i965: Delete old metaops code now that there are no remaining consumers.
[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 memset(key, 0, sizeof(*key));
128
129 /* CACHE_NEW_SF_PROG */
130 key->total_grf = brw->sf.prog_data->total_grf;
131 key->urb_entry_read_length = brw->sf.prog_data->urb_read_length;
132
133 /* BRW_NEW_URB_FENCE */
134 key->nr_urb_entries = brw->urb.nr_sf_entries;
135 key->urb_size = brw->urb.vsize;
136 key->sfsize = brw->urb.sfsize;
137
138 key->scissor = brw->attribs.Scissor->Enabled;
139 key->front_face = brw->attribs.Polygon->FrontFace;
140
141 if (brw->attribs.Polygon->CullFlag)
142 key->cull_face = brw->attribs.Polygon->CullFaceMode;
143 else
144 key->cull_face = GL_NONE;
145
146 key->line_width = brw->attribs.Line->Width;
147 key->line_smooth = brw->attribs.Line->SmoothFlag;
148
149 key->point_sprite = brw->attribs.Point->PointSprite;
150 key->point_size = brw->attribs.Point->Size;
151 key->point_attenuated = brw->attribs.Point->_Attenuated;
152
153 key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
154 }
155
156 static dri_bo *
157 sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
158 dri_bo **reloc_bufs)
159 {
160 struct brw_sf_unit_state sf;
161 dri_bo *bo;
162
163 memset(&sf, 0, sizeof(sf));
164
165 sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
166 sf.thread0.kernel_start_pointer = brw->sf.prog_bo->offset >> 6; /* reloc */
167
168 sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
169
170 sf.thread3.dispatch_grf_start_reg = 3;
171 sf.thread3.urb_entry_read_offset = 1;
172 sf.thread3.urb_entry_read_length = key->urb_entry_read_length;
173
174 sf.thread4.nr_urb_entries = key->nr_urb_entries;
175 sf.thread4.urb_entry_allocation_size = key->sfsize - 1;
176 /* Each SF thread produces 1 PUE, and there can be up to 24 threads */
177 sf.thread4.max_threads = MIN2(24, key->nr_urb_entries) - 1;
178
179 if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
180 sf.thread4.max_threads = 0;
181
182 if (INTEL_DEBUG & DEBUG_STATS)
183 sf.thread4.stats_enable = 1;
184
185 /* CACHE_NEW_SF_VP */
186 sf.sf5.sf_viewport_state_offset = brw->sf.vp_bo->offset >> 5; /* reloc */
187
188 sf.sf5.viewport_transform = 1;
189
190 /* _NEW_SCISSOR */
191 if (key->scissor)
192 sf.sf6.scissor = 1;
193
194 /* _NEW_POLYGON */
195 if (key->front_face == GL_CCW)
196 sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
197 else
198 sf.sf5.front_winding = BRW_FRONTWINDING_CW;
199
200 /* The viewport is inverted for rendering to a FBO, and that inverts
201 * polygon front/back orientation.
202 */
203 sf.sf5.front_winding ^= key->render_to_fbo;
204
205 switch (key->cull_face) {
206 case GL_FRONT:
207 sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
208 break;
209 case GL_BACK:
210 sf.sf6.cull_mode = BRW_CULLMODE_BACK;
211 break;
212 case GL_FRONT_AND_BACK:
213 sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
214 break;
215 case GL_NONE:
216 sf.sf6.cull_mode = BRW_CULLMODE_NONE;
217 break;
218 default:
219 assert(0);
220 break;
221 }
222
223 /* _NEW_LINE */
224 /* XXX use ctx->Const.Min/MaxLineWidth here */
225 sf.sf6.line_width = CLAMP(key->line_width, 1.0, 5.0) * (1<<1);
226
227 sf.sf6.line_endcap_aa_region_width = 1;
228 if (key->line_smooth)
229 sf.sf6.aa_enable = 1;
230 else if (sf.sf6.line_width <= 0x2)
231 sf.sf6.line_width = 0;
232
233 /* _NEW_POINT */
234 sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT; /* opengl conventions */
235 /* XXX clamp max depends on AA vs. non-AA */
236
237 sf.sf7.sprite_point = key->point_sprite;
238 sf.sf7.point_size = CLAMP(rint(key->point_size), 1, 255) * (1<<3);
239 sf.sf7.use_point_size_state = !key->point_attenuated;
240 sf.sf7.aa_line_distance_mode = 0;
241
242 /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
243 */
244 sf.sf7.trifan_pv = 2;
245 sf.sf7.linestrip_pv = 1;
246 sf.sf7.tristrip_pv = 2;
247 sf.sf7.line_last_pixel_enable = 0;
248
249 /* Set bias for OpenGL rasterization rules:
250 */
251 sf.sf6.dest_org_vbias = 0x8;
252 sf.sf6.dest_org_hbias = 0x8;
253
254 bo = brw_upload_cache(&brw->cache, BRW_SF_UNIT,
255 key, sizeof(*key),
256 reloc_bufs, 2,
257 &sf, sizeof(sf),
258 NULL, NULL);
259
260 /* Emit SF program relocation */
261 dri_bo_emit_reloc(bo,
262 I915_GEM_DOMAIN_INSTRUCTION, 0,
263 sf.thread0.grf_reg_count << 1,
264 offsetof(struct brw_sf_unit_state, thread0),
265 brw->sf.prog_bo);
266
267 /* Emit SF viewport relocation */
268 dri_bo_emit_reloc(bo,
269 I915_GEM_DOMAIN_INSTRUCTION, 0,
270 sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
271 offsetof(struct brw_sf_unit_state, sf5),
272 brw->sf.vp_bo);
273
274 return bo;
275 }
276
277 static void upload_sf_unit( struct brw_context *brw )
278 {
279 struct brw_sf_unit_key key;
280 dri_bo *reloc_bufs[2];
281
282 sf_unit_populate_key(brw, &key);
283
284 reloc_bufs[0] = brw->sf.prog_bo;
285 reloc_bufs[1] = brw->sf.vp_bo;
286
287 dri_bo_unreference(brw->sf.state_bo);
288 brw->sf.state_bo = brw_search_cache(&brw->cache, BRW_SF_UNIT,
289 &key, sizeof(key),
290 reloc_bufs, 2,
291 NULL);
292 if (brw->sf.state_bo == NULL) {
293 brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs);
294 }
295 }
296
297 const struct brw_tracked_state brw_sf_unit = {
298 .dirty = {
299 .mesa = (_NEW_POLYGON |
300 _NEW_LINE |
301 _NEW_POINT |
302 _NEW_SCISSOR),
303 .brw = BRW_NEW_URB_FENCE,
304 .cache = (CACHE_NEW_SF_VP |
305 CACHE_NEW_SF_PROG)
306 },
307 .prepare = upload_sf_unit,
308 };