etnaviv: stencil fix
[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 #include "util/u_upload_mgr.h"
46
47 static void
48 etna_set_stencil_ref(struct pipe_context *pctx, const struct pipe_stencil_ref *sr)
49 {
50 struct etna_context *ctx = etna_context(pctx);
51 struct compiled_stencil_ref *cs = &ctx->stencil_ref;
52
53 ctx->stencil_ref_s = *sr;
54
55 for (unsigned i = 0; i < 2; i++) {
56 cs->PE_STENCIL_CONFIG[i] =
57 VIVS_PE_STENCIL_CONFIG_REF_FRONT(sr->ref_value[i]);
58 cs->PE_STENCIL_CONFIG_EXT[i] =
59 VIVS_PE_STENCIL_CONFIG_EXT_REF_BACK(sr->ref_value[!i]);
60 }
61 ctx->dirty |= ETNA_DIRTY_STENCIL_REF;
62 }
63
64 static void
65 etna_set_clip_state(struct pipe_context *pctx, const struct pipe_clip_state *pcs)
66 {
67 /* NOOP */
68 }
69
70 static void
71 etna_set_sample_mask(struct pipe_context *pctx, unsigned sample_mask)
72 {
73 struct etna_context *ctx = etna_context(pctx);
74
75 ctx->sample_mask = sample_mask;
76 ctx->dirty |= ETNA_DIRTY_SAMPLE_MASK;
77 }
78
79 static void
80 etna_set_constant_buffer(struct pipe_context *pctx,
81 enum pipe_shader_type shader, uint index,
82 const struct pipe_constant_buffer *cb)
83 {
84 struct etna_context *ctx = etna_context(pctx);
85
86 if (unlikely(index > 0)) {
87 DBG("Unhandled buffer index %i", index);
88 return;
89 }
90
91
92 util_copy_constant_buffer(&ctx->constant_buffer[shader], cb);
93
94 /* Note that the state tracker can unbind constant buffers by
95 * passing NULL here. */
96 if (unlikely(!cb || (!cb->buffer && !cb->user_buffer)))
97 return;
98
99 /* there is no support for ARB_uniform_buffer_object */
100 assert(cb->buffer == NULL && cb->user_buffer != NULL);
101
102 if (!cb->buffer) {
103 struct pipe_constant_buffer *cb = &ctx->constant_buffer[shader];
104 u_upload_data(pctx->const_uploader, 0, cb->buffer_size, 16, cb->user_buffer, &cb->buffer_offset, &cb->buffer);
105 }
106
107 ctx->dirty |= ETNA_DIRTY_CONSTBUF;
108 }
109
110 static void
111 etna_update_render_resource(struct pipe_context *pctx, struct etna_resource *base)
112 {
113 struct etna_resource *to = base, *from = base;
114
115 if (base->texture && etna_resource_newer(etna_resource(base->texture), base))
116 from = etna_resource(base->texture);
117
118 if (base->render)
119 to = etna_resource(base->render);
120
121 if ((to != from) && etna_resource_older(to, from)) {
122 etna_copy_resource(pctx, &to->base, &from->base, 0, base->base.last_level);
123 to->seqno = from->seqno;
124 }
125 }
126
127 static void
128 etna_set_framebuffer_state(struct pipe_context *pctx,
129 const struct pipe_framebuffer_state *sv)
130 {
131 struct etna_context *ctx = etna_context(pctx);
132 struct compiled_framebuffer_state *cs = &ctx->framebuffer;
133 int nr_samples_color = -1;
134 int nr_samples_depth = -1;
135
136 /* Set up TS as well. Warning: this state is used by both the RS and PE */
137 uint32_t ts_mem_config = 0;
138 uint32_t pe_mem_config = 0;
139
140 if (sv->nr_cbufs > 0) { /* at least one color buffer? */
141 struct etna_surface *cbuf = etna_surface(sv->cbufs[0]);
142 struct etna_resource *res = etna_resource(cbuf->base.texture);
143 bool color_supertiled = (res->layout & ETNA_LAYOUT_BIT_SUPER) != 0;
144
145 assert(res->layout & ETNA_LAYOUT_BIT_TILE); /* Cannot render to linear surfaces */
146 etna_update_render_resource(pctx, etna_resource(cbuf->prsc));
147
148 cs->PE_COLOR_FORMAT =
149 VIVS_PE_COLOR_FORMAT_FORMAT(translate_rs_format(cbuf->base.format)) |
150 VIVS_PE_COLOR_FORMAT_COMPONENTS__MASK |
151 VIVS_PE_COLOR_FORMAT_OVERWRITE |
152 COND(color_supertiled, VIVS_PE_COLOR_FORMAT_SUPER_TILED) |
153 COND(color_supertiled && ctx->specs.halti >= 5, VIVS_PE_COLOR_FORMAT_SUPER_TILED_NEW);
154 /* VIVS_PE_COLOR_FORMAT_COMPONENTS() and
155 * VIVS_PE_COLOR_FORMAT_OVERWRITE comes from blend_state
156 * but only if we set the bits above. */
157 /* merged with depth_stencil_alpha */
158 if ((cbuf->surf.offset & 63) ||
159 (((cbuf->surf.stride * 4) & 63) && cbuf->surf.height > 4)) {
160 /* XXX Must make temporary surface here.
161 * Need the same mechanism on gc2000 when we want to do mipmap
162 * generation by
163 * rendering to levels > 1 due to multitiled / tiled conversion. */
164 BUG("Alignment error, trying to render to offset %08x with tile "
165 "stride %i",
166 cbuf->surf.offset, cbuf->surf.stride * 4);
167 }
168
169 if (ctx->specs.pixel_pipes == 1) {
170 cs->PE_COLOR_ADDR = cbuf->reloc[0];
171 cs->PE_COLOR_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
172 } else {
173 /* Rendered textures must always be multi-tiled, or single-buffer mode must be supported */
174 assert((res->layout & ETNA_LAYOUT_BIT_MULTI) || ctx->specs.single_buffer);
175 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
176 cs->PE_PIPE_COLOR_ADDR[i] = cbuf->reloc[i];
177 cs->PE_PIPE_COLOR_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
178 }
179 }
180 cs->PE_COLOR_STRIDE = cbuf->surf.stride;
181
182 if (cbuf->surf.ts_size) {
183 cs->TS_COLOR_CLEAR_VALUE = cbuf->level->clear_value;
184
185 cs->TS_COLOR_STATUS_BASE = cbuf->ts_reloc;
186 cs->TS_COLOR_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
187
188 cs->TS_COLOR_SURFACE_BASE = cbuf->reloc[0];
189 cs->TS_COLOR_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
190
191 pe_mem_config |= VIVS_PE_MEM_CONFIG_COLOR_TS_MODE(cbuf->level->ts_mode);
192
193 if (cbuf->level->ts_compress_fmt >= 0) {
194 /* overwrite bit breaks v1/v2 compression */
195 if (!ctx->specs.v4_compression)
196 cs->PE_COLOR_FORMAT &= ~VIVS_PE_COLOR_FORMAT_OVERWRITE;
197
198 ts_mem_config |=
199 VIVS_TS_MEM_CONFIG_COLOR_COMPRESSION |
200 VIVS_TS_MEM_CONFIG_COLOR_COMPRESSION_FORMAT(cbuf->level->ts_compress_fmt);
201 }
202 }
203
204 nr_samples_color = cbuf->base.texture->nr_samples;
205 } else {
206 /* Clearing VIVS_PE_COLOR_FORMAT_COMPONENTS__MASK and
207 * VIVS_PE_COLOR_FORMAT_OVERWRITE prevents us from overwriting the
208 * color target */
209 cs->PE_COLOR_FORMAT = VIVS_PE_COLOR_FORMAT_OVERWRITE;
210 cs->PE_COLOR_STRIDE = 0;
211 cs->TS_COLOR_STATUS_BASE.bo = NULL;
212 cs->TS_COLOR_SURFACE_BASE.bo = NULL;
213
214 cs->PE_COLOR_ADDR = ctx->dummy_rt_reloc;
215 for (int i = 0; i < ctx->specs.pixel_pipes; i++)
216 cs->PE_PIPE_COLOR_ADDR[i] = ctx->dummy_rt_reloc;
217 }
218
219 if (sv->zsbuf != NULL) {
220 struct etna_surface *zsbuf = etna_surface(sv->zsbuf);
221 struct etna_resource *res = etna_resource(zsbuf->base.texture);
222
223 etna_update_render_resource(pctx, etna_resource(zsbuf->prsc));
224
225 assert(res->layout &ETNA_LAYOUT_BIT_TILE); /* Cannot render to linear surfaces */
226
227 uint32_t depth_format = translate_depth_format(zsbuf->base.format);
228 unsigned depth_bits =
229 depth_format == VIVS_PE_DEPTH_CONFIG_DEPTH_FORMAT_D16 ? 16 : 24;
230 bool depth_supertiled = (res->layout & ETNA_LAYOUT_BIT_SUPER) != 0;
231
232 cs->PE_DEPTH_CONFIG =
233 depth_format |
234 COND(depth_supertiled, VIVS_PE_DEPTH_CONFIG_SUPER_TILED) |
235 VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_Z |
236 VIVS_PE_DEPTH_CONFIG_UNK18 | /* something to do with clipping? */
237 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 */
238 ;
239 /* VIVS_PE_DEPTH_CONFIG_ONLY_DEPTH */
240 /* merged with depth_stencil_alpha */
241
242 if (ctx->specs.pixel_pipes == 1) {
243 cs->PE_DEPTH_ADDR = zsbuf->reloc[0];
244 cs->PE_DEPTH_ADDR.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
245 } else {
246 for (int i = 0; i < ctx->specs.pixel_pipes; i++) {
247 cs->PE_PIPE_DEPTH_ADDR[i] = zsbuf->reloc[i];
248 cs->PE_PIPE_DEPTH_ADDR[i].flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
249 }
250 }
251
252 cs->PE_DEPTH_STRIDE = zsbuf->surf.stride;
253 cs->PE_HDEPTH_CONTROL = VIVS_PE_HDEPTH_CONTROL_FORMAT_DISABLED;
254 cs->PE_DEPTH_NORMALIZE = fui(exp2f(depth_bits) - 1.0f);
255
256 if (zsbuf->surf.ts_size) {
257 cs->TS_DEPTH_CLEAR_VALUE = zsbuf->level->clear_value;
258
259 cs->TS_DEPTH_STATUS_BASE = zsbuf->ts_reloc;
260 cs->TS_DEPTH_STATUS_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
261
262 cs->TS_DEPTH_SURFACE_BASE = zsbuf->reloc[0];
263 cs->TS_DEPTH_SURFACE_BASE.flags = ETNA_RELOC_READ | ETNA_RELOC_WRITE;
264
265 pe_mem_config |= VIVS_PE_MEM_CONFIG_DEPTH_TS_MODE(zsbuf->level->ts_mode);
266
267 if (zsbuf->level->ts_compress_fmt >= 0) {
268 ts_mem_config |=
269 VIVS_TS_MEM_CONFIG_DEPTH_COMPRESSION |
270 COND(zsbuf->level->ts_compress_fmt == COMPRESSION_FORMAT_D24S8,
271 VIVS_TS_MEM_CONFIG_STENCIL_ENABLE);
272 }
273 }
274
275 ts_mem_config |= COND(depth_bits == 16, VIVS_TS_MEM_CONFIG_DEPTH_16BPP);
276
277 nr_samples_depth = zsbuf->base.texture->nr_samples;
278 } else {
279 cs->PE_DEPTH_CONFIG = VIVS_PE_DEPTH_CONFIG_DEPTH_MODE_NONE;
280 cs->PE_DEPTH_ADDR.bo = NULL;
281 cs->PE_DEPTH_STRIDE = 0;
282 cs->TS_DEPTH_STATUS_BASE.bo = NULL;
283 cs->TS_DEPTH_SURFACE_BASE.bo = NULL;
284
285 for (int i = 0; i < ETNA_MAX_PIXELPIPES; i++)
286 cs->PE_PIPE_DEPTH_ADDR[i].bo = NULL;
287 }
288
289 /* MSAA setup */
290 if (nr_samples_depth != -1 && nr_samples_color != -1 &&
291 nr_samples_depth != nr_samples_color) {
292 BUG("Number of samples in color and depth texture must match (%i and %i respectively)",
293 nr_samples_color, nr_samples_depth);
294 }
295
296 switch (MAX2(nr_samples_depth, nr_samples_color)) {
297 case 0:
298 case 1: /* Are 0 and 1 samples allowed? */
299 cs->GL_MULTI_SAMPLE_CONFIG =
300 VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_NONE;
301 cs->msaa_mode = false;
302 break;
303 case 2:
304 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_2X;
305 cs->msaa_mode = true; /* Add input to PS */
306 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
307 cs->RA_MULTISAMPLE_UNK00E10[0] = 0x0000aa22;
308 cs->RA_CENTROID_TABLE[0] = 0x66aa2288;
309 cs->RA_CENTROID_TABLE[1] = 0x88558800;
310 cs->RA_CENTROID_TABLE[2] = 0x88881100;
311 cs->RA_CENTROID_TABLE[3] = 0x33888800;
312 break;
313 case 4:
314 cs->GL_MULTI_SAMPLE_CONFIG = VIVS_GL_MULTI_SAMPLE_CONFIG_MSAA_SAMPLES_4X;
315 cs->msaa_mode = true; /* Add input to PS */
316 cs->RA_MULTISAMPLE_UNK00E04 = 0x0;
317 cs->RA_MULTISAMPLE_UNK00E10[0] = 0xeaa26e26;
318 cs->RA_MULTISAMPLE_UNK00E10[1] = 0xe6ae622a;
319 cs->RA_MULTISAMPLE_UNK00E10[2] = 0xaaa22a22;
320 cs->RA_CENTROID_TABLE[0] = 0x4a6e2688;
321 cs->RA_CENTROID_TABLE[1] = 0x888888a2;
322 cs->RA_CENTROID_TABLE[2] = 0x888888ea;
323 cs->RA_CENTROID_TABLE[3] = 0x888888c6;
324 cs->RA_CENTROID_TABLE[4] = 0x46622a88;
325 cs->RA_CENTROID_TABLE[5] = 0x888888ae;
326 cs->RA_CENTROID_TABLE[6] = 0x888888e6;
327 cs->RA_CENTROID_TABLE[7] = 0x888888ca;
328 cs->RA_CENTROID_TABLE[8] = 0x262a2288;
329 cs->RA_CENTROID_TABLE[9] = 0x886688a2;
330 cs->RA_CENTROID_TABLE[10] = 0x888866aa;
331 cs->RA_CENTROID_TABLE[11] = 0x668888a6;
332 break;
333 }
334
335 /* Scissor setup */
336 cs->SE_SCISSOR_LEFT = 0; /* affected by rasterizer and scissor state as well */
337 cs->SE_SCISSOR_TOP = 0;
338 cs->SE_SCISSOR_RIGHT = (sv->width << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
339 cs->SE_SCISSOR_BOTTOM = (sv->height << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
340 cs->SE_CLIP_RIGHT = (sv->width << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
341 cs->SE_CLIP_BOTTOM = (sv->height << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
342
343 cs->TS_MEM_CONFIG = ts_mem_config;
344 cs->PE_MEM_CONFIG = pe_mem_config;
345
346 /* Single buffer setup. There is only one switch for this, not a separate
347 * one per color buffer / depth buffer. To keep the logic simple always use
348 * single buffer when this feature is available.
349 */
350 cs->PE_LOGIC_OP = VIVS_PE_LOGIC_OP_SINGLE_BUFFER(ctx->specs.single_buffer ? 3 : 0);
351
352 /* keep copy of original structure */
353 util_copy_framebuffer_state(&ctx->framebuffer_s, sv);
354 ctx->dirty |= ETNA_DIRTY_FRAMEBUFFER | ETNA_DIRTY_DERIVE_TS;
355 }
356
357 static void
358 etna_set_polygon_stipple(struct pipe_context *pctx,
359 const struct pipe_poly_stipple *stipple)
360 {
361 /* NOP */
362 }
363
364 static void
365 etna_set_scissor_states(struct pipe_context *pctx, unsigned start_slot,
366 unsigned num_scissors, const struct pipe_scissor_state *ss)
367 {
368 struct etna_context *ctx = etna_context(pctx);
369 struct compiled_scissor_state *cs = &ctx->scissor;
370 assert(ss->minx <= ss->maxx);
371 assert(ss->miny <= ss->maxy);
372
373 /* note that this state is only used when rasterizer_state->scissor is on */
374 ctx->scissor_s = *ss;
375 cs->SE_SCISSOR_LEFT = (ss->minx << 16);
376 cs->SE_SCISSOR_TOP = (ss->miny << 16);
377 cs->SE_SCISSOR_RIGHT = (ss->maxx << 16) + ETNA_SE_SCISSOR_MARGIN_RIGHT;
378 cs->SE_SCISSOR_BOTTOM = (ss->maxy << 16) + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
379 cs->SE_CLIP_RIGHT = (ss->maxx << 16) + ETNA_SE_CLIP_MARGIN_RIGHT;
380 cs->SE_CLIP_BOTTOM = (ss->maxy << 16) + ETNA_SE_CLIP_MARGIN_BOTTOM;
381
382 ctx->dirty |= ETNA_DIRTY_SCISSOR;
383 }
384
385 static void
386 etna_set_viewport_states(struct pipe_context *pctx, unsigned start_slot,
387 unsigned num_scissors, const struct pipe_viewport_state *vs)
388 {
389 struct etna_context *ctx = etna_context(pctx);
390 struct compiled_viewport_state *cs = &ctx->viewport;
391
392 ctx->viewport_s = *vs;
393 /**
394 * For Vivante GPU, viewport z transformation is 0..1 to 0..1 instead of
395 * -1..1 to 0..1.
396 * scaling and translation to 0..1 already happened, so remove that
397 *
398 * z' = (z * 2 - 1) * scale + translate
399 * = z * (2 * scale) + (translate - scale)
400 *
401 * scale' = 2 * scale
402 * translate' = translate - scale
403 */
404
405 /* must be fixp as v4 state deltas assume it is */
406 cs->PA_VIEWPORT_SCALE_X = etna_f32_to_fixp16(vs->scale[0]);
407 cs->PA_VIEWPORT_SCALE_Y = etna_f32_to_fixp16(vs->scale[1]);
408 cs->PA_VIEWPORT_SCALE_Z = fui(vs->scale[2] * 2.0f);
409 cs->PA_VIEWPORT_OFFSET_X = etna_f32_to_fixp16(vs->translate[0]);
410 cs->PA_VIEWPORT_OFFSET_Y = etna_f32_to_fixp16(vs->translate[1]);
411 cs->PA_VIEWPORT_OFFSET_Z = fui(vs->translate[2] - vs->scale[2]);
412
413 /* Compute scissor rectangle (fixp) from viewport.
414 * Make sure left is always < right and top always < bottom.
415 */
416 cs->SE_SCISSOR_LEFT = etna_f32_to_fixp16(MAX2(vs->translate[0] - fabsf(vs->scale[0]), 0.0f));
417 cs->SE_SCISSOR_TOP = etna_f32_to_fixp16(MAX2(vs->translate[1] - fabsf(vs->scale[1]), 0.0f));
418 uint32_t right_fixp = etna_f32_to_fixp16(MAX2(vs->translate[0] + fabsf(vs->scale[0]), 0.0f));
419 uint32_t bottom_fixp = etna_f32_to_fixp16(MAX2(vs->translate[1] + fabsf(vs->scale[1]), 0.0f));
420 cs->SE_SCISSOR_RIGHT = right_fixp + ETNA_SE_SCISSOR_MARGIN_RIGHT;
421 cs->SE_SCISSOR_BOTTOM = bottom_fixp + ETNA_SE_SCISSOR_MARGIN_BOTTOM;
422 cs->SE_CLIP_RIGHT = right_fixp + ETNA_SE_CLIP_MARGIN_RIGHT;
423 cs->SE_CLIP_BOTTOM = bottom_fixp + ETNA_SE_CLIP_MARGIN_BOTTOM;
424
425 cs->PE_DEPTH_NEAR = fui(0.0); /* not affected if depth mode is Z (as in GL) */
426 cs->PE_DEPTH_FAR = fui(1.0);
427 ctx->dirty |= ETNA_DIRTY_VIEWPORT;
428 }
429
430 static void
431 etna_set_vertex_buffers(struct pipe_context *pctx, unsigned start_slot,
432 unsigned num_buffers, const struct pipe_vertex_buffer *vb)
433 {
434 struct etna_context *ctx = etna_context(pctx);
435 struct etna_vertexbuf_state *so = &ctx->vertex_buffer;
436
437 util_set_vertex_buffers_mask(so->vb, &so->enabled_mask, vb, start_slot, num_buffers);
438 so->count = util_last_bit(so->enabled_mask);
439
440 for (unsigned idx = start_slot; idx < start_slot + num_buffers; ++idx) {
441 struct compiled_set_vertex_buffer *cs = &so->cvb[idx];
442 struct pipe_vertex_buffer *vbi = &so->vb[idx];
443
444 assert(!vbi->is_user_buffer); /* XXX support user_buffer using
445 etna_usermem_map */
446
447 if (vbi->buffer.resource) { /* GPU buffer */
448 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = etna_resource(vbi->buffer.resource)->bo;
449 cs->FE_VERTEX_STREAM_BASE_ADDR.offset = vbi->buffer_offset;
450 cs->FE_VERTEX_STREAM_BASE_ADDR.flags = ETNA_RELOC_READ;
451 cs->FE_VERTEX_STREAM_CONTROL =
452 FE_VERTEX_STREAM_CONTROL_VERTEX_STRIDE(vbi->stride);
453 } else {
454 cs->FE_VERTEX_STREAM_BASE_ADDR.bo = NULL;
455 cs->FE_VERTEX_STREAM_CONTROL = 0;
456 }
457 }
458
459 ctx->dirty |= ETNA_DIRTY_VERTEX_BUFFERS;
460 }
461
462 static void
463 etna_blend_state_bind(struct pipe_context *pctx, void *bs)
464 {
465 struct etna_context *ctx = etna_context(pctx);
466
467 ctx->blend = bs;
468 ctx->dirty |= ETNA_DIRTY_BLEND;
469 }
470
471 static void
472 etna_blend_state_delete(struct pipe_context *pctx, void *bs)
473 {
474 FREE(bs);
475 }
476
477 static void
478 etna_rasterizer_state_bind(struct pipe_context *pctx, void *rs)
479 {
480 struct etna_context *ctx = etna_context(pctx);
481
482 ctx->rasterizer = rs;
483 ctx->dirty |= ETNA_DIRTY_RASTERIZER;
484 }
485
486 static void
487 etna_rasterizer_state_delete(struct pipe_context *pctx, void *rs)
488 {
489 FREE(rs);
490 }
491
492 static void
493 etna_zsa_state_bind(struct pipe_context *pctx, void *zs)
494 {
495 struct etna_context *ctx = etna_context(pctx);
496
497 ctx->zsa = zs;
498 ctx->dirty |= ETNA_DIRTY_ZSA;
499 }
500
501 static void
502 etna_zsa_state_delete(struct pipe_context *pctx, void *zs)
503 {
504 FREE(zs);
505 }
506
507 /** Create vertex element states, which define a layout for fetching
508 * vertices for rendering.
509 */
510 static void *
511 etna_vertex_elements_state_create(struct pipe_context *pctx,
512 unsigned num_elements, const struct pipe_vertex_element *elements)
513 {
514 struct etna_context *ctx = etna_context(pctx);
515 struct compiled_vertex_elements_state *cs = CALLOC_STRUCT(compiled_vertex_elements_state);
516
517 if (!cs)
518 return NULL;
519
520 if (num_elements > ctx->specs.vertex_max_elements) {
521 BUG("number of elements (%u) exceeds chip maximum (%u)", num_elements,
522 ctx->specs.vertex_max_elements);
523 return NULL;
524 }
525
526 /* XXX could minimize number of consecutive stretches here by sorting, and
527 * permuting the inputs in shader or does Mesa do this already? */
528
529 /* Check that vertex element binding is compatible with hardware; thus
530 * elements[idx].vertex_buffer_index are < stream_count. If not, the binding
531 * uses more streams than is supported, and u_vbuf should have done some
532 * reorganization for compatibility. */
533
534 /* TODO: does mesa this for us? */
535 bool incompatible = false;
536 for (unsigned idx = 0; idx < num_elements; ++idx) {
537 if (elements[idx].vertex_buffer_index >= ctx->specs.stream_count || elements[idx].instance_divisor > 0)
538 incompatible = true;
539 }
540
541 cs->num_elements = num_elements;
542 if (incompatible || num_elements == 0) {
543 DBG("Error: zero vertex elements, or more vertex buffers used than supported");
544 FREE(cs);
545 return NULL;
546 }
547
548 unsigned start_offset = 0; /* start of current consecutive stretch */
549 bool nonconsecutive = true; /* previous value of nonconsecutive */
550
551 for (unsigned idx = 0; idx < num_elements; ++idx) {
552 unsigned element_size = util_format_get_blocksize(elements[idx].src_format);
553 unsigned end_offset = elements[idx].src_offset + element_size;
554 uint32_t format_type, normalize;
555
556 if (nonconsecutive)
557 start_offset = elements[idx].src_offset;
558
559 /* maximum vertex size is 256 bytes */
560 assert(element_size != 0 && end_offset <= 256);
561
562 /* check whether next element is consecutive to this one */
563 nonconsecutive = (idx == (num_elements - 1)) ||
564 elements[idx + 1].vertex_buffer_index != elements[idx].vertex_buffer_index ||
565 end_offset != elements[idx + 1].src_offset;
566
567 format_type = translate_vertex_format_type(elements[idx].src_format);
568 normalize = translate_vertex_format_normalize(elements[idx].src_format);
569
570 assert(format_type != ETNA_NO_MATCH);
571 assert(normalize != ETNA_NO_MATCH);
572
573 if (ctx->specs.halti < 5) {
574 cs->FE_VERTEX_ELEMENT_CONFIG[idx] =
575 COND(nonconsecutive, VIVS_FE_VERTEX_ELEMENT_CONFIG_NONCONSECUTIVE) |
576 format_type |
577 VIVS_FE_VERTEX_ELEMENT_CONFIG_NUM(util_format_get_nr_components(elements[idx].src_format)) |
578 normalize | VIVS_FE_VERTEX_ELEMENT_CONFIG_ENDIAN(ENDIAN_MODE_NO_SWAP) |
579 VIVS_FE_VERTEX_ELEMENT_CONFIG_STREAM(elements[idx].vertex_buffer_index) |
580 VIVS_FE_VERTEX_ELEMENT_CONFIG_START(elements[idx].src_offset) |
581 VIVS_FE_VERTEX_ELEMENT_CONFIG_END(end_offset - start_offset);
582 } else { /* HALTI5 spread vertex attrib config over two registers */
583 cs->NFE_GENERIC_ATTRIB_CONFIG0[idx] =
584 format_type |
585 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_NUM(util_format_get_nr_components(elements[idx].src_format)) |
586 normalize | VIVS_NFE_GENERIC_ATTRIB_CONFIG0_ENDIAN(ENDIAN_MODE_NO_SWAP) |
587 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_STREAM(elements[idx].vertex_buffer_index) |
588 VIVS_NFE_GENERIC_ATTRIB_CONFIG0_START(elements[idx].src_offset);
589 cs->NFE_GENERIC_ATTRIB_CONFIG1[idx] =
590 COND(nonconsecutive, VIVS_NFE_GENERIC_ATTRIB_CONFIG1_NONCONSECUTIVE) |
591 VIVS_NFE_GENERIC_ATTRIB_CONFIG1_END(end_offset - start_offset);
592 }
593 cs->NFE_GENERIC_ATTRIB_SCALE[idx] = 0x3f800000; /* 1 for integer, 1.0 for float */
594 }
595
596 return cs;
597 }
598
599 static void
600 etna_vertex_elements_state_delete(struct pipe_context *pctx, void *ve)
601 {
602 FREE(ve);
603 }
604
605 static void
606 etna_vertex_elements_state_bind(struct pipe_context *pctx, void *ve)
607 {
608 struct etna_context *ctx = etna_context(pctx);
609
610 ctx->vertex_elements = ve;
611 ctx->dirty |= ETNA_DIRTY_VERTEX_ELEMENTS;
612 }
613
614 static bool
615 etna_update_ts_config(struct etna_context *ctx)
616 {
617 uint32_t new_ts_config = ctx->framebuffer.TS_MEM_CONFIG;
618
619 if (ctx->framebuffer_s.nr_cbufs > 0) {
620 struct etna_surface *c_surf = etna_surface(ctx->framebuffer_s.cbufs[0]);
621
622 if(c_surf->level->ts_size && c_surf->level->ts_valid) {
623 new_ts_config |= VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
624 } else {
625 new_ts_config &= ~VIVS_TS_MEM_CONFIG_COLOR_FAST_CLEAR;
626 }
627 }
628
629 if (ctx->framebuffer_s.zsbuf) {
630 struct etna_surface *zs_surf = etna_surface(ctx->framebuffer_s.zsbuf);
631
632 if(zs_surf->level->ts_size && zs_surf->level->ts_valid) {
633 new_ts_config |= VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
634 } else {
635 new_ts_config &= ~VIVS_TS_MEM_CONFIG_DEPTH_FAST_CLEAR;
636 }
637 }
638
639 if (new_ts_config != ctx->framebuffer.TS_MEM_CONFIG ||
640 (ctx->dirty & ETNA_DIRTY_FRAMEBUFFER)) {
641 ctx->framebuffer.TS_MEM_CONFIG = new_ts_config;
642 ctx->dirty |= ETNA_DIRTY_TS;
643 }
644
645 ctx->dirty &= ~ETNA_DIRTY_DERIVE_TS;
646
647 return true;
648 }
649
650 struct etna_state_updater {
651 bool (*update)(struct etna_context *ctx);
652 uint32_t dirty;
653 };
654
655 static const struct etna_state_updater etna_state_updates[] = {
656 {
657 etna_shader_update_vertex, ETNA_DIRTY_SHADER | ETNA_DIRTY_VERTEX_ELEMENTS,
658 },
659 {
660 etna_shader_link, ETNA_DIRTY_SHADER,
661 },
662 {
663 etna_update_blend, ETNA_DIRTY_BLEND | ETNA_DIRTY_FRAMEBUFFER
664 },
665 {
666 etna_update_blend_color, ETNA_DIRTY_BLEND_COLOR | ETNA_DIRTY_FRAMEBUFFER,
667 },
668 {
669 etna_update_ts_config, ETNA_DIRTY_DERIVE_TS,
670 }
671 };
672
673 bool
674 etna_state_update(struct etna_context *ctx)
675 {
676 for (unsigned int i = 0; i < ARRAY_SIZE(etna_state_updates); i++)
677 if (ctx->dirty & etna_state_updates[i].dirty)
678 if (!etna_state_updates[i].update(ctx))
679 return false;
680
681 return true;
682 }
683
684 void
685 etna_state_init(struct pipe_context *pctx)
686 {
687 pctx->set_blend_color = etna_set_blend_color;
688 pctx->set_stencil_ref = etna_set_stencil_ref;
689 pctx->set_clip_state = etna_set_clip_state;
690 pctx->set_sample_mask = etna_set_sample_mask;
691 pctx->set_constant_buffer = etna_set_constant_buffer;
692 pctx->set_framebuffer_state = etna_set_framebuffer_state;
693 pctx->set_polygon_stipple = etna_set_polygon_stipple;
694 pctx->set_scissor_states = etna_set_scissor_states;
695 pctx->set_viewport_states = etna_set_viewport_states;
696
697 pctx->set_vertex_buffers = etna_set_vertex_buffers;
698
699 pctx->bind_blend_state = etna_blend_state_bind;
700 pctx->delete_blend_state = etna_blend_state_delete;
701
702 pctx->bind_rasterizer_state = etna_rasterizer_state_bind;
703 pctx->delete_rasterizer_state = etna_rasterizer_state_delete;
704
705 pctx->bind_depth_stencil_alpha_state = etna_zsa_state_bind;
706 pctx->delete_depth_stencil_alpha_state = etna_zsa_state_delete;
707
708 pctx->create_vertex_elements_state = etna_vertex_elements_state_create;
709 pctx->delete_vertex_elements_state = etna_vertex_elements_state_delete;
710 pctx->bind_vertex_elements_state = etna_vertex_elements_state_bind;
711 }