nv50/ir: Fold IMM into MAD
[mesa.git] / src / gallium / drivers / nouveau / nv50 / nv50_tex.c
1 /*
2 * Copyright 2008 Ben Skeggs
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 */
22
23 #include "nv50/nv50_context.h"
24 #include "nv50/nv50_resource.h"
25 #include "nv50/nv50_texture.xml.h"
26 #include "nv50/nv50_defs.xml.h"
27
28 #include "util/u_format.h"
29
30 #define NV50_TIC_0_SWIZZLE__MASK \
31 (NV50_TIC_0_MAPA__MASK | NV50_TIC_0_MAPB__MASK | \
32 NV50_TIC_0_MAPG__MASK | NV50_TIC_0_MAPR__MASK)
33
34 static INLINE uint32_t
35 nv50_tic_swizzle(uint32_t tc, unsigned swz, boolean tex_int)
36 {
37 switch (swz) {
38 case PIPE_SWIZZLE_RED:
39 return (tc & NV50_TIC_0_MAPR__MASK) >> NV50_TIC_0_MAPR__SHIFT;
40 case PIPE_SWIZZLE_GREEN:
41 return (tc & NV50_TIC_0_MAPG__MASK) >> NV50_TIC_0_MAPG__SHIFT;
42 case PIPE_SWIZZLE_BLUE:
43 return (tc & NV50_TIC_0_MAPB__MASK) >> NV50_TIC_0_MAPB__SHIFT;
44 case PIPE_SWIZZLE_ALPHA:
45 return (tc & NV50_TIC_0_MAPA__MASK) >> NV50_TIC_0_MAPA__SHIFT;
46 case PIPE_SWIZZLE_ONE:
47 return tex_int ? NV50_TIC_MAP_ONE_INT : NV50_TIC_MAP_ONE_FLOAT;
48 case PIPE_SWIZZLE_ZERO:
49 default:
50 return NV50_TIC_MAP_ZERO;
51 }
52 }
53
54 struct pipe_sampler_view *
55 nv50_create_sampler_view(struct pipe_context *pipe,
56 struct pipe_resource *res,
57 const struct pipe_sampler_view *templ)
58 {
59 uint32_t flags = 0;
60
61 if (templ->target == PIPE_TEXTURE_RECT || templ->target == PIPE_BUFFER)
62 flags |= NV50_TEXVIEW_SCALED_COORDS;
63
64 return nv50_create_texture_view(pipe, res, templ, flags, templ->target);
65 }
66
67 struct pipe_sampler_view *
68 nv50_create_texture_view(struct pipe_context *pipe,
69 struct pipe_resource *texture,
70 const struct pipe_sampler_view *templ,
71 uint32_t flags,
72 enum pipe_texture_target target)
73 {
74 const struct util_format_description *desc;
75 uint64_t addr;
76 uint32_t *tic;
77 uint32_t swz[4];
78 uint32_t depth;
79 struct nv50_tic_entry *view;
80 struct nv50_miptree *mt = nv50_miptree(texture);
81 boolean tex_int;
82
83 view = MALLOC_STRUCT(nv50_tic_entry);
84 if (!view)
85 return NULL;
86
87 view->pipe = *templ;
88 view->pipe.reference.count = 1;
89 view->pipe.texture = NULL;
90 view->pipe.context = pipe;
91
92 view->id = -1;
93
94 pipe_resource_reference(&view->pipe.texture, texture);
95
96 tic = &view->tic[0];
97
98 desc = util_format_description(view->pipe.format);
99
100 /* TIC[0] */
101
102 tic[0] = nv50_format_table[view->pipe.format].tic;
103
104 tex_int = util_format_is_pure_integer(view->pipe.format);
105
106 swz[0] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_r, tex_int);
107 swz[1] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_g, tex_int);
108 swz[2] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_b, tex_int);
109 swz[3] = nv50_tic_swizzle(tic[0], view->pipe.swizzle_a, tex_int);
110 tic[0] = (tic[0] & ~NV50_TIC_0_SWIZZLE__MASK) |
111 (swz[0] << NV50_TIC_0_MAPR__SHIFT) |
112 (swz[1] << NV50_TIC_0_MAPG__SHIFT) |
113 (swz[2] << NV50_TIC_0_MAPB__SHIFT) |
114 (swz[3] << NV50_TIC_0_MAPA__SHIFT);
115
116 addr = mt->base.address;
117
118 depth = MAX2(mt->base.base.array_size, mt->base.base.depth0);
119
120 if (mt->base.base.array_size > 1) {
121 /* there doesn't seem to be a base layer field in TIC */
122 addr += view->pipe.u.tex.first_layer * mt->layer_stride;
123 depth = view->pipe.u.tex.last_layer - view->pipe.u.tex.first_layer + 1;
124 }
125
126 tic[2] = 0x10001000 | NV50_TIC_2_NO_BORDER;
127
128 if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB)
129 tic[2] |= NV50_TIC_2_COLORSPACE_SRGB;
130
131 if (!(flags & NV50_TEXVIEW_SCALED_COORDS))
132 tic[2] |= NV50_TIC_2_NORMALIZED_COORDS;
133
134 if (unlikely(!nouveau_bo_memtype(nv04_resource(texture)->bo))) {
135 if (target == PIPE_BUFFER) {
136 addr += view->pipe.u.buf.first_element * desc->block.bits / 8;
137 tic[2] |= NV50_TIC_2_LINEAR | NV50_TIC_2_TARGET_BUFFER;
138 tic[3] = 0;
139 tic[4] = /* width */
140 view->pipe.u.buf.last_element - view->pipe.u.buf.first_element + 1;
141 tic[5] = 0;
142 } else {
143 tic[2] |= NV50_TIC_2_LINEAR | NV50_TIC_2_TARGET_RECT;
144 tic[3] = mt->level[0].pitch;
145 tic[4] = mt->base.base.width0;
146 tic[5] = (1 << 16) | (mt->base.base.height0);
147 }
148 tic[6] =
149 tic[7] = 0;
150 tic[1] = addr;
151 tic[2] |= addr >> 32;
152 return &view->pipe;
153 }
154
155 tic[1] = addr;
156 tic[2] |= (addr >> 32) & 0xff;
157
158 tic[2] |=
159 ((mt->level[0].tile_mode & 0x0f0) << (22 - 4)) |
160 ((mt->level[0].tile_mode & 0xf00) << (25 - 8));
161
162 switch (target) {
163 case PIPE_TEXTURE_1D:
164 tic[2] |= NV50_TIC_2_TARGET_1D;
165 break;
166 case PIPE_TEXTURE_2D:
167 tic[2] |= NV50_TIC_2_TARGET_2D;
168 break;
169 case PIPE_TEXTURE_RECT:
170 tic[2] |= NV50_TIC_2_TARGET_RECT;
171 break;
172 case PIPE_TEXTURE_3D:
173 tic[2] |= NV50_TIC_2_TARGET_3D;
174 break;
175 case PIPE_TEXTURE_CUBE:
176 depth /= 6;
177 tic[2] |= NV50_TIC_2_TARGET_CUBE;
178 break;
179 case PIPE_TEXTURE_1D_ARRAY:
180 tic[2] |= NV50_TIC_2_TARGET_1D_ARRAY;
181 break;
182 case PIPE_TEXTURE_2D_ARRAY:
183 tic[2] |= NV50_TIC_2_TARGET_2D_ARRAY;
184 break;
185 case PIPE_TEXTURE_CUBE_ARRAY:
186 depth /= 6;
187 tic[2] |= NV50_TIC_2_TARGET_CUBE_ARRAY;
188 break;
189 case PIPE_BUFFER:
190 assert(0); /* should be linear and handled above ! */
191 tic[2] |= NV50_TIC_2_TARGET_BUFFER | NV50_TIC_2_LINEAR;
192 break;
193 default:
194 NOUVEAU_ERR("invalid texture target: %d\n", mt->base.base.target);
195 return FALSE;
196 }
197
198 tic[3] = (flags & NV50_TEXVIEW_FILTER_MSAA8) ? 0x20000000 : 0x00300000;
199
200 tic[4] = (1 << 31) | (mt->base.base.width0 << mt->ms_x);
201
202 tic[5] = (mt->base.base.height0 << mt->ms_y) & 0xffff;
203 tic[5] |= depth << 16;
204 tic[5] |= mt->base.base.last_level << NV50_TIC_5_LAST_LEVEL__SHIFT;
205
206 tic[6] = (mt->ms_x > 1) ? 0x88000000 : 0x03000000; /* sampling points */
207
208 tic[7] = (view->pipe.u.tex.last_level << 4) | view->pipe.u.tex.first_level;
209
210 if (unlikely(!(tic[2] & NV50_TIC_2_NORMALIZED_COORDS)))
211 if (mt->base.base.last_level)
212 tic[5] &= ~NV50_TIC_5_LAST_LEVEL__MASK;
213
214 return &view->pipe;
215 }
216
217 static boolean
218 nv50_validate_tic(struct nv50_context *nv50, int s)
219 {
220 struct nouveau_pushbuf *push = nv50->base.pushbuf;
221 struct nouveau_bo *txc = nv50->screen->txc;
222 unsigned i;
223 boolean need_flush = FALSE;
224
225 assert(nv50->num_textures[s] <= PIPE_MAX_SAMPLERS);
226 for (i = 0; i < nv50->num_textures[s]; ++i) {
227 struct nv50_tic_entry *tic = nv50_tic_entry(nv50->textures[s][i]);
228 struct nv04_resource *res;
229
230 if (!tic) {
231 BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1);
232 PUSH_DATA (push, (i << 1) | 0);
233 continue;
234 }
235 res = &nv50_miptree(tic->pipe.texture)->base;
236
237 if (tic->id < 0) {
238 tic->id = nv50_screen_tic_alloc(nv50->screen, tic);
239
240 BEGIN_NV04(push, NV50_2D(DST_FORMAT), 2);
241 PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM);
242 PUSH_DATA (push, 1);
243 BEGIN_NV04(push, NV50_2D(DST_PITCH), 5);
244 PUSH_DATA (push, 262144);
245 PUSH_DATA (push, 65536);
246 PUSH_DATA (push, 1);
247 PUSH_DATAh(push, txc->offset);
248 PUSH_DATA (push, txc->offset);
249 BEGIN_NV04(push, NV50_2D(SIFC_BITMAP_ENABLE), 2);
250 PUSH_DATA (push, 0);
251 PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM);
252 BEGIN_NV04(push, NV50_2D(SIFC_WIDTH), 10);
253 PUSH_DATA (push, 32);
254 PUSH_DATA (push, 1);
255 PUSH_DATA (push, 0);
256 PUSH_DATA (push, 1);
257 PUSH_DATA (push, 0);
258 PUSH_DATA (push, 1);
259 PUSH_DATA (push, 0);
260 PUSH_DATA (push, tic->id * 32);
261 PUSH_DATA (push, 0);
262 PUSH_DATA (push, 0);
263 BEGIN_NI04(push, NV50_2D(SIFC_DATA), 8);
264 PUSH_DATAp(push, &tic->tic[0], 8);
265
266 need_flush = TRUE;
267 } else
268 if (res->status & NOUVEAU_BUFFER_STATUS_GPU_WRITING) {
269 BEGIN_NV04(push, NV50_3D(TEX_CACHE_CTL), 1);
270 PUSH_DATA (push, 0x20);
271 }
272
273 nv50->screen->tic.lock[tic->id / 32] |= 1 << (tic->id % 32);
274
275 res->status &= ~NOUVEAU_BUFFER_STATUS_GPU_WRITING;
276 res->status |= NOUVEAU_BUFFER_STATUS_GPU_READING;
277
278 BCTX_REFN(nv50->bufctx_3d, TEXTURES, res, RD);
279
280 BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1);
281 PUSH_DATA (push, (tic->id << 9) | (i << 1) | 1);
282 }
283 for (; i < nv50->state.num_textures[s]; ++i) {
284 BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1);
285 PUSH_DATA (push, (i << 1) | 0);
286 }
287 if (nv50->num_textures[s]) {
288 BEGIN_NV04(push, NV50_3D(CB_ADDR), 1);
289 PUSH_DATA (push, ((NV50_CB_AUX_TEX_MS_OFFSET + 16 * s * 2 * 4) << (8 - 2)) | NV50_CB_AUX);
290 BEGIN_NI04(push, NV50_3D(CB_DATA(0)), nv50->num_textures[s] * 2);
291 for (i = 0; i < nv50->num_textures[s]; i++) {
292 struct nv50_tic_entry *tic = nv50_tic_entry(nv50->textures[s][i]);
293 struct nv50_miptree *res;
294
295 if (!tic) {
296 PUSH_DATA (push, 0);
297 PUSH_DATA (push, 0);
298 continue;
299 }
300 res = nv50_miptree(tic->pipe.texture);
301 PUSH_DATA (push, res->ms_x);
302 PUSH_DATA (push, res->ms_y);
303 }
304 }
305 nv50->state.num_textures[s] = nv50->num_textures[s];
306
307 return need_flush;
308 }
309
310 void nv50_validate_textures(struct nv50_context *nv50)
311 {
312 boolean need_flush;
313
314 need_flush = nv50_validate_tic(nv50, 0);
315 need_flush |= nv50_validate_tic(nv50, 1);
316 need_flush |= nv50_validate_tic(nv50, 2);
317
318 if (need_flush) {
319 BEGIN_NV04(nv50->base.pushbuf, NV50_3D(TIC_FLUSH), 1);
320 PUSH_DATA (nv50->base.pushbuf, 0);
321 }
322 }
323
324 static boolean
325 nv50_validate_tsc(struct nv50_context *nv50, int s)
326 {
327 struct nouveau_pushbuf *push = nv50->base.pushbuf;
328 unsigned i;
329 boolean need_flush = FALSE;
330
331 assert(nv50->num_samplers[s] <= PIPE_MAX_SAMPLERS);
332 for (i = 0; i < nv50->num_samplers[s]; ++i) {
333 struct nv50_tsc_entry *tsc = nv50_tsc_entry(nv50->samplers[s][i]);
334
335 if (!tsc) {
336 BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1);
337 PUSH_DATA (push, (i << 4) | 0);
338 continue;
339 }
340 if (tsc->id < 0) {
341 tsc->id = nv50_screen_tsc_alloc(nv50->screen, tsc);
342
343 nv50_sifc_linear_u8(&nv50->base, nv50->screen->txc,
344 65536 + tsc->id * 32,
345 NOUVEAU_BO_VRAM, 32, tsc->tsc);
346 need_flush = TRUE;
347 }
348 nv50->screen->tsc.lock[tsc->id / 32] |= 1 << (tsc->id % 32);
349
350 BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1);
351 PUSH_DATA (push, (tsc->id << 12) | (i << 4) | 1);
352 }
353 for (; i < nv50->state.num_samplers[s]; ++i) {
354 BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1);
355 PUSH_DATA (push, (i << 4) | 0);
356 }
357 nv50->state.num_samplers[s] = nv50->num_samplers[s];
358
359 return need_flush;
360 }
361
362 void nv50_validate_samplers(struct nv50_context *nv50)
363 {
364 boolean need_flush;
365
366 need_flush = nv50_validate_tsc(nv50, 0);
367 need_flush |= nv50_validate_tsc(nv50, 1);
368 need_flush |= nv50_validate_tsc(nv50, 2);
369
370 if (need_flush) {
371 BEGIN_NV04(nv50->base.pushbuf, NV50_3D(TSC_FLUSH), 1);
372 PUSH_DATA (nv50->base.pushbuf, 0);
373 }
374 }
375
376 /* There can be up to 4 different MS levels (1, 2, 4, 8). To simplify the
377 * shader logic, allow each one to take up 8 offsets.
378 */
379 #define COMBINE(x, y) x, y
380 #define DUMMY 0, 0
381 static const uint32_t msaa_sample_xy_offsets[] = {
382 /* MS1 */
383 COMBINE(0, 0),
384 DUMMY,
385 DUMMY,
386 DUMMY,
387 DUMMY,
388 DUMMY,
389 DUMMY,
390 DUMMY,
391
392 /* MS2 */
393 COMBINE(0, 0),
394 COMBINE(1, 0),
395 DUMMY,
396 DUMMY,
397 DUMMY,
398 DUMMY,
399 DUMMY,
400 DUMMY,
401
402 /* MS4 */
403 COMBINE(0, 0),
404 COMBINE(1, 0),
405 COMBINE(0, 1),
406 COMBINE(1, 1),
407 DUMMY,
408 DUMMY,
409 DUMMY,
410 DUMMY,
411
412 /* MS8 */
413 COMBINE(0, 0),
414 COMBINE(1, 0),
415 COMBINE(0, 1),
416 COMBINE(1, 1),
417 COMBINE(2, 0),
418 COMBINE(3, 0),
419 COMBINE(2, 1),
420 COMBINE(3, 1),
421 };
422
423 void nv50_upload_ms_info(struct nouveau_pushbuf *push)
424 {
425 BEGIN_NV04(push, NV50_3D(CB_ADDR), 1);
426 PUSH_DATA (push, (NV50_CB_AUX_MS_OFFSET << (8 - 2)) | NV50_CB_AUX);
427 BEGIN_NI04(push, NV50_3D(CB_DATA(0)), Elements(msaa_sample_xy_offsets));
428 PUSH_DATAp(push, msaa_sample_xy_offsets, Elements(msaa_sample_xy_offsets));
429 }