projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
gallium: split depth_clip into depth_clip_near & depth_clip_far
[mesa.git] / src / gallium / drivers / freedreno / a3xx / fd3_program.c
1 /* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
2
3 /*
4 * Copyright (C) 2013 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_math.h"
32 #include "util/u_memory.h"
33 #include "util/u_inlines.h"
34 #include "util/u_format.h"
35
36 #include "freedreno_program.h"
37
38 #include "fd3_program.h"
39 #include "fd3_emit.h"
40 #include "fd3_texture.h"
41 #include "fd3_format.h"
42
43 static struct ir3_shader *
44 create_shader_stateobj(struct pipe_context *pctx, const struct pipe_shader_state *cso,
45 enum shader_t type)
46 {
47 struct fd_context *ctx = fd_context(pctx);
48 struct ir3_compiler *compiler = ctx->screen->compiler;
49 return ir3_shader_create(compiler, cso, type, &ctx->debug);
50 }
51
52 static void *
53 fd3_fp_state_create(struct pipe_context *pctx,
54 const struct pipe_shader_state *cso)
55 {
56 return create_shader_stateobj(pctx, cso, SHADER_FRAGMENT);
57 }
58
59 static void
60 fd3_fp_state_delete(struct pipe_context *pctx, void *hwcso)
61 {
62 struct ir3_shader *so = hwcso;
63 ir3_shader_destroy(so);
64 }
65
66 static void *
67 fd3_vp_state_create(struct pipe_context *pctx,
68 const struct pipe_shader_state *cso)
69 {
70 return create_shader_stateobj(pctx, cso, SHADER_VERTEX);
71 }
72
73 static void
74 fd3_vp_state_delete(struct pipe_context *pctx, void *hwcso)
75 {
76 struct ir3_shader *so = hwcso;
77 ir3_shader_destroy(so);
78 }
79
80 bool
81 fd3_needs_manual_clipping(const struct ir3_shader *shader,
82 const struct pipe_rasterizer_state *rast)
83 {
84 uint64_t outputs = ir3_shader_outputs(shader);
85
86 return (!rast->depth_clip_near ||
87 util_bitcount(rast->clip_plane_enable) > 6 ||
88 outputs & ((1ULL << VARYING_SLOT_CLIP_VERTEX) |
89 (1ULL << VARYING_SLOT_CLIP_DIST0) |
90 (1ULL << VARYING_SLOT_CLIP_DIST1)));
91 }
92
93
94 static void
95 emit_shader(struct fd_ringbuffer *ring, const struct ir3_shader_variant *so)
96 {
97 const struct ir3_info *si = &so->info;
98 enum adreno_state_block sb;
99 enum adreno_state_src src;
100 uint32_t i, sz, *bin;
101
102 if (so->type == SHADER_VERTEX) {
103 sb = SB_VERT_SHADER;
104 } else {
105 sb = SB_FRAG_SHADER;
106 }
107
108 if (fd_mesa_debug & FD_DBG_DIRECT) {
109 sz = si->sizedwords;
110 src = SS_DIRECT;
111 bin = fd_bo_map(so->bo);
112 } else {
113 sz = 0;
114 src = SS_INDIRECT;
115 bin = NULL;
116 }
117
118 OUT_PKT3(ring, CP_LOAD_STATE, 2 + sz);
119 OUT_RING(ring, CP_LOAD_STATE_0_DST_OFF(0) |
120 CP_LOAD_STATE_0_STATE_SRC(src) |
121 CP_LOAD_STATE_0_STATE_BLOCK(sb) |
122 CP_LOAD_STATE_0_NUM_UNIT(so->instrlen));
123 if (bin) {
124 OUT_RING(ring, CP_LOAD_STATE_1_EXT_SRC_ADDR(0) |
125 CP_LOAD_STATE_1_STATE_TYPE(ST_SHADER));
126 } else {
127 OUT_RELOC(ring, so->bo, 0,
128 CP_LOAD_STATE_1_STATE_TYPE(ST_SHADER), 0);
129 }
130 for (i = 0; i < sz; i++) {
131 OUT_RING(ring, bin[i]);
132 }
133 }
134
135 void
136 fd3_program_emit(struct fd_ringbuffer *ring, struct fd3_emit *emit,
137 int nr, struct pipe_surface **bufs)
138 {
139 const struct ir3_shader_variant *vp, *fp;
140 const struct ir3_info *vsi, *fsi;
141 enum a3xx_instrbuffermode fpbuffer, vpbuffer;
142 uint32_t fpbuffersz, vpbuffersz, fsoff;
143 uint32_t pos_regid, posz_regid, psize_regid;
144 uint32_t vcoord_regid, face_regid, coord_regid, zwcoord_regid;
145 uint32_t color_regid[4] = {0};
146 int constmode;
147 int i, j;
148
149 debug_assert(nr <= ARRAY_SIZE(color_regid));
150
151 vp = fd3_emit_get_vp(emit);
152 fp = fd3_emit_get_fp(emit);
153
154 vsi = &vp->info;
155 fsi = &fp->info;
156
157 fpbuffer = BUFFER;
158 vpbuffer = BUFFER;
159 fpbuffersz = fp->instrlen;
160 vpbuffersz = vp->instrlen;
161
162 /*
163 * Decide whether to use BUFFER or CACHE mode for VS and FS. It
164 * appears like 256 is the hard limit, but when the combined size
165 * exceeds 128 then blob will try to keep FS in BUFFER mode and
166 * switch to CACHE for VS until VS is too large. The blob seems
167 * to switch FS out of BUFFER mode at slightly under 128. But
168 * a bit fuzzy on the decision tree, so use slightly conservative
169 * limits.
170 *
171 * TODO check if these thresholds for BUFFER vs CACHE mode are the
172 * same for all a3xx or whether we need to consider the gpuid
173 */
174
175 if ((fpbuffersz + vpbuffersz) > 128) {
176 if (fpbuffersz < 112) {
177 /* FP:BUFFER VP:CACHE */
178 vpbuffer = CACHE;
179 vpbuffersz = 256 - fpbuffersz;
180 } else if (vpbuffersz < 112) {
181 /* FP:CACHE VP:BUFFER */
182 fpbuffer = CACHE;
183 fpbuffersz = 256 - vpbuffersz;
184 } else {
185 /* FP:CACHE VP:CACHE */
186 vpbuffer = fpbuffer = CACHE;
187 vpbuffersz = fpbuffersz = 192;
188 }
189 }
190
191 if (fpbuffer == BUFFER) {
192 fsoff = 128 - fpbuffersz;
193 } else {
194 fsoff = 256 - fpbuffersz;
195 }
196
197 /* seems like vs->constlen + fs->constlen > 256, then CONSTMODE=1 */
198 constmode = ((vp->constlen + fp->constlen) > 256) ? 1 : 0;
199
200 pos_regid = ir3_find_output_regid(vp, VARYING_SLOT_POS);
201 posz_regid = ir3_find_output_regid(fp, FRAG_RESULT_DEPTH);
202 psize_regid = ir3_find_output_regid(vp, VARYING_SLOT_PSIZ);
203 if (fp->color0_mrt) {
204 color_regid[0] = color_regid[1] = color_regid[2] = color_regid[3] =
205 ir3_find_output_regid(fp, FRAG_RESULT_COLOR);
206 } else {
207 color_regid[0] = ir3_find_output_regid(fp, FRAG_RESULT_DATA0);
208 color_regid[1] = ir3_find_output_regid(fp, FRAG_RESULT_DATA1);
209 color_regid[2] = ir3_find_output_regid(fp, FRAG_RESULT_DATA2);
210 color_regid[3] = ir3_find_output_regid(fp, FRAG_RESULT_DATA3);
211 }
212
213 face_regid = ir3_find_sysval_regid(fp, SYSTEM_VALUE_FRONT_FACE);
214 coord_regid = ir3_find_sysval_regid(fp, SYSTEM_VALUE_FRAG_COORD);
215 zwcoord_regid = (coord_regid == regid(63,0)) ? regid(63,0) : (coord_regid + 2);
216 vcoord_regid = ir3_find_sysval_regid(fp, SYSTEM_VALUE_VARYING_COORD);
217
218 /* adjust regids for alpha output formats. there is no alpha render
219 * format, so it's just treated like red
220 */
221 for (i = 0; i < nr; i++)
222 if (util_format_is_alpha(pipe_surface_format(bufs[i])))
223 color_regid[i] += 3;
224
225 /* we could probably divide this up into things that need to be
226 * emitted if frag-prog is dirty vs if vert-prog is dirty..
227 */
228
229 OUT_PKT0(ring, REG_A3XX_HLSQ_CONTROL_0_REG, 6);
230 OUT_RING(ring, A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(FOUR_QUADS) |
231 A3XX_HLSQ_CONTROL_0_REG_CONSTMODE(constmode) |
232 /* NOTE: I guess SHADERRESTART and CONSTFULLUPDATE maybe
233 * flush some caches? I think we only need to set those
234 * bits if we have updated const or shader..
235 */
236 A3XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART |
237 A3XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE);
238 OUT_RING(ring, A3XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE(TWO_QUADS) |
239 A3XX_HLSQ_CONTROL_1_REG_VSSUPERTHREADENABLE |
240 A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDXYREGID(coord_regid) |
241 A3XX_HLSQ_CONTROL_1_REG_FRAGCOORDZWREGID(zwcoord_regid));
242 OUT_RING(ring, A3XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD(31) |
243 A3XX_HLSQ_CONTROL_2_REG_FACENESSREGID(face_regid));
244 OUT_RING(ring, A3XX_HLSQ_CONTROL_3_REG_REGID(vcoord_regid));
245 OUT_RING(ring, A3XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH(vp->constlen) |
246 A3XX_HLSQ_VS_CONTROL_REG_CONSTSTARTOFFSET(0) |
247 A3XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH(vpbuffersz));
248 OUT_RING(ring, A3XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH(fp->constlen) |
249 A3XX_HLSQ_FS_CONTROL_REG_CONSTSTARTOFFSET(128) |
250 A3XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH(fpbuffersz));
251
252 OUT_PKT0(ring, REG_A3XX_SP_SP_CTRL_REG, 1);
253 OUT_RING(ring, A3XX_SP_SP_CTRL_REG_CONSTMODE(constmode) |
254 COND(emit->key.binning_pass, A3XX_SP_SP_CTRL_REG_BINNING) |
255 A3XX_SP_SP_CTRL_REG_SLEEPMODE(1) |
256 A3XX_SP_SP_CTRL_REG_L0MODE(0));
257
258 OUT_PKT0(ring, REG_A3XX_SP_VS_LENGTH_REG, 1);
259 OUT_RING(ring, A3XX_SP_VS_LENGTH_REG_SHADERLENGTH(vp->instrlen));
260
261 OUT_PKT0(ring, REG_A3XX_SP_VS_CTRL_REG0, 3);
262 OUT_RING(ring, A3XX_SP_VS_CTRL_REG0_THREADMODE(MULTI) |
263 A3XX_SP_VS_CTRL_REG0_INSTRBUFFERMODE(vpbuffer) |
264 COND(vpbuffer == CACHE, A3XX_SP_VS_CTRL_REG0_CACHEINVALID) |
265 A3XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(vsi->max_half_reg + 1) |
266 A3XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(vsi->max_reg + 1) |
267 A3XX_SP_VS_CTRL_REG0_THREADSIZE(TWO_QUADS) |
268 A3XX_SP_VS_CTRL_REG0_SUPERTHREADMODE |
269 A3XX_SP_VS_CTRL_REG0_LENGTH(vpbuffersz));
270 OUT_RING(ring, A3XX_SP_VS_CTRL_REG1_CONSTLENGTH(vp->constlen) |
271 A3XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING(vp->total_in) |
272 A3XX_SP_VS_CTRL_REG1_CONSTFOOTPRINT(MAX2(vp->constlen + 1, 0)));
273 OUT_RING(ring, A3XX_SP_VS_PARAM_REG_POSREGID(pos_regid) |
274 A3XX_SP_VS_PARAM_REG_PSIZEREGID(psize_regid) |
275 A3XX_SP_VS_PARAM_REG_TOTALVSOUTVAR(fp->varying_in));
276
277 struct ir3_shader_linkage l = {0};
278 ir3_link_shaders(&l, vp, fp);
279
280 for (i = 0, j = 0; (i < 16) && (j < l.cnt); i++) {
281 uint32_t reg = 0;
282
283 OUT_PKT0(ring, REG_A3XX_SP_VS_OUT_REG(i), 1);
284
285 reg |= A3XX_SP_VS_OUT_REG_A_REGID(l.var[j].regid);
286 reg |= A3XX_SP_VS_OUT_REG_A_COMPMASK(l.var[j].compmask);
287 j++;
288
289 reg |= A3XX_SP_VS_OUT_REG_B_REGID(l.var[j].regid);
290 reg |= A3XX_SP_VS_OUT_REG_B_COMPMASK(l.var[j].compmask);
291 j++;
292
293 OUT_RING(ring, reg);
294 }
295
296 for (i = 0, j = 0; (i < 8) && (j < l.cnt); i++) {
297 uint32_t reg = 0;
298
299 OUT_PKT0(ring, REG_A3XX_SP_VS_VPC_DST_REG(i), 1);
300
301 reg |= A3XX_SP_VS_VPC_DST_REG_OUTLOC0(l.var[j++].loc + 8);
302 reg |= A3XX_SP_VS_VPC_DST_REG_OUTLOC1(l.var[j++].loc + 8);
303 reg |= A3XX_SP_VS_VPC_DST_REG_OUTLOC2(l.var[j++].loc + 8);
304 reg |= A3XX_SP_VS_VPC_DST_REG_OUTLOC3(l.var[j++].loc + 8);
305
306 OUT_RING(ring, reg);
307 }
308
309 OUT_PKT0(ring, REG_A3XX_SP_VS_OBJ_OFFSET_REG, 2);
310 OUT_RING(ring, A3XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(0) |
311 A3XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET(0));
312 OUT_RELOC(ring, vp->bo, 0, 0, 0); /* SP_VS_OBJ_START_REG */
313
314 if (emit->key.binning_pass) {
315 OUT_PKT0(ring, REG_A3XX_SP_FS_LENGTH_REG, 1);
316 OUT_RING(ring, 0x00000000);
317
318 OUT_PKT0(ring, REG_A3XX_SP_FS_CTRL_REG0, 2);
319 OUT_RING(ring, A3XX_SP_FS_CTRL_REG0_THREADMODE(MULTI) |
320 A3XX_SP_FS_CTRL_REG0_INSTRBUFFERMODE(BUFFER));
321 OUT_RING(ring, 0x00000000);
322
323 OUT_PKT0(ring, REG_A3XX_SP_FS_OBJ_OFFSET_REG, 1);
324 OUT_RING(ring, A3XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(128) |
325 A3XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET(0));
326 } else {
327 OUT_PKT0(ring, REG_A3XX_SP_FS_LENGTH_REG, 1);
328 OUT_RING(ring, A3XX_SP_FS_LENGTH_REG_SHADERLENGTH(fp->instrlen));
329
330 OUT_PKT0(ring, REG_A3XX_SP_FS_CTRL_REG0, 2);
331 OUT_RING(ring, A3XX_SP_FS_CTRL_REG0_THREADMODE(MULTI) |
332 A3XX_SP_FS_CTRL_REG0_INSTRBUFFERMODE(fpbuffer) |
333 COND(fpbuffer == CACHE, A3XX_SP_FS_CTRL_REG0_CACHEINVALID) |
334 A3XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(fsi->max_half_reg + 1) |
335 A3XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(fsi->max_reg + 1) |
336 A3XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP |
337 A3XX_SP_FS_CTRL_REG0_THREADSIZE(FOUR_QUADS) |
338 A3XX_SP_FS_CTRL_REG0_SUPERTHREADMODE |
339 COND(fp->has_samp > 0, A3XX_SP_FS_CTRL_REG0_PIXLODENABLE) |
340 A3XX_SP_FS_CTRL_REG0_LENGTH(fpbuffersz));
341 OUT_RING(ring, A3XX_SP_FS_CTRL_REG1_CONSTLENGTH(fp->constlen) |
342 A3XX_SP_FS_CTRL_REG1_INITIALOUTSTANDING(fp->total_in) |
343 A3XX_SP_FS_CTRL_REG1_CONSTFOOTPRINT(MAX2(fp->constlen + 1, 0)) |
344 A3XX_SP_FS_CTRL_REG1_HALFPRECVAROFFSET(63));
345
346 OUT_PKT0(ring, REG_A3XX_SP_FS_OBJ_OFFSET_REG, 2);
347 OUT_RING(ring, A3XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(
348 MAX2(128, vp->constlen)) |
349 A3XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET(fsoff));
350 OUT_RELOC(ring, fp->bo, 0, 0, 0); /* SP_FS_OBJ_START_REG */
351 }
352
353 OUT_PKT0(ring, REG_A3XX_SP_FS_OUTPUT_REG, 1);
354 OUT_RING(ring,
355 COND(fp->writes_pos, A3XX_SP_FS_OUTPUT_REG_DEPTH_ENABLE) |
356 A3XX_SP_FS_OUTPUT_REG_DEPTH_REGID(posz_regid) |
357 A3XX_SP_FS_OUTPUT_REG_MRT(MAX2(1, nr) - 1));
358
359 OUT_PKT0(ring, REG_A3XX_SP_FS_MRT_REG(0), 4);
360 for (i = 0; i < 4; i++) {
361 uint32_t mrt_reg = A3XX_SP_FS_MRT_REG_REGID(color_regid[i]) |
362 COND(fp->key.half_precision, A3XX_SP_FS_MRT_REG_HALF_PRECISION);
363
364 if (i < nr) {
365 enum pipe_format fmt = pipe_surface_format(bufs[i]);
366 mrt_reg |= COND(util_format_is_pure_uint(fmt), A3XX_SP_FS_MRT_REG_UINT) |
367 COND(util_format_is_pure_sint(fmt), A3XX_SP_FS_MRT_REG_SINT);
368 }
369 OUT_RING(ring, mrt_reg);
370 }
371
372 if (emit->key.binning_pass) {
373 OUT_PKT0(ring, REG_A3XX_VPC_ATTR, 2);
374 OUT_RING(ring, A3XX_VPC_ATTR_THRDASSIGN(1) |
375 A3XX_VPC_ATTR_LMSIZE(1) |
376 COND(vp->writes_psize, A3XX_VPC_ATTR_PSIZE));
377 OUT_RING(ring, 0x00000000);
378 } else {
379 uint32_t vinterp[4], flatshade[2], vpsrepl[4];
380
381 memset(vinterp, 0, sizeof(vinterp));
382 memset(flatshade, 0, sizeof(flatshade));
383 memset(vpsrepl, 0, sizeof(vpsrepl));
384
385 /* figure out VARYING_INTERP / FLAT_SHAD register values: */
386 for (j = -1; (j = ir3_next_varying(fp, j)) < (int)fp->inputs_count; ) {
387 /* NOTE: varyings are packed, so if compmask is 0xb
388 * then first, third, and fourth component occupy
389 * three consecutive varying slots:
390 */
391 unsigned compmask = fp->inputs[j].compmask;
392
393 uint32_t inloc = fp->inputs[j].inloc;
394
395 if ((fp->inputs[j].interpolate == INTERP_MODE_FLAT) ||
396 (fp->inputs[j].rasterflat && emit->rasterflat)) {
397 uint32_t loc = inloc;
398
399 for (i = 0; i < 4; i++) {
400 if (compmask & (1 << i)) {
401 vinterp[loc / 16] |= FLAT << ((loc % 16) * 2);
402 flatshade[loc / 32] |= 1 << (loc % 32);
403 loc++;
404 }
405 }
406 }
407
408 gl_varying_slot slot = fp->inputs[j].slot;
409
410 /* since we don't enable PIPE_CAP_TGSI_TEXCOORD: */
411 if (slot >= VARYING_SLOT_VAR0) {
412 unsigned texmask = 1 << (slot - VARYING_SLOT_VAR0);
413 /* Replace the .xy coordinates with S/T from the point sprite. Set
414 * interpolation bits for .zw such that they become .01
415 */
416 if (emit->sprite_coord_enable & texmask) {
417 /* mask is two 2-bit fields, where:
418 * '01' -> S
419 * '10' -> T
420 * '11' -> 1 - T (flip mode)
421 */
422 unsigned mask = emit->sprite_coord_mode ? 0b1101 : 0b1001;
423 uint32_t loc = inloc;
424 if (compmask & 0x1) {
425 vpsrepl[loc / 16] |= ((mask >> 0) & 0x3) << ((loc % 16) * 2);
426 loc++;
427 }
428 if (compmask & 0x2) {
429 vpsrepl[loc / 16] |= ((mask >> 2) & 0x3) << ((loc % 16) * 2);
430 loc++;
431 }
432 if (compmask & 0x4) {
433 /* .z <- 0.0f */
434 vinterp[loc / 16] |= 0b10 << ((loc % 16) * 2);
435 loc++;
436 }
437 if (compmask & 0x8) {
438 /* .w <- 1.0f */
439 vinterp[loc / 16] |= 0b11 << ((loc % 16) * 2);
440 loc++;
441 }
442 }
443 }
444 }
445
446 OUT_PKT0(ring, REG_A3XX_VPC_ATTR, 2);
447 OUT_RING(ring, A3XX_VPC_ATTR_TOTALATTR(fp->total_in) |
448 A3XX_VPC_ATTR_THRDASSIGN(1) |
449 A3XX_VPC_ATTR_LMSIZE(1) |
450 COND(vp->writes_psize, A3XX_VPC_ATTR_PSIZE));
451 OUT_RING(ring, A3XX_VPC_PACK_NUMFPNONPOSVAR(fp->total_in) |
452 A3XX_VPC_PACK_NUMNONPOSVSVAR(fp->total_in));
453
454 OUT_PKT0(ring, REG_A3XX_VPC_VARYING_INTERP_MODE(0), 4);
455 OUT_RING(ring, vinterp[0]); /* VPC_VARYING_INTERP[0].MODE */
456 OUT_RING(ring, vinterp[1]); /* VPC_VARYING_INTERP[1].MODE */
457 OUT_RING(ring, vinterp[2]); /* VPC_VARYING_INTERP[2].MODE */
458 OUT_RING(ring, vinterp[3]); /* VPC_VARYING_INTERP[3].MODE */
459
460 OUT_PKT0(ring, REG_A3XX_VPC_VARYING_PS_REPL_MODE(0), 4);
461 OUT_RING(ring, vpsrepl[0]); /* VPC_VARYING_PS_REPL[0].MODE */
462 OUT_RING(ring, vpsrepl[1]); /* VPC_VARYING_PS_REPL[1].MODE */
463 OUT_RING(ring, vpsrepl[2]); /* VPC_VARYING_PS_REPL[2].MODE */
464 OUT_RING(ring, vpsrepl[3]); /* VPC_VARYING_PS_REPL[3].MODE */
465
466 OUT_PKT0(ring, REG_A3XX_SP_FS_FLAT_SHAD_MODE_REG_0, 2);
467 OUT_RING(ring, flatshade[0]); /* SP_FS_FLAT_SHAD_MODE_REG_0 */
468 OUT_RING(ring, flatshade[1]); /* SP_FS_FLAT_SHAD_MODE_REG_1 */
469 }
470
471 if (vpbuffer == BUFFER)
472 emit_shader(ring, vp);
473
474 OUT_PKT0(ring, REG_A3XX_VFD_PERFCOUNTER0_SELECT, 1);
475 OUT_RING(ring, 0x00000000); /* VFD_PERFCOUNTER0_SELECT */
476
477 if (!emit->key.binning_pass) {
478 if (fpbuffer == BUFFER)
479 emit_shader(ring, fp);
480
481 OUT_PKT0(ring, REG_A3XX_VFD_PERFCOUNTER0_SELECT, 1);
482 OUT_RING(ring, 0x00000000); /* VFD_PERFCOUNTER0_SELECT */
483 }
484 }
485
486 void
487 fd3_prog_init(struct pipe_context *pctx)
488 {
489 pctx->create_fs_state = fd3_fp_state_create;
490 pctx->delete_fs_state = fd3_fp_state_delete;
491
492 pctx->create_vs_state = fd3_vp_state_create;
493 pctx->delete_vs_state = fd3_vp_state_delete;
494
495 fd_prog_init(pctx);
496 }