i965: Remove brw->attribs now that we can just always look in the GLcontext.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_sf_state.c
1 /*
2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
5
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:
13
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.
17
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.
25
26 **********************************************************************/
27 /*
28 * Authors:
29 * Keith Whitwell <keith@tungstengraphics.com>
30 */
31
32
33
34 #include "brw_context.h"
35 #include "brw_state.h"
36 #include "brw_defines.h"
37 #include "main/macros.h"
38 #include "intel_fbo.h"
39
40 static void upload_sf_vp(struct brw_context *brw)
41 {
42 GLcontext *ctx = &brw->intel.ctx;
43 const GLfloat depth_scale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
44 struct brw_sf_viewport sfv;
45 GLfloat y_scale, y_bias;
46 const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
47
48 memset(&sfv, 0, sizeof(sfv));
49
50 if (render_to_fbo) {
51 y_scale = 1.0;
52 y_bias = 0;
53 }
54 else {
55 y_scale = -1.0;
56 y_bias = ctx->DrawBuffer->Height;
57 }
58
59 /* _NEW_VIEWPORT */
60
61 const GLfloat *v = ctx->Viewport._WindowMap.m;
62
63 sfv.viewport.m00 = v[MAT_SX];
64 sfv.viewport.m11 = v[MAT_SY] * y_scale;
65 sfv.viewport.m22 = v[MAT_SZ] * depth_scale;
66 sfv.viewport.m30 = v[MAT_TX];
67 sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias;
68 sfv.viewport.m32 = v[MAT_TZ] * depth_scale;
69
70 /* _NEW_SCISSOR */
71
72 /* The scissor only needs to handle the intersection of drawable and
73 * scissor rect. Clipping to the boundaries of static shared buffers
74 * for front/back/depth is covered by looping over cliprects in brw_draw.c.
75 *
76 * Note that the hardware's coordinates are inclusive, while Mesa's min is
77 * inclusive but max is exclusive.
78 */
79 if (render_to_fbo) {
80 /* texmemory: Y=0=bottom */
81 sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
82 sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
83 sfv.scissor.ymin = ctx->DrawBuffer->_Ymin;
84 sfv.scissor.ymax = ctx->DrawBuffer->_Ymax - 1;
85 }
86 else {
87 /* memory: Y=0=top */
88 sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
89 sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
90 sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax;
91 sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1;
92 }
93
94 dri_bo_unreference(brw->sf.vp_bo);
95 brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 );
96 }
97
98 const struct brw_tracked_state brw_sf_vp = {
99 .dirty = {
100 .mesa = (_NEW_VIEWPORT |
101 _NEW_SCISSOR),
102 .brw = 0,
103 .cache = 0
104 },
105 .prepare = upload_sf_vp
106 };
107
108 struct brw_sf_unit_key {
109 unsigned int total_grf;
110 unsigned int urb_entry_read_length;
111
112 unsigned int nr_urb_entries, urb_size, sfsize;
113
114 GLenum front_face, cull_face;
115 unsigned scissor:1;
116 unsigned line_smooth:1;
117 unsigned point_sprite:1;
118 unsigned point_attenuated:1;
119 unsigned render_to_fbo:1;
120 float line_width;
121 float point_size;
122 };
123
124 static void
125 sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
126 {
127 GLcontext *ctx = &brw->intel.ctx;
128 memset(key, 0, sizeof(*key));
129
130 /* CACHE_NEW_SF_PROG */
131 key->total_grf = brw->sf.prog_data->total_grf;
132 key->urb_entry_read_length = brw->sf.prog_data->urb_read_length;
133
134 /* BRW_NEW_URB_FENCE */
135 key->nr_urb_entries = brw->urb.nr_sf_entries;
136 key->urb_size = brw->urb.vsize;
137 key->sfsize = brw->urb.sfsize;
138
139 key->scissor = ctx->Scissor.Enabled;
140 key->front_face = ctx->Polygon.FrontFace;
141
142 if (ctx->Polygon.CullFlag)
143 key->cull_face = ctx->Polygon.CullFaceMode;
144 else
145 key->cull_face = GL_NONE;
146
147 key->line_width = ctx->Line.Width;
148 key->line_smooth = ctx->Line.SmoothFlag;
149
150 key->point_sprite = ctx->Point.PointSprite;
151 key->point_size = ctx->Point.Size;
152 key->point_attenuated = ctx->Point._Attenuated;
153
154 key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
155 }
156
157 static dri_bo *
158 sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
159 dri_bo **reloc_bufs)
160 {
161 struct brw_sf_unit_state sf;
162 dri_bo *bo;
163
164 memset(&sf, 0, sizeof(sf));
165
166 sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
167 sf.thread0.kernel_start_pointer = brw->sf.prog_bo->offset >> 6; /* reloc */
168
169 sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
170
171 sf.thread3.dispatch_grf_start_reg = 3;
172 sf.thread3.urb_entry_read_offset = 1;
173 sf.thread3.urb_entry_read_length = key->urb_entry_read_length;
174
175 sf.thread4.nr_urb_entries = key->nr_urb_entries;
176 sf.thread4.urb_entry_allocation_size = key->sfsize - 1;
177 /* Each SF thread produces 1 PUE, and there can be up to 24 threads */
178 sf.thread4.max_threads = MIN2(24, key->nr_urb_entries) - 1;
179
180 if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
181 sf.thread4.max_threads = 0;
182
183 if (INTEL_DEBUG & DEBUG_STATS)
184 sf.thread4.stats_enable = 1;
185
186 /* CACHE_NEW_SF_VP */
187 sf.sf5.sf_viewport_state_offset = brw->sf.vp_bo->offset >> 5; /* reloc */
188
189 sf.sf5.viewport_transform = 1;
190
191 /* _NEW_SCISSOR */
192 if (key->scissor)
193 sf.sf6.scissor = 1;
194
195 /* _NEW_POLYGON */
196 if (key->front_face == GL_CCW)
197 sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
198 else
199 sf.sf5.front_winding = BRW_FRONTWINDING_CW;
200
201 /* The viewport is inverted for rendering to a FBO, and that inverts
202 * polygon front/back orientation.
203 */
204 sf.sf5.front_winding ^= key->render_to_fbo;
205
206 switch (key->cull_face) {
207 case GL_FRONT:
208 sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
209 break;
210 case GL_BACK:
211 sf.sf6.cull_mode = BRW_CULLMODE_BACK;
212 break;
213 case GL_FRONT_AND_BACK:
214 sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
215 break;
216 case GL_NONE:
217 sf.sf6.cull_mode = BRW_CULLMODE_NONE;
218 break;
219 default:
220 assert(0);
221 break;
222 }
223
224 /* _NEW_LINE */
225 /* XXX use ctx->Const.Min/MaxLineWidth here */
226 sf.sf6.line_width = CLAMP(key->line_width, 1.0, 5.0) * (1<<1);
227
228 sf.sf6.line_endcap_aa_region_width = 1;
229 if (key->line_smooth)
230 sf.sf6.aa_enable = 1;
231 else if (sf.sf6.line_width <= 0x2)
232 sf.sf6.line_width = 0;
233
234 /* _NEW_POINT */
235 sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT; /* opengl conventions */
236 /* XXX clamp max depends on AA vs. non-AA */
237
238 sf.sf7.sprite_point = key->point_sprite;
239 sf.sf7.point_size = CLAMP(rint(key->point_size), 1, 255) * (1<<3);
240 sf.sf7.use_point_size_state = !key->point_attenuated;
241 sf.sf7.aa_line_distance_mode = 0;
242
243 /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
244 */
245 sf.sf7.trifan_pv = 2;
246 sf.sf7.linestrip_pv = 1;
247 sf.sf7.tristrip_pv = 2;
248 sf.sf7.line_last_pixel_enable = 0;
249
250 /* Set bias for OpenGL rasterization rules:
251 */
252 sf.sf6.dest_org_vbias = 0x8;
253 sf.sf6.dest_org_hbias = 0x8;
254
255 bo = brw_upload_cache(&brw->cache, BRW_SF_UNIT,
256 key, sizeof(*key),
257 reloc_bufs, 2,
258 &sf, sizeof(sf),
259 NULL, NULL);
260
261 /* Emit SF program relocation */
262 dri_bo_emit_reloc(bo,
263 I915_GEM_DOMAIN_INSTRUCTION, 0,
264 sf.thread0.grf_reg_count << 1,
265 offsetof(struct brw_sf_unit_state, thread0),
266 brw->sf.prog_bo);
267
268 /* Emit SF viewport relocation */
269 dri_bo_emit_reloc(bo,
270 I915_GEM_DOMAIN_INSTRUCTION, 0,
271 sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
272 offsetof(struct brw_sf_unit_state, sf5),
273 brw->sf.vp_bo);
274
275 return bo;
276 }
277
278 static void upload_sf_unit( struct brw_context *brw )
279 {
280 struct brw_sf_unit_key key;
281 dri_bo *reloc_bufs[2];
282
283 sf_unit_populate_key(brw, &key);
284
285 reloc_bufs[0] = brw->sf.prog_bo;
286 reloc_bufs[1] = brw->sf.vp_bo;
287
288 dri_bo_unreference(brw->sf.state_bo);
289 brw->sf.state_bo = brw_search_cache(&brw->cache, BRW_SF_UNIT,
290 &key, sizeof(key),
291 reloc_bufs, 2,
292 NULL);
293 if (brw->sf.state_bo == NULL) {
294 brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs);
295 }
296 }
297
298 const struct brw_tracked_state brw_sf_unit = {
299 .dirty = {
300 .mesa = (_NEW_POLYGON |
301 _NEW_LINE |
302 _NEW_POINT |
303 _NEW_SCISSOR),
304 .brw = BRW_NEW_URB_FENCE,
305 .cache = (CACHE_NEW_SF_VP |
306 CACHE_NEW_SF_PROG)
307 },
308 .prepare = upload_sf_unit,
309 };