Merge commit 'origin/master' into drm-gem
[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 "macros.h"
38 #include "intel_fbo.h"
39
40 static int 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 return dri_bufmgr_check_aperture_space(brw->sf.vp_bo);
103 }
104
105 const struct brw_tracked_state brw_sf_vp = {
106 .dirty = {
107 .mesa = (_NEW_VIEWPORT |
108 _NEW_SCISSOR),
109 .brw = BRW_NEW_METAOPS,
110 .cache = 0
111 },
112 .prepare = upload_sf_vp
113 };
114
115 struct brw_sf_unit_key {
116 unsigned int total_grf;
117 unsigned int urb_entry_read_length;
118
119 unsigned int nr_urb_entries, urb_size, sfsize;
120
121 GLenum front_face, cull_face;
122 GLboolean scissor, line_smooth, point_sprite, point_attenuated;
123 float line_width;
124 float point_size;
125 };
126
127 static void
128 sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
129 {
130 memset(key, 0, sizeof(*key));
131
132 /* CACHE_NEW_SF_PROG */
133 key->total_grf = brw->sf.prog_data->total_grf;
134 key->urb_entry_read_length = brw->sf.prog_data->urb_read_length;
135
136 /* BRW_NEW_URB_FENCE */
137 key->nr_urb_entries = brw->urb.nr_sf_entries;
138 key->urb_size = brw->urb.vsize;
139 key->sfsize = brw->urb.sfsize;
140
141 key->scissor = brw->attribs.Scissor->Enabled;
142 key->front_face = brw->attribs.Polygon->FrontFace;
143
144 if (brw->attribs.Polygon->CullFlag)
145 key->cull_face = brw->attribs.Polygon->CullFaceMode;
146 else
147 key->cull_face = GL_NONE;
148
149 key->line_width = brw->attribs.Line->Width;
150 key->line_smooth = brw->attribs.Line->SmoothFlag;
151
152 key->point_sprite = brw->attribs.Point->PointSprite;
153 key->point_size = brw->attribs.Point->Size;
154 key->point_attenuated = brw->attribs.Point->_Attenuated;
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 sf.thread4.max_threads = MIN2(12, key->nr_urb_entries / 2) - 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 switch (key->cull_face) {
201 case GL_FRONT:
202 sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
203 break;
204 case GL_BACK:
205 sf.sf6.cull_mode = BRW_CULLMODE_BACK;
206 break;
207 case GL_FRONT_AND_BACK:
208 sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
209 break;
210 case GL_NONE:
211 sf.sf6.cull_mode = BRW_CULLMODE_NONE;
212 break;
213 default:
214 assert(0);
215 break;
216 }
217
218 /* _NEW_LINE */
219 /* XXX use ctx->Const.Min/MaxLineWidth here */
220 sf.sf6.line_width = CLAMP(key->line_width, 1.0, 5.0) * (1<<1);
221
222 sf.sf6.line_endcap_aa_region_width = 1;
223 if (key->line_smooth)
224 sf.sf6.aa_enable = 1;
225 else if (sf.sf6.line_width <= 0x2)
226 sf.sf6.line_width = 0;
227
228 /* _NEW_POINT */
229 sf.sf6.point_rast_rule = BRW_RASTRULE_UPPER_RIGHT; /* opengl conventions */
230 /* XXX clamp max depends on AA vs. non-AA */
231
232 sf.sf7.sprite_point = key->point_sprite;
233 sf.sf7.point_size = CLAMP(nearbyint(key->point_size), 1, 255) * (1<<3);
234 sf.sf7.use_point_size_state = !key->point_attenuated;
235 sf.sf7.aa_line_distance_mode = 0;
236
237 /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
238 */
239 sf.sf7.trifan_pv = 2;
240 sf.sf7.linestrip_pv = 1;
241 sf.sf7.tristrip_pv = 2;
242 sf.sf7.line_last_pixel_enable = 0;
243
244 /* Set bias for OpenGL rasterization rules:
245 */
246 sf.sf6.dest_org_vbias = 0x8;
247 sf.sf6.dest_org_hbias = 0x8;
248
249 bo = brw_upload_cache(&brw->cache, BRW_SF_UNIT,
250 key, sizeof(*key),
251 reloc_bufs, 2,
252 &sf, sizeof(sf),
253 NULL, NULL);
254
255 /* Emit SF program relocation */
256 intel_bo_emit_reloc(bo,
257 DRM_GEM_DOMAIN_I915_INSTRUCTION, 0,
258 sf.thread0.grf_reg_count << 1,
259 offsetof(struct brw_sf_unit_state, thread0),
260 brw->sf.prog_bo);
261
262 /* Emit SF viewport relocation */
263 intel_bo_emit_reloc(bo,
264 DRM_GEM_DOMAIN_I915_INSTRUCTION, 0,
265 sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
266 offsetof(struct brw_sf_unit_state, sf5),
267 brw->sf.vp_bo);
268
269 return bo;
270 }
271
272 static int upload_sf_unit( struct brw_context *brw )
273 {
274 struct brw_sf_unit_key key;
275 dri_bo *reloc_bufs[2];
276 int ret = 0;
277
278 sf_unit_populate_key(brw, &key);
279
280 reloc_bufs[0] = brw->sf.prog_bo;
281 reloc_bufs[1] = brw->sf.vp_bo;
282
283 dri_bo_unreference(brw->sf.state_bo);
284 brw->sf.state_bo = brw_search_cache(&brw->cache, BRW_SF_UNIT,
285 &key, sizeof(key),
286 reloc_bufs, 2,
287 NULL);
288 if (brw->sf.state_bo == NULL) {
289 brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs);
290 }
291
292 if (reloc_bufs[0])
293 ret |= dri_bufmgr_check_aperture_space(reloc_bufs[0]);
294
295 if (reloc_bufs[1])
296 ret |= dri_bufmgr_check_aperture_space(reloc_bufs[1]);
297
298 ret |= dri_bufmgr_check_aperture_space(brw->sf.state_bo);
299 return ret;
300 }
301
302 const struct brw_tracked_state brw_sf_unit = {
303 .dirty = {
304 .mesa = (_NEW_POLYGON |
305 _NEW_LINE |
306 _NEW_POINT |
307 _NEW_SCISSOR),
308 .brw = (BRW_NEW_URB_FENCE |
309 BRW_NEW_METAOPS),
310 .cache = (CACHE_NEW_SF_VP |
311 CACHE_NEW_SF_PROG)
312 },
313 .prepare = upload_sf_unit,
314 };