etnaviv: Consolidate buffer references from framebuffers
[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_framebuffer.h"
41 #include "util/u_helpers.h"
42 #include "util/u_inlines.h"
43 #include "util/u_math.h"
44 #include "util/u_memory.h"
45
46 static void
47 etna_set_stencil_ref(struct pipe_context *pctx, const struct pipe_stencil_ref *sr)
48 {
49 struct etna_context *ctx = etna_context(pctx);
50 struct compiled_stencil_ref *cs = &ctx->stencil_ref;
51
52 ctx->stencil_ref_s = *sr;
53
54 cs->PE_STENCIL_CONFIG = VIVS_PE_STENCIL_CONFIG_REF_FRONT(sr->ref_value[0]);
55 /* rest of bits weaved in from depth_stencil_alpha */
56 cs->PE_STENCIL_CONFIG_EXT =
57 VIVS_PE_STENCIL_CONFIG_EXT_REF_BACK(sr->ref_value[0]);
58 ctx->dirty |= ETNA_DIRTY_STENCIL_REF;
59 }
60
61 static void
62 etna_set_clip_state(struct pipe_context *pctx, const struct pipe_clip_state *pcs)
63 {
64 /* NOOP */
65 }
66
67 static void
68 etna_set_sample_mask(struct pipe_context *pctx, unsigned sample_mask)
69 {
70 struct etna_context *ctx = etna_context(pctx);
71
72 ctx->sample_mask = sample_mask;
73 ctx->dirty |= ETNA_DIRTY_SAMPLE_MASK;
74 }
75
76 static void
77 etna_set_constant_buffer(struct pipe_context *pctx,
78 enum pipe_shader_type shader, uint index,
79 const struct pipe_constant_buffer *cb)
80 {
81 struct etna_context *ctx = etna_context(pctx);
82
83 if (unlikely(index > 0)) {
84 DBG("Unhandled buffer index %i", index);
85 return;
86 }
87
88
89 util_copy_constant_buffer(&ctx->constant_buffer[shader], cb);
90
91 /* Note that the state tracker can unbind constant buffers by
92 * passing NULL here. */
93 if (unlikely(!cb || (!cb->buffer && !cb->user_buffer)))
94 return;
95
96 /* there is no support for ARB_uniform_buffer_object */
97 assert(cb->buffer == NULL && cb->user_buffer != NULL);
98
99 ctx->dirty |= ETNA_DIRTY_CONSTBUF;
100 }
101
102 static void
103 etna_update_render_resource(struct pipe_context *pctx, struct pipe_resource *pres)
104 {
105 struct etna_resource *res = etna_resource(pres);
106
107 if (res->texture && etna_resource_older(res, etna_resource(res->texture))) {
108 /* The render buffer is older than the texture buffer. Copy it over. */
109 etna_copy_resource(pctx, pres, res->texture, 0, pres->last_level);
110 res->seqno = etna_resource(res->texture)->seqno;
111 }
112 }
113
114 static void
115 etna_set_framebuffer_state(struct pipe_context *pctx,
116 const struct pipe_framebuffer_state *sv)
117 {
118 struct etna_context *ctx = etna_context(pctx);
119 struct compiled_framebuffer_state *cs = &ctx->framebuffer;
120 int nr_samples_color = -1;
121 int nr_samples_depth = -1;
122
123 /* Set up TS as well. Warning: this state is used by both the RS and PE */
124 uint32_t ts_mem_config = 0;
125
126 if (sv->nr_cbufs > 0) { /* at least one color buffer? */
127 struct etna_surface *cbuf = etna_surface(sv->cbufs[0]);
128 struct etna_resource *res = etna_resource(cbuf->base.texture);
129 bool color_supertiled = (res->layout & ETNA_LAYOUT_BIT_SUPER) != 0;
130
131 assert(res->layout & ETNA_LAYOUT_BIT_TILE); /* Cannot render to linear surfaces */
132 etna_update_render_resource(pctx, cbuf->base.texture);
133
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 COND(color_supertiled && ctx->specs.halti >= 5, VIVS_PE_COLOR_FORMAT_SUPER_TILED_NEW);
140 /* VIVS_PE_COLOR_FORMAT_COMPONENTS() and
141 * VIVS_PE_COLOR_FORMAT_OVERWRITE comes from blend_state
142 * but only if we set the bits above. */
143 /* merged with depth_stencil_alpha */
144 if ((cbuf->surf.offset & 63) ||
145 (((cbuf->surf.stride * 4) & 63) && cbuf->surf.height > 4)) {
146 /* XXX Must make temporary surface here.
147 * Need the same mechanism on gc2000 when we want to do mipmap
148 * generation by
149 * rendering to levels > 1 due to multitiled / tiled conversion. */
150 BUG("Alignment error, trying to render to offset %08x with tile "
151 "stride %i",
152 cbuf->surf.offset, cbuf->surf.stride * 4);
153 }
154
155 if (ctx->specs.pixel_pipes == 1) {
156 cs->PE_COLOR_ADDR = cbuf->reloc[0];
157 cs->PE_COLOR_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
158 } else {
159 /* Rendered textures must always be multi-tiled, or single-buffer mode must be supported */
160 assert((res->layout & ETNA_LAYOUT_BIT_MULTI) || ctx->specs.single_buffer);
161 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
162 cs->PE_PIPE_COLOR_ADDR[i] = cbuf->reloc[i];
163 cs->PE_PIPE_COLOR_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
164 }
165 }
166 cs->PE_COLOR_STRIDE = cbuf->surf.stride;
167
168 if (cbuf->surf.ts_size) {
169 cs->TS_COLOR_CLEAR_VALUE = cbuf->level->clear_value;
170
171 cs->TS_COLOR_STATUS_BASE = cbuf->ts_reloc;
172 cs->TS_COLOR_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
173
174 cs->TS_COLOR_SURFACE_BASE = cbuf->reloc[0];
175 cs->TS_COLOR_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
176 }
177
178 /* MSAA */
179 if (cbuf->base.texture->nr_samples > 1)
180 ts_mem_config |=
181 VIVS_TS_MEM_CONFIG_COLOR_COMPRESSION | translate_msaa_format(cbuf->base.format);
182
183 nr_samples_color = cbuf->base.texture->nr_samples;
184 } else {
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 = VIVS_PE_COLOR_FORMAT_OVERWRITE;
189 cs->PE_COLOR_STRIDE = 0;
190 cs->TS_COLOR_STATUS_BASE.bo = NULL;
191 cs->TS_COLOR_SURFACE_BASE.bo = NULL;
192
193 cs->PE_COLOR_ADDR = ctx->dummy_rt_reloc;
194 for (int i = 0; i < ctx->specs.pixel_pipes; i++)
195 cs->PE_PIPE_COLOR_ADDR[i] = ctx->dummy_rt_reloc;
196 }
197
198 if (sv->zsbuf != NULL) {
199 struct etna_surface *zsbuf = etna_surface(sv->zsbuf);
200 struct etna_resource *res = etna_resource(zsbuf->base.texture);
201
202 etna_update_render_resource(pctx, zsbuf->base.texture);
203
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 COND(ctx->specs.halti >= 5, VIVS_PE_DEPTH_CONFIG_DISABLE_ZS) /* Needs to be enabled on GC7000, otherwise depth writes hang w/ TS - apparently it does something else now */
216 ;
217 /* VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH */
218 /* merged with depth_stencil_alpha */
219
220 if (ctx->specs.pixel_pipes == 1) {
221 cs->PE_DEPTH_ADDR = zsbuf->reloc[0];
222 cs->PE_DEPTH_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
223 } else {
224 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
225 cs->PE_PIPE_DEPTH_ADDR[i] = zsbuf->reloc[i];
226 cs->PE_PIPE_DEPTH_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
227 }
228 }
229
230 cs->PE_DEPTH_STRIDE = zsbuf->surf.stride;
231 cs->PE_HDEPTH_CONTROL = VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED;
232 cs->PE_DEPTH_NORMALIZE = fui(exp2f(depth_bits) - 1.0f);
233
234 if (zsbuf->surf.ts_size) {
235 cs->TS_DEPTH_CLEAR_VALUE = zsbuf->level->clear_value;
236
237 cs->TS_DEPTH_STATUS_BASE = zsbuf->ts_reloc;
238 cs->TS_DEPTH_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
239
240 cs->TS_DEPTH_SURFACE_BASE = zsbuf->reloc[0];
241 cs->TS_DEPTH_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
242 }
243
244 ts_mem_config |= COND(depth_bits == 16, VIVS_TS_MEM_CONFIG_DEPTH_16BPP);
245
246 /* MSAA */
247 if (zsbuf->base.texture->nr_samples > 1)
248 /* XXX VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION;
249 * Disable without MSAA for now, as it causes corruption in glquake. */
250 ts_mem_config |= VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION;
251
252 nr_samples_depth = zsbuf->base.texture->nr_samples;
253 } else {
254 cs->PE_DEPTH_CONFIG = VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_NONE;
255 cs->PE_DEPTH_ADDR.bo = NULL;
256 cs->PE_DEPTH_STRIDE = 0;
257 cs->TS_DEPTH_STATUS_BASE.bo = NULL;
258 cs->TS_DEPTH_SURFACE_BASE.bo = NULL;
259
260 for (int i = 0; i < ETNA_MAX_PIXELPIPES; i++)
261 cs->PE_PIPE_DEPTH_ADDR[i].bo = NULL;
262 }
263
264 /* MSAA setup */
265 if (nr_samples_depth != -1 && nr_samples_color != -1 &&
266 nr_samples_depth != nr_samples_color) {
267 BUG("Number of samples in color and depth texture must match (%i and %i respectively)",
268 nr_samples_color, nr_samples_depth);
269 }
270
271 switch (MAX2(nr_samples_depth, nr_samples_color)) {
272 case 0:
273 case 1: /* Are 0 and 1 samples allowed? */
274 cs->GL_MULTI_SAMPLE_CONFIG =
275 VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_NONE;
276 cs->msaa_mode = false;
277 break;
278 case 2:
279 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_2X;
280 cs->msaa_mode = true; /* Add input to PS */
281 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
282 cs->RA_MULTISAMPLE_UNK00E10[0] = 0x0000aa22;
283 cs->RA_CENTROID_TABLE[0] = 0x66aa2288;
284 cs->RA_CENTROID_TABLE[1] = 0x88558800;
285 cs->RA_CENTROID_TABLE[2] = 0x88881100;
286 cs->RA_CENTROID_TABLE[3] = 0x33888800;
287 break;
288 case 4:
289 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_4X;
290 cs->msaa_mode = true; /* Add input to PS */
291 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
292 cs->RA_MULTISAMPLE_UNK00E10[0] = 0xeaa26e26;
293 cs->RA_MULTISAMPLE_UNK00E10[1] = 0xe6ae622a;
294 cs->RA_MULTISAMPLE_UNK00E10[2] = 0xaaa22a22;
295 cs->RA_CENTROID_TABLE[0] = 0x4a6e2688;
296 cs->RA_CENTROID_TABLE[1] = 0x888888a2;
297 cs->RA_CENTROID_TABLE[2] = 0x888888ea;
298 cs->RA_CENTROID_TABLE[3] = 0x888888c6;
299 cs->RA_CENTROID_TABLE[4] = 0x46622a88;
300 cs->RA_CENTROID_TABLE[5] = 0x888888ae;
301 cs->RA_CENTROID_TABLE[6] = 0x888888e6;
302 cs->RA_CENTROID_TABLE[7] = 0x888888ca;
303 cs->RA_CENTROID_TABLE[8] = 0x262a2288;
304 cs->RA_CENTROID_TABLE[9] = 0x886688a2;
305 cs->RA_CENTROID_TABLE[10] = 0x888866aa;
306 cs->RA_CENTROID_TABLE[11] = 0x668888a6;
307 break;
308 }
309
310 /* Scissor setup */
311 cs->SE_SCISSOR_LEFT = 0; /* affected by rasterizer and scissor state as well */
312 cs->SE_SCISSOR_TOP = 0;
313 cs->SE_SCISSOR_RIGHT = (sv->width << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
314 cs->SE_SCISSOR_BOTTOM = (sv->height << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
315 cs->SE_CLIP_RIGHT = (sv->width << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
316 cs->SE_CLIP_BOTTOM = (sv->height << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
317
318 cs->TS_MEM_CONFIG = ts_mem_config;
319
320 /* Single buffer setup. There is only one switch for this, not a separate
321 * one per color buffer / depth buffer. To keep the logic simple always use
322 * single buffer when this feature is available.
323 */
324 cs->PE_LOGIC_OP = VIVS_PE_LOGIC_OP_SINGLE_BUFFER(ctx->specs.single_buffer ? 3 : 0);
325
326 /* keep copy of original structure */
327 util_copy_framebuffer_state(&ctx->framebuffer_s, sv);
328 ctx->dirty |= ETNA_DIRTY_FRAMEBUFFER | ETNA_DIRTY_DERIVE_TS;
329 }
330
331 static void
332 etna_set_polygon_stipple(struct pipe_context *pctx,
333 const struct pipe_poly_stipple *stipple)
334 {
335 /* NOP */
336 }
337
338 static void
339 etna_set_scissor_states(struct pipe_context *pctx, unsigned start_slot,
340 unsigned num_scissors, const struct pipe_scissor_state *ss)
341 {
342 struct etna_context *ctx = etna_context(pctx);
343 struct compiled_scissor_state *cs = &ctx->scissor;
344 assert(ss->minx <= ss->maxx);
345 assert(ss->miny <= ss->maxy);
346
347 /* note that this state is only used when rasterizer_state->scissor is on */
348 ctx->scissor_s = *ss;
349 cs->SE_SCISSOR_LEFT = (ss->minx << 16);
350 cs->SE_SCISSOR_TOP = (ss->miny << 16);
351 cs->SE_SCISSOR_RIGHT = (ss->maxx << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
352 cs->SE_SCISSOR_BOTTOM = (ss->maxy << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
353 cs->SE_CLIP_RIGHT = (ss->maxx << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
354 cs->SE_CLIP_BOTTOM = (ss->maxy << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
355
356 ctx->dirty |= ETNA_DIRTY_SCISSOR;
357 }
358
359 static void
360 etna_set_viewport_states(struct pipe_context *pctx, unsigned start_slot,
361 unsigned num_scissors, const struct pipe_viewport_state *vs)
362 {
363 struct etna_context *ctx = etna_context(pctx);
364 struct compiled_viewport_state *cs = &ctx->viewport;
365
366 ctx->viewport_s = *vs;
367 /**
368 * For Vivante GPU, viewport z transformation is 0..1 to 0..1 instead of
369 * -1..1 to 0..1.
370 * scaling and translation to 0..1 already happened, so remove that
371 *
372 * z' = (z * 2 - 1) * scale + translate
373 * = z * (2 * scale) + (translate - scale)
374 *
375 * scale' = 2 * scale
376 * translate' = translate - scale
377 */
378
379 /* must be fixp as v4 state deltas assume it is */
380 cs->PA_VIEWPORT_SCALE_X = etna_f32_to_fixp16(vs->scale[0]);
381 cs->PA_VIEWPORT_SCALE_Y = etna_f32_to_fixp16(vs->scale[1]);
382 cs->PA_VIEWPORT_SCALE_Z = fui(vs->scale[2] * 2.0f);
383 cs->PA_VIEWPORT_OFFSET_X = etna_f32_to_fixp16(vs->translate[0]);
384 cs->PA_VIEWPORT_OFFSET_Y = etna_f32_to_fixp16(vs->translate[1]);
385 cs->PA_VIEWPORT_OFFSET_Z = fui(vs->translate[2] - vs->scale[2]);
386
387 /* Compute scissor rectangle (fixp) from viewport.
388 * Make sure left is always < right and top always < bottom.
389 */
390 cs->SE_SCISSOR_LEFT = etna_f32_to_fixp16(MAX2(vs->translate[0] - fabsf(vs->scale[0]), 0.0f));
391 cs->SE_SCISSOR_TOP = etna_f32_to_fixp16(MAX2(vs->translate[1] - fabsf(vs->scale[1]), 0.0f));
392 uint32_t right_fixp = etna_f32_to_fixp16(MAX2(vs->translate[0] + fabsf(vs->scale[0]), 0.0f));
393 uint32_t bottom_fixp = etna_f32_to_fixp16(MAX2(vs->translate[1] + fabsf(vs->scale[1]), 0.0f));
394 cs->SE_SCISSOR_RIGHT = right_fixp + ETNA_SE_SCISSOR_MARGIN_RIGHT;
395 cs->SE_SCISSOR_BOTTOM = bottom_fixp + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
396 cs->SE_CLIP_RIGHT = right_fixp + ETNA_SE_CLIP_MARGIN_RIGHT;
397 cs->SE_CLIP_BOTTOM = bottom_fixp + ETNA_SE_CLIP_MARGIN_BOTTOM;
398
399 cs->PE_DEPTH_NEAR = fui(0.0); /* not affected if depth mode is Z (as in GL) */
400 cs->PE_DEPTH_FAR = fui(1.0);
401 ctx->dirty |= ETNA_DIRTY_VIEWPORT;
402 }
403
404 static void
405 etna_set_vertex_buffers(struct pipe_context *pctx, unsigned start_slot,
406 unsigned num_buffers, const struct pipe_vertex_buffer *vb)
407 {
408 struct etna_context *ctx = etna_context(pctx);
409 struct etna_vertexbuf_state *so = &ctx->vertex_buffer;
410
411 util_set_vertex_buffers_mask(so->vb, &so->enabled_mask, vb, start_slot, num_buffers);
412 so->count = util_last_bit(so->enabled_mask);
413
414 for (unsigned idx = start_slot; idx < start_slot + num_buffers; ++idx) {
415 struct compiled_set_vertex_buffer *cs = &so->cvb[idx];
416 struct pipe_vertex_buffer *vbi = &so->vb[idx];
417
418 assert(!vbi->is_user_buffer); /* XXX support user_buffer using
419 etna_usermem_map */
420
421 if (vbi->buffer.resource) { /* GPU buffer */
422 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = etna_resource(vbi->buffer.resource)->bo;
423 cs->FE_VERTEX_STREAM_BASE_ADDR.offset = vbi->buffer_offset;
424 cs->FE_VERTEX_STREAM_BASE_ADDR.flags = ETNA_RELOC_READ;
425 cs->FE_VERTEX_STREAM_CONTROL =
426 FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE(vbi->stride);
427 } else {
428 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = NULL;
429 cs->FE_VERTEX_STREAM_CONTROL = 0;
430 }
431 }
432
433 ctx->dirty |= ETNA_DIRTY_VERTEX_BUFFERS;
434 }
435
436 static void
437 etna_blend_state_bind(struct pipe_context *pctx, void *bs)
438 {
439 struct etna_context *ctx = etna_context(pctx);
440
441 ctx->blend = bs;
442 ctx->dirty |= ETNA_DIRTY_BLEND;
443 }
444
445 static void
446 etna_blend_state_delete(struct pipe_context *pctx, void *bs)
447 {
448 FREE(bs);
449 }
450
451 static void
452 etna_rasterizer_state_bind(struct pipe_context *pctx, void *rs)
453 {
454 struct etna_context *ctx = etna_context(pctx);
455
456 ctx->rasterizer = rs;
457 ctx->dirty |= ETNA_DIRTY_RASTERIZER;
458 }
459
460 static void
461 etna_rasterizer_state_delete(struct pipe_context *pctx, void *rs)
462 {
463 FREE(rs);
464 }
465
466 static void
467 etna_zsa_state_bind(struct pipe_context *pctx, void *zs)
468 {
469 struct etna_context *ctx = etna_context(pctx);
470
471 ctx->zsa = zs;
472 ctx->dirty |= ETNA_DIRTY_ZSA;
473 }
474
475 static void
476 etna_zsa_state_delete(struct pipe_context *pctx, void *zs)
477 {
478 FREE(zs);
479 }
480
481 /** Create vertex element states, which define a layout for fetching
482 * vertices for rendering.
483 */
484 static void *
485 etna_vertex_elements_state_create(struct pipe_context *pctx,
486 unsigned num_elements, const struct pipe_vertex_element *elements)
487 {
488 struct etna_context *ctx = etna_context(pctx);
489 struct compiled_vertex_elements_state *cs = CALLOC_STRUCT(compiled_vertex_elements_state);
490
491 if (!cs)
492 return NULL;
493
494 if (num_elements > ctx->specs.vertex_max_elements) {
495 BUG("number of elements (%u) exceeds chip maximum (%u)", num_elements,
496 ctx->specs.vertex_max_elements);
497 return NULL;
498 }
499
500 /* XXX could minimize number of consecutive stretches here by sorting, and
501 * permuting the inputs in shader or does Mesa do this already? */
502
503 /* Check that vertex element binding is compatible with hardware; thus
504 * elements[idx].vertex_buffer_index are < stream_count. If not, the binding
505 * uses more streams than is supported, and u_vbuf should have done some
506 * reorganization for compatibility. */
507
508 /* TODO: does mesa this for us? */
509 bool incompatible = false;
510 for (unsigned idx = 0; idx < num_elements; ++idx) {
511 if (elements[idx].vertex_buffer_index >= ctx->specs.stream_count || elements[idx].instance_divisor > 0)
512 incompatible = true;
513 }
514
515 cs->num_elements = num_elements;
516 if (incompatible || num_elements == 0) {
517 DBG("Error: zero vertex elements, or more vertex buffers used than supported");
518 FREE(cs);
519 return NULL;
520 }
521
522 unsigned start_offset = 0; /* start of current consecutive stretch */
523 bool nonconsecutive = true; /* previous value of nonconsecutive */
524
525 for (unsigned idx = 0; idx < num_elements; ++idx) {
526 unsigned element_size = util_format_get_blocksize(elements[idx].src_format);
527 unsigned end_offset = elements[idx].src_offset + element_size;
528 uint32_t format_type, normalize;
529
530 if (nonconsecutive)
531 start_offset = elements[idx].src_offset;
532
533 /* maximum vertex size is 256 bytes */
534 assert(element_size != 0 && end_offset <= 256);
535
536 /* check whether next element is consecutive to this one */
537 nonconsecutive = (idx == (num_elements - 1)) ||
538 elements[idx + 1].vertex_buffer_index != elements[idx].vertex_buffer_index ||
539 end_offset != elements[idx + 1].src_offset;
540
541 format_type = translate_vertex_format_type(elements[idx].src_format);
542 normalize = translate_vertex_format_normalize(elements[idx].src_format);
543
544 assert(format_type != ETNA_NO_MATCH);
545 assert(normalize != ETNA_NO_MATCH);
546
547 if (ctx->specs.halti < 5) {
548 cs->FE_VERTEX_ELEMENT_CONFIG[idx] =
549 COND(nonconsecutive, VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE) |
550 format_type |
551 VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(util_format_get_nr_components(elements[idx].src_format)) |
552 normalize | VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN(ENDIAN_MODE_NO_SWAP) |
553 VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(elements[idx].vertex_buffer_index) |
554 VIVS_FE_VERTEX_ELEMENT_CONFIG_START(elements[idx].src_offset) |
555 VIVS_FE_VERTEX_ELEMENT_CONFIG_END(end_offset - start_offset);
556 } else { /* HALTI5 spread vertex attrib config over two registers */
557 cs->NFE_GENERIC_ATTRIB_CONFIG0[idx] =
558 format_type |
559 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_NUM(util_format_get_nr_components(elements[idx].src_format)) |
560 normalize | VIVS_NFE_GENERIC_ATTRIB_CONFIG0_ENDIAN(ENDIAN_MODE_NO_SWAP) |
561 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_STREAM(elements[idx].vertex_buffer_index) |
562 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_START(elements[idx].src_offset);
563 cs->NFE_GENERIC_ATTRIB_CONFIG1[idx] =
564 COND(nonconsecutive, VIVS_NFE_GENERIC_ATTRIB_CONFIG1_NONCONSECUTIVE) |
565 VIVS_NFE_GENERIC_ATTRIB_CONFIG1_END(end_offset - start_offset);
566 }
567 cs->NFE_GENERIC_ATTRIB_SCALE[idx] = 0x3f800000; /* 1 for integer, 1.0 for float */
568 }
569
570 return cs;
571 }
572
573 static void
574 etna_vertex_elements_state_delete(struct pipe_context *pctx, void *ve)
575 {
576 FREE(ve);
577 }
578
579 static void
580 etna_vertex_elements_state_bind(struct pipe_context *pctx, void *ve)
581 {
582 struct etna_context *ctx = etna_context(pctx);
583
584 ctx->vertex_elements = ve;
585 ctx->dirty |= ETNA_DIRTY_VERTEX_ELEMENTS;
586 }
587
588 static bool
589 etna_update_ts_config(struct etna_context *ctx)
590 {
591 uint32_t new_ts_config = ctx->framebuffer.TS_MEM_CONFIG;
592
593 if (ctx->framebuffer_s.nr_cbufs > 0) {
594 struct etna_surface *c_surf = etna_surface(ctx->framebuffer_s.cbufs[0]);
595
596 if(c_surf->level->ts_size && c_surf->level->ts_valid) {
597 new_ts_config |= VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
598 } else {
599 new_ts_config &= ~VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
600 }
601 }
602
603 if (ctx->framebuffer_s.zsbuf) {
604 struct etna_surface *zs_surf = etna_surface(ctx->framebuffer_s.zsbuf);
605
606 if(zs_surf->level->ts_size && zs_surf->level->ts_valid) {
607 new_ts_config |= VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
608 } else {
609 new_ts_config &= ~VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
610 }
611 }
612
613 if (new_ts_config != ctx->framebuffer.TS_MEM_CONFIG ||
614 (ctx->dirty & ETNA_DIRTY_FRAMEBUFFER)) {
615 ctx->framebuffer.TS_MEM_CONFIG = new_ts_config;
616 ctx->dirty |= ETNA_DIRTY_TS;
617 }
618
619 ctx->dirty &= ~ETNA_DIRTY_DERIVE_TS;
620
621 return true;
622 }
623
624 struct etna_state_updater {
625 bool (*update)(struct etna_context *ctx);
626 uint32_t dirty;
627 };
628
629 static const struct etna_state_updater etna_state_updates[] = {
630 {
631 etna_shader_update_vertex, ETNA_DIRTY_SHADER | ETNA_DIRTY_VERTEX_ELEMENTS,
632 },
633 {
634 etna_shader_link, ETNA_DIRTY_SHADER,
635 },
636 {
637 etna_update_blend, ETNA_DIRTY_BLEND | ETNA_DIRTY_FRAMEBUFFER
638 },
639 {
640 etna_update_blend_color, ETNA_DIRTY_BLEND_COLOR | ETNA_DIRTY_FRAMEBUFFER,
641 },
642 {
643 etna_update_ts_config, ETNA_DIRTY_DERIVE_TS,
644 }
645 };
646
647 bool
648 etna_state_update(struct etna_context *ctx)
649 {
650 for (unsigned int i = 0; i < ARRAY_SIZE(etna_state_updates); i++)
651 if (ctx->dirty & etna_state_updates[i].dirty)
652 if (!etna_state_updates[i].update(ctx))
653 return false;
654
655 return true;
656 }
657
658 void
659 etna_state_init(struct pipe_context *pctx)
660 {
661 pctx->set_blend_color = etna_set_blend_color;
662 pctx->set_stencil_ref = etna_set_stencil_ref;
663 pctx->set_clip_state = etna_set_clip_state;
664 pctx->set_sample_mask = etna_set_sample_mask;
665 pctx->set_constant_buffer = etna_set_constant_buffer;
666 pctx->set_framebuffer_state = etna_set_framebuffer_state;
667 pctx->set_polygon_stipple = etna_set_polygon_stipple;
668 pctx->set_scissor_states = etna_set_scissor_states;
669 pctx->set_viewport_states = etna_set_viewport_states;
670
671 pctx->set_vertex_buffers = etna_set_vertex_buffers;
672
673 pctx->bind_blend_state = etna_blend_state_bind;
674 pctx->delete_blend_state = etna_blend_state_delete;
675
676 pctx->bind_rasterizer_state = etna_rasterizer_state_bind;
677 pctx->delete_rasterizer_state = etna_rasterizer_state_delete;
678
679 pctx->bind_depth_stencil_alpha_state = etna_zsa_state_bind;
680 pctx->delete_depth_stencil_alpha_state = etna_zsa_state_delete;
681
682 pctx->create_vertex_elements_state = etna_vertex_elements_state_create;
683 pctx->delete_vertex_elements_state = etna_vertex_elements_state_delete;
684 pctx->bind_vertex_elements_state = etna_vertex_elements_state_bind;
685 }