Merge commit 'origin/gallium-0.1' into gallium-0.2
[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 struct intel_renderbuffer *irb =
46 intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
47 GLfloat y_scale, y_bias;
48
49 memset(&sfv, 0, sizeof(sfv));
50
51 if (ctx->DrawBuffer->Name) {
52 /* User-created FBO */
53 if (irb && !irb->RenderToTexture) {
54 y_scale = -1.0;
55 y_bias = ctx->DrawBuffer->Height;
56 } else {
57 y_scale = 1.0;
58 y_bias = 0;
59 }
60 } else {
61 y_scale = -1.0;
62 y_bias = ctx->DrawBuffer->Height;
63 }
64
65 /* _NEW_VIEWPORT, BRW_NEW_METAOPS */
66
67 if (!brw->metaops.active) {
68 const GLfloat *v = ctx->Viewport._WindowMap.m;
69
70 sfv.viewport.m00 = v[MAT_SX];
71 sfv.viewport.m11 = v[MAT_SY] * y_scale;
72 sfv.viewport.m22 = v[MAT_SZ] * depth_scale;
73 sfv.viewport.m30 = v[MAT_TX];
74 sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias;
75 sfv.viewport.m32 = v[MAT_TZ] * depth_scale;
76 } else {
77 sfv.viewport.m00 = 1;
78 sfv.viewport.m11 = - 1;
79 sfv.viewport.m22 = 1;
80 sfv.viewport.m30 = 0;
81 sfv.viewport.m31 = ctx->DrawBuffer->Height;
82 sfv.viewport.m32 = 0;
83 }
84
85 /* _NEW_SCISSOR */
86
87 /* The scissor only needs to handle the intersection of drawable and
88 * scissor rect. Clipping to the boundaries of static shared buffers
89 * for front/back/depth is covered by looping over cliprects in brw_draw.c.
90 *
91 * Note that the hardware's coordinates are inclusive, while Mesa's min is
92 * inclusive but max is exclusive.
93 */
94 sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
95 sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
96 sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax;
97 sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1;
98
99 dri_bo_unreference(brw->sf.vp_bo);
100 brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 );
101 }
102
103 const struct brw_tracked_state brw_sf_vp = {
104 .dirty = {
105 .mesa = (_NEW_VIEWPORT |
106 _NEW_SCISSOR),
107 .brw = BRW_NEW_METAOPS,
108 .cache = 0
109 },
110 .prepare = upload_sf_vp
111 };
112
113 struct brw_sf_unit_key {
114 unsigned int total_grf;
115 unsigned int urb_entry_read_length;
116
117 unsigned int nr_urb_entries, urb_size, sfsize;
118
119 GLenum front_face, cull_face;
120 GLboolean scissor, line_smooth, point_sprite, point_attenuated;
121 float line_width;
122 float point_size;
123 };
124
125 static void
126 sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
127 {
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 = brw->attribs.Scissor->Enabled;
140 key->front_face = brw->attribs.Polygon->FrontFace;
141
142 if (brw->attribs.Polygon->CullFlag)
143 key->cull_face = brw->attribs.Polygon->CullFaceMode;
144 else
145 key->cull_face = GL_NONE;
146
147 key->line_width = brw->attribs.Line->Width;
148 key->line_smooth = brw->attribs.Line->SmoothFlag;
149
150 key->point_sprite = brw->attribs.Point->PointSprite;
151 key->point_size = brw->attribs.Point->Size;
152 key->point_attenuated = brw->attribs.Point->_Attenuated;
153 }
154
155 static dri_bo *
156 sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
157 dri_bo **reloc_bufs)
158 {
159 struct brw_sf_unit_state sf;
160 dri_bo *bo;
161
162 memset(&sf, 0, sizeof(sf));
163
164 sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
165 sf.thread0.kernel_start_pointer = brw->sf.prog_bo->offset >> 6; /* reloc */
166
167 sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
168
169 sf.thread3.dispatch_grf_start_reg = 3;
170 sf.thread3.urb_entry_read_offset = 1;
171 sf.thread3.urb_entry_read_length = key->urb_entry_read_length;
172
173 sf.thread4.nr_urb_entries = key->nr_urb_entries;
174 sf.thread4.urb_entry_allocation_size = key->sfsize - 1;
175 /* Each SF thread produces 1 PUE, and there can be up to 24 threads */
176 sf.thread4.max_threads = MIN2(24, key->nr_urb_entries) - 1;
177
178 if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
179 sf.thread4.max_threads = 0;
180
181 if (INTEL_DEBUG & DEBUG_STATS)
182 sf.thread4.stats_enable = 1;
183
184 /* CACHE_NEW_SF_VP */
185 sf.sf5.sf_viewport_state_offset = brw->sf.vp_bo->offset >> 5; /* reloc */
186
187 sf.sf5.viewport_transform = 1;
188
189 /* _NEW_SCISSOR */
190 if (key->scissor)
191 sf.sf6.scissor = 1;
192
193 /* _NEW_POLYGON */
194 if (key->front_face == GL_CCW)
195 sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
196 else
197 sf.sf5.front_winding = BRW_FRONTWINDING_CW;
198
199 switch (key->cull_face) {
200 case GL_FRONT:
201 sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
202 break;
203 case GL_BACK:
204 sf.sf6.cull_mode = BRW_CULLMODE_BACK;
205 break;
206 case GL_FRONT_AND_BACK:
207 sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
208 break;
209 case GL_NONE:
210 sf.sf6.cull_mode = BRW_CULLMODE_NONE;
211 break;
212 default:
213 assert(0);
214 break;
215 }
216
217 /* _NEW_LINE */
218 /* XXX use ctx->Const.Min/MaxLineWidth here */
219 sf.sf6.line_width = CLAMP(key->line_width, 1.0, 5.0) * (1<<1);
220
221 sf.sf6.line_endcap_aa_region_width = 1;
222 if (key->line_smooth)
223 sf.sf6.aa_enable = 1;
224 else if (sf.sf6.line_width <= 0x2)
225 sf.sf6.line_width = 0;
226
227 /* _NEW_POINT */
228 sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT; /* opengl conventions */
229 /* XXX clamp max depends on AA vs. non-AA */
230
231 sf.sf7.sprite_point = key->point_sprite;
232 sf.sf7.point_size = CLAMP(nearbyint(key->point_size), 1, 255) * (1<<3);
233 sf.sf7.use_point_size_state = !key->point_attenuated;
234 sf.sf7.aa_line_distance_mode = 0;
235
236 /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
237 */
238 sf.sf7.trifan_pv = 2;
239 sf.sf7.linestrip_pv = 1;
240 sf.sf7.tristrip_pv = 2;
241 sf.sf7.line_last_pixel_enable = 0;
242
243 /* Set bias for OpenGL rasterization rules:
244 */
245 sf.sf6.dest_org_vbias = 0x8;
246 sf.sf6.dest_org_hbias = 0x8;
247
248 bo = brw_upload_cache(&brw->cache, BRW_SF_UNIT,
249 key, sizeof(*key),
250 reloc_bufs, 2,
251 &sf, sizeof(sf),
252 NULL, NULL);
253
254 /* Emit SF program relocation */
255 dri_bo_emit_reloc(bo,
256 I915_GEM_DOMAIN_INSTRUCTION, 0,
257 sf.thread0.grf_reg_count << 1,
258 offsetof(struct brw_sf_unit_state, thread0),
259 brw->sf.prog_bo);
260
261 /* Emit SF viewport relocation */
262 dri_bo_emit_reloc(bo,
263 I915_GEM_DOMAIN_INSTRUCTION, 0,
264 sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
265 offsetof(struct brw_sf_unit_state, sf5),
266 brw->sf.vp_bo);
267
268 return bo;
269 }
270
271 static void upload_sf_unit( struct brw_context *brw )
272 {
273 struct brw_sf_unit_key key;
274 dri_bo *reloc_bufs[2];
275
276 sf_unit_populate_key(brw, &key);
277
278 reloc_bufs[0] = brw->sf.prog_bo;
279 reloc_bufs[1] = brw->sf.vp_bo;
280
281 dri_bo_unreference(brw->sf.state_bo);
282 brw->sf.state_bo = brw_search_cache(&brw->cache, BRW_SF_UNIT,
283 &key, sizeof(key),
284 reloc_bufs, 2,
285 NULL);
286 if (brw->sf.state_bo == NULL) {
287 brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs);
288 }
289 }
290
291 const struct brw_tracked_state brw_sf_unit = {
292 .dirty = {
293 .mesa = (_NEW_POLYGON |
294 _NEW_LINE |
295 _NEW_POINT |
296 _NEW_SCISSOR),
297 .brw = (BRW_NEW_URB_FENCE |
298 BRW_NEW_METAOPS),
299 .cache = (CACHE_NEW_SF_VP |
300 CACHE_NEW_SF_PROG)
301 },
302 .prepare = upload_sf_unit,
303 };