Merge remote branch 'main/master' into radeon-rewrite
[mesa.git] / src / mesa / drivers / dri / i965 / brw_wm_sampler_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 #include "brw_context.h"
34 #include "brw_state.h"
35 #include "brw_defines.h"
36
37 #include "main/macros.h"
38
39
40
41 /* Samplers aren't strictly wm state from the hardware's perspective,
42 * but that is the only situation in which we use them in this driver.
43 */
44
45
46
47 /* The brw (and related graphics cores) do not support GL_CLAMP. The
48 * Intel drivers for "other operating systems" implement GL_CLAMP as
49 * GL_CLAMP_TO_EDGE, so the same is done here.
50 */
51 static GLuint translate_wrap_mode( GLenum wrap )
52 {
53 switch( wrap ) {
54 case GL_REPEAT:
55 return BRW_TEXCOORDMODE_WRAP;
56 case GL_CLAMP:
57 return BRW_TEXCOORDMODE_CLAMP;
58 case GL_CLAMP_TO_EDGE:
59 return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */
60 case GL_CLAMP_TO_BORDER:
61 return BRW_TEXCOORDMODE_CLAMP_BORDER;
62 case GL_MIRRORED_REPEAT:
63 return BRW_TEXCOORDMODE_MIRROR;
64 default:
65 return BRW_TEXCOORDMODE_WRAP;
66 }
67 }
68
69
70 static GLuint U_FIXED(GLfloat value, GLuint frac_bits)
71 {
72 value *= (1<<frac_bits);
73 return value < 0 ? 0 : value;
74 }
75
76 static GLint S_FIXED(GLfloat value, GLuint frac_bits)
77 {
78 return value * (1<<frac_bits);
79 }
80
81
82 static dri_bo *upload_default_color( struct brw_context *brw,
83 const GLfloat *color )
84 {
85 struct brw_sampler_default_color sdc;
86
87 COPY_4V(sdc.color, color);
88
89 return brw_cache_data( &brw->cache, BRW_SAMPLER_DEFAULT_COLOR, &sdc,
90 NULL, 0 );
91 }
92
93
94 struct wm_sampler_key {
95 int sampler_count;
96
97 struct wm_sampler_entry {
98 GLenum tex_target;
99 GLenum wrap_r, wrap_s, wrap_t;
100 float maxlod, minlod;
101 float lod_bias;
102 float max_aniso;
103 GLenum minfilter, magfilter;
104 GLenum comparemode, comparefunc;
105 dri_bo *sdc_bo;
106 } sampler[BRW_MAX_TEX_UNIT];
107 };
108
109 /**
110 * Sets the sampler state for a single unit based off of the sampler key
111 * entry.
112 */
113 static void brw_update_sampler_state(struct wm_sampler_entry *key,
114 dri_bo *sdc_bo,
115 struct brw_sampler_state *sampler)
116 {
117 _mesa_memset(sampler, 0, sizeof(*sampler));
118
119 switch (key->minfilter) {
120 case GL_NEAREST:
121 sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
122 sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
123 break;
124 case GL_LINEAR:
125 sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
126 sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
127 break;
128 case GL_NEAREST_MIPMAP_NEAREST:
129 sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
130 sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
131 break;
132 case GL_LINEAR_MIPMAP_NEAREST:
133 sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
134 sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
135 break;
136 case GL_NEAREST_MIPMAP_LINEAR:
137 sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
138 sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
139 break;
140 case GL_LINEAR_MIPMAP_LINEAR:
141 sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
142 sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
143 break;
144 default:
145 break;
146 }
147
148 /* Set Anisotropy:
149 */
150 if (key->max_aniso > 1.0) {
151 sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC;
152 sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
153
154 if (key->max_aniso > 2.0) {
155 sampler->ss3.max_aniso = MAX2((key->max_aniso - 2) / 2,
156 BRW_ANISORATIO_16);
157 }
158 }
159 else {
160 switch (key->magfilter) {
161 case GL_NEAREST:
162 sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
163 break;
164 case GL_LINEAR:
165 sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
166 break;
167 default:
168 break;
169 }
170 }
171
172 if (key->tex_target == GL_TEXTURE_CUBE_MAP &&
173 (key->minfilter != GL_NEAREST || key->magfilter != GL_NEAREST)) {
174 /* If we're using anything but nearest sampling for a cube map, we
175 * need to set this wrap mode to avoid GPU lock-ups.
176 */
177 sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
178 sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
179 sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
180 }
181 else {
182 sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
183 sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
184 sampler->ss1.t_wrap_mode = translate_wrap_mode(key->wrap_t);
185 }
186
187 /* Set shadow function:
188 */
189 if (key->comparemode == GL_COMPARE_R_TO_TEXTURE_ARB) {
190 /* Shadowing is "enabled" by emitting a particular sampler
191 * message (sample_c). So need to recompile WM program when
192 * shadow comparison is enabled on each/any texture unit.
193 */
194 sampler->ss0.shadow_function =
195 intel_translate_shadow_compare_func(key->comparefunc);
196 }
197
198 /* Set LOD bias:
199 */
200 sampler->ss0.lod_bias = S_FIXED(CLAMP(key->lod_bias, -16, 15), 6);
201
202 sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
203 sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
204
205 /* Set BaseMipLevel, MaxLOD, MinLOD:
206 *
207 * XXX: I don't think that using firstLevel, lastLevel works,
208 * because we always setup the surface state as if firstLevel ==
209 * level zero. Probably have to subtract firstLevel from each of
210 * these:
211 */
212 sampler->ss0.base_level = U_FIXED(0, 1);
213
214 sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(key->maxlod, 0), 13), 6);
215 sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(key->minlod, 0), 13), 6);
216
217 sampler->ss2.default_color_pointer = sdc_bo->offset >> 5; /* reloc */
218 }
219
220 /** Sets up the cache key for sampler state for all texture units */
221 static void
222 brw_wm_sampler_populate_key(struct brw_context *brw,
223 struct wm_sampler_key *key)
224 {
225 GLcontext *ctx = &brw->intel.ctx;
226 int unit;
227
228 memset(key, 0, sizeof(*key));
229
230 for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) {
231 if (ctx->Texture.Unit[unit]._ReallyEnabled) {
232 struct wm_sampler_entry *entry = &key->sampler[unit];
233 struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
234 struct gl_texture_object *texObj = texUnit->_Current;
235 struct intel_texture_object *intelObj = intel_texture_object(texObj);
236 struct gl_texture_image *firstImage =
237 texObj->Image[0][intelObj->firstLevel];
238
239 entry->tex_target = texObj->Target;
240
241 entry->wrap_r = texObj->WrapR;
242 entry->wrap_s = texObj->WrapS;
243 entry->wrap_t = texObj->WrapT;
244
245 entry->maxlod = texObj->MaxLod;
246 entry->minlod = texObj->MinLod;
247 entry->lod_bias = texUnit->LodBias + texObj->LodBias;
248 entry->max_aniso = texObj->MaxAnisotropy;
249 entry->minfilter = texObj->MinFilter;
250 entry->magfilter = texObj->MagFilter;
251 entry->comparemode = texObj->CompareMode;
252 entry->comparefunc = texObj->CompareFunc;
253
254 dri_bo_unreference(brw->wm.sdc_bo[unit]);
255 if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
256 float bordercolor[4] = {
257 texObj->BorderColor[0],
258 texObj->BorderColor[0],
259 texObj->BorderColor[0],
260 texObj->BorderColor[0]
261 };
262 /* GL specs that border color for depth textures is taken from the
263 * R channel, while the hardware uses A. Spam R into all the
264 * channels for safety.
265 */
266 brw->wm.sdc_bo[unit] = upload_default_color(brw, bordercolor);
267 } else {
268 brw->wm.sdc_bo[unit] = upload_default_color(brw,
269 texObj->BorderColor);
270 }
271 key->sampler_count = unit + 1;
272 }
273 }
274 }
275
276 /* All samplers must be uploaded in a single contiguous array, which
277 * complicates various things. However, this is still too confusing -
278 * FIXME: simplify all the different new texture state flags.
279 */
280 static void upload_wm_samplers( struct brw_context *brw )
281 {
282 GLcontext *ctx = &brw->intel.ctx;
283 struct wm_sampler_key key;
284 int i;
285
286 brw_wm_sampler_populate_key(brw, &key);
287
288 if (brw->wm.sampler_count != key.sampler_count) {
289 brw->wm.sampler_count = key.sampler_count;
290 brw->state.dirty.cache |= CACHE_NEW_SAMPLER;
291 }
292
293 dri_bo_unreference(brw->wm.sampler_bo);
294 brw->wm.sampler_bo = NULL;
295 if (brw->wm.sampler_count == 0)
296 return;
297
298 brw->wm.sampler_bo = brw_search_cache(&brw->cache, BRW_SAMPLER,
299 &key, sizeof(key),
300 brw->wm.sdc_bo, key.sampler_count,
301 NULL);
302
303 /* If we didnt find it in the cache, compute the state and put it in the
304 * cache.
305 */
306 if (brw->wm.sampler_bo == NULL) {
307 struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT];
308
309 memset(sampler, 0, sizeof(sampler));
310 for (i = 0; i < key.sampler_count; i++) {
311 if (brw->wm.sdc_bo[i] == NULL)
312 continue;
313
314 brw_update_sampler_state(&key.sampler[i], brw->wm.sdc_bo[i],
315 &sampler[i]);
316 }
317
318 brw->wm.sampler_bo = brw_upload_cache(&brw->cache, BRW_SAMPLER,
319 &key, sizeof(key),
320 brw->wm.sdc_bo, key.sampler_count,
321 &sampler, sizeof(sampler),
322 NULL, NULL);
323
324 /* Emit SDC relocations */
325 for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
326 if (!ctx->Texture.Unit[i]._ReallyEnabled)
327 continue;
328
329 dri_bo_emit_reloc(brw->wm.sampler_bo,
330 I915_GEM_DOMAIN_SAMPLER, 0,
331 0,
332 i * sizeof(struct brw_sampler_state) +
333 offsetof(struct brw_sampler_state, ss2),
334 brw->wm.sdc_bo[i]);
335 }
336 }
337 }
338
339 const struct brw_tracked_state brw_wm_samplers = {
340 .dirty = {
341 .mesa = _NEW_TEXTURE,
342 .brw = 0,
343 .cache = 0
344 },
345 .prepare = upload_wm_samplers,
346 };
347
348