1 /**************************************************************************
3 * Copyright 2011 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 VMWARE 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 **************************************************************************/
30 #include "pipe/p_screen.h"
31 #include "pipe/p_context.h"
33 #include "util/u_draw.h"
34 #include "util/u_sampler.h"
35 #include "util/u_inlines.h"
36 #include "util/u_memory.h"
38 #include "tgsi/tgsi_ureg.h"
40 #include "vl_defines.h"
44 #include "vl_vertex_buffers.h"
52 const int vl_zscan_normal_16
[] =
54 /* Zig-Zag scan pattern */
55 0, 1, 4, 8, 5, 2, 3, 6,
56 9,12,13,10, 7,11,14,15
59 const int vl_zscan_linear
[] =
61 /* Linear scan pattern */
62 0, 1, 2, 3, 4, 5, 6, 7,
63 8, 9,10,11,12,13,14,15,
64 16,17,18,19,20,21,22,23,
65 24,25,26,27,28,29,30,31,
66 32,33,34,35,36,37,38,39,
67 40,41,42,43,44,45,46,47,
68 48,49,50,51,52,53,54,55,
69 56,57,58,59,60,61,62,63
72 const int vl_zscan_normal
[] =
74 /* Zig-Zag scan pattern */
75 0, 1, 8,16, 9, 2, 3,10,
76 17,24,32,25,18,11, 4, 5,
77 12,19,26,33,40,48,41,34,
78 27,20,13, 6, 7,14,21,28,
79 35,42,49,56,57,50,43,36,
80 29,22,15,23,30,37,44,51,
81 58,59,52,45,38,31,39,46,
82 53,60,61,54,47,55,62,63
85 const int vl_zscan_alternate
[] =
87 /* Alternate scan pattern */
88 0, 8,16,24, 1, 9, 2,10,
89 17,25,32,40,48,56,57,49,
90 41,33,26,18, 3,11, 4,12,
91 19,27,34,42,50,58,35,43,
92 51,59,20,28, 5,13, 6,14,
93 21,29,36,44,52,60,37,45,
94 53,61,22,30, 7,15,23,31,
95 38,46,54,62,39,47,55,63
99 create_vert_shader(struct vl_zscan
*zscan
)
101 struct ureg_program
*shader
;
102 struct ureg_src scale
;
103 struct ureg_src vrect
, vpos
, block_num
;
105 struct ureg_dst o_vpos
;
106 struct ureg_dst
*o_vtex
;
109 shader
= ureg_create(PIPE_SHADER_VERTEX
);
113 o_vtex
= MALLOC(zscan
->num_channels
* sizeof(struct ureg_dst
));
115 scale
= ureg_imm2f(shader
,
116 (float)VL_BLOCK_WIDTH
/ zscan
->buffer_width
,
117 (float)VL_BLOCK_HEIGHT
/ zscan
->buffer_height
);
119 vrect
= ureg_DECL_vs_input(shader
, VS_I_RECT
);
120 vpos
= ureg_DECL_vs_input(shader
, VS_I_VPOS
);
121 block_num
= ureg_DECL_vs_input(shader
, VS_I_BLOCK_NUM
);
123 tmp
= ureg_DECL_temporary(shader
);
125 o_vpos
= ureg_DECL_output(shader
, TGSI_SEMANTIC_POSITION
, VS_O_VPOS
);
127 for (i
= 0; i
< zscan
->num_channels
; ++i
)
128 o_vtex
[i
] = ureg_DECL_output(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_VTEX
+ i
);
131 * o_vpos.xy = (vpos + vrect) * scale
134 * tmp.xy = InstanceID / blocks_per_line
135 * tmp.x = frac(tmp.x)
136 * tmp.y = floor(tmp.y)
138 * o_vtex.x = vrect.x / blocks_per_line + tmp.x
140 * o_vtex.z = tmp.z * blocks_per_line / blocks_total
142 ureg_ADD(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_XY
), vpos
, vrect
);
143 ureg_MUL(shader
, ureg_writemask(o_vpos
, TGSI_WRITEMASK_XY
), ureg_src(tmp
), scale
);
144 ureg_MOV(shader
, ureg_writemask(o_vpos
, TGSI_WRITEMASK_ZW
), ureg_imm1f(shader
, 1.0f
));
146 ureg_MUL(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_XW
), ureg_scalar(block_num
, TGSI_SWIZZLE_X
),
147 ureg_imm1f(shader
, 1.0f
/ zscan
->blocks_per_line
));
149 ureg_FRC(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_Y
), ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_X
));
150 ureg_FLR(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_W
), ureg_src(tmp
));
152 for (i
= 0; i
< zscan
->num_channels
; ++i
) {
153 ureg_ADD(shader
, ureg_writemask(tmp
, TGSI_WRITEMASK_X
), ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_Y
),
154 ureg_imm1f(shader
, 1.0f
/ (zscan
->blocks_per_line
* VL_BLOCK_WIDTH
)
155 * (i
- (signed)zscan
->num_channels
/ 2)));
157 ureg_MAD(shader
, ureg_writemask(o_vtex
[i
], TGSI_WRITEMASK_X
), vrect
,
158 ureg_imm1f(shader
, 1.0f
/ zscan
->blocks_per_line
), ureg_src(tmp
));
159 ureg_MOV(shader
, ureg_writemask(o_vtex
[i
], TGSI_WRITEMASK_Y
), vrect
);
160 ureg_MOV(shader
, ureg_writemask(o_vtex
[i
], TGSI_WRITEMASK_Z
), vpos
);
161 ureg_MUL(shader
, ureg_writemask(o_vtex
[i
], TGSI_WRITEMASK_W
), ureg_src(tmp
),
162 ureg_imm1f(shader
, (float)zscan
->blocks_per_line
/ zscan
->blocks_total
));
165 ureg_release_temporary(shader
, tmp
);
170 return ureg_create_shader_and_destroy(shader
, zscan
->pipe
);
174 create_frag_shader(struct vl_zscan
*zscan
)
176 struct ureg_program
*shader
;
177 struct ureg_src
*vtex
;
179 struct ureg_src samp_src
, samp_scan
, samp_quant
;
181 struct ureg_dst
*tmp
;
182 struct ureg_dst quant
, fragment
;
186 shader
= ureg_create(PIPE_SHADER_FRAGMENT
);
190 vtex
= MALLOC(zscan
->num_channels
* sizeof(struct ureg_src
));
191 tmp
= MALLOC(zscan
->num_channels
* sizeof(struct ureg_dst
));
193 for (i
= 0; i
< zscan
->num_channels
; ++i
)
194 vtex
[i
] = ureg_DECL_fs_input(shader
, TGSI_SEMANTIC_GENERIC
, VS_O_VTEX
+ i
, TGSI_INTERPOLATE_LINEAR
);
196 samp_src
= ureg_DECL_sampler(shader
, 0);
197 samp_scan
= ureg_DECL_sampler(shader
, 1);
198 samp_quant
= ureg_DECL_sampler(shader
, 2);
200 for (i
= 0; i
< zscan
->num_channels
; ++i
)
201 tmp
[i
] = ureg_DECL_temporary(shader
);
202 quant
= ureg_DECL_temporary(shader
);
204 fragment
= ureg_DECL_output(shader
, TGSI_SEMANTIC_COLOR
, 0);
207 * tmp.x = tex(vtex, 1)
209 * fragment = tex(tmp, 0) * quant
211 for (i
= 0; i
< zscan
->num_channels
; ++i
)
212 ureg_TEX(shader
, ureg_writemask(tmp
[i
], TGSI_WRITEMASK_X
), TGSI_TEXTURE_2D
, vtex
[i
], samp_scan
);
214 for (i
= 0; i
< zscan
->num_channels
; ++i
)
215 ureg_MOV(shader
, ureg_writemask(tmp
[i
], TGSI_WRITEMASK_Y
), ureg_scalar(vtex
[i
], TGSI_SWIZZLE_W
));
217 for (i
= 0; i
< zscan
->num_channels
; ++i
) {
218 ureg_TEX(shader
, ureg_writemask(tmp
[0], TGSI_WRITEMASK_X
<< i
), TGSI_TEXTURE_2D
, ureg_src(tmp
[i
]), samp_src
);
219 ureg_TEX(shader
, ureg_writemask(quant
, TGSI_WRITEMASK_X
<< i
), TGSI_TEXTURE_3D
, vtex
[i
], samp_quant
);
222 ureg_MUL(shader
, quant
, ureg_src(quant
), ureg_imm1f(shader
, 16.0f
));
223 ureg_MUL(shader
, fragment
, ureg_src(tmp
[0]), ureg_src(quant
));
225 for (i
= 0; i
< zscan
->num_channels
; ++i
)
226 ureg_release_temporary(shader
, tmp
[i
]);
232 return ureg_create_shader_and_destroy(shader
, zscan
->pipe
);
236 init_shaders(struct vl_zscan
*zscan
)
240 zscan
->vs
= create_vert_shader(zscan
);
244 zscan
->fs
= create_frag_shader(zscan
);
251 zscan
->pipe
->delete_vs_state(zscan
->pipe
, zscan
->vs
);
258 cleanup_shaders(struct vl_zscan
*zscan
)
262 zscan
->pipe
->delete_vs_state(zscan
->pipe
, zscan
->vs
);
263 zscan
->pipe
->delete_fs_state(zscan
->pipe
, zscan
->fs
);
267 init_state(struct vl_zscan
*zscan
)
269 struct pipe_blend_state blend
;
270 struct pipe_rasterizer_state rs_state
;
271 struct pipe_sampler_state sampler
;
276 memset(&rs_state
, 0, sizeof(rs_state
));
277 rs_state
.half_pixel_center
= true;
278 rs_state
.bottom_edge_rule
= true;
279 rs_state
.depth_clip
= 1;
280 zscan
->rs_state
= zscan
->pipe
->create_rasterizer_state(zscan
->pipe
, &rs_state
);
281 if (!zscan
->rs_state
)
284 memset(&blend
, 0, sizeof blend
);
286 blend
.independent_blend_enable
= 0;
287 blend
.rt
[0].blend_enable
= 0;
288 blend
.rt
[0].rgb_func
= PIPE_BLEND_ADD
;
289 blend
.rt
[0].rgb_src_factor
= PIPE_BLENDFACTOR_ONE
;
290 blend
.rt
[0].rgb_dst_factor
= PIPE_BLENDFACTOR_ONE
;
291 blend
.rt
[0].alpha_func
= PIPE_BLEND_ADD
;
292 blend
.rt
[0].alpha_src_factor
= PIPE_BLENDFACTOR_ONE
;
293 blend
.rt
[0].alpha_dst_factor
= PIPE_BLENDFACTOR_ONE
;
294 blend
.logicop_enable
= 0;
295 blend
.logicop_func
= PIPE_LOGICOP_CLEAR
;
296 /* Needed to allow color writes to FB, even if blending disabled */
297 blend
.rt
[0].colormask
= PIPE_MASK_RGBA
;
299 zscan
->blend
= zscan
->pipe
->create_blend_state(zscan
->pipe
, &blend
);
303 for (i
= 0; i
< 3; ++i
) {
304 memset(&sampler
, 0, sizeof(sampler
));
305 sampler
.wrap_s
= PIPE_TEX_WRAP_REPEAT
;
306 sampler
.wrap_t
= PIPE_TEX_WRAP_REPEAT
;
307 sampler
.wrap_r
= PIPE_TEX_WRAP_CLAMP_TO_EDGE
;
308 sampler
.min_img_filter
= PIPE_TEX_FILTER_NEAREST
;
309 sampler
.min_mip_filter
= PIPE_TEX_MIPFILTER_NONE
;
310 sampler
.mag_img_filter
= PIPE_TEX_FILTER_NEAREST
;
311 sampler
.compare_mode
= PIPE_TEX_COMPARE_NONE
;
312 sampler
.compare_func
= PIPE_FUNC_ALWAYS
;
313 sampler
.normalized_coords
= 1;
314 zscan
->samplers
[i
] = zscan
->pipe
->create_sampler_state(zscan
->pipe
, &sampler
);
315 if (!zscan
->samplers
[i
])
322 for (i
= 0; i
< 2; ++i
)
323 if (zscan
->samplers
[i
])
324 zscan
->pipe
->delete_sampler_state(zscan
->pipe
, zscan
->samplers
[i
]);
326 zscan
->pipe
->delete_rasterizer_state(zscan
->pipe
, zscan
->rs_state
);
329 zscan
->pipe
->delete_blend_state(zscan
->pipe
, zscan
->blend
);
336 cleanup_state(struct vl_zscan
*zscan
)
342 for (i
= 0; i
< 3; ++i
)
343 zscan
->pipe
->delete_sampler_state(zscan
->pipe
, zscan
->samplers
[i
]);
345 zscan
->pipe
->delete_rasterizer_state(zscan
->pipe
, zscan
->rs_state
);
346 zscan
->pipe
->delete_blend_state(zscan
->pipe
, zscan
->blend
);
349 struct pipe_sampler_view
*
350 vl_zscan_layout(struct pipe_context
*pipe
, const int layout
[64], unsigned blocks_per_line
)
352 const unsigned total_size
= blocks_per_line
* VL_BLOCK_WIDTH
* VL_BLOCK_HEIGHT
;
354 int patched_layout
[64];
356 struct pipe_resource res_tmpl
, *res
;
357 struct pipe_sampler_view sv_tmpl
, *sv
;
358 struct pipe_transfer
*buf_transfer
;
359 unsigned x
, y
, i
, pitch
;
362 struct pipe_box rect
=
365 VL_BLOCK_WIDTH
* blocks_per_line
,
370 assert(pipe
&& layout
&& blocks_per_line
);
372 for (i
= 0; i
< 64; ++i
)
373 patched_layout
[layout
[i
]] = i
;
375 memset(&res_tmpl
, 0, sizeof(res_tmpl
));
376 res_tmpl
.target
= PIPE_TEXTURE_2D
;
377 res_tmpl
.format
= PIPE_FORMAT_R32_FLOAT
;
378 res_tmpl
.width0
= VL_BLOCK_WIDTH
* blocks_per_line
;
379 res_tmpl
.height0
= VL_BLOCK_HEIGHT
;
381 res_tmpl
.array_size
= 1;
382 res_tmpl
.usage
= PIPE_USAGE_IMMUTABLE
;
383 res_tmpl
.bind
= PIPE_BIND_SAMPLER_VIEW
;
385 res
= pipe
->screen
->resource_create(pipe
->screen
, &res_tmpl
);
389 f
= pipe
->transfer_map(pipe
, res
,
390 0, PIPE_TRANSFER_WRITE
| PIPE_TRANSFER_DISCARD_RANGE
,
391 &rect
, &buf_transfer
);
395 pitch
= buf_transfer
->stride
/ sizeof(float);
397 for (i
= 0; i
< blocks_per_line
; ++i
)
398 for (y
= 0; y
< VL_BLOCK_HEIGHT
; ++y
)
399 for (x
= 0; x
< VL_BLOCK_WIDTH
; ++x
) {
400 float addr
= patched_layout
[x
+ y
* VL_BLOCK_WIDTH
] +
401 i
* VL_BLOCK_WIDTH
* VL_BLOCK_HEIGHT
;
405 f
[i
* VL_BLOCK_WIDTH
+ y
* pitch
+ x
] = addr
;
408 pipe
->transfer_unmap(pipe
, buf_transfer
);
410 memset(&sv_tmpl
, 0, sizeof(sv_tmpl
));
411 u_sampler_view_default_template(&sv_tmpl
, res
, res
->format
);
412 sv
= pipe
->create_sampler_view(pipe
, res
, &sv_tmpl
);
413 pipe_resource_reference(&res
, NULL
);
420 pipe_resource_reference(&res
, NULL
);
427 vl_zscan_init(struct vl_zscan
*zscan
, struct pipe_context
*pipe
,
428 unsigned buffer_width
, unsigned buffer_height
,
429 unsigned blocks_per_line
, unsigned blocks_total
,
430 unsigned num_channels
)
432 assert(zscan
&& pipe
);
435 zscan
->buffer_width
= buffer_width
;
436 zscan
->buffer_height
= buffer_height
;
437 zscan
->num_channels
= num_channels
;
438 zscan
->blocks_per_line
= blocks_per_line
;
439 zscan
->blocks_total
= blocks_total
;
441 if(!init_shaders(zscan
))
444 if(!init_state(zscan
)) {
445 cleanup_shaders(zscan
);
453 vl_zscan_cleanup(struct vl_zscan
*zscan
)
457 cleanup_shaders(zscan
);
458 cleanup_state(zscan
);
462 vl_zscan_init_buffer(struct vl_zscan
*zscan
, struct vl_zscan_buffer
*buffer
,
463 struct pipe_sampler_view
*src
, struct pipe_surface
*dst
)
465 struct pipe_resource res_tmpl
, *res
;
466 struct pipe_sampler_view sv_tmpl
;
468 assert(zscan
&& buffer
);
470 memset(buffer
, 0, sizeof(struct vl_zscan_buffer
));
472 pipe_sampler_view_reference(&buffer
->src
, src
);
474 buffer
->viewport
.scale
[0] = dst
->width
;
475 buffer
->viewport
.scale
[1] = dst
->height
;
476 buffer
->viewport
.scale
[2] = 1;
477 buffer
->viewport
.translate
[0] = 0;
478 buffer
->viewport
.translate
[1] = 0;
479 buffer
->viewport
.translate
[2] = 0;
481 buffer
->fb_state
.width
= dst
->width
;
482 buffer
->fb_state
.height
= dst
->height
;
483 buffer
->fb_state
.nr_cbufs
= 1;
484 pipe_surface_reference(&buffer
->fb_state
.cbufs
[0], dst
);
486 memset(&res_tmpl
, 0, sizeof(res_tmpl
));
487 res_tmpl
.target
= PIPE_TEXTURE_3D
;
488 res_tmpl
.format
= PIPE_FORMAT_R8_UNORM
;
489 res_tmpl
.width0
= VL_BLOCK_WIDTH
* zscan
->blocks_per_line
;
490 res_tmpl
.height0
= VL_BLOCK_HEIGHT
;
492 res_tmpl
.array_size
= 1;
493 res_tmpl
.usage
= PIPE_USAGE_IMMUTABLE
;
494 res_tmpl
.bind
= PIPE_BIND_SAMPLER_VIEW
;
496 res
= zscan
->pipe
->screen
->resource_create(zscan
->pipe
->screen
, &res_tmpl
);
500 memset(&sv_tmpl
, 0, sizeof(sv_tmpl
));
501 u_sampler_view_default_template(&sv_tmpl
, res
, res
->format
);
502 sv_tmpl
.swizzle_r
= sv_tmpl
.swizzle_g
= sv_tmpl
.swizzle_b
= sv_tmpl
.swizzle_a
= TGSI_SWIZZLE_X
;
503 buffer
->quant
= zscan
->pipe
->create_sampler_view(zscan
->pipe
, res
, &sv_tmpl
);
504 pipe_resource_reference(&res
, NULL
);
512 vl_zscan_cleanup_buffer(struct vl_zscan_buffer
*buffer
)
516 pipe_sampler_view_reference(&buffer
->src
, NULL
);
517 pipe_sampler_view_reference(&buffer
->layout
, NULL
);
518 pipe_sampler_view_reference(&buffer
->quant
, NULL
);
519 pipe_surface_reference(&buffer
->fb_state
.cbufs
[0], NULL
);
523 vl_zscan_set_layout(struct vl_zscan_buffer
*buffer
, struct pipe_sampler_view
*layout
)
528 pipe_sampler_view_reference(&buffer
->layout
, layout
);
532 vl_zscan_upload_quant(struct vl_zscan
*zscan
, struct vl_zscan_buffer
*buffer
,
533 const uint8_t matrix
[64], bool intra
)
535 struct pipe_context
*pipe
;
536 struct pipe_transfer
*buf_transfer
;
537 unsigned x
, y
, i
, pitch
;
540 struct pipe_box rect
=
553 rect
.width
*= zscan
->blocks_per_line
;
555 data
= pipe
->transfer_map(pipe
, buffer
->quant
->texture
,
556 0, PIPE_TRANSFER_WRITE
|
557 PIPE_TRANSFER_DISCARD_RANGE
,
558 &rect
, &buf_transfer
);
562 pitch
= buf_transfer
->stride
;
564 for (i
= 0; i
< zscan
->blocks_per_line
; ++i
)
565 for (y
= 0; y
< VL_BLOCK_HEIGHT
; ++y
)
566 for (x
= 0; x
< VL_BLOCK_WIDTH
; ++x
)
567 data
[i
* VL_BLOCK_WIDTH
+ y
* pitch
+ x
] = matrix
[x
+ y
* VL_BLOCK_WIDTH
];
569 pipe
->transfer_unmap(pipe
, buf_transfer
);
573 vl_zscan_render(struct vl_zscan
*zscan
, struct vl_zscan_buffer
*buffer
, unsigned num_instances
)
577 zscan
->pipe
->bind_rasterizer_state(zscan
->pipe
, zscan
->rs_state
);
578 zscan
->pipe
->bind_blend_state(zscan
->pipe
, zscan
->blend
);
579 zscan
->pipe
->bind_sampler_states(zscan
->pipe
, PIPE_SHADER_FRAGMENT
,
580 0, 3, zscan
->samplers
);
581 zscan
->pipe
->set_framebuffer_state(zscan
->pipe
, &buffer
->fb_state
);
582 zscan
->pipe
->set_viewport_states(zscan
->pipe
, 0, 1, &buffer
->viewport
);
583 zscan
->pipe
->set_sampler_views(zscan
->pipe
, PIPE_SHADER_FRAGMENT
,
585 zscan
->pipe
->bind_vs_state(zscan
->pipe
, zscan
->vs
);
586 zscan
->pipe
->bind_fs_state(zscan
->pipe
, zscan
->fs
);
587 util_draw_arrays_instanced(zscan
->pipe
, PIPE_PRIM_QUADS
, 0, 4, 0, num_instances
);