2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
6 Permission is hereby granted, free of charge, to any person obtaining
7 a 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, sublicense, 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
16 portions of the Software.
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **********************************************************************/
29 * Keith Whitwell <keith@tungstengraphics.com>
31 #include "tgsi/tgsi_info.h"
33 #include "brw_context.h"
34 #include "brw_screen.h"
37 #include "brw_state.h"
38 #include "brw_debug.h"
39 #include "brw_pipe_rast.h"
42 /** Return number of src args for given instruction */
43 GLuint
brw_wm_nr_args( GLuint opcode
)
60 assert(opcode
< MAX_OPCODE
);
61 return tgsi_get_opcode_info(opcode
)->num_src
;
66 GLuint
brw_wm_is_scalar_result( GLuint opcode
)
89 * Do GPU code generation for shaders without flow control. Shaders
90 * without flow control instructions can more readily be analysed for
91 * SSA-style optimizations.
94 brw_wm_linear_shader_emit(struct brw_context
*brw
, struct brw_wm_compile
*c
)
96 /* Augment fragment program. Add instructions for pre- and
97 * post-fragment-program tasks such as interpolation and fogging.
101 /* Translate to intermediate representation. Build register usage
106 /* Dead code removal.
110 /* Register allocation.
111 * Divide by two because we operate on 16 pixels at a time and require
112 * two GRF entries for each logical shader register.
114 c
->grf_limit
= BRW_WM_MAX_GRF
/ 2;
118 /* how many general-purpose registers are used */
119 c
->prog_data
.total_grf
= c
->max_wm_grf
;
121 /* Scratch space is used for register spilling */
122 if (c
->last_scratch
) {
123 c
->prog_data
.total_scratch
= c
->last_scratch
+ 0x40;
126 c
->prog_data
.total_scratch
= 0;
136 * All Mesa program -> GPU code generation goes through this function.
137 * Depending on the instructions used (i.e. flow control instructions)
138 * we'll use one of two code generators.
140 static int do_wm_prog( struct brw_context
*brw
,
141 struct brw_fragment_shader
*fp
,
142 struct brw_wm_prog_key
*key
)
144 struct brw_wm_compile
*c
;
145 const GLuint
*program
;
148 c
= brw
->wm
.compile_data
;
150 brw
->wm
.compile_data
= calloc(1, sizeof(*brw
->wm
.compile_data
));
151 c
= brw
->wm
.compile_data
;
153 /* Ouch - big out of memory problem. Can't continue
154 * without triggering a segfault, no way to signal,
157 return PIPE_ERROR_OUT_OF_MEMORY
;
160 memset(c
, 0, sizeof(*brw
->wm
.compile_data
));
162 memcpy(&c
->key
, key
, sizeof(*key
));
165 c
->env_param
= NULL
; /*brw->intel.ctx.FragmentProgram.Parameters;*/
167 brw_init_compile(brw
, &c
->func
);
169 /* temporary sanity check assertion */
170 assert(fp
->has_flow_control
== brw_wm_has_flow_control(c
->fp
));
173 * Shader which use GLSL features such as flow control are handled
174 * differently from "simple" shaders.
176 if (fp
->has_flow_control
) {
177 c
->dispatch_width
= 8;
181 //brw_wm_branching_shader_emit(brw, c);
184 c
->dispatch_width
= 16;
185 brw_wm_linear_shader_emit(brw
, c
);
188 if (BRW_DEBUG
& DEBUG_WM
)
193 program
= brw_get_program(&c
->func
, &program_size
);
195 brw
->sws
->bo_unreference(brw
->wm
.prog_bo
);
196 brw
->wm
.prog_bo
= brw_upload_cache( &brw
->cache
, BRW_WM_PROG
,
197 &c
->key
, sizeof(c
->key
),
199 program
, program_size
,
201 &brw
->wm
.prog_data
);
208 static void brw_wm_populate_key( struct brw_context
*brw
,
209 struct brw_wm_prog_key
*key
)
211 unsigned lookup
, line_aa
;
214 memset(key
, 0, sizeof(*key
));
216 /* PIPE_NEW_FRAGMENT_SHADER
217 * PIPE_NEW_DEPTH_STENCIL_ALPHA
219 lookup
= (brw
->curr
.zstencil
->iz_lookup
|
220 brw
->curr
.fragment_shader
->iz_lookup
);
224 * BRW_NEW_REDUCED_PRIMITIVE
226 switch (brw
->reduced_primitive
) {
227 case PIPE_PRIM_POINTS
:
230 case PIPE_PRIM_LINES
:
234 line_aa
= brw
->curr
.rast
->unfilled_aa_line
;
238 brw_wm_lookup_iz(line_aa
,
240 brw
->curr
.fragment_shader
->uses_depth
,
244 key
->flat_shade
= brw
->curr
.rast
->templ
.flatshade
;
247 /* PIPE_NEW_BOUND_TEXTURES */
248 for (i
= 0; i
< brw
->curr
.num_textures
; i
++) {
249 const struct brw_texture
*tex
= brw_texture(brw
->curr
.texture
[i
]);
251 if (tex
->base
.format
== PIPE_FORMAT_YCBCR
)
252 key
->yuvtex_mask
|= 1 << i
;
254 if (tex
->base
.format
== PIPE_FORMAT_YCBCR_REV
)
255 key
->yuvtex_swap_mask
|= 1 << i
;
257 /* XXX: shadow texture
259 /* key->shadowtex_mask |= 1<<i; */
262 /* CACHE_NEW_VS_PROG */
263 key
->vp_nr_outputs
= brw
->vs
.prog_data
->nr_outputs
;
265 /* The unique fragment program ID */
266 key
->program_string_id
= brw
->curr
.fragment_shader
->id
;
270 static int brw_prepare_wm_prog(struct brw_context
*brw
)
272 struct brw_wm_prog_key key
;
273 struct brw_fragment_shader
*fs
= brw
->curr
.fragment_shader
;
275 brw_wm_populate_key(brw
, &key
);
277 /* Make an early check for the key.
279 brw
->sws
->bo_unreference(brw
->wm
.prog_bo
);
280 brw
->wm
.prog_bo
= brw_search_cache(&brw
->cache
, BRW_WM_PROG
,
284 if (brw
->wm
.prog_bo
== NULL
)
285 return do_wm_prog(brw
, fs
, &key
);
291 const struct brw_tracked_state brw_wm_prog
= {
293 .mesa
= (PIPE_NEW_FRAGMENT_SHADER
|
294 PIPE_NEW_DEPTH_STENCIL_ALPHA
|
296 PIPE_NEW_BOUND_TEXTURES
),
297 .brw
= (BRW_NEW_WM_INPUT_DIMENSIONS
|
298 BRW_NEW_REDUCED_PRIMITIVE
),
299 .cache
= CACHE_NEW_VS_PROG
,
301 .prepare
= brw_prepare_wm_prog