gm107/ir: allow indirect inputs to be loaded by frag shader
[mesa.git] / src / gallium / drivers / nouveau / nvc0 / nvc0_miptree.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 "pipe/p_state.h"
24 #include "pipe/p_defines.h"
25 #include "util/u_inlines.h"
26 #include "util/u_format.h"
27
28 #include "nvc0/nvc0_context.h"
29 #include "nvc0/nvc0_resource.h"
30
31 static uint32_t
32 nvc0_tex_choose_tile_dims(unsigned nx, unsigned ny, unsigned nz, bool is_3d)
33 {
34 return nv50_tex_choose_tile_dims_helper(nx, ny, nz, is_3d);
35 }
36
37 static uint32_t
38 nvc0_mt_choose_storage_type(struct nv50_miptree *mt, bool compressed)
39 {
40 const unsigned ms = util_logbase2(mt->base.base.nr_samples);
41
42 uint32_t tile_flags;
43
44 if (unlikely(mt->base.base.bind & PIPE_BIND_CURSOR))
45 return 0;
46 if (unlikely(mt->base.base.flags & NOUVEAU_RESOURCE_FLAG_LINEAR))
47 return 0;
48
49 switch (mt->base.base.format) {
50 case PIPE_FORMAT_Z16_UNORM:
51 if (compressed)
52 tile_flags = 0x02 + ms;
53 else
54 tile_flags = 0x01;
55 break;
56 case PIPE_FORMAT_X8Z24_UNORM:
57 case PIPE_FORMAT_S8X24_UINT:
58 case PIPE_FORMAT_S8_UINT_Z24_UNORM:
59 if (compressed)
60 tile_flags = 0x51 + ms;
61 else
62 tile_flags = 0x46;
63 break;
64 case PIPE_FORMAT_X24S8_UINT:
65 case PIPE_FORMAT_Z24X8_UNORM:
66 case PIPE_FORMAT_Z24_UNORM_S8_UINT:
67 if (compressed)
68 tile_flags = 0x17 + ms;
69 else
70 tile_flags = 0x11;
71 break;
72 case PIPE_FORMAT_Z32_FLOAT:
73 if (compressed)
74 tile_flags = 0x86 + ms;
75 else
76 tile_flags = 0x7b;
77 break;
78 case PIPE_FORMAT_X32_S8X24_UINT:
79 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
80 if (compressed)
81 tile_flags = 0xce + ms;
82 else
83 tile_flags = 0xc3;
84 break;
85 default:
86 switch (util_format_get_blocksizebits(mt->base.base.format)) {
87 case 128:
88 if (compressed)
89 tile_flags = 0xf4 + ms * 2;
90 else
91 tile_flags = 0xfe;
92 break;
93 case 64:
94 if (compressed) {
95 switch (ms) {
96 case 0: tile_flags = 0xe6; break;
97 case 1: tile_flags = 0xeb; break;
98 case 2: tile_flags = 0xed; break;
99 case 3: tile_flags = 0xf2; break;
100 default:
101 return 0;
102 }
103 } else {
104 tile_flags = 0xfe;
105 }
106 break;
107 case 32:
108 if (compressed && ms) {
109 switch (ms) {
110 /* This one makes things blurry:
111 case 0: tile_flags = 0xdb; break;
112 */
113 case 1: tile_flags = 0xdd; break;
114 case 2: tile_flags = 0xdf; break;
115 case 3: tile_flags = 0xe4; break;
116 default:
117 return 0;
118 }
119 } else {
120 tile_flags = 0xfe;
121 }
122 break;
123 case 16:
124 case 8:
125 tile_flags = 0xfe;
126 break;
127 default:
128 return 0;
129 }
130 break;
131 }
132
133 return tile_flags;
134 }
135
136 static inline bool
137 nvc0_miptree_init_ms_mode(struct nv50_miptree *mt)
138 {
139 switch (mt->base.base.nr_samples) {
140 case 8:
141 mt->ms_mode = NVC0_3D_MULTISAMPLE_MODE_MS8;
142 mt->ms_x = 2;
143 mt->ms_y = 1;
144 break;
145 case 4:
146 mt->ms_mode = NVC0_3D_MULTISAMPLE_MODE_MS4;
147 mt->ms_x = 1;
148 mt->ms_y = 1;
149 break;
150 case 2:
151 mt->ms_mode = NVC0_3D_MULTISAMPLE_MODE_MS2;
152 mt->ms_x = 1;
153 break;
154 case 1:
155 case 0:
156 mt->ms_mode = NVC0_3D_MULTISAMPLE_MODE_MS1;
157 break;
158 default:
159 NOUVEAU_ERR("invalid nr_samples: %u\n", mt->base.base.nr_samples);
160 return false;
161 }
162 return true;
163 }
164
165 static void
166 nvc0_miptree_init_layout_video(struct nv50_miptree *mt)
167 {
168 const struct pipe_resource *pt = &mt->base.base;
169 const unsigned blocksize = util_format_get_blocksize(pt->format);
170
171 assert(pt->last_level == 0);
172 assert(mt->ms_x == 0 && mt->ms_y == 0);
173 assert(!util_format_is_compressed(pt->format));
174
175 mt->layout_3d = pt->target == PIPE_TEXTURE_3D;
176
177 mt->level[0].tile_mode = 0x10;
178 mt->level[0].pitch = align(pt->width0 * blocksize, 64);
179 mt->total_size = align(pt->height0, 16) * mt->level[0].pitch * (mt->layout_3d ? pt->depth0 : 1);
180
181 if (pt->array_size > 1) {
182 mt->layer_stride = align(mt->total_size, NVC0_TILE_SIZE(0x10));
183 mt->total_size = mt->layer_stride * pt->array_size;
184 }
185 }
186
187 static void
188 nvc0_miptree_init_layout_tiled(struct nv50_miptree *mt)
189 {
190 struct pipe_resource *pt = &mt->base.base;
191 unsigned w, h, d, l;
192 const unsigned blocksize = util_format_get_blocksize(pt->format);
193
194 mt->layout_3d = pt->target == PIPE_TEXTURE_3D;
195
196 w = pt->width0 << mt->ms_x;
197 h = pt->height0 << mt->ms_y;
198
199 /* For 3D textures, a mipmap is spanned by all the layers, for array
200 * textures and cube maps, each layer contains its own mipmaps.
201 */
202 d = mt->layout_3d ? pt->depth0 : 1;
203
204 assert(!mt->ms_mode || !pt->last_level);
205
206 for (l = 0; l <= pt->last_level; ++l) {
207 struct nv50_miptree_level *lvl = &mt->level[l];
208 unsigned tsx, tsy, tsz;
209 unsigned nbx = util_format_get_nblocksx(pt->format, w);
210 unsigned nby = util_format_get_nblocksy(pt->format, h);
211
212 lvl->offset = mt->total_size;
213
214 lvl->tile_mode = nvc0_tex_choose_tile_dims(nbx, nby, d, mt->layout_3d);
215
216 tsx = NVC0_TILE_SIZE_X(lvl->tile_mode); /* x is tile row pitch in bytes */
217 tsy = NVC0_TILE_SIZE_Y(lvl->tile_mode);
218 tsz = NVC0_TILE_SIZE_Z(lvl->tile_mode);
219
220 lvl->pitch = align(nbx * blocksize, tsx);
221
222 mt->total_size += lvl->pitch * align(nby, tsy) * align(d, tsz);
223
224 w = u_minify(w, 1);
225 h = u_minify(h, 1);
226 d = u_minify(d, 1);
227 }
228
229 if (pt->array_size > 1) {
230 mt->layer_stride = align(mt->total_size,
231 NVC0_TILE_SIZE(mt->level[0].tile_mode));
232 mt->total_size = mt->layer_stride * pt->array_size;
233 }
234 }
235
236 const struct u_resource_vtbl nvc0_miptree_vtbl =
237 {
238 nv50_miptree_get_handle, /* get_handle */
239 nv50_miptree_destroy, /* resource_destroy */
240 nvc0_miptree_transfer_map, /* transfer_map */
241 u_default_transfer_flush_region, /* transfer_flush_region */
242 nvc0_miptree_transfer_unmap, /* transfer_unmap */
243 };
244
245 struct pipe_resource *
246 nvc0_miptree_create(struct pipe_screen *pscreen,
247 const struct pipe_resource *templ)
248 {
249 struct nouveau_device *dev = nouveau_screen(pscreen)->device;
250 struct nouveau_drm *drm = nouveau_screen(pscreen)->drm;
251 struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
252 struct pipe_resource *pt = &mt->base.base;
253 bool compressed = drm->version >= 0x01000101;
254 int ret;
255 union nouveau_bo_config bo_config;
256 uint32_t bo_flags;
257
258 if (!mt)
259 return NULL;
260
261 mt->base.vtbl = &nvc0_miptree_vtbl;
262 *pt = *templ;
263 pipe_reference_init(&pt->reference, 1);
264 pt->screen = pscreen;
265
266 if (pt->usage == PIPE_USAGE_STAGING) {
267 switch (pt->target) {
268 case PIPE_TEXTURE_2D:
269 case PIPE_TEXTURE_RECT:
270 if (pt->last_level == 0 &&
271 !util_format_is_depth_or_stencil(pt->format) &&
272 pt->nr_samples <= 1)
273 pt->flags |= NOUVEAU_RESOURCE_FLAG_LINEAR;
274 break;
275 default:
276 break;
277 }
278 }
279
280 if (pt->bind & PIPE_BIND_LINEAR)
281 pt->flags |= NOUVEAU_RESOURCE_FLAG_LINEAR;
282
283 bo_config.nvc0.memtype = nvc0_mt_choose_storage_type(mt, compressed);
284
285 if (!nvc0_miptree_init_ms_mode(mt)) {
286 FREE(mt);
287 return NULL;
288 }
289
290 if (unlikely(pt->flags & NVC0_RESOURCE_FLAG_VIDEO)) {
291 nvc0_miptree_init_layout_video(mt);
292 } else
293 if (likely(bo_config.nvc0.memtype)) {
294 nvc0_miptree_init_layout_tiled(mt);
295 } else
296 if (!nv50_miptree_init_layout_linear(mt, 128)) {
297 FREE(mt);
298 return NULL;
299 }
300 bo_config.nvc0.tile_mode = mt->level[0].tile_mode;
301
302 if (!bo_config.nvc0.memtype && (pt->usage == PIPE_USAGE_STAGING || pt->bind & PIPE_BIND_SHARED))
303 mt->base.domain = NOUVEAU_BO_GART;
304 else
305 mt->base.domain = NV_VRAM_DOMAIN(nouveau_screen(pscreen));
306
307 bo_flags = mt->base.domain | NOUVEAU_BO_NOSNOOP;
308
309 if (mt->base.base.bind & (PIPE_BIND_CURSOR | PIPE_BIND_DISPLAY_TARGET))
310 bo_flags |= NOUVEAU_BO_CONTIG;
311
312 ret = nouveau_bo_new(dev, bo_flags, 4096, mt->total_size, &bo_config,
313 &mt->base.bo);
314 if (ret) {
315 FREE(mt);
316 return NULL;
317 }
318 mt->base.address = mt->base.bo->offset;
319
320 NOUVEAU_DRV_STAT(nouveau_screen(pscreen), tex_obj_current_count, 1);
321 NOUVEAU_DRV_STAT(nouveau_screen(pscreen), tex_obj_current_bytes,
322 mt->total_size);
323
324 return pt;
325 }
326
327 /* Offset of zslice @z from start of level @l. */
328 inline unsigned
329 nvc0_mt_zslice_offset(const struct nv50_miptree *mt, unsigned l, unsigned z)
330 {
331 const struct pipe_resource *pt = &mt->base.base;
332
333 unsigned tds = NVC0_TILE_SHIFT_Z(mt->level[l].tile_mode);
334 unsigned ths = NVC0_TILE_SHIFT_Y(mt->level[l].tile_mode);
335
336 unsigned nby = util_format_get_nblocksy(pt->format,
337 u_minify(pt->height0, l));
338
339 /* to next 2D tile slice within a 3D tile */
340 unsigned stride_2d = NVC0_TILE_SIZE_2D(mt->level[l].tile_mode);
341
342 /* to slice in the next (in z direction) 3D tile */
343 unsigned stride_3d = (align(nby, (1 << ths)) * mt->level[l].pitch) << tds;
344
345 return (z & (1 << (tds - 1))) * stride_2d + (z >> tds) * stride_3d;
346 }
347
348 /* Surface functions.
349 */
350
351 struct pipe_surface *
352 nvc0_miptree_surface_new(struct pipe_context *pipe,
353 struct pipe_resource *pt,
354 const struct pipe_surface *templ)
355 {
356 struct nv50_surface *ns = nv50_surface_from_miptree(nv50_miptree(pt), templ);
357 if (!ns)
358 return NULL;
359 ns->base.context = pipe;
360 return &ns->base;
361 }