1 /* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
2
3 /*
4 * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice (including the next
14 * paragraph) shall be included in all copies or substantial portions of the
15 * Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
23 * SOFTWARE.
24 *
25 * Authors:
26 * Rob Clark <robclark@freedesktop.org>
27 */
28
29 #include "pipe/p_state.h"
30 #include "util/u_string.h"
31 #include "util/u_memory.h"
32 #include "util/u_helpers.h"
33 #include "util/u_format.h"
34 #include "util/u_viewport.h"
35
36 #include "freedreno_resource.h"
37 #include "freedreno_query_hw.h"
38
39 #include "fd4_emit.h"
40 #include "fd4_blend.h"
41 #include "fd4_context.h"
42 #include "fd4_program.h"
43 #include "fd4_rasterizer.h"
44 #include "fd4_texture.h"
45 #include "fd4_format.h"
46 #include "fd4_zsa.h"
47
48 static const enum adreno_state_block sb[] = {
49 [SHADER_VERTEX] = SB_VERT_SHADER,
50 [SHADER_FRAGMENT] = SB_FRAG_SHADER,
51 };
52
53 /* regid: base const register
54 * prsc or dwords: buffer containing constant values
55 * sizedwords: size of const value buffer
56 */
57 static void
fd4_emit_const(struct fd_ringbuffer * ring,enum shader_t type,uint32_t regid,uint32_t offset,uint32_t sizedwords,const uint32_t * dwords,struct pipe_resource * prsc)58 fd4_emit_const(struct fd_ringbuffer *ring, enum shader_t type,
59 uint32_t regid, uint32_t offset, uint32_t sizedwords,
60 const uint32_t *dwords, struct pipe_resource *prsc)
61 {
62 uint32_t i, sz;
63 enum adreno_state_src src;
64
65 debug_assert((regid % 4) == 0);
66 debug_assert((sizedwords % 4) == 0);
67
68 if (prsc) {
69 sz = 0;
70 src = 0x2; // TODO ??
71 } else {
72 sz = sizedwords;
73 src = SS_DIRECT;
74 }
75
76 OUT_PKT3(ring, CP_LOAD_STATE, 2 + sz);
77 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(regid/4) |
78 CP_LOAD_STATE_0_STATE_SRC(src) |
79 CP_LOAD_STATE_0_STATE_BLOCK(sb[type]) |
80 CP_LOAD_STATE_0_NUM_UNIT(sizedwords/4));
81 if (prsc) {
82 struct fd_bo *bo = fd_resource(prsc)->bo;
83 OUT_RELOC(ring, bo, offset,
84 CP_LOAD_STATE_1_STATE_TYPE(ST_CONSTANTS), 0);
85 } else {
86 OUT_RING(ring, CP_LOAD_STATE_1_EXT_SRC_ADDR(0) |
87 CP_LOAD_STATE_1_STATE_TYPE(ST_CONSTANTS));
88 dwords = (uint32_t *)&((uint8_t *)dwords)[offset];
89 }
90 for (i = 0; i < sz; i++) {
91 OUT_RING(ring, dwords[i]);
92 }
93 }
94
95 static void
fd4_emit_const_bo(struct fd_ringbuffer * ring,enum shader_t type,boolean write,uint32_t regid,uint32_t num,struct pipe_resource ** prscs,uint32_t * offsets)96 fd4_emit_const_bo(struct fd_ringbuffer *ring, enum shader_t type, boolean write,
97 uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets)
98 {
99 uint32_t anum = align(num, 4);
100 uint32_t i;
101
102 debug_assert((regid % 4) == 0);
103
104 OUT_PKT3(ring, CP_LOAD_STATE, 2 + anum);
105 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(regid/4) |
106 CP_LOAD_STATE_0_STATE_SRC(SS_DIRECT) |
107 CP_LOAD_STATE_0_STATE_BLOCK(sb[type]) |
108 CP_LOAD_STATE_0_NUM_UNIT(anum/4));
109 OUT_RING(ring, CP_LOAD_STATE_1_EXT_SRC_ADDR(0) |
110 CP_LOAD_STATE_1_STATE_TYPE(ST_CONSTANTS));
111
112 for (i = 0; i < num; i++) {
113 if (prscs[i]) {
114 if (write) {
115 OUT_RELOCW(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
116 } else {
117 OUT_RELOC(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
118 }
119 } else {
120 OUT_RING(ring, 0xbad00000 | (i << 16));
121 }
122 }
123
124 for (; i < anum; i++)
125 OUT_RING(ring, 0xffffffff);
126 }
127
128 static void
emit_textures(struct fd_context * ctx,struct fd_ringbuffer * ring,enum adreno_state_block sb,struct fd_texture_stateobj * tex,const struct ir3_shader_variant * v)129 emit_textures(struct fd_context *ctx, struct fd_ringbuffer *ring,
130 enum adreno_state_block sb, struct fd_texture_stateobj *tex,
131 const struct ir3_shader_variant *v)
132 {
133 static const uint32_t bcolor_reg[] = {
134 [SB_VERT_TEX] = REG_A4XX_TPL1_TP_VS_BORDER_COLOR_BASE_ADDR,
135 [SB_FRAG_TEX] = REG_A4XX_TPL1_TP_FS_BORDER_COLOR_BASE_ADDR,
136 };
137 struct fd4_context *fd4_ctx = fd4_context(ctx);
138 bool needs_border = false;
139 unsigned i;
140
141 if (tex->num_samplers > 0) {
142 int num_samplers;
143
144 /* not sure if this is an a420.0 workaround, but we seem
145 * to need to emit these in pairs.. emit a final dummy
146 * entry if odd # of samplers:
147 */
148 num_samplers = align(tex->num_samplers, 2);
149
150 /* output sampler state: */
151 OUT_PKT3(ring, CP_LOAD_STATE, 2 + (2 * num_samplers));
152 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(0) |
153 CP_LOAD_STATE_0_STATE_SRC(SS_DIRECT) |
154 CP_LOAD_STATE_0_STATE_BLOCK(sb) |
155 CP_LOAD_STATE_0_NUM_UNIT(num_samplers));
156 OUT_RING(ring, CP_LOAD_STATE_1_STATE_TYPE(ST_SHADER) |
157 CP_LOAD_STATE_1_EXT_SRC_ADDR(0));
158 for (i = 0; i < tex->num_samplers; i++) {
159 static const struct fd4_sampler_stateobj dummy_sampler = {};
160 const struct fd4_sampler_stateobj *sampler = tex->samplers[i] ?
161 fd4_sampler_stateobj(tex->samplers[i]) :
162 &dummy_sampler;
163 OUT_RING(ring, sampler->texsamp0);
164 OUT_RING(ring, sampler->texsamp1);
165
166 needs_border |= sampler->needs_border;
167 }
168
169 for (; i < num_samplers; i++) {
170 OUT_RING(ring, 0x00000000);
171 OUT_RING(ring, 0x00000000);
172 }
173 }
174
175 if (tex->num_textures > 0) {
176 unsigned num_textures = tex->num_textures + v->astc_srgb.count;
177
178 /* emit texture state: */
179 OUT_PKT3(ring, CP_LOAD_STATE, 2 + (8 * num_textures));
180 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(0) |
181 CP_LOAD_STATE_0_STATE_SRC(SS_DIRECT) |
182 CP_LOAD_STATE_0_STATE_BLOCK(sb) |
183 CP_LOAD_STATE_0_NUM_UNIT(num_textures));
184 OUT_RING(ring, CP_LOAD_STATE_1_STATE_TYPE(ST_CONSTANTS) |
185 CP_LOAD_STATE_1_EXT_SRC_ADDR(0));
186 for (i = 0; i < tex->num_textures; i++) {
187 static const struct fd4_pipe_sampler_view dummy_view = {};
188 const struct fd4_pipe_sampler_view *view = tex->textures[i] ?
189 fd4_pipe_sampler_view(tex->textures[i]) :
190 &dummy_view;
191
192 OUT_RING(ring, view->texconst0);
193 OUT_RING(ring, view->texconst1);
194 OUT_RING(ring, view->texconst2);
195 OUT_RING(ring, view->texconst3);
196 if (view->base.texture) {
197 struct fd_resource *rsc = fd_resource(view->base.texture);
198 OUT_RELOC(ring, rsc->bo, view->offset, view->texconst4, 0);
199 } else {
200 OUT_RING(ring, 0x00000000);
201 }
202 OUT_RING(ring, 0x00000000);
203 OUT_RING(ring, 0x00000000);
204 OUT_RING(ring, 0x00000000);
205 }
206
207 for (i = 0; i < v->astc_srgb.count; i++) {
208 static const struct fd4_pipe_sampler_view dummy_view = {};
209 const struct fd4_pipe_sampler_view *view;
210 unsigned idx = v->astc_srgb.orig_idx[i];
211
212 view = tex->textures[idx] ?
213 fd4_pipe_sampler_view(tex->textures[idx]) :
214 &dummy_view;
215
216 debug_assert(view->texconst0 & A4XX_TEX_CONST_0_SRGB);
217
218 OUT_RING(ring, view->texconst0 & ~A4XX_TEX_CONST_0_SRGB);
219 OUT_RING(ring, view->texconst1);
220 OUT_RING(ring, view->texconst2);
221 OUT_RING(ring, view->texconst3);
222 if (view->base.texture) {
223 struct fd_resource *rsc = fd_resource(view->base.texture);
224 OUT_RELOC(ring, rsc->bo, view->offset, view->texconst4, 0);
225 } else {
226 OUT_RING(ring, 0x00000000);
227 }
228 OUT_RING(ring, 0x00000000);
229 OUT_RING(ring, 0x00000000);
230 OUT_RING(ring, 0x00000000);
231 }
232 } else {
233 debug_assert(v->astc_srgb.count == 0);
234 }
235
236 if (needs_border) {
237 unsigned off;
238 void *ptr;
239
240 u_upload_alloc(fd4_ctx->border_color_uploader,
241 0, BORDER_COLOR_UPLOAD_SIZE,
242 BORDER_COLOR_UPLOAD_SIZE, &off,
243 &fd4_ctx->border_color_buf,
244 &ptr);
245
246 fd_setup_border_colors(tex, ptr, 0);
247 OUT_PKT0(ring, bcolor_reg[sb], 1);
248 OUT_RELOC(ring, fd_resource(fd4_ctx->border_color_buf)->bo, off, 0, 0);
249
250 u_upload_unmap(fd4_ctx->border_color_uploader);
251 }
252 }
253
254 /* emit texture state for mem->gmem restore operation.. eventually it would
255 * be good to get rid of this and use normal CSO/etc state for more of these
256 * special cases..
257 */
258 void
fd4_emit_gmem_restore_tex(struct fd_ringbuffer * ring,unsigned nr_bufs,struct pipe_surface ** bufs)259 fd4_emit_gmem_restore_tex(struct fd_ringbuffer *ring, unsigned nr_bufs,
260 struct pipe_surface **bufs)
261 {
262 unsigned char mrt_comp[A4XX_MAX_RENDER_TARGETS];
263 int i;
264
265 for (i = 0; i < A4XX_MAX_RENDER_TARGETS; i++) {
266 mrt_comp[i] = (i < nr_bufs) ? 0xf : 0;
267 }
268
269 /* output sampler state: */
270 OUT_PKT3(ring, CP_LOAD_STATE, 2 + (2 * nr_bufs));
271 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(0) |
272 CP_LOAD_STATE_0_STATE_SRC(SS_DIRECT) |
273 CP_LOAD_STATE_0_STATE_BLOCK(SB_FRAG_TEX) |
274 CP_LOAD_STATE_0_NUM_UNIT(nr_bufs));
275 OUT_RING(ring, CP_LOAD_STATE_1_STATE_TYPE(ST_SHADER) |
276 CP_LOAD_STATE_1_EXT_SRC_ADDR(0));
277 for (i = 0; i < nr_bufs; i++) {
278 OUT_RING(ring, A4XX_TEX_SAMP_0_XY_MAG(A4XX_TEX_NEAREST) |
279 A4XX_TEX_SAMP_0_XY_MIN(A4XX_TEX_NEAREST) |
280 A4XX_TEX_SAMP_0_WRAP_S(A4XX_TEX_CLAMP_TO_EDGE) |
281 A4XX_TEX_SAMP_0_WRAP_T(A4XX_TEX_CLAMP_TO_EDGE) |
282 A4XX_TEX_SAMP_0_WRAP_R(A4XX_TEX_REPEAT));
283 OUT_RING(ring, 0x00000000);
284 }
285
286 /* emit texture state: */
287 OUT_PKT3(ring, CP_LOAD_STATE, 2 + (8 * nr_bufs));
288 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(0) |
289 CP_LOAD_STATE_0_STATE_SRC(SS_DIRECT) |
290 CP_LOAD_STATE_0_STATE_BLOCK(SB_FRAG_TEX) |
291 CP_LOAD_STATE_0_NUM_UNIT(nr_bufs));
292 OUT_RING(ring, CP_LOAD_STATE_1_STATE_TYPE(ST_CONSTANTS) |
293 CP_LOAD_STATE_1_EXT_SRC_ADDR(0));
294 for (i = 0; i < nr_bufs; i++) {
295 if (bufs[i]) {
296 struct fd_resource *rsc = fd_resource(bufs[i]->texture);
297 enum pipe_format format = fd_gmem_restore_format(bufs[i]->format);
298
299 /* The restore blit_zs shader expects stencil in sampler 0,
300 * and depth in sampler 1
301 */
302 if (rsc->stencil && (i == 0)) {
303 rsc = rsc->stencil;
304 format = fd_gmem_restore_format(rsc->base.b.format);
305 }
306
307 /* note: PIPE_BUFFER disallowed for surfaces */
308 unsigned lvl = bufs[i]->u.tex.level;
309 struct fd_resource_slice *slice = fd_resource_slice(rsc, lvl);
310 unsigned offset = fd_resource_offset(rsc, lvl, bufs[i]->u.tex.first_layer);
311
312 /* z32 restore is accomplished using depth write. If there is
313 * no stencil component (ie. PIPE_FORMAT_Z32_FLOAT_S8X24_UINT)
314 * then no render target:
315 *
316 * (The same applies for z32_s8x24, since for stencil sampler
317 * state the above 'if' will replace 'format' with s8)
318 */
319 if ((format == PIPE_FORMAT_Z32_FLOAT) ||
320 (format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT))
321 mrt_comp[i] = 0;
322
323 debug_assert(bufs[i]->u.tex.first_layer == bufs[i]->u.tex.last_layer);
324
325 OUT_RING(ring, A4XX_TEX_CONST_0_FMT(fd4_pipe2tex(format)) |
326 A4XX_TEX_CONST_0_TYPE(A4XX_TEX_2D) |
327 fd4_tex_swiz(format, PIPE_SWIZZLE_X, PIPE_SWIZZLE_Y,
328 PIPE_SWIZZLE_Z, PIPE_SWIZZLE_W));
329 OUT_RING(ring, A4XX_TEX_CONST_1_WIDTH(bufs[i]->width) |
330 A4XX_TEX_CONST_1_HEIGHT(bufs[i]->height));
331 OUT_RING(ring, A4XX_TEX_CONST_2_PITCH(slice->pitch * rsc->cpp) |
332 A4XX_TEX_CONST_2_FETCHSIZE(fd4_pipe2fetchsize(format)));
333 OUT_RING(ring, 0x00000000);
334 OUT_RELOC(ring, rsc->bo, offset, 0, 0);
335 OUT_RING(ring, 0x00000000);
336 OUT_RING(ring, 0x00000000);
337 OUT_RING(ring, 0x00000000);
338 } else {
339 OUT_RING(ring, A4XX_TEX_CONST_0_FMT(0) |
340 A4XX_TEX_CONST_0_TYPE(A4XX_TEX_2D) |
341 A4XX_TEX_CONST_0_SWIZ_X(A4XX_TEX_ONE) |
342 A4XX_TEX_CONST_0_SWIZ_Y(A4XX_TEX_ONE) |
343 A4XX_TEX_CONST_0_SWIZ_Z(A4XX_TEX_ONE) |
344 A4XX_TEX_CONST_0_SWIZ_W(A4XX_TEX_ONE));
345 OUT_RING(ring, A4XX_TEX_CONST_1_WIDTH(0) |
346 A4XX_TEX_CONST_1_HEIGHT(0));
347 OUT_RING(ring, A4XX_TEX_CONST_2_PITCH(0));
348 OUT_RING(ring, 0x00000000);
349 OUT_RING(ring, 0x00000000);
350 OUT_RING(ring, 0x00000000);
351 OUT_RING(ring, 0x00000000);
352 OUT_RING(ring, 0x00000000);
353 }
354 }
355
356 OUT_PKT0(ring, REG_A4XX_RB_RENDER_COMPONENTS, 1);
357 OUT_RING(ring, A4XX_RB_RENDER_COMPONENTS_RT0(mrt_comp[0]) |
358 A4XX_RB_RENDER_COMPONENTS_RT1(mrt_comp[1]) |
359 A4XX_RB_RENDER_COMPONENTS_RT2(mrt_comp[2]) |
360 A4XX_RB_RENDER_COMPONENTS_RT3(mrt_comp[3]) |
361 A4XX_RB_RENDER_COMPONENTS_RT4(mrt_comp[4]) |
362 A4XX_RB_RENDER_COMPONENTS_RT5(mrt_comp[5]) |
363 A4XX_RB_RENDER_COMPONENTS_RT6(mrt_comp[6]) |
364 A4XX_RB_RENDER_COMPONENTS_RT7(mrt_comp[7]));
365 }
366
367 void
fd4_emit_vertex_bufs(struct fd_ringbuffer * ring,struct fd4_emit * emit)368 fd4_emit_vertex_bufs(struct fd_ringbuffer *ring, struct fd4_emit *emit)
369 {
370 int32_t i, j, last = -1;
371 uint32_t total_in = 0;
372 const struct fd_vertex_state *vtx = emit->vtx;
373 const struct ir3_shader_variant *vp = fd4_emit_get_vp(emit);
374 unsigned vertex_regid = regid(63, 0);
375 unsigned instance_regid = regid(63, 0);
376 unsigned vtxcnt_regid = regid(63, 0);
377
378 /* Note that sysvals come *after* normal inputs: */
379 for (i = 0; i < vp->inputs_count; i++) {
380 if (!vp->inputs[i].compmask)
381 continue;
382 if (vp->inputs[i].sysval) {
383 switch(vp->inputs[i].slot) {
384 case SYSTEM_VALUE_BASE_VERTEX:
385 /* handled elsewhere */
386 break;
387 case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
388 vertex_regid = vp->inputs[i].regid;
389 break;
390 case SYSTEM_VALUE_INSTANCE_ID:
391 instance_regid = vp->inputs[i].regid;
392 break;
393 case SYSTEM_VALUE_VERTEX_CNT:
394 vtxcnt_regid = vp->inputs[i].regid;
395 break;
396 default:
397 unreachable("invalid system value");
398 break;
399 }
400 } else if (i < vtx->vtx->num_elements) {
401 last = i;
402 }
403 }
404
405 for (i = 0, j = 0; i <= last; i++) {
406 assert(!vp->inputs[i].sysval);
407 if (vp->inputs[i].compmask) {
408 struct pipe_vertex_element *elem = &vtx->vtx->pipe[i];
409 const struct pipe_vertex_buffer *vb =
410 &vtx->vertexbuf.vb[elem->vertex_buffer_index];
411 struct fd_resource *rsc = fd_resource(vb->buffer);
412 enum pipe_format pfmt = elem->src_format;
413 enum a4xx_vtx_fmt fmt = fd4_pipe2vtx(pfmt);
414 bool switchnext = (i != last) ||
415 (vertex_regid != regid(63, 0)) ||
416 (instance_regid != regid(63, 0)) ||
417 (vtxcnt_regid != regid(63, 0));
418 bool isint = util_format_is_pure_integer(pfmt);
419 uint32_t fs = util_format_get_blocksize(pfmt);
420 uint32_t off = vb->buffer_offset + elem->src_offset;
421 uint32_t size = fd_bo_size(rsc->bo) - off;
422 debug_assert(fmt != ~0);
423
424 OUT_PKT0(ring, REG_A4XX_VFD_FETCH(j), 4);
425 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_0_FETCHSIZE(fs - 1) |
426 A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE(vb->stride) |
427 COND(elem->instance_divisor, A4XX_VFD_FETCH_INSTR_0_INSTANCED) |
428 COND(switchnext, A4XX_VFD_FETCH_INSTR_0_SWITCHNEXT));
429 OUT_RELOC(ring, rsc->bo, off, 0, 0);
430 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_2_SIZE(size));
431 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_3_STEPRATE(MAX2(1, elem->instance_divisor)));
432
433 OUT_PKT0(ring, REG_A4XX_VFD_DECODE_INSTR(j), 1);
434 OUT_RING(ring, A4XX_VFD_DECODE_INSTR_CONSTFILL |
435 A4XX_VFD_DECODE_INSTR_WRITEMASK(vp->inputs[i].compmask) |
436 A4XX_VFD_DECODE_INSTR_FORMAT(fmt) |
437 A4XX_VFD_DECODE_INSTR_SWAP(fd4_pipe2swap(pfmt)) |
438 A4XX_VFD_DECODE_INSTR_REGID(vp->inputs[i].regid) |
439 A4XX_VFD_DECODE_INSTR_SHIFTCNT(fs) |
440 A4XX_VFD_DECODE_INSTR_LASTCOMPVALID |
441 COND(isint, A4XX_VFD_DECODE_INSTR_INT) |
442 COND(switchnext, A4XX_VFD_DECODE_INSTR_SWITCHNEXT));
443
444 total_in += vp->inputs[i].ncomp;
445 j++;
446 }
447 }
448
449 /* hw doesn't like to be configured for zero vbo's, it seems: */
450 if (last < 0) {
451 /* just recycle the shader bo, we just need to point to *something*
452 * valid:
453 */
454 struct fd_bo *dummy_vbo = vp->bo;
455 bool switchnext = (vertex_regid != regid(63, 0)) ||
456 (instance_regid != regid(63, 0)) ||
457 (vtxcnt_regid != regid(63, 0));
458
459 OUT_PKT0(ring, REG_A4XX_VFD_FETCH(0), 4);
460 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_0_FETCHSIZE(0) |
461 A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE(0) |
462 COND(switchnext, A4XX_VFD_FETCH_INSTR_0_SWITCHNEXT));
463 OUT_RELOC(ring, dummy_vbo, 0, 0, 0);
464 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_2_SIZE(1));
465 OUT_RING(ring, A4XX_VFD_FETCH_INSTR_3_STEPRATE(1));
466
467 OUT_PKT0(ring, REG_A4XX_VFD_DECODE_INSTR(0), 1);
468 OUT_RING(ring, A4XX_VFD_DECODE_INSTR_CONSTFILL |
469 A4XX_VFD_DECODE_INSTR_WRITEMASK(0x1) |
470 A4XX_VFD_DECODE_INSTR_FORMAT(VFMT4_8_UNORM) |
471 A4XX_VFD_DECODE_INSTR_SWAP(XYZW) |
472 A4XX_VFD_DECODE_INSTR_REGID(regid(0,0)) |
473 A4XX_VFD_DECODE_INSTR_SHIFTCNT(1) |
474 A4XX_VFD_DECODE_INSTR_LASTCOMPVALID |
475 COND(switchnext, A4XX_VFD_DECODE_INSTR_SWITCHNEXT));
476
477 total_in = 1;
478 j = 1;
479 }
480
481 OUT_PKT0(ring, REG_A4XX_VFD_CONTROL_0, 5);
482 OUT_RING(ring, A4XX_VFD_CONTROL_0_TOTALATTRTOVS(total_in) |
483 0xa0000 | /* XXX */
484 A4XX_VFD_CONTROL_0_STRMDECINSTRCNT(j) |
485 A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT(j));
486 OUT_RING(ring, A4XX_VFD_CONTROL_1_MAXSTORAGE(129) | // XXX
487 A4XX_VFD_CONTROL_1_REGID4VTX(vertex_regid) |
488 A4XX_VFD_CONTROL_1_REGID4INST(instance_regid));
489 OUT_RING(ring, 0x00000000); /* XXX VFD_CONTROL_2 */
490 OUT_RING(ring, A4XX_VFD_CONTROL_3_REGID_VTXCNT(vtxcnt_regid));
491 OUT_RING(ring, 0x00000000); /* XXX VFD_CONTROL_4 */
492
493 /* cache invalidate, otherwise vertex fetch could see
494 * stale vbo contents:
495 */
496 OUT_PKT0(ring, REG_A4XX_UCHE_INVALIDATE0, 2);
497 OUT_RING(ring, 0x00000000);
498 OUT_RING(ring, 0x00000012);
499 }
500
501 void
fd4_emit_state(struct fd_context * ctx,struct fd_ringbuffer * ring,struct fd4_emit * emit)502 fd4_emit_state(struct fd_context *ctx, struct fd_ringbuffer *ring,
503 struct fd4_emit *emit)
504 {
505 const struct ir3_shader_variant *vp = fd4_emit_get_vp(emit);
506 const struct ir3_shader_variant *fp = fd4_emit_get_fp(emit);
507 uint32_t dirty = emit->dirty;
508
509 emit_marker(ring, 5);
510
511 if ((dirty & FD_DIRTY_FRAMEBUFFER) && !emit->key.binning_pass) {
512 struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer;
513 unsigned char mrt_comp[A4XX_MAX_RENDER_TARGETS] = {0};
514
515 for (unsigned i = 0; i < A4XX_MAX_RENDER_TARGETS; i++) {
516 mrt_comp[i] = ((i < pfb->nr_cbufs) && pfb->cbufs[i]) ? 0xf : 0;
517 }
518
519 OUT_PKT0(ring, REG_A4XX_RB_RENDER_COMPONENTS, 1);
520 OUT_RING(ring, A4XX_RB_RENDER_COMPONENTS_RT0(mrt_comp[0]) |
521 A4XX_RB_RENDER_COMPONENTS_RT1(mrt_comp[1]) |
522 A4XX_RB_RENDER_COMPONENTS_RT2(mrt_comp[2]) |
523 A4XX_RB_RENDER_COMPONENTS_RT3(mrt_comp[3]) |
524 A4XX_RB_RENDER_COMPONENTS_RT4(mrt_comp[4]) |
525 A4XX_RB_RENDER_COMPONENTS_RT5(mrt_comp[5]) |
526 A4XX_RB_RENDER_COMPONENTS_RT6(mrt_comp[6]) |
527 A4XX_RB_RENDER_COMPONENTS_RT7(mrt_comp[7]));
528 }
529
530 if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_FRAMEBUFFER)) {
531 struct fd4_zsa_stateobj *zsa = fd4_zsa_stateobj(ctx->zsa);
532 struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer;
533 uint32_t rb_alpha_control = zsa->rb_alpha_control;
534
535 if (util_format_is_pure_integer(pipe_surface_format(pfb->cbufs[0])))
536 rb_alpha_control &= ~A4XX_RB_ALPHA_CONTROL_ALPHA_TEST;
537
538 OUT_PKT0(ring, REG_A4XX_RB_ALPHA_CONTROL, 1);
539 OUT_RING(ring, rb_alpha_control);
540
541 OUT_PKT0(ring, REG_A4XX_RB_STENCIL_CONTROL, 2);
542 OUT_RING(ring, zsa->rb_stencil_control);
543 OUT_RING(ring, zsa->rb_stencil_control2);
544 }
545
546 if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_STENCIL_REF)) {
547 struct fd4_zsa_stateobj *zsa = fd4_zsa_stateobj(ctx->zsa);
548 struct pipe_stencil_ref *sr = &ctx->stencil_ref;
549
550 OUT_PKT0(ring, REG_A4XX_RB_STENCILREFMASK, 2);
551 OUT_RING(ring, zsa->rb_stencilrefmask |
552 A4XX_RB_STENCILREFMASK_STENCILREF(sr->ref_value[0]));
553 OUT_RING(ring, zsa->rb_stencilrefmask_bf |
554 A4XX_RB_STENCILREFMASK_BF_STENCILREF(sr->ref_value[1]));
555 }
556
557 if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_RASTERIZER | FD_DIRTY_PROG)) {
558 struct fd4_zsa_stateobj *zsa = fd4_zsa_stateobj(ctx->zsa);
559 bool fragz = fp->has_kill | fp->writes_pos;
560 bool clamp = !ctx->rasterizer->depth_clip;
561
562 OUT_PKT0(ring, REG_A4XX_RB_DEPTH_CONTROL, 1);
563 OUT_RING(ring, zsa->rb_depth_control |
564 COND(clamp, A4XX_RB_DEPTH_CONTROL_Z_CLAMP_ENABLE) |
565 COND(fragz, A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE) |
566 COND(fragz && fp->frag_coord, A4XX_RB_DEPTH_CONTROL_FORCE_FRAGZ_TO_FS));
567
568 /* maybe this register/bitfield needs a better name.. this
569 * appears to be just disabling early-z
570 */
571 OUT_PKT0(ring, REG_A4XX_GRAS_ALPHA_CONTROL, 1);
572 OUT_RING(ring, zsa->gras_alpha_control |
573 COND(fragz, A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE) |
574 COND(fragz && fp->frag_coord, A4XX_GRAS_ALPHA_CONTROL_FORCE_FRAGZ_TO_FS));
575 }
576
577 if (dirty & FD_DIRTY_RASTERIZER) {
578 struct fd4_rasterizer_stateobj *rasterizer =
579 fd4_rasterizer_stateobj(ctx->rasterizer);
580
581 OUT_PKT0(ring, REG_A4XX_GRAS_SU_MODE_CONTROL, 1);
582 OUT_RING(ring, rasterizer->gras_su_mode_control |
583 A4XX_GRAS_SU_MODE_CONTROL_RENDERING_PASS);
584
585 OUT_PKT0(ring, REG_A4XX_GRAS_SU_POINT_MINMAX, 2);
586 OUT_RING(ring, rasterizer->gras_su_point_minmax);
587 OUT_RING(ring, rasterizer->gras_su_point_size);
588
589 OUT_PKT0(ring, REG_A4XX_GRAS_SU_POLY_OFFSET_SCALE, 2);
590 OUT_RING(ring, rasterizer->gras_su_poly_offset_scale);
591 OUT_RING(ring, rasterizer->gras_su_poly_offset_offset);
592
593 OUT_PKT0(ring, REG_A4XX_GRAS_CL_CLIP_CNTL, 1);
594 OUT_RING(ring, rasterizer->gras_cl_clip_cntl);
595 }
596
597 /* NOTE: since primitive_restart is not actually part of any
598 * state object, we need to make sure that we always emit
599 * PRIM_VTX_CNTL.. either that or be more clever and detect
600 * when it changes.
601 */
602 if (emit->info) {
603 const struct pipe_draw_info *info = emit->info;
604 struct fd4_rasterizer_stateobj *rast =
605 fd4_rasterizer_stateobj(ctx->rasterizer);
606 uint32_t val = rast->pc_prim_vtx_cntl;
607
608 if (info->indexed && info->primitive_restart)
609 val |= A4XX_PC_PRIM_VTX_CNTL_PRIMITIVE_RESTART;
610
611 val |= COND(vp->writes_psize, A4XX_PC_PRIM_VTX_CNTL_PSIZE);
612
613 if (fp->total_in > 0) {
614 uint32_t varout = align(fp->total_in, 16) / 16;
615 if (varout > 1)
616 varout = align(varout, 2);
617 val |= A4XX_PC_PRIM_VTX_CNTL_VAROUT(varout);
618 }
619
620 OUT_PKT0(ring, REG_A4XX_PC_PRIM_VTX_CNTL, 2);
621 OUT_RING(ring, val);
622 OUT_RING(ring, rast->pc_prim_vtx_cntl2);
623 }
624
625 if (dirty & FD_DIRTY_SCISSOR) {
626 struct pipe_scissor_state *scissor = fd_context_get_scissor(ctx);
627
628 OUT_PKT0(ring, REG_A4XX_GRAS_SC_WINDOW_SCISSOR_BR, 2);
629 OUT_RING(ring, A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X(scissor->maxx - 1) |
630 A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(scissor->maxy - 1));
631 OUT_RING(ring, A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X(scissor->minx) |
632 A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(scissor->miny));
633
634 ctx->batch->max_scissor.minx = MIN2(ctx->batch->max_scissor.minx, scissor->minx);
635 ctx->batch->max_scissor.miny = MIN2(ctx->batch->max_scissor.miny, scissor->miny);
636 ctx->batch->max_scissor.maxx = MAX2(ctx->batch->max_scissor.maxx, scissor->maxx);
637 ctx->batch->max_scissor.maxy = MAX2(ctx->batch->max_scissor.maxy, scissor->maxy);
638 }
639
640 if (dirty & FD_DIRTY_VIEWPORT) {
641 fd_wfi(ctx->batch, ring);
642 OUT_PKT0(ring, REG_A4XX_GRAS_CL_VPORT_XOFFSET_0, 6);
643 OUT_RING(ring, A4XX_GRAS_CL_VPORT_XOFFSET_0(ctx->viewport.translate[0]));
644 OUT_RING(ring, A4XX_GRAS_CL_VPORT_XSCALE_0(ctx->viewport.scale[0]));
645 OUT_RING(ring, A4XX_GRAS_CL_VPORT_YOFFSET_0(ctx->viewport.translate[1]));
646 OUT_RING(ring, A4XX_GRAS_CL_VPORT_YSCALE_0(ctx->viewport.scale[1]));
647 OUT_RING(ring, A4XX_GRAS_CL_VPORT_ZOFFSET_0(ctx->viewport.translate[2]));
648 OUT_RING(ring, A4XX_GRAS_CL_VPORT_ZSCALE_0(ctx->viewport.scale[2]));
649 }
650
651 if (dirty & (FD_DIRTY_VIEWPORT | FD_DIRTY_RASTERIZER | FD_DIRTY_FRAMEBUFFER)) {
652 float zmin, zmax;
653 int depth = 24;
654 if (ctx->batch->framebuffer.zsbuf) {
655 depth = util_format_get_component_bits(
656 pipe_surface_format(ctx->batch->framebuffer.zsbuf),
657 UTIL_FORMAT_COLORSPACE_ZS, 0);
658 }
659 util_viewport_zmin_zmax(&ctx->viewport, ctx->rasterizer->clip_halfz,
660 &zmin, &zmax);
661
662 OUT_PKT0(ring, REG_A4XX_RB_VPORT_Z_CLAMP(0), 2);
663 if (depth == 32) {
664 OUT_RING(ring, fui(zmin));
665 OUT_RING(ring, fui(zmax));
666 } else if (depth == 16) {
667 OUT_RING(ring, (uint32_t)(zmin * 0xffff));
668 OUT_RING(ring, (uint32_t)(zmax * 0xffff));
669 } else {
670 OUT_RING(ring, (uint32_t)(zmin * 0xffffff));
671 OUT_RING(ring, (uint32_t)(zmax * 0xffffff));
672 }
673 }
674
675 if (dirty & (FD_DIRTY_PROG | FD_DIRTY_FRAMEBUFFER)) {
676 struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer;
677 unsigned n = pfb->nr_cbufs;
678 /* if we have depth/stencil, we need at least on MRT: */
679 if (pfb->zsbuf)
680 n = MAX2(1, n);
681 fd4_program_emit(ring, emit, n, pfb->cbufs);
682 }
683
684 if (emit->prog == &ctx->prog) { /* evil hack to deal sanely with clear path */
685 ir3_emit_consts(vp, ring, ctx, emit->info, dirty);
686 if (!emit->key.binning_pass)
687 ir3_emit_consts(fp, ring, ctx, emit->info, dirty);
688 }
689
690 if ((dirty & FD_DIRTY_BLEND)) {
691 struct fd4_blend_stateobj *blend = fd4_blend_stateobj(ctx->blend);
692 uint32_t i;
693
694 for (i = 0; i < A4XX_MAX_RENDER_TARGETS; i++) {
695 enum pipe_format format = pipe_surface_format(
696 ctx->batch->framebuffer.cbufs[i]);
697 bool is_int = util_format_is_pure_integer(format);
698 bool has_alpha = util_format_has_alpha(format);
699 uint32_t control = blend->rb_mrt[i].control;
700 uint32_t blend_control = blend->rb_mrt[i].blend_control_alpha;
701
702 if (is_int) {
703 control &= A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK;
704 control |= A4XX_RB_MRT_CONTROL_ROP_CODE(ROP_COPY);
705 }
706
707 if (has_alpha) {
708 blend_control |= blend->rb_mrt[i].blend_control_rgb;
709 } else {
710 blend_control |= blend->rb_mrt[i].blend_control_no_alpha_rgb;
711 control &= ~A4XX_RB_MRT_CONTROL_BLEND2;
712 }
713
714 OUT_PKT0(ring, REG_A4XX_RB_MRT_CONTROL(i), 1);
715 OUT_RING(ring, control);
716
717 OUT_PKT0(ring, REG_A4XX_RB_MRT_BLEND_CONTROL(i), 1);
718 OUT_RING(ring, blend_control);
719 }
720
721 OUT_PKT0(ring, REG_A4XX_RB_FS_OUTPUT, 1);
722 OUT_RING(ring, blend->rb_fs_output |
723 A4XX_RB_FS_OUTPUT_SAMPLE_MASK(0xffff));
724 }
725
726 if (dirty & FD_DIRTY_BLEND_COLOR) {
727 struct pipe_blend_color *bcolor = &ctx->blend_color;
728
729 OUT_PKT0(ring, REG_A4XX_RB_BLEND_RED, 8);
730 OUT_RING(ring, A4XX_RB_BLEND_RED_FLOAT(bcolor->color[0]) |
731 A4XX_RB_BLEND_RED_UINT(bcolor->color[0] * 0xff) |
732 A4XX_RB_BLEND_RED_SINT(bcolor->color[0] * 0x7f));
733 OUT_RING(ring, A4XX_RB_BLEND_RED_F32(bcolor->color[0]));
734 OUT_RING(ring, A4XX_RB_BLEND_GREEN_FLOAT(bcolor->color[1]) |
735 A4XX_RB_BLEND_GREEN_UINT(bcolor->color[1] * 0xff) |
736 A4XX_RB_BLEND_GREEN_SINT(bcolor->color[1] * 0x7f));
737 OUT_RING(ring, A4XX_RB_BLEND_RED_F32(bcolor->color[1]));
738 OUT_RING(ring, A4XX_RB_BLEND_BLUE_FLOAT(bcolor->color[2]) |
739 A4XX_RB_BLEND_BLUE_UINT(bcolor->color[2] * 0xff) |
740 A4XX_RB_BLEND_BLUE_SINT(bcolor->color[2] * 0x7f));
741 OUT_RING(ring, A4XX_RB_BLEND_BLUE_F32(bcolor->color[2]));
742 OUT_RING(ring, A4XX_RB_BLEND_ALPHA_FLOAT(bcolor->color[3]) |
743 A4XX_RB_BLEND_ALPHA_UINT(bcolor->color[3] * 0xff) |
744 A4XX_RB_BLEND_ALPHA_SINT(bcolor->color[3] * 0x7f));
745 OUT_RING(ring, A4XX_RB_BLEND_ALPHA_F32(bcolor->color[3]));
746 }
747
748 if (dirty & FD_DIRTY_VERTTEX) {
749 if (vp->has_samp)
750 emit_textures(ctx, ring, SB_VERT_TEX, &ctx->verttex, vp);
751 else
752 dirty &= ~FD_DIRTY_VERTTEX;
753 }
754
755 if (dirty & FD_DIRTY_FRAGTEX) {
756 if (fp->has_samp)
757 emit_textures(ctx, ring, SB_FRAG_TEX, &ctx->fragtex, fp);
758 else
759 dirty &= ~FD_DIRTY_FRAGTEX;
760 }
761
762 ctx->dirty &= ~dirty;
763 }
764
765 /* emit setup at begin of new cmdstream buffer (don't rely on previous
766 * state, there could have been a context switch between ioctls):
767 */
768 void
fd4_emit_restore(struct fd_batch * batch,struct fd_ringbuffer * ring)769 fd4_emit_restore(struct fd_batch *batch, struct fd_ringbuffer *ring)
770 {
771 struct fd_context *ctx = batch->ctx;
772 struct fd4_context *fd4_ctx = fd4_context(ctx);
773
774 OUT_PKT0(ring, REG_A4XX_RBBM_PERFCTR_CTL, 1);
775 OUT_RING(ring, 0x00000001);
776
777 OUT_PKT0(ring, REG_A4XX_GRAS_DEBUG_ECO_CONTROL, 1);
778 OUT_RING(ring, 0x00000000);
779
780 OUT_PKT0(ring, REG_A4XX_SP_MODE_CONTROL, 1);
781 OUT_RING(ring, 0x00000006);
782
783 OUT_PKT0(ring, REG_A4XX_TPL1_TP_MODE_CONTROL, 1);
784 OUT_RING(ring, 0x0000003a);
785
786 OUT_PKT0(ring, REG_A4XX_UNKNOWN_0D01, 1);
787 OUT_RING(ring, 0x00000001);
788
789 OUT_PKT0(ring, REG_A4XX_UNKNOWN_0E42, 1);
790 OUT_RING(ring, 0x00000000);
791
792 OUT_PKT0(ring, REG_A4XX_UCHE_CACHE_WAYS_VFD, 1);
793 OUT_RING(ring, 0x00000007);
794
795 OUT_PKT0(ring, REG_A4XX_UCHE_CACHE_MODE_CONTROL, 1);
796 OUT_RING(ring, 0x00000000);
797
798 OUT_PKT0(ring, REG_A4XX_UCHE_INVALIDATE0, 2);
799 OUT_RING(ring, 0x00000000);
800 OUT_RING(ring, 0x00000012);
801
802 OUT_PKT0(ring, REG_A4XX_HLSQ_MODE_CONTROL, 1);
803 OUT_RING(ring, 0x00000000);
804
805 OUT_PKT0(ring, REG_A4XX_UNKNOWN_0CC5, 1);
806 OUT_RING(ring, 0x00000006);
807
808 OUT_PKT0(ring, REG_A4XX_UNKNOWN_0CC6, 1);
809 OUT_RING(ring, 0x00000000);
810
811 OUT_PKT0(ring, REG_A4XX_UNKNOWN_0EC2, 1);
812 OUT_RING(ring, 0x00040000);
813
814 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2001, 1);
815 OUT_RING(ring, 0x00000000);
816
817 OUT_PKT3(ring, CP_INVALIDATE_STATE, 1);
818 OUT_RING(ring, 0x00001000);
819
820 OUT_PKT0(ring, REG_A4XX_UNKNOWN_20EF, 1);
821 OUT_RING(ring, 0x00000000);
822
823 OUT_PKT0(ring, REG_A4XX_RB_BLEND_RED, 4);
824 OUT_RING(ring, A4XX_RB_BLEND_RED_UINT(0) |
825 A4XX_RB_BLEND_RED_FLOAT(0.0));
826 OUT_RING(ring, A4XX_RB_BLEND_GREEN_UINT(0) |
827 A4XX_RB_BLEND_GREEN_FLOAT(0.0));
828 OUT_RING(ring, A4XX_RB_BLEND_BLUE_UINT(0) |
829 A4XX_RB_BLEND_BLUE_FLOAT(0.0));
830 OUT_RING(ring, A4XX_RB_BLEND_ALPHA_UINT(0x7fff) |
831 A4XX_RB_BLEND_ALPHA_FLOAT(1.0));
832
833 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2152, 1);
834 OUT_RING(ring, 0x00000000);
835
836 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2153, 1);
837 OUT_RING(ring, 0x00000000);
838
839 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2154, 1);
840 OUT_RING(ring, 0x00000000);
841
842 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2155, 1);
843 OUT_RING(ring, 0x00000000);
844
845 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2156, 1);
846 OUT_RING(ring, 0x00000000);
847
848 OUT_PKT0(ring, REG_A4XX_UNKNOWN_2157, 1);
849 OUT_RING(ring, 0x00000000);
850
851 OUT_PKT0(ring, REG_A4XX_UNKNOWN_21C3, 1);
852 OUT_RING(ring, 0x0000001d);
853
854 OUT_PKT0(ring, REG_A4XX_PC_GS_PARAM, 1);
855 OUT_RING(ring, 0x00000000);
856
857 OUT_PKT0(ring, REG_A4XX_UNKNOWN_21E6, 1);
858 OUT_RING(ring, 0x00000001);
859
860 OUT_PKT0(ring, REG_A4XX_PC_HS_PARAM, 1);
861 OUT_RING(ring, 0x00000000);
862
863 OUT_PKT0(ring, REG_A4XX_UNKNOWN_22D7, 1);
864 OUT_RING(ring, 0x00000000);
865
866 OUT_PKT0(ring, REG_A4XX_TPL1_TP_TEX_OFFSET, 1);
867 OUT_RING(ring, 0x00000000);
868
869 OUT_PKT0(ring, REG_A4XX_TPL1_TP_TEX_COUNT, 1);
870 OUT_RING(ring, A4XX_TPL1_TP_TEX_COUNT_VS(16) |
871 A4XX_TPL1_TP_TEX_COUNT_HS(0) |
872 A4XX_TPL1_TP_TEX_COUNT_DS(0) |
873 A4XX_TPL1_TP_TEX_COUNT_GS(0));
874
875 OUT_PKT0(ring, REG_A4XX_TPL1_TP_FS_TEX_COUNT, 1);
876 OUT_RING(ring, 16);
877
878 /* we don't use this yet.. probably best to disable.. */
879 OUT_PKT3(ring, CP_SET_DRAW_STATE, 2);
880 OUT_RING(ring, CP_SET_DRAW_STATE__0_COUNT(0) |
881 CP_SET_DRAW_STATE__0_DISABLE_ALL_GROUPS |
882 CP_SET_DRAW_STATE__0_GROUP_ID(0));
883 OUT_RING(ring, CP_SET_DRAW_STATE__1_ADDR_LO(0));
884
885 OUT_PKT0(ring, REG_A4XX_SP_VS_PVT_MEM_PARAM, 2);
886 OUT_RING(ring, 0x08000001); /* SP_VS_PVT_MEM_PARAM */
887 OUT_RELOC(ring, fd4_ctx->vs_pvt_mem, 0,0,0); /* SP_VS_PVT_MEM_ADDR */
888
889 OUT_PKT0(ring, REG_A4XX_SP_FS_PVT_MEM_PARAM, 2);
890 OUT_RING(ring, 0x08000001); /* SP_FS_PVT_MEM_PARAM */
891 OUT_RELOC(ring, fd4_ctx->fs_pvt_mem, 0,0,0); /* SP_FS_PVT_MEM_ADDR */
892
893 OUT_PKT0(ring, REG_A4XX_GRAS_SC_CONTROL, 1);
894 OUT_RING(ring, A4XX_GRAS_SC_CONTROL_RENDER_MODE(RB_RENDERING_PASS) |
895 A4XX_GRAS_SC_CONTROL_MSAA_DISABLE |
896 A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES(MSAA_ONE) |
897 A4XX_GRAS_SC_CONTROL_RASTER_MODE(0));
898
899 OUT_PKT0(ring, REG_A4XX_RB_MSAA_CONTROL, 1);
900 OUT_RING(ring, A4XX_RB_MSAA_CONTROL_DISABLE |
901 A4XX_RB_MSAA_CONTROL_SAMPLES(MSAA_ONE));
902
903 OUT_PKT0(ring, REG_A4XX_GRAS_CL_GB_CLIP_ADJ, 1);
904 OUT_RING(ring, A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ(0) |
905 A4XX_GRAS_CL_GB_CLIP_ADJ_VERT(0));
906
907 OUT_PKT0(ring, REG_A4XX_RB_ALPHA_CONTROL, 1);
908 OUT_RING(ring, A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC(FUNC_ALWAYS));
909
910 OUT_PKT0(ring, REG_A4XX_RB_FS_OUTPUT, 1);
911 OUT_RING(ring, A4XX_RB_FS_OUTPUT_SAMPLE_MASK(0xffff));
912
913 OUT_PKT0(ring, REG_A4XX_GRAS_CLEAR_CNTL, 1);
914 OUT_RING(ring, A4XX_GRAS_CLEAR_CNTL_NOT_FASTCLEAR);
915
916 OUT_PKT0(ring, REG_A4XX_GRAS_ALPHA_CONTROL, 1);
917 OUT_RING(ring, 0x0);
918
919 fd_hw_query_enable(batch, ring);
920 }
921
922 static void
fd4_emit_ib(struct fd_ringbuffer * ring,struct fd_ringbuffer * target)923 fd4_emit_ib(struct fd_ringbuffer *ring, struct fd_ringbuffer *target)
924 {
925 __OUT_IB(ring, true, target);
926 }
927
928 void
fd4_emit_init(struct pipe_context * pctx)929 fd4_emit_init(struct pipe_context *pctx)
930 {
931 struct fd_context *ctx = fd_context(pctx);
932 ctx->emit_const = fd4_emit_const;
933 ctx->emit_const_bo = fd4_emit_const_bo;
934 ctx->emit_ib = fd4_emit_ib;
935 }
936