1 /**************************************************************************
3 * Copyright 2009 VMware, Inc.
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 **************************************************************************/
29 * The rast code is concerned with rasterization of command bins.
30 * Each screen tile has a bin associated with it. To render the
31 * scene we iterate over the tile bins and execute the commands
33 * We'll do that with multiple threads...
40 #include "pipe/p_compiler.h"
41 #include "util/u_pack_color.h"
50 #define FIXED_TYPE_WIDTH 64
51 /** For sub-pixel positioning */
53 #define FIXED_ONE (1<<FIXED_ORDER)
54 #define FIXED_SHIFT (FIXED_TYPE_WIDTH - 1)
55 /** Maximum length of an edge in a primitive in pixels.
56 * If the framebuffer is large we have to think about fixed-point
57 * integer overflow. Coordinates need ((FIXED_TYPE_WIDTH/2) - 1) bits
58 * to be able to fit product of two such coordinates inside
59 * FIXED_TYPE_WIDTH, any larger and we could overflow a
60 * FIXED_TYPE_WIDTH_-bit int.
62 #define MAX_FIXED_LENGTH (1 << (((FIXED_TYPE_WIDTH/2) - 1) - FIXED_ORDER))
64 #define MAX_FIXED_LENGTH32 (1 << (((32/2) - 1) - FIXED_ORDER))
66 /* Rasterizer output size going to jit fs, width/height */
67 #define LP_RASTER_BLOCK_SIZE 4
69 #define LP_MAX_ACTIVE_BINNED_QUERIES 64
71 #define IMUL64(a, b) (((int64_t)(a)) * ((int64_t)(b)))
73 struct lp_rasterizer_task
;
77 * Rasterization state.
78 * Objects of this type are put into the shared data bin and pointed
79 * to by commands in the per-tile bins.
81 struct lp_rast_state
{
82 /* State for the shader. This also contains state which feeds into
83 * the fragment shader, such as blend color and alpha ref value.
85 struct lp_jit_context jit_context
;
87 /* The shader itself. Probably we also need to pass a pointer to
88 * the tile color/z/stencil data somehow
90 struct lp_fragment_shader_variant
*variant
;
95 * Coefficients necessary to run the shader at a given location.
96 * First coefficient is position.
97 * These pointers point into the bin data buffer.
99 struct lp_rast_shader_inputs
{
100 unsigned frontfacing
:1; /** True for front-facing */
101 unsigned disable
:1; /** Partially binned, disable this command */
102 unsigned opaque
:1; /** Is opaque */
103 unsigned pad0
:29; /* wasted space */
104 unsigned stride
; /* how much to advance data between a0, dadx, dady */
105 unsigned layer
; /* the layer to render to (from gs, already clamped) */
106 unsigned viewport_index
; /* the active viewport index (from gs, already clamped) */
107 /* followed by a0, dadx, dady and planes[] */
110 struct lp_rast_plane
{
111 /* edge function values at minx,miny ?? */
117 /* one-pixel sized trivial reject offsets for each plane */
122 * Rasterization information for a triangle known to be in this bin,
123 * plus inputs to run the shader:
124 * These fields are tile- and bin-independent.
125 * Objects of this type are put into the lp_setup_context::data buffer.
127 struct lp_rast_triangle
{
134 /* inputs for the shader */
135 struct lp_rast_shader_inputs inputs
;
136 /* planes are also allocated here */
140 struct lp_rast_clear_rb
{
141 union util_color color_val
;
146 #define GET_A0(inputs) ((float (*)[4])((inputs)+1))
147 #define GET_DADX(inputs) ((float (*)[4])((char *)((inputs) + 1) + (inputs)->stride))
148 #define GET_DADY(inputs) ((float (*)[4])((char *)((inputs) + 1) + 2 * (inputs)->stride))
149 #define GET_PLANES(tri) ((struct lp_rast_plane *)((char *)(&(tri)->inputs + 1) + 3 * (tri)->inputs.stride))
153 struct lp_rasterizer
*
154 lp_rast_create( unsigned num_threads
);
157 lp_rast_destroy( struct lp_rasterizer
* );
160 lp_rast_queue_scene( struct lp_rasterizer
*rast
,
161 struct lp_scene
*scene
);
164 lp_rast_finish( struct lp_rasterizer
*rast
);
167 union lp_rast_cmd_arg
{
168 const struct lp_rast_shader_inputs
*shade_tile
;
170 const struct lp_rast_triangle
*tri
;
173 const struct lp_rast_state
*set_state
;
174 const struct lp_rast_clear_rb
*clear_rb
;
179 const struct lp_rast_state
*state
;
180 struct lp_fence
*fence
;
181 struct llvmpipe_query
*query_obj
;
185 /* Cast wrappers. Hopefully these compile to noops!
187 static inline union lp_rast_cmd_arg
188 lp_rast_arg_inputs( const struct lp_rast_shader_inputs
*shade_tile
)
190 union lp_rast_cmd_arg arg
;
191 arg
.shade_tile
= shade_tile
;
195 static inline union lp_rast_cmd_arg
196 lp_rast_arg_triangle( const struct lp_rast_triangle
*triangle
,
199 union lp_rast_cmd_arg arg
;
200 arg
.triangle
.tri
= triangle
;
201 arg
.triangle
.plane_mask
= plane_mask
;
206 * Build argument for a contained triangle.
208 * All planes are enabled, so instead of the plane mask we pass the upper
209 * left coordinates of the a block that fully encloses the triangle.
211 static inline union lp_rast_cmd_arg
212 lp_rast_arg_triangle_contained( const struct lp_rast_triangle
*triangle
,
213 unsigned x
, unsigned y
)
215 union lp_rast_cmd_arg arg
;
216 arg
.triangle
.tri
= triangle
;
217 arg
.triangle
.plane_mask
= x
| (y
<< 8);
221 static inline union lp_rast_cmd_arg
222 lp_rast_arg_state( const struct lp_rast_state
*state
)
224 union lp_rast_cmd_arg arg
;
225 arg
.set_state
= state
;
229 static inline union lp_rast_cmd_arg
230 lp_rast_arg_fence( struct lp_fence
*fence
)
232 union lp_rast_cmd_arg arg
;
238 static inline union lp_rast_cmd_arg
239 lp_rast_arg_clearzs( uint64_t value
, uint64_t mask
)
241 union lp_rast_cmd_arg arg
;
242 arg
.clear_zstencil
.value
= value
;
243 arg
.clear_zstencil
.mask
= mask
;
248 static inline union lp_rast_cmd_arg
249 lp_rast_arg_query( struct llvmpipe_query
*pq
)
251 union lp_rast_cmd_arg arg
;
256 static inline union lp_rast_cmd_arg
257 lp_rast_arg_null( void )
259 union lp_rast_cmd_arg arg
;
260 arg
.set_state
= NULL
;
267 * These get put into bins by the setup code and are called when
268 * the bins are executed.
270 #define LP_RAST_OP_CLEAR_COLOR 0x0
271 #define LP_RAST_OP_CLEAR_ZSTENCIL 0x1
272 #define LP_RAST_OP_TRIANGLE_1 0x2
273 #define LP_RAST_OP_TRIANGLE_2 0x3
274 #define LP_RAST_OP_TRIANGLE_3 0x4
275 #define LP_RAST_OP_TRIANGLE_4 0x5
276 #define LP_RAST_OP_TRIANGLE_5 0x6
277 #define LP_RAST_OP_TRIANGLE_6 0x7
278 #define LP_RAST_OP_TRIANGLE_7 0x8
279 #define LP_RAST_OP_TRIANGLE_8 0x9
280 #define LP_RAST_OP_TRIANGLE_3_4 0xa
281 #define LP_RAST_OP_TRIANGLE_3_16 0xb
282 #define LP_RAST_OP_TRIANGLE_4_16 0xc
283 #define LP_RAST_OP_SHADE_TILE 0xd
284 #define LP_RAST_OP_SHADE_TILE_OPAQUE 0xe
285 #define LP_RAST_OP_BEGIN_QUERY 0xf
286 #define LP_RAST_OP_END_QUERY 0x10
287 #define LP_RAST_OP_SET_STATE 0x11
288 #define LP_RAST_OP_TRIANGLE_32_1 0x12
289 #define LP_RAST_OP_TRIANGLE_32_2 0x13
290 #define LP_RAST_OP_TRIANGLE_32_3 0x14
291 #define LP_RAST_OP_TRIANGLE_32_4 0x15
292 #define LP_RAST_OP_TRIANGLE_32_5 0x16
293 #define LP_RAST_OP_TRIANGLE_32_6 0x17
294 #define LP_RAST_OP_TRIANGLE_32_7 0x18
295 #define LP_RAST_OP_TRIANGLE_32_8 0x19
296 #define LP_RAST_OP_TRIANGLE_32_3_4 0x1a
297 #define LP_RAST_OP_TRIANGLE_32_3_16 0x1b
298 #define LP_RAST_OP_TRIANGLE_32_4_16 0x1c
300 #define LP_RAST_OP_MAX 0x1d
301 #define LP_RAST_OP_MASK 0xff
304 lp_debug_bins( struct lp_scene
*scene
);
306 lp_debug_draw_bins_by_cmd_length( struct lp_scene
*scene
);
308 lp_debug_draw_bins_by_coverage( struct lp_scene
*scene
);
312 #include <emmintrin.h>
313 #include "util/u_sse.h"
315 static inline __m128i
316 lp_plane_to_m128i(const struct lp_rast_plane
*plane
)
318 return _mm_setr_epi32((int32_t)plane
->c
, (int32_t)plane
->dcdx
,
319 (int32_t)plane
->dcdy
, (int32_t)plane
->eo
);