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 enum pipe_error
do_wm_prog( struct brw_context
*brw
,
141 struct brw_fragment_shader
*fp
,
142 struct brw_wm_prog_key
*key
,
143 struct brw_winsys_buffer
**bo_out
)
146 struct brw_wm_compile
*c
;
147 const GLuint
*program
;
150 if (brw
->wm
.compile_data
== NULL
) {
151 brw
->wm
.compile_data
= MALLOC(sizeof(*brw
->wm
.compile_data
));
152 if (!brw
->wm
.compile_data
)
153 return PIPE_ERROR_OUT_OF_MEMORY
;
156 c
= brw
->wm
.compile_data
;
157 memset(c
, 0, sizeof *c
);
161 c
->env_param
= NULL
; /*brw->intel.ctx.FragmentProgram.Parameters;*/
163 brw_init_compile(brw
, &c
->func
);
166 * Shader which use GLSL features such as flow control are handled
167 * differently from "simple" shaders.
169 if (fp
->has_flow_control
) {
170 c
->dispatch_width
= 8;
174 //brw_wm_branching_shader_emit(brw, c);
177 c
->dispatch_width
= 16;
178 brw_wm_linear_shader_emit(brw
, c
);
181 if (BRW_DEBUG
& DEBUG_WM
)
186 ret
= brw_get_program(&c
->func
, &program
, &program_size
);
190 ret
= brw_upload_cache( &brw
->cache
, BRW_WM_PROG
,
191 &c
->key
, sizeof(c
->key
),
193 program
, program_size
,
205 static void brw_wm_populate_key( struct brw_context
*brw
,
206 struct brw_wm_prog_key
*key
)
208 unsigned lookup
, line_aa
;
211 memset(key
, 0, sizeof(*key
));
213 /* PIPE_NEW_FRAGMENT_SHADER
214 * PIPE_NEW_DEPTH_STENCIL_ALPHA
216 lookup
= (brw
->curr
.zstencil
->iz_lookup
|
217 brw
->curr
.fragment_shader
->iz_lookup
);
221 * BRW_NEW_REDUCED_PRIMITIVE
223 switch (brw
->reduced_primitive
) {
224 case PIPE_PRIM_POINTS
:
227 case PIPE_PRIM_LINES
:
228 line_aa
= (brw
->curr
.rast
->templ
.line_smooth
?
229 AA_ALWAYS
: AA_NEVER
);
232 line_aa
= brw
->curr
.rast
->unfilled_aa_line
;
236 brw_wm_lookup_iz(line_aa
,
238 brw
->curr
.fragment_shader
->uses_depth
,
242 key
->flat_shade
= brw
->curr
.rast
->templ
.flatshade
;
245 /* PIPE_NEW_BOUND_TEXTURES */
246 for (i
= 0; i
< brw
->curr
.num_textures
; i
++) {
247 const struct brw_texture
*tex
= brw_texture(brw
->curr
.texture
[i
]);
249 if (tex
->base
.format
== PIPE_FORMAT_YCBCR
)
250 key
->yuvtex_mask
|= 1 << i
;
252 if (tex
->base
.format
== PIPE_FORMAT_YCBCR_REV
)
253 key
->yuvtex_swap_mask
|= 1 << i
;
255 /* XXX: shadow texture
257 /* key->shadowtex_mask |= 1<<i; */
260 /* CACHE_NEW_VS_PROG */
261 key
->vp_nr_outputs
= brw
->vs
.prog_data
->nr_outputs
;
263 key
->nr_cbufs
= brw
->curr
.fb
.nr_cbufs
;
265 key
->nr_inputs
= brw
->curr
.fragment_shader
->info
.num_inputs
;
267 /* The unique fragment program ID */
268 key
->program_string_id
= brw
->curr
.fragment_shader
->id
;
272 static enum pipe_error
brw_prepare_wm_prog(struct brw_context
*brw
)
274 struct brw_wm_prog_key key
;
275 struct brw_fragment_shader
*fs
= brw
->curr
.fragment_shader
;
278 brw_wm_populate_key(brw
, &key
);
280 /* Make an early check for the key.
282 if (brw_search_cache(&brw
->cache
, BRW_WM_PROG
,
289 ret
= do_wm_prog(brw
, fs
, &key
, &brw
->wm
.prog_bo
);
297 const struct brw_tracked_state brw_wm_prog
= {
299 .mesa
= (PIPE_NEW_FRAGMENT_SHADER
|
300 PIPE_NEW_DEPTH_STENCIL_ALPHA
|
303 PIPE_NEW_BOUND_TEXTURES
),
304 .brw
= (BRW_NEW_WM_INPUT_DIMENSIONS
|
305 BRW_NEW_REDUCED_PRIMITIVE
),
306 .cache
= CACHE_NEW_VS_PROG
,
308 .prepare
= brw_prepare_wm_prog