tc
));
90 ureg_MOV(shader
, ureg_writemask(addr
[1], TGSI_WRITEMASK_Z
), tc
);
94 create_vert_shader(struct vl_idct
*idct
, bool matrix_stage
, int color_swizzle
)
96 struct ureg_program
*shader
;
97 struct ureg_src vrect
, vpos
, vblock
, eb
[4];
98 struct ureg_src scale
, blocks_xy
, t_eb
;
99 struct ureg_dst t_tex
, t_start
;
100 struct ureg_dst o_vpos
, o_l_addr
[2], o_r_addr
[2];
103 shader
= ureg_create(TGSI_PROCESSOR_VERTEX
);
107 t_tex
= ureg_DECL_temporary(shader
);
108 t_start
= ureg_DECL_temporary(shader
);
110 vrect
= ureg_DECL_vs_input(shader
, VS_I_RECT
);
111 vpos
= ureg_DECL_vs_input(shader
, VS_I_VPOS
);
112 vblock
= ureg_swizzle(vrect
, TGSI_SWIZZLE_Z
, TGSI_SWIZZLE_W
, TGSI_SWIZZLE_X
, TGSI_SWIZZLE_X
);
114 o_vpos
= ureg_DECL_output(shader
, TGSI_SEMANTIC_POSITION
, VS_O_VPOS
);
116 eb
[0] = ureg_DECL_vs_input(shader
, VS_I_EB_0_0
);
117 eb
[1] = ureg_DECL_vs_input(shader
, VS_I_EB_1_0
);
118 eb
[2] = ureg_DECL_vs_input(shader
, VS_I_EB_0_1
);
119 eb
[3] = ureg_DECL_vs_input(shader
, VS_I_EB_1_1
);
121 o_l_addr
[0] = ureg_DECL_output(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR0
);
122 o_l_addr
[1] = ureg_DECL_output(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR1
);
124 o_r_addr
[0] = ureg_DECL_output(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR0
);
125 o_r_addr
[1] = ureg_DECL_output(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR1
);
128 * scale = (BLOCK_WIDTH, BLOCK_HEIGHT) / (dst.width, dst.height)
129 * blocks_xy = (blocks_x, blocks_y)
131 * ar = vblock.y * blocks.x + vblock.x
132 * if eb[ar].(color_swizzle)
135 * t_tex = vpos * blocks_xy + vblock
136 * t_start = t_tex * scale
137 * t_tex = t_tex + vrect
138 * o_vpos.xy = t_tex * scale
140 * o_l_addr = calc_addr(...)
141 * o_r_addr = calc_addr(...)
147 scale
= ureg_imm2f(shader
,
148 (float)BLOCK_WIDTH
/ idct
->buffer_width
,
149 (float)BLOCK_HEIGHT
/ idct
->buffer_height
);
151 blocks_xy
= ureg_imm2f(shader
, idct
->blocks_x
, idct
->blocks_y
);
153 if (idct
->blocks_x
> 1 || idct
->blocks_y
> 1) {
154 struct ureg_dst ar
= ureg_DECL_address(shader
);
156 ureg_MAD(shader
, ureg_writemask(t_tex
, TGSI_WRITEMASK_X
),
157 ureg_scalar(vblock
, TGSI_SWIZZLE_Y
), blocks_xy
, vblock
);
159 ureg_ARL(shader
, ureg_writemask(ar
, TGSI_WRITEMASK_X
), ureg_src(t_tex
));
160 t_eb
= ureg_src_indirect(eb
[0], ureg_src(ar
));
165 ureg_IF(shader
, ureg_scalar(t_eb
, color_swizzle
), &label
);
167 ureg_MOV(shader
, o_vpos
, ureg_imm1f(shader
, -1.0f
));
169 ureg_fixup_label(shader
, label
, ureg_get_instruction_number(shader
));
170 ureg_ELSE(shader
, &label
);
172 ureg_MAD(shader
, ureg_writemask(t_tex
, TGSI_WRITEMASK_XY
), vpos
, blocks_xy
, vblock
);
173 ureg_MUL(shader
, ureg_writemask(t_start
, TGSI_WRITEMASK_XY
), ureg_src(t_tex
), scale
);
175 ureg_ADD(shader
, ureg_writemask(t_tex
, TGSI_WRITEMASK_XY
), ureg_src(t_tex
), vrect
);
177 ureg_MUL(shader
, ureg_writemask(t_tex
, TGSI_WRITEMASK_XY
), ureg_src(t_tex
), scale
);
178 ureg_MUL(shader
, ureg_writemask(t_tex
, TGSI_WRITEMASK_Z
),
179 ureg_scalar(vrect
, TGSI_SWIZZLE_X
),
180 ureg_imm1f(shader
, BLOCK_WIDTH
/ NR_RENDER_TARGETS
));
182 ureg_MOV(shader
, ureg_writemask(o_vpos
, TGSI_WRITEMASK_XY
), ureg_src(t_tex
));
185 calc_addr(shader
, o_l_addr
, ureg_src(t_tex
), ureg_src(t_start
), false, false, idct
->buffer_width
/ 4);
186 calc_addr(shader
, o_r_addr
, vrect
, ureg_imm1f(shader
, 0.0f
), true, true, BLOCK_WIDTH
/ 4);
188 calc_addr(shader
, o_l_addr
, vrect
, ureg_imm1f(shader
, 0.0f
), false, false, BLOCK_WIDTH
/ 4);
189 calc_addr(shader
, o_r_addr
, ureg_src(t_tex
), ureg_src(t_start
), true, false, idct
->buffer_height
/ 4);
192 ureg_fixup_label(shader
, label
, ureg_get_instruction_number(shader
));
195 ureg_MOV(shader
, ureg_writemask(o_vpos
, TGSI_WRITEMASK_ZW
), vpos
);
197 ureg_release_temporary(shader
, t_tex
);
198 ureg_release_temporary(shader
, t_start
);
202 return ureg_create_shader_and_destroy(shader
, idct
->pipe
);
206 increment_addr(struct ureg_program
*shader
, struct ureg_dst daddr
[2],
207 struct ureg_src saddr
[2], bool right_side
, bool transposed
,
210 unsigned wm_start
= (right_side
== transposed
) ? TGSI_WRITEMASK_X
: TGSI_WRITEMASK_Y
;
211 unsigned wm_tc
= (right_side
== transposed
) ? TGSI_WRITEMASK_Y
: TGSI_WRITEMASK_X
;
214 * daddr[0..1].(start) = saddr[0..1].(start)
215 * daddr[0..1].(tc) = saddr[0..1].(tc)
218 ureg_MOV(shader
, ureg_writemask(daddr
[0], wm_start
), saddr
[0]);
219 ureg_ADD(shader
, ureg_writemask(daddr
[0], wm_tc
), saddr
[0], ureg_imm1f(shader
, pos
/ size
));
220 ureg_MOV(shader
, ureg_writemask(daddr
[1], wm_start
), saddr
[1]);
221 ureg_ADD(shader
, ureg_writemask(daddr
[1], wm_tc
), saddr
[1], ureg_imm1f(shader
, pos
/ size
));
225 fetch_four(struct ureg_program
*shader
, struct ureg_dst m
[2], struct ureg_src addr
[2], struct ureg_src sampler
)
227 ureg_TEX(shader
, m
[0], TGSI_TEXTURE_3D
, addr
[0], sampler
);
228 ureg_TEX(shader
, m
[1], TGSI_TEXTURE_3D
, addr
[1], sampler
);
232 matrix_mul(struct ureg_program
*shader
, struct ureg_dst dst
, struct ureg_dst l
[2], struct ureg_dst r
[2])
236 tmp
= ureg_DECL_temporary(shader
);
239 * tmp.xy = dot4(m[0][0..1], m[1][0..1])
240 * dst = tmp.x + tmp.y
242 ureg_DP4(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_X
), ureg_src(l
[0]), ureg_src(r
[0]));
243 ureg_DP4(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_Y
), ureg_src(l
[1]), ureg_src(r
[1]));
244 ureg_ADD(shader
, dst
,
245 ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_X
),
246 ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_Y
));
248 ureg_release_temporary(shader
, tmp
);
252 create_matrix_frag_shader(struct vl_idct
*idct
)
254 struct ureg_program
*shader
;
256 struct ureg_src l_addr
[2], r_addr
[2];
258 struct ureg_dst l
[4][2], r
[2];
259 struct ureg_dst fragment
[NR_RENDER_TARGETS
];
263 shader
= ureg_create(TGSI_PROCESSOR_FRAGMENT
);
267 l_addr
[0] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR0
, TGSI_INTERPOLATE_LINEAR
);
268 l_addr
[1] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR1
, TGSI_INTERPOLATE_LINEAR
);
270 r_addr
[0] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR0
, TGSI_INTERPOLATE_LINEAR
);
271 r_addr
[1] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR1
, TGSI_INTERPOLATE_LINEAR
);
273 for (i
= 0; i
< NR_RENDER_TARGETS
; ++i
)
274 fragment
[i
] = ureg_DECL_output(shader
, TGSI_SEMANTIC_COLOR
, i
);
276 for (i
= 0; i
< 4; ++i
) {
277 l
[i
][0] = ureg_DECL_temporary(shader
);
278 l
[i
][1] = ureg_DECL_temporary(shader
);
281 r
[0] = ureg_DECL_temporary(shader
);
282 r
[1] = ureg_DECL_temporary(shader
);
284 for (i
= 1; i
< 4; ++i
) {
285 increment_addr(shader
, l
[i
], l_addr
, false, false, i
, idct
->buffer_height
);
288 for (i
= 0; i
< 4; ++i
) {
289 struct ureg_src s_addr
[2];
290 s_addr
[0] = i
== 0 ? l_addr
[0] : ureg_src(l
[i
][0]);
291 s_addr
[1] = i
== 0 ? l_addr
[1] : ureg_src(l
[i
][1]);
292 fetch_four(shader
, l
[i
], s_addr
, ureg_DECL_sampler(shader
, 1));
295 for (i
= 0; i
< NR_RENDER_TARGETS
; ++i
) {
297 increment_addr(shader
, r
, r_addr
, true, true, i
, BLOCK_HEIGHT
);
299 struct ureg_src s_addr
[2] = { ureg_src(r
[0]), ureg_src(r
[1]) };
300 s_addr
[0] = i
== 0 ? r_addr
[0] : ureg_src(r
[0]);
301 s_addr
[1] = i
== 0 ? r_addr
[1] : ureg_src(r
[1]);
302 fetch_four(shader
, r
, s_addr
, ureg_DECL_sampler(shader
, 0));
304 for (j
= 0; j
< 4; ++j
) {
305 matrix_mul(shader
, ureg_writemask(fragment
[i
], TGSI_WRITEMASK_X
<< j
), l
[j
], r
);
309 for (i
= 0; i
< 4; ++i
) {
310 ureg_release_temporary(shader
, l
[i
][0]);
311 ureg_release_temporary(shader
, l
[i
][1]);
313 ureg_release_temporary(shader
, r
[0]);
314 ureg_release_temporary(shader
, r
[1]);
318 return ureg_create_shader_and_destroy(shader
, idct
->pipe
);
322 create_transpose_frag_shader(struct vl_idct
*idct
)
324 struct ureg_program
*shader
;
326 struct ureg_src l_addr
[2], r_addr
[2];
328 struct ureg_dst l
[2], r
[2];
329 struct ureg_dst fragment
;
331 shader
= ureg_create(TGSI_PROCESSOR_FRAGMENT
);
335 l_addr
[0] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR0
, TGSI_INTERPOLATE_LINEAR
);
336 l_addr
[1] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_L_ADDR1
, TGSI_INTERPOLATE_LINEAR
);
338 r_addr
[0] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR0
, TGSI_INTERPOLATE_LINEAR
);
339 r_addr
[1] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_R_ADDR1
, TGSI_INTERPOLATE_LINEAR
);
341 l
[0] = ureg_DECL_temporary(shader
);
342 l
[1] = ureg_DECL_temporary(shader
);
343 r
[0] = ureg_DECL_temporary(shader
);
344 r
[1] = ureg_DECL_temporary(shader
);
346 fetch_four(shader
, l
, l_addr
, ureg_DECL_sampler(shader
, 0));
347 fetch_four(shader
, r
, r_addr
, ureg_DECL_sampler(shader
, 1));
349 fragment
= ureg_DECL_output(shader
, TGSI_SEMANTIC_COLOR
, 0);
351 matrix_mul(shader
, ureg_writemask(fragment
, TGSI_WRITEMASK_X
), l
, r
);
353 ureg_release_temporary(shader
, l
[0]);
354 ureg_release_temporary(shader
, l
[1]);
355 ureg_release_temporary(shader
, r
[0]);
356 ureg_release_temporary(shader
, r
[1]);
360 return ureg_create_shader_and_destroy(shader
, idct
->pipe
);
364 init_shaders(struct vl_idct
*idct
, int color_swizzle
)
366 idct
->matrix_vs
= create_vert_shader(idct
, true, color_swizzle
);
367 idct
->matrix_fs
= create_matrix_frag_shader(idct
);
369 idct
->transpose_vs
= create_vert_shader(idct
, false, color_swizzle
);
370 idct
->transpose_fs
= create_transpose_frag_shader(idct
);
373 idct
->matrix_vs
!= NULL
&&
374 idct
->matrix_fs
!= NULL
&&
375 idct
->transpose_vs
!= NULL
&&
376 idct
->transpose_fs
!= NULL
;
380 cleanup_shaders(struct vl_idct
*idct
)
382 idct
->pipe
->delete_vs_state(idct
->pipe
, idct
->matrix_vs
);
383 idct
->pipe
->delete_fs_state(idct
->pipe
, idct
->matrix_fs
);
384 idct
->pipe
->delete_vs_state(idct
->pipe
, idct
->transpose_vs
);
385 idct
->pipe
->delete_fs_state(idct
->pipe
, idct
->transpose_fs
);
389 init_state(struct vl_idct
*idct
)
391 struct pipe_sampler_state sampler
;
392 struct pipe_rasterizer_state rs_state
;
397 for (i
= 0; i
< 4; ++i
) {
398 memset(&sampler
, 0, sizeof(sampler
));
399 sampler
.wrap_s
= PIPE_TEX_WRAP_REPEAT
;
400 sampler
.wrap_t
= PIPE_TEX_WRAP_REPEAT
;
401 sampler
.wrap_r
= PIPE_TEX_WRAP_REPEAT
;
402 sampler
.min_img_filter
= PIPE_TEX_FILTER_NEAREST
;
403 sampler
.min_mip_filter
= PIPE_TEX_MIPFILTER_NONE
;
404 sampler
.mag_img_filter
= PIPE_TEX_FILTER_NEAREST
;
405 sampler
.compare_mode
= PIPE_TEX_COMPARE_NONE
;
406 sampler
.compare_func
= PIPE_FUNC_ALWAYS
;
407 sampler
.normalized_coords
= 1;
408 /*sampler.shadow_ambient = ; */
409 /*sampler.lod_bias = ; */
411 /*sampler.max_lod = ; */
412 /*sampler.border_color[0] = ; */
413 /*sampler.max_anisotropy = ; */
414 idct
->samplers
.all
[i
] = idct
->pipe
->create_sampler_state(idct
->pipe
, &sampler
);
417 memset(&rs_state
, 0, sizeof(rs_state
));
418 /*rs_state.sprite_coord_enable */
419 rs_state
.sprite_coord_mode
= PIPE_SPRITE_COORD_UPPER_LEFT
;
420 rs_state
.point_quad_rasterization
= true;
421 rs_state
.point_size
= BLOCK_WIDTH
;
422 rs_state
.gl_rasterization_rules
= false;
423 idct
->rs_state
= idct
->pipe
->create_rasterizer_state(idct
->pipe
, &rs_state
);
429 cleanup_state(struct vl_idct
*idct
)
433 for (i
= 0; i
< 4; ++i
)
434 idct
->pipe
->delete_sampler_state(idct
->pipe
, idct
->samplers
.all
[i
]);
436 idct
->pipe
->delete_rasterizer_state(idct
->pipe
, idct
->rs_state
);
440 init_textures(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
442 struct pipe_resource
template;
443 struct pipe_sampler_view sampler_view
;
446 assert(idct
&& buffer
);
448 /* create textures */
449 memset(&template, 0, sizeof(struct pipe_resource
));
450 template.last_level
= 0;
451 template.bind
= PIPE_BIND_SAMPLER_VIEW
;
454 template.target
= PIPE_TEXTURE_2D
;
455 template.format
= PIPE_FORMAT_R16G16B16A16_SNORM
;
456 template.width0
= idct
->buffer_width
/ 4;
457 template.height0
= idct
->buffer_height
;
459 template.array_size
= 1;
460 template.usage
= PIPE_USAGE_STREAM
;
461 buffer
->textures
.individual
.source
= idct
->pipe
->screen
->resource_create(idct
->pipe
->screen
, &template);
463 template.target
= PIPE_TEXTURE_3D
;
464 template.format
= PIPE_FORMAT_R16G16B16A16_SNORM
;
465 template.width0
= idct
->buffer_width
/ NR_RENDER_TARGETS
;
466 template.height0
= idct
->buffer_height
/ 4;
467 template.depth0
= NR_RENDER_TARGETS
;
468 template.usage
= PIPE_USAGE_STATIC
;
469 buffer
->textures
.individual
.intermediate
= idct
->pipe
->screen
->resource_create(idct
->pipe
->screen
, &template);
471 for (i
= 0; i
< 4; ++i
) {
472 if(buffer
->textures
.all
[i
] == NULL
)
473 return false; /* a texture failed to allocate */
475 u_sampler_view_default_template(&sampler_view
, buffer
->textures
.all
[i
], buffer
->textures
.all
[i
]->format
);
476 buffer
->sampler_views
.all
[i
] = idct
->pipe
->create_sampler_view(idct
->pipe
, buffer
->textures
.all
[i
], &sampler_view
);
479 template.target
= PIPE_TEXTURE_2D
;
480 /* TODO: Accomodate HW that can't do this and also for cases when this isn't precise enough */
481 template.format
= PIPE_FORMAT_R16_SNORM
;
482 template.width0
= idct
->buffer_width
;
483 template.height0
= idct
->buffer_height
;
486 buffer
->destination
= idct
->pipe
->screen
->resource_create(idct
->pipe
->screen
, &template);
492 cleanup_textures(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
496 assert(idct
&& buffer
);
498 for (i
= 0; i
< 4; ++i
) {
499 pipe_sampler_view_reference(&buffer
->sampler_views
.all
[i
], NULL
);
500 pipe_resource_reference(&buffer
->textures
.all
[i
], NULL
);
504 struct pipe_resource
*
505 vl_idct_upload_matrix(struct pipe_context
*pipe
)
507 struct pipe_resource
template, *matrix
;
508 struct pipe_transfer
*buf_transfer
;
509 unsigned i
, j
, pitch
;
512 struct pipe_box rect
=
520 memset(&template, 0, sizeof(struct pipe_resource
));
521 template.target
= PIPE_TEXTURE_2D
;
522 template.format
= PIPE_FORMAT_R32G32B32A32_FLOAT
;
523 template.last_level
= 0;
525 template.height0
= 8;
527 template.array_size
= 1;
528 template.usage
= PIPE_USAGE_IMMUTABLE
;
529 template.bind
= PIPE_BIND_SAMPLER_VIEW
;
532 matrix
= pipe
->screen
->resource_create(pipe
->screen
, &template);
535 buf_transfer
= pipe
->get_transfer
538 0, PIPE_TRANSFER_WRITE
| PIPE_TRANSFER_DISCARD
,
541 pitch
= buf_transfer
->stride
/ sizeof(float);
543 f
= pipe
->transfer_map(pipe
, buf_transfer
);
544 for(i
= 0; i
< BLOCK_HEIGHT
; ++i
)
545 for(j
= 0; j
< BLOCK_WIDTH
; ++j
)
546 // transpose and scale
547 f
[i
* pitch
+ j
] = const_matrix
[j
][i
] * sqrtf(SCALE_FACTOR_16_TO_9
);
549 pipe
->transfer_unmap(pipe
, buf_transfer
);
550 pipe
->transfer_destroy(pipe
, buf_transfer
);
555 bool vl_idct_init(struct vl_idct
*idct
, struct pipe_context
*pipe
,
556 unsigned buffer_width
, unsigned buffer_height
,
557 unsigned blocks_x
, unsigned blocks_y
,
558 int color_swizzle
, struct pipe_resource
*matrix
)
560 assert(idct
&& pipe
&& matrix
);
563 idct
->buffer_width
= buffer_width
;
564 idct
->buffer_height
= buffer_height
;
565 idct
->blocks_x
= blocks_x
;
566 idct
->blocks_y
= blocks_y
;
567 pipe_resource_reference(&idct
->matrix
, matrix
);
569 if(!init_shaders(idct
, color_swizzle
))
572 if(!init_state(idct
)) {
573 cleanup_shaders(idct
);
581 vl_idct_cleanup(struct vl_idct
*idct
)
583 cleanup_shaders(idct
);
586 pipe_resource_reference(&idct
->matrix
, NULL
);
589 struct pipe_resource
*
590 vl_idct_init_buffer(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
592 struct pipe_surface
template;
599 pipe_resource_reference(&buffer
->textures
.individual
.matrix
, idct
->matrix
);
600 pipe_resource_reference(&buffer
->textures
.individual
.transpose
, idct
->matrix
);
602 if (!init_textures(idct
, buffer
))
606 buffer
->viewport
[0].scale
[0] = buffer
->textures
.individual
.intermediate
->width0
;
607 buffer
->viewport
[0].scale
[1] = buffer
->textures
.individual
.intermediate
->height0
;
609 buffer
->viewport
[1].scale
[0] = buffer
->destination
->width0
;
610 buffer
->viewport
[1].scale
[1] = buffer
->destination
->height0
;
612 buffer
->fb_state
[0].width
= buffer
->textures
.individual
.intermediate
->width0
;
613 buffer
->fb_state
[0].height
= buffer
->textures
.individual
.intermediate
->height0
;
615 buffer
->fb_state
[0].nr_cbufs
= NR_RENDER_TARGETS
;
616 for(i
= 0; i
< NR_RENDER_TARGETS
; ++i
) {
617 memset(&template, 0, sizeof(template));
618 template.format
= buffer
->textures
.individual
.intermediate
->format
;
619 template.u
.tex
.first_layer
= i
;
620 template.u
.tex
.last_layer
= i
;
621 template.usage
= PIPE_BIND_RENDER_TARGET
;
622 buffer
->fb_state
[0].cbufs
[i
] = idct
->pipe
->create_surface(
623 idct
->pipe
, buffer
->textures
.individual
.intermediate
,
627 buffer
->fb_state
[1].width
= buffer
->destination
->width0
;
628 buffer
->fb_state
[1].height
= buffer
->destination
->height0
;
630 buffer
->fb_state
[1].nr_cbufs
= 1;
632 memset(&template, 0, sizeof(template));
633 template.format
= buffer
->destination
->format
;
634 template.usage
= PIPE_BIND_RENDER_TARGET
;
635 buffer
->fb_state
[1].cbufs
[0] = idct
->pipe
->create_surface(
636 idct
->pipe
, buffer
->destination
, &template);
638 for(i
= 0; i
< 2; ++i
) {
639 buffer
->viewport
[i
].scale
[2] = 1;
640 buffer
->viewport
[i
].scale
[3] = 1;
641 buffer
->viewport
[i
].translate
[0] = 0;
642 buffer
->viewport
[i
].translate
[1] = 0;
643 buffer
->viewport
[i
].translate
[2] = 0;
644 buffer
->viewport
[i
].translate
[3] = 0;
646 buffer
->fb_state
[i
].zsbuf
= NULL
;
649 return buffer
->destination
;
653 vl_idct_cleanup_buffer(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
659 for(i
= 0; i
< NR_RENDER_TARGETS
; ++i
) {
660 idct
->pipe
->surface_destroy(idct
->pipe
, buffer
->fb_state
[0].cbufs
[i
]);
663 idct
->pipe
->surface_destroy(idct
->pipe
, buffer
->fb_state
[1].cbufs
[0]);
665 cleanup_textures(idct
, buffer
);
669 vl_idct_map_buffers(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
673 struct pipe_box rect
=
676 buffer
->textures
.individual
.source
->width0
,
677 buffer
->textures
.individual
.source
->height0
,
681 buffer
->tex_transfer
= idct
->pipe
->get_transfer
683 idct
->pipe
, buffer
->textures
.individual
.source
,
684 0, PIPE_TRANSFER_WRITE
| PIPE_TRANSFER_DISCARD
,
688 buffer
->texels
= idct
->pipe
->transfer_map(idct
->pipe
, buffer
->tex_transfer
);
692 vl_idct_add_block(struct vl_idct_buffer
*buffer
, unsigned x
, unsigned y
, short *block
)
701 tex_pitch
= buffer
->tex_transfer
->stride
/ sizeof(short);
702 texels
= buffer
->texels
+ y
* tex_pitch
* BLOCK_HEIGHT
+ x
* BLOCK_WIDTH
;
704 for (i
= 0; i
< BLOCK_HEIGHT
; ++i
)
705 memcpy(texels
+ i
* tex_pitch
, block
+ i
* BLOCK_WIDTH
, BLOCK_WIDTH
* sizeof(short));
709 vl_idct_unmap_buffers(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
)
711 assert(idct
&& buffer
);
713 idct
->pipe
->transfer_unmap(idct
->pipe
, buffer
->tex_transfer
);
714 idct
->pipe
->transfer_destroy(idct
->pipe
, buffer
->tex_transfer
);
718 vl_idct_flush(struct vl_idct
*idct
, struct vl_idct_buffer
*buffer
, unsigned num_instances
)
725 if(num_instances
> 0) {
726 num_verts
= idct
->blocks_x
* idct
->blocks_y
* 4;
728 idct
->pipe
->bind_rasterizer_state(idct
->pipe
, idct
->rs_state
);
731 idct
->pipe
->set_framebuffer_state(idct
->pipe
, &buffer
->fb_state
[0]);
732 idct
->pipe
->set_viewport_state(idct
->pipe
, &buffer
->viewport
[0]);
733 idct
->pipe
->set_fragment_sampler_views(idct
->pipe
, 2, buffer
->sampler_views
.stage
[0]);
734 idct
->pipe
->bind_fragment_sampler_states(idct
->pipe
, 2, idct
->samplers
.stage
[0]);
735 idct
->pipe
->bind_vs_state(idct
->pipe
, idct
->matrix_vs
);
736 idct
->pipe
->bind_fs_state(idct
->pipe
, idct
->matrix_fs
);
737 util_draw_arrays_instanced(idct
->pipe
, PIPE_PRIM_QUADS
, 0, num_verts
, 0, num_instances
);
740 idct
->pipe
->set_framebuffer_state(idct
->pipe
, &buffer
->fb_state
[1]);
741 idct
->pipe
->set_viewport_state(idct
->pipe
, &buffer
->viewport
[1]);
742 idct
->pipe
->set_fragment_sampler_views(idct
->pipe
, 2, buffer
->sampler_views
.stage
[1]);
743 idct
->pipe
->bind_fragment_sampler_states(idct
->pipe
, 2, idct
->samplers
.stage
[1]);
744 idct
->pipe
->bind_vs_state(idct
->pipe
, idct
->transpose_vs
);
745 idct
->pipe
->bind_fs_state(idct
->pipe
, idct
->transpose_fs
);
746 util_draw_arrays_instanced(idct
->pipe
, PIPE_PRIM_QUADS
, 0, num_verts
, 0, num_instances
);