etnaviv: rework TS enable to be a derived state
[mesa.git] / src / gallium / drivers / etnaviv / etnaviv_state.c
1 /*
2 * Copyright (c) 2012-2015 Etnaviv Project
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, sub license,
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 (including the
12 * next paragraph) shall be included in all copies or substantial portions
13 * of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
21 * DEALINGS IN THE SOFTWARE.
22 *
23 * Authors:
24 * Wladimir J. van der Laan <laanwj@gmail.com>
25 * Christian Gmeiner <christian.gmeiner@gmail.com>
26 */
27
28 #include "etnaviv_state.h"
29
30 #include "hw/common.xml.h"
31
32 #include "etnaviv_blend.h"
33 #include "etnaviv_clear_blit.h"
34 #include "etnaviv_context.h"
35 #include "etnaviv_format.h"
36 #include "etnaviv_shader.h"
37 #include "etnaviv_surface.h"
38 #include "etnaviv_translate.h"
39 #include "etnaviv_util.h"
40 #include "util/u_helpers.h"
41 #include "util/u_inlines.h"
42 #include "util/u_math.h"
43 #include "util/u_memory.h"
44
45 static void
46 etna_set_stencil_ref(struct pipe_context *pctx, const struct pipe_stencil_ref *sr)
47 {
48 struct etna_context *ctx = etna_context(pctx);
49 struct compiled_stencil_ref *cs = &ctx->stencil_ref;
50
51 ctx->stencil_ref_s = *sr;
52
53 cs->PE_STENCIL_CONFIG = VIVS_PE_STENCIL_CONFIG_REF_FRONT(sr->ref_value[0]);
54 /* rest of bits weaved in from depth_stencil_alpha */
55 cs->PE_STENCIL_CONFIG_EXT =
56 VIVS_PE_STENCIL_CONFIG_EXT_REF_BACK(sr->ref_value[0]);
57 ctx->dirty |= ETNA_DIRTY_STENCIL_REF;
58 }
59
60 static void
61 etna_set_clip_state(struct pipe_context *pctx, const struct pipe_clip_state *pcs)
62 {
63 /* NOOP */
64 }
65
66 static void
67 etna_set_sample_mask(struct pipe_context *pctx, unsigned sample_mask)
68 {
69 struct etna_context *ctx = etna_context(pctx);
70
71 ctx->sample_mask = sample_mask;
72 ctx->dirty |= ETNA_DIRTY_SAMPLE_MASK;
73 }
74
75 static void
76 etna_set_constant_buffer(struct pipe_context *pctx,
77 enum pipe_shader_type shader, uint index,
78 const struct pipe_constant_buffer *cb)
79 {
80 struct etna_context *ctx = etna_context(pctx);
81
82 if (unlikely(index > 0)) {
83 DBG("Unhandled buffer index %i", index);
84 return;
85 }
86
87
88 util_copy_constant_buffer(&ctx->constant_buffer[shader], cb);
89
90 /* Note that the state tracker can unbind constant buffers by
91 * passing NULL here. */
92 if (unlikely(!cb))
93 return;
94
95 /* there is no support for ARB_uniform_buffer_object */
96 assert(cb->buffer == NULL && cb->user_buffer != NULL);
97
98 ctx->dirty |= ETNA_DIRTY_CONSTBUF;
99 }
100
101 static void
102 etna_update_render_resource(struct pipe_context *pctx, struct pipe_resource *pres)
103 {
104 struct etna_resource *res = etna_resource(pres);
105
106 if (res->texture && etna_resource_older(res, etna_resource(res->texture))) {
107 /* The render buffer is older than the texture buffer. Copy it over. */
108 etna_copy_resource(pctx, pres, res->texture, 0, pres->last_level);
109 res->seqno = etna_resource(res->texture)->seqno;
110 }
111 }
112
113 static void
114 etna_set_framebuffer_state(struct pipe_context *pctx,
115 const struct pipe_framebuffer_state *sv)
116 {
117 struct etna_context *ctx = etna_context(pctx);
118 struct compiled_framebuffer_state *cs = &ctx->framebuffer;
119 int nr_samples_color = -1;
120 int nr_samples_depth = -1;
121
122 /* Set up TS as well. Warning: this state is used by both the RS and PE */
123 uint32_t ts_mem_config = 0;
124
125 if (sv->nr_cbufs > 0) { /* at least one color buffer? */
126 struct etna_surface *cbuf = etna_surface(sv->cbufs[0]);
127 struct etna_resource *res = etna_resource(cbuf->base.texture);
128 bool color_supertiled = (res->layout & ETNA_LAYOUT_BIT_SUPER) != 0;
129
130 assert(res->layout & ETNA_LAYOUT_BIT_TILE); /* Cannot render to linear surfaces */
131 etna_update_render_resource(pctx, cbuf->base.texture);
132
133 pipe_surface_reference(&cs->cbuf, &cbuf->base);
134 cs->PE_COLOR_FORMAT =
135 VIVS_PE_COLOR_FORMAT_FORMAT(translate_rs_format(cbuf->base.format)) |
136 VIVS_PE_COLOR_FORMAT_COMPONENTS__MASK |
137 VIVS_PE_COLOR_FORMAT_OVERWRITE |
138 COND(color_supertiled, VIVS_PE_COLOR_FORMAT_SUPER_TILED);
139 /* VIVS_PE_COLOR_FORMAT_COMPONENTS() and
140 * VIVS_PE_COLOR_FORMAT_OVERWRITE comes from blend_state
141 * but only if we set the bits above. */
142 /* merged with depth_stencil_alpha */
143 if ((cbuf->surf.offset & 63) ||
144 (((cbuf->surf.stride * 4) & 63) && cbuf->surf.height > 4)) {
145 /* XXX Must make temporary surface here.
146 * Need the same mechanism on gc2000 when we want to do mipmap
147 * generation by
148 * rendering to levels > 1 due to multitiled / tiled conversion. */
149 BUG("Alignment error, trying to render to offset %08x with tile "
150 "stride %i",
151 cbuf->surf.offset, cbuf->surf.stride * 4);
152 }
153
154 if (ctx->specs.pixel_pipes == 1) {
155 cs->PE_COLOR_ADDR = cbuf->reloc[0];
156 cs->PE_COLOR_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
157 } else {
158 /* Rendered textures must always be multi-tiled, or single-buffer mode must be supported */
159 assert((res->layout & ETNA_LAYOUT_BIT_MULTI) || ctx->specs.single_buffer);
160 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
161 cs->PE_PIPE_COLOR_ADDR[i] = cbuf->reloc[i];
162 cs->PE_PIPE_COLOR_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
163 }
164 }
165 cs->PE_COLOR_STRIDE = cbuf->surf.stride;
166
167 if (cbuf->surf.ts_size) {
168 cs->TS_COLOR_CLEAR_VALUE = cbuf->level->clear_value;
169
170 cs->TS_COLOR_STATUS_BASE = cbuf->ts_reloc;
171 cs->TS_COLOR_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
172
173 cs->TS_COLOR_SURFACE_BASE = cbuf->reloc[0];
174 cs->TS_COLOR_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
175 }
176
177 /* MSAA */
178 if (cbuf->base.texture->nr_samples > 1)
179 ts_mem_config |=
180 VIVS_TS_MEM_CONFIG_MSAA | translate_msaa_format(cbuf->base.format);
181
182 nr_samples_color = cbuf->base.texture->nr_samples;
183 } else {
184 pipe_surface_reference(&cs->cbuf, NULL);
185 /* Clearing VIVS_PE_COLOR_FORMAT_COMPONENTS__MASK and
186 * VIVS_PE_COLOR_FORMAT_OVERWRITE prevents us from overwriting the
187 * color target */
188 cs->PE_COLOR_FORMAT = 0;
189 cs->PE_COLOR_STRIDE = 0;
190 cs->TS_COLOR_STATUS_BASE.bo = NULL;
191 cs->TS_COLOR_SURFACE_BASE.bo = NULL;
192
193 for (int i = 0; i < ETNA_MAX_PIXELPIPES; i++)
194 cs->PE_PIPE_COLOR_ADDR[i].bo = NULL;
195 }
196
197 if (sv->zsbuf != NULL) {
198 struct etna_surface *zsbuf = etna_surface(sv->zsbuf);
199 struct etna_resource *res = etna_resource(zsbuf->base.texture);
200
201 etna_update_render_resource(pctx, zsbuf->base.texture);
202
203 pipe_surface_reference(&cs->zsbuf, &zsbuf->base);
204 assert(res->layout &ETNA_LAYOUT_BIT_TILE); /* Cannot render to linear surfaces */
205
206 uint32_t depth_format = translate_depth_format(zsbuf->base.format);
207 unsigned depth_bits =
208 depth_format == VIVS_PE_DEPTH_CONFIG_DEPTH_FORMAT_D16 ? 16 : 24;
209 bool depth_supertiled = (res->layout & ETNA_LAYOUT_BIT_SUPER) != 0;
210
211 cs->PE_DEPTH_CONFIG =
212 depth_format |
213 COND(depth_supertiled, VIVS_PE_DEPTH_CONFIG_SUPER_TILED) |
214 VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_Z;
215 /* VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH */
216 /* merged with depth_stencil_alpha */
217
218 if (ctx->specs.pixel_pipes == 1) {
219 cs->PE_DEPTH_ADDR = zsbuf->reloc[0];
220 cs->PE_DEPTH_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
221 } else {
222 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
223 cs->PE_PIPE_DEPTH_ADDR[i] = zsbuf->reloc[i];
224 cs->PE_PIPE_DEPTH_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
225 }
226 }
227
228 cs->PE_DEPTH_STRIDE = zsbuf->surf.stride;
229 cs->PE_HDEPTH_CONTROL = VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED;
230 cs->PE_DEPTH_NORMALIZE = fui(exp2f(depth_bits) - 1.0f);
231
232 if (zsbuf->surf.ts_size) {
233 cs->TS_DEPTH_CLEAR_VALUE = zsbuf->level->clear_value;
234
235 cs->TS_DEPTH_STATUS_BASE = zsbuf->ts_reloc;
236 cs->TS_DEPTH_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
237
238 cs->TS_DEPTH_SURFACE_BASE = zsbuf->reloc[0];
239 cs->TS_DEPTH_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
240 }
241
242 ts_mem_config |= COND(depth_bits == 16, VIVS_TS_MEM_CONFIG_DEPTH_16BPP);
243
244 /* MSAA */
245 if (zsbuf->base.texture->nr_samples > 1)
246 /* XXX VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION;
247 * Disable without MSAA for now, as it causes corruption in glquake. */
248 ts_mem_config |= VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION;
249
250 nr_samples_depth = zsbuf->base.texture->nr_samples;
251 } else {
252 pipe_surface_reference(&cs->zsbuf, NULL);
253 cs->PE_DEPTH_CONFIG = VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_NONE;
254 cs->PE_DEPTH_ADDR.bo = NULL;
255 cs->PE_DEPTH_STRIDE = 0;
256 cs->TS_DEPTH_STATUS_BASE.bo = NULL;
257 cs->TS_DEPTH_SURFACE_BASE.bo = NULL;
258
259 for (int i = 0; i < ETNA_MAX_PIXELPIPES; i++)
260 cs->PE_PIPE_DEPTH_ADDR[i].bo = NULL;
261 }
262
263 /* MSAA setup */
264 if (nr_samples_depth != -1 && nr_samples_color != -1 &&
265 nr_samples_depth != nr_samples_color) {
266 BUG("Number of samples in color and depth texture must match (%i and %i respectively)",
267 nr_samples_color, nr_samples_depth);
268 }
269
270 switch (MAX2(nr_samples_depth, nr_samples_color)) {
271 case 0:
272 case 1: /* Are 0 and 1 samples allowed? */
273 cs->GL_MULTI_SAMPLE_CONFIG =
274 VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_NONE;
275 cs->msaa_mode = false;
276 break;
277 case 2:
278 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_2X;
279 cs->msaa_mode = true; /* Add input to PS */
280 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
281 cs->RA_MULTISAMPLE_UNK00E10[0] = 0x0000aa22;
282 cs->RA_CENTROID_TABLE[0] = 0x66aa2288;
283 cs->RA_CENTROID_TABLE[1] = 0x88558800;
284 cs->RA_CENTROID_TABLE[2] = 0x88881100;
285 cs->RA_CENTROID_TABLE[3] = 0x33888800;
286 break;
287 case 4:
288 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_4X;
289 cs->msaa_mode = true; /* Add input to PS */
290 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
291 cs->RA_MULTISAMPLE_UNK00E10[0] = 0xeaa26e26;
292 cs->RA_MULTISAMPLE_UNK00E10[1] = 0xe6ae622a;
293 cs->RA_MULTISAMPLE_UNK00E10[2] = 0xaaa22a22;
294 cs->RA_CENTROID_TABLE[0] = 0x4a6e2688;
295 cs->RA_CENTROID_TABLE[1] = 0x888888a2;
296 cs->RA_CENTROID_TABLE[2] = 0x888888ea;
297 cs->RA_CENTROID_TABLE[3] = 0x888888c6;
298 cs->RA_CENTROID_TABLE[4] = 0x46622a88;
299 cs->RA_CENTROID_TABLE[5] = 0x888888ae;
300 cs->RA_CENTROID_TABLE[6] = 0x888888e6;
301 cs->RA_CENTROID_TABLE[7] = 0x888888ca;
302 cs->RA_CENTROID_TABLE[8] = 0x262a2288;
303 cs->RA_CENTROID_TABLE[9] = 0x886688a2;
304 cs->RA_CENTROID_TABLE[10] = 0x888866aa;
305 cs->RA_CENTROID_TABLE[11] = 0x668888a6;
306 break;
307 }
308
309 /* Scissor setup */
310 cs->SE_SCISSOR_LEFT = 0; /* affected by rasterizer and scissor state as well */
311 cs->SE_SCISSOR_TOP = 0;
312 cs->SE_SCISSOR_RIGHT = (sv->width << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
313 cs->SE_SCISSOR_BOTTOM = (sv->height << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
314 cs->SE_CLIP_RIGHT = (sv->width << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
315 cs->SE_CLIP_BOTTOM = (sv->height << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
316
317 cs->TS_MEM_CONFIG = ts_mem_config;
318
319 /* Single buffer setup. There is only one switch for this, not a separate
320 * one per color buffer / depth buffer. To keep the logic simple always use
321 * single buffer when this feature is available.
322 */
323 cs->PE_LOGIC_OP = VIVS_PE_LOGIC_OP_SINGLE_BUFFER(ctx->specs.single_buffer ? 2 : 0);
324
325 ctx->framebuffer_s = *sv; /* keep copy of original structure */
326 ctx->dirty |= ETNA_DIRTY_FRAMEBUFFER | ETNA_DIRTY_DERIVE_TS;
327 }
328
329 static void
330 etna_set_polygon_stipple(struct pipe_context *pctx,
331 const struct pipe_poly_stipple *stipple)
332 {
333 /* NOP */
334 }
335
336 static void
337 etna_set_scissor_states(struct pipe_context *pctx, unsigned start_slot,
338 unsigned num_scissors, const struct pipe_scissor_state *ss)
339 {
340 struct etna_context *ctx = etna_context(pctx);
341 struct compiled_scissor_state *cs = &ctx->scissor;
342 assert(ss->minx <= ss->maxx);
343 assert(ss->miny <= ss->maxy);
344
345 /* note that this state is only used when rasterizer_state->scissor is on */
346 ctx->scissor_s = *ss;
347 cs->SE_SCISSOR_LEFT = (ss->minx << 16);
348 cs->SE_SCISSOR_TOP = (ss->miny << 16);
349 cs->SE_SCISSOR_RIGHT = (ss->maxx << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
350 cs->SE_SCISSOR_BOTTOM = (ss->maxy << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
351 cs->SE_CLIP_RIGHT = (ss->maxx << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
352 cs->SE_CLIP_BOTTOM = (ss->maxy << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
353
354 ctx->dirty |= ETNA_DIRTY_SCISSOR;
355 }
356
357 static void
358 etna_set_viewport_states(struct pipe_context *pctx, unsigned start_slot,
359 unsigned num_scissors, const struct pipe_viewport_state *vs)
360 {
361 struct etna_context *ctx = etna_context(pctx);
362 struct compiled_viewport_state *cs = &ctx->viewport;
363
364 ctx->viewport_s = *vs;
365 /**
366 * For Vivante GPU, viewport z transformation is 0..1 to 0..1 instead of
367 * -1..1 to 0..1.
368 * scaling and translation to 0..1 already happened, so remove that
369 *
370 * z' = (z * 2 - 1) * scale + translate
371 * = z * (2 * scale) + (translate - scale)
372 *
373 * scale' = 2 * scale
374 * translate' = translate - scale
375 */
376
377 /* must be fixp as v4 state deltas assume it is */
378 cs->PA_VIEWPORT_SCALE_X = etna_f32_to_fixp16(vs->scale[0]);
379 cs->PA_VIEWPORT_SCALE_Y = etna_f32_to_fixp16(vs->scale[1]);
380 cs->PA_VIEWPORT_SCALE_Z = fui(vs->scale[2] * 2.0f);
381 cs->PA_VIEWPORT_OFFSET_X = etna_f32_to_fixp16(vs->translate[0]);
382 cs->PA_VIEWPORT_OFFSET_Y = etna_f32_to_fixp16(vs->translate[1]);
383 cs->PA_VIEWPORT_OFFSET_Z = fui(vs->translate[2] - vs->scale[2]);
384
385 /* Compute scissor rectangle (fixp) from viewport.
386 * Make sure left is always < right and top always < bottom.
387 */
388 cs->SE_SCISSOR_LEFT = etna_f32_to_fixp16(MAX2(vs->translate[0] - fabsf(vs->scale[0]), 0.0f));
389 cs->SE_SCISSOR_TOP = etna_f32_to_fixp16(MAX2(vs->translate[1] - fabsf(vs->scale[1]), 0.0f));
390 uint32_t right_fixp = etna_f32_to_fixp16(MAX2(vs->translate[0] + fabsf(vs->scale[0]), 0.0f));
391 uint32_t bottom_fixp = etna_f32_to_fixp16(MAX2(vs->translate[1] + fabsf(vs->scale[1]), 0.0f));
392 cs->SE_SCISSOR_RIGHT = right_fixp + ETNA_SE_SCISSOR_MARGIN_RIGHT;
393 cs->SE_SCISSOR_BOTTOM = bottom_fixp + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
394 cs->SE_CLIP_RIGHT = right_fixp + ETNA_SE_CLIP_MARGIN_RIGHT;
395 cs->SE_CLIP_BOTTOM = bottom_fixp + ETNA_SE_CLIP_MARGIN_BOTTOM;
396
397 cs->PE_DEPTH_NEAR = fui(0.0); /* not affected if depth mode is Z (as in GL) */
398 cs->PE_DEPTH_FAR = fui(1.0);
399 ctx->dirty |= ETNA_DIRTY_VIEWPORT;
400 }
401
402 static void
403 etna_set_vertex_buffers(struct pipe_context *pctx, unsigned start_slot,
404 unsigned num_buffers, const struct pipe_vertex_buffer *vb)
405 {
406 struct etna_context *ctx = etna_context(pctx);
407 struct etna_vertexbuf_state *so = &ctx->vertex_buffer;
408
409 util_set_vertex_buffers_mask(so->vb, &so->enabled_mask, vb, start_slot, num_buffers);
410 so->count = util_last_bit(so->enabled_mask);
411
412 for (unsigned idx = start_slot; idx < start_slot + num_buffers; ++idx) {
413 struct compiled_set_vertex_buffer *cs = &so->cvb[idx];
414 struct pipe_vertex_buffer *vbi = &so->vb[idx];
415
416 assert(!vbi->is_user_buffer); /* XXX support user_buffer using
417 etna_usermem_map */
418
419 if (vbi->buffer.resource) { /* GPU buffer */
420 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = etna_resource(vbi->buffer.resource)->bo;
421 cs->FE_VERTEX_STREAM_BASE_ADDR.offset = vbi->buffer_offset;
422 cs->FE_VERTEX_STREAM_BASE_ADDR.flags = ETNA_RELOC_READ;
423 cs->FE_VERTEX_STREAM_CONTROL =
424 FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE(vbi->stride);
425 } else {
426 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = NULL;
427 cs->FE_VERTEX_STREAM_CONTROL = 0;
428 }
429 }
430
431 ctx->dirty |= ETNA_DIRTY_VERTEX_BUFFERS;
432 }
433
434 static void
435 etna_blend_state_bind(struct pipe_context *pctx, void *bs)
436 {
437 struct etna_context *ctx = etna_context(pctx);
438
439 ctx->blend = bs;
440 ctx->dirty |= ETNA_DIRTY_BLEND;
441 }
442
443 static void
444 etna_blend_state_delete(struct pipe_context *pctx, void *bs)
445 {
446 FREE(bs);
447 }
448
449 static void
450 etna_rasterizer_state_bind(struct pipe_context *pctx, void *rs)
451 {
452 struct etna_context *ctx = etna_context(pctx);
453
454 ctx->rasterizer = rs;
455 ctx->dirty |= ETNA_DIRTY_RASTERIZER;
456 }
457
458 static void
459 etna_rasterizer_state_delete(struct pipe_context *pctx, void *rs)
460 {
461 FREE(rs);
462 }
463
464 static void
465 etna_zsa_state_bind(struct pipe_context *pctx, void *zs)
466 {
467 struct etna_context *ctx = etna_context(pctx);
468
469 ctx->zsa = zs;
470 ctx->dirty |= ETNA_DIRTY_ZSA;
471 }
472
473 static void
474 etna_zsa_state_delete(struct pipe_context *pctx, void *zs)
475 {
476 FREE(zs);
477 }
478
479 /** Create vertex element states, which define a layout for fetching
480 * vertices for rendering.
481 */
482 static void *
483 etna_vertex_elements_state_create(struct pipe_context *pctx,
484 unsigned num_elements, const struct pipe_vertex_element *elements)
485 {
486 struct etna_context *ctx = etna_context(pctx);
487 struct compiled_vertex_elements_state *cs = CALLOC_STRUCT(compiled_vertex_elements_state);
488
489 if (!cs)
490 return NULL;
491
492 if (num_elements > ctx->specs.vertex_max_elements) {
493 BUG("number of elements (%u) exceeds chip maximum (%u)", num_elements,
494 ctx->specs.vertex_max_elements);
495 return NULL;
496 }
497
498 /* XXX could minimize number of consecutive stretches here by sorting, and
499 * permuting the inputs in shader or does Mesa do this already? */
500
501 /* Check that vertex element binding is compatible with hardware; thus
502 * elements[idx].vertex_buffer_index are < stream_count. If not, the binding
503 * uses more streams than is supported, and u_vbuf should have done some
504 * reorganization for compatibility. */
505
506 /* TODO: does mesa this for us? */
507 bool incompatible = false;
508 for (unsigned idx = 0; idx < num_elements; ++idx) {
509 if (elements[idx].vertex_buffer_index >= ctx->specs.stream_count || elements[idx].instance_divisor > 0)
510 incompatible = true;
511 }
512
513 cs->num_elements = num_elements;
514 if (incompatible || num_elements == 0) {
515 DBG("Error: zero vertex elements, or more vertex buffers used than supported");
516 FREE(cs);
517 return NULL;
518 }
519
520 unsigned start_offset = 0; /* start of current consecutive stretch */
521 bool nonconsecutive = true; /* previous value of nonconsecutive */
522
523 for (unsigned idx = 0; idx < num_elements; ++idx) {
524 unsigned element_size = util_format_get_blocksize(elements[idx].src_format);
525 unsigned end_offset = elements[idx].src_offset + element_size;
526 uint32_t format_type, normalize;
527
528 if (nonconsecutive)
529 start_offset = elements[idx].src_offset;
530
531 /* maximum vertex size is 256 bytes */
532 assert(element_size != 0 && end_offset <= 256);
533
534 /* check whether next element is consecutive to this one */
535 nonconsecutive = (idx == (num_elements - 1)) ||
536 elements[idx + 1].vertex_buffer_index != elements[idx].vertex_buffer_index ||
537 end_offset != elements[idx + 1].src_offset;
538
539 format_type = translate_vertex_format_type(elements[idx].src_format);
540 normalize = translate_vertex_format_normalize(elements[idx].src_format);
541
542 assert(format_type != ETNA_NO_MATCH);
543 assert(normalize != ETNA_NO_MATCH);
544
545 cs->FE_VERTEX_ELEMENT_CONFIG[idx] =
546 COND(nonconsecutive, VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE) |
547 format_type |
548 VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(util_format_get_nr_components(elements[idx].src_format)) |
549 normalize | VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN(ENDIAN_MODE_NO_SWAP) |
550 VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(elements[idx].vertex_buffer_index) |
551 VIVS_FE_VERTEX_ELEMENT_CONFIG_START(elements[idx].src_offset) |
552 VIVS_FE_VERTEX_ELEMENT_CONFIG_END(end_offset - start_offset);
553 }
554
555 return cs;
556 }
557
558 static void
559 etna_vertex_elements_state_delete(struct pipe_context *pctx, void *ve)
560 {
561 FREE(ve);
562 }
563
564 static void
565 etna_vertex_elements_state_bind(struct pipe_context *pctx, void *ve)
566 {
567 struct etna_context *ctx = etna_context(pctx);
568
569 ctx->vertex_elements = ve;
570 ctx->dirty |= ETNA_DIRTY_VERTEX_ELEMENTS;
571 }
572
573 static bool
574 etna_update_ts_config(struct etna_context *ctx)
575 {
576 uint32_t new_ts_config = ctx->framebuffer.TS_MEM_CONFIG;
577
578 if (ctx->framebuffer_s.nr_cbufs > 0) {
579 struct etna_surface *c_surf = etna_surface(ctx->framebuffer_s.cbufs[0]);
580
581 if(c_surf->level->ts_size && c_surf->level->ts_valid) {
582 new_ts_config |= VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
583 } else {
584 new_ts_config &= ~VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
585 }
586 }
587
588 if (ctx->framebuffer_s.zsbuf) {
589 struct etna_surface *zs_surf = etna_surface(ctx->framebuffer_s.zsbuf);
590
591 if(zs_surf->level->ts_size && zs_surf->level->ts_valid) {
592 new_ts_config |= VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
593 } else {
594 new_ts_config &= ~VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
595 }
596 }
597
598 if (new_ts_config != ctx->framebuffer.TS_MEM_CONFIG) {
599 ctx->framebuffer.TS_MEM_CONFIG = new_ts_config;
600 ctx->dirty |= ETNA_DIRTY_TS;
601 }
602
603 ctx->dirty &= ~ETNA_DIRTY_DERIVE_TS;
604
605 return true;
606 }
607
608 struct etna_state_updater {
609 bool (*update)(struct etna_context *ctx);
610 uint32_t dirty;
611 };
612
613 static const struct etna_state_updater etna_state_updates[] = {
614 {
615 etna_shader_update_vertex, ETNA_DIRTY_SHADER | ETNA_DIRTY_VERTEX_ELEMENTS,
616 },
617 {
618 etna_shader_link, ETNA_DIRTY_SHADER,
619 },
620 {
621 etna_update_blend, ETNA_DIRTY_BLEND | ETNA_DIRTY_FRAMEBUFFER
622 },
623 {
624 etna_update_blend_color, ETNA_DIRTY_BLEND_COLOR | ETNA_DIRTY_FRAMEBUFFER,
625 },
626 {
627 etna_update_ts_config, ETNA_DIRTY_DERIVE_TS,
628 }
629 };
630
631 bool
632 etna_state_update(struct etna_context *ctx)
633 {
634 for (unsigned int i = 0; i < ARRAY_SIZE(etna_state_updates); i++)
635 if (ctx->dirty & etna_state_updates[i].dirty)
636 if (!etna_state_updates[i].update(ctx))
637 return false;
638
639 return true;
640 }
641
642 void
643 etna_state_init(struct pipe_context *pctx)
644 {
645 pctx->set_blend_color = etna_set_blend_color;
646 pctx->set_stencil_ref = etna_set_stencil_ref;
647 pctx->set_clip_state = etna_set_clip_state;
648 pctx->set_sample_mask = etna_set_sample_mask;
649 pctx->set_constant_buffer = etna_set_constant_buffer;
650 pctx->set_framebuffer_state = etna_set_framebuffer_state;
651 pctx->set_polygon_stipple = etna_set_polygon_stipple;
652 pctx->set_scissor_states = etna_set_scissor_states;
653 pctx->set_viewport_states = etna_set_viewport_states;
654
655 pctx->set_vertex_buffers = etna_set_vertex_buffers;
656
657 pctx->bind_blend_state = etna_blend_state_bind;
658 pctx->delete_blend_state = etna_blend_state_delete;
659
660 pctx->bind_rasterizer_state = etna_rasterizer_state_bind;
661 pctx->delete_rasterizer_state = etna_rasterizer_state_delete;
662
663 pctx->bind_depth_stencil_alpha_state = etna_zsa_state_bind;
664 pctx->delete_depth_stencil_alpha_state = etna_zsa_state_delete;
665
666 pctx->create_vertex_elements_state = etna_vertex_elements_state_create;
667 pctx->delete_vertex_elements_state = etna_vertex_elements_state_delete;
668 pctx->bind_vertex_elements_state = etna_vertex_elements_state_bind;
669 }