blob: 9b5c9746958868b98152446a03cacb78b3d3d010 [file] [log] [blame]
Ben Skeggs054b93e2009-12-15 22:02:47 +10001/*
2 * Copyright 2009 Red Hat Inc.
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, sublicense,
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 shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * Authors: Ben Skeggs
23 */
24
25/* NVIDIA context programs handle a number of other conditions which are
26 * not implemented in our versions. It's not clear why NVIDIA context
27 * programs have this code, nor whether it's strictly necessary for
28 * correct operation. We'll implement additional handling if/when we
29 * discover it's necessary.
30 *
31 * - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state"
32 * flag is set, this gets saved into the context.
33 * - On context save, the context program for all cards load nsource
34 * into a flag register and check for ILLEGAL_MTHD. If it's set,
35 * opcode 0x60000d is called before resuming normal operation.
36 * - Some context programs check more conditions than the above. NV44
37 * checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001))
38 * and calls 0x60000d before resuming normal operation.
39 * - At the very beginning of NVIDIA's context programs, flag 9 is checked
40 * and if true 0x800001 is called with count=0, pos=0, the flag is cleared
41 * and then the ctxprog is aborted. It looks like a complicated NOP,
42 * its purpose is unknown.
43 * - In the section of code that loads the per-vs state, NVIDIA check
44 * flag 10. If it's set, they only transfer the small 0x300 byte block
45 * of state + the state for a single vs as opposed to the state for
46 * all vs units. It doesn't seem likely that it'll occur in normal
47 * operation, especially seeing as it appears NVIDIA may have screwed
48 * up the ctxprogs for some cards and have an invalid instruction
49 * rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction.
50 * - There's a number of places where context offset 0 (where we place
51 * the PRAMIN offset of the context) is loaded into either 0x408000,
52 * 0x408004 or 0x408008. Not sure what's up there either.
53 * - The ctxprogs for some cards save 0x400a00 again during the cleanup
54 * path for auto-loadctx.
55 */
56
57#define CP_FLAG_CLEAR 0
58#define CP_FLAG_SET 1
59#define CP_FLAG_SWAP_DIRECTION ((0 * 32) + 0)
60#define CP_FLAG_SWAP_DIRECTION_LOAD 0
61#define CP_FLAG_SWAP_DIRECTION_SAVE 1
62#define CP_FLAG_USER_SAVE ((0 * 32) + 5)
63#define CP_FLAG_USER_SAVE_NOT_PENDING 0
64#define CP_FLAG_USER_SAVE_PENDING 1
65#define CP_FLAG_USER_LOAD ((0 * 32) + 6)
66#define CP_FLAG_USER_LOAD_NOT_PENDING 0
67#define CP_FLAG_USER_LOAD_PENDING 1
68#define CP_FLAG_STATUS ((3 * 32) + 0)
69#define CP_FLAG_STATUS_IDLE 0
70#define CP_FLAG_STATUS_BUSY 1
71#define CP_FLAG_AUTO_SAVE ((3 * 32) + 4)
72#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
73#define CP_FLAG_AUTO_SAVE_PENDING 1
74#define CP_FLAG_AUTO_LOAD ((3 * 32) + 5)
75#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
76#define CP_FLAG_AUTO_LOAD_PENDING 1
77#define CP_FLAG_UNK54 ((3 * 32) + 6)
78#define CP_FLAG_UNK54_CLEAR 0
79#define CP_FLAG_UNK54_SET 1
80#define CP_FLAG_ALWAYS ((3 * 32) + 8)
81#define CP_FLAG_ALWAYS_FALSE 0
82#define CP_FLAG_ALWAYS_TRUE 1
83#define CP_FLAG_UNK57 ((3 * 32) + 9)
84#define CP_FLAG_UNK57_CLEAR 0
85#define CP_FLAG_UNK57_SET 1
86
87#define CP_CTX 0x00100000
88#define CP_CTX_COUNT 0x000fc000
89#define CP_CTX_COUNT_SHIFT 14
90#define CP_CTX_REG 0x00003fff
91#define CP_LOAD_SR 0x00200000
92#define CP_LOAD_SR_VALUE 0x000fffff
93#define CP_BRA 0x00400000
94#define CP_BRA_IP 0x0000ff00
95#define CP_BRA_IP_SHIFT 8
96#define CP_BRA_IF_CLEAR 0x00000080
97#define CP_BRA_FLAG 0x0000007f
98#define CP_WAIT 0x00500000
99#define CP_WAIT_SET 0x00000080
100#define CP_WAIT_FLAG 0x0000007f
101#define CP_SET 0x00700000
102#define CP_SET_1 0x00000080
103#define CP_SET_FLAG 0x0000007f
104#define CP_NEXT_TO_SWAP 0x00600007
105#define CP_NEXT_TO_CURRENT 0x00600009
106#define CP_SET_CONTEXT_POINTER 0x0060000a
107#define CP_END 0x0060000e
108#define CP_LOAD_MAGIC_UNK01 0x00800001 /* unknown */
109#define CP_LOAD_MAGIC_NV44TCL 0x00800029 /* per-vs state (0x4497) */
110#define CP_LOAD_MAGIC_NV40TCL 0x00800041 /* per-vs state (0x4097) */
111
112#include "drmP.h"
113#include "nouveau_drv.h"
114#include "nouveau_grctx.h"
115
116/* TODO:
117 * - get vs count from 0x1540
Ben Skeggs054b93e2009-12-15 22:02:47 +1000118 */
119
120static int
121nv40_graph_4097(struct drm_device *dev)
122{
123 struct drm_nouveau_private *dev_priv = dev->dev_private;
124
125 if ((dev_priv->chipset & 0xf0) == 0x60)
126 return 0;
127
128 return !!(0x0baf & (1 << dev_priv->chipset));
129}
130
131static int
132nv40_graph_vs_count(struct drm_device *dev)
133{
134 struct drm_nouveau_private *dev_priv = dev->dev_private;
135
136 switch (dev_priv->chipset) {
137 case 0x47:
138 case 0x49:
139 case 0x4b:
140 return 8;
141 case 0x40:
142 return 6;
143 case 0x41:
144 case 0x42:
145 return 5;
146 case 0x43:
147 case 0x44:
148 case 0x46:
149 case 0x4a:
150 return 3;
151 case 0x4c:
152 case 0x4e:
153 case 0x67:
154 default:
155 return 1;
156 }
157}
158
159
160enum cp_label {
161 cp_check_load = 1,
162 cp_setup_auto_load,
163 cp_setup_load,
164 cp_setup_save,
165 cp_swap_state,
166 cp_swap_state3d_3_is_save,
167 cp_prepare_exit,
168 cp_exit,
169};
170
171static void
172nv40_graph_construct_general(struct nouveau_grctx *ctx)
173{
174 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
175 int i;
176
177 cp_ctx(ctx, 0x4000a4, 1);
178 gr_def(ctx, 0x4000a4, 0x00000008);
179 cp_ctx(ctx, 0x400144, 58);
180 gr_def(ctx, 0x400144, 0x00000001);
181 cp_ctx(ctx, 0x400314, 1);
182 gr_def(ctx, 0x400314, 0x00000000);
183 cp_ctx(ctx, 0x400400, 10);
184 cp_ctx(ctx, 0x400480, 10);
185 cp_ctx(ctx, 0x400500, 19);
186 gr_def(ctx, 0x400514, 0x00040000);
187 gr_def(ctx, 0x400524, 0x55555555);
188 gr_def(ctx, 0x400528, 0x55555555);
189 gr_def(ctx, 0x40052c, 0x55555555);
190 gr_def(ctx, 0x400530, 0x55555555);
191 cp_ctx(ctx, 0x400560, 6);
192 gr_def(ctx, 0x400568, 0x0000ffff);
193 gr_def(ctx, 0x40056c, 0x0000ffff);
194 cp_ctx(ctx, 0x40057c, 5);
195 cp_ctx(ctx, 0x400710, 3);
196 gr_def(ctx, 0x400710, 0x20010001);
197 gr_def(ctx, 0x400714, 0x0f73ef00);
198 cp_ctx(ctx, 0x400724, 1);
199 gr_def(ctx, 0x400724, 0x02008821);
200 cp_ctx(ctx, 0x400770, 3);
201 if (dev_priv->chipset == 0x40) {
202 cp_ctx(ctx, 0x400814, 4);
203 cp_ctx(ctx, 0x400828, 5);
204 cp_ctx(ctx, 0x400840, 5);
205 gr_def(ctx, 0x400850, 0x00000040);
206 cp_ctx(ctx, 0x400858, 4);
207 gr_def(ctx, 0x400858, 0x00000040);
208 gr_def(ctx, 0x40085c, 0x00000040);
209 gr_def(ctx, 0x400864, 0x80000000);
210 cp_ctx(ctx, 0x40086c, 9);
211 gr_def(ctx, 0x40086c, 0x80000000);
212 gr_def(ctx, 0x400870, 0x80000000);
213 gr_def(ctx, 0x400874, 0x80000000);
214 gr_def(ctx, 0x400878, 0x80000000);
215 gr_def(ctx, 0x400888, 0x00000040);
216 gr_def(ctx, 0x40088c, 0x80000000);
217 cp_ctx(ctx, 0x4009c0, 8);
218 gr_def(ctx, 0x4009cc, 0x80000000);
219 gr_def(ctx, 0x4009dc, 0x80000000);
220 } else {
221 cp_ctx(ctx, 0x400840, 20);
222 if (!nv40_graph_4097(ctx->dev)) {
223 for (i = 0; i < 8; i++)
224 gr_def(ctx, 0x400860 + (i * 4), 0x00000001);
225 }
226 gr_def(ctx, 0x400880, 0x00000040);
227 gr_def(ctx, 0x400884, 0x00000040);
228 gr_def(ctx, 0x400888, 0x00000040);
229 cp_ctx(ctx, 0x400894, 11);
230 gr_def(ctx, 0x400894, 0x00000040);
231 if (nv40_graph_4097(ctx->dev)) {
232 for (i = 0; i < 8; i++)
233 gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000);
234 }
235 cp_ctx(ctx, 0x4008e0, 2);
236 cp_ctx(ctx, 0x4008f8, 2);
237 if (dev_priv->chipset == 0x4c ||
238 (dev_priv->chipset & 0xf0) == 0x60)
239 cp_ctx(ctx, 0x4009f8, 1);
240 }
241 cp_ctx(ctx, 0x400a00, 73);
242 gr_def(ctx, 0x400b0c, 0x0b0b0b0c);
243 cp_ctx(ctx, 0x401000, 4);
244 cp_ctx(ctx, 0x405004, 1);
245 switch (dev_priv->chipset) {
246 case 0x47:
247 case 0x49:
248 case 0x4b:
249 cp_ctx(ctx, 0x403448, 1);
250 gr_def(ctx, 0x403448, 0x00001010);
251 break;
252 default:
253 cp_ctx(ctx, 0x403440, 1);
254 switch (dev_priv->chipset) {
255 case 0x40:
256 gr_def(ctx, 0x403440, 0x00000010);
257 break;
258 case 0x44:
259 case 0x46:
260 case 0x4a:
261 gr_def(ctx, 0x403440, 0x00003010);
262 break;
263 case 0x41:
264 case 0x42:
265 case 0x43:
266 case 0x4c:
267 case 0x4e:
268 case 0x67:
269 default:
270 gr_def(ctx, 0x403440, 0x00001010);
271 break;
272 }
273 break;
274 }
275}
276
277static void
278nv40_graph_construct_state3d(struct nouveau_grctx *ctx)
279{
280 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
281 int i;
282
283 if (dev_priv->chipset == 0x40) {
284 cp_ctx(ctx, 0x401880, 51);
285 gr_def(ctx, 0x401940, 0x00000100);
286 } else
287 if (dev_priv->chipset == 0x46 || dev_priv->chipset == 0x47 ||
288 dev_priv->chipset == 0x49 || dev_priv->chipset == 0x4b) {
289 cp_ctx(ctx, 0x401880, 32);
290 for (i = 0; i < 16; i++)
291 gr_def(ctx, 0x401880 + (i * 4), 0x00000111);
292 if (dev_priv->chipset == 0x46)
293 cp_ctx(ctx, 0x401900, 16);
294 cp_ctx(ctx, 0x401940, 3);
295 }
296 cp_ctx(ctx, 0x40194c, 18);
297 gr_def(ctx, 0x401954, 0x00000111);
298 gr_def(ctx, 0x401958, 0x00080060);
299 gr_def(ctx, 0x401974, 0x00000080);
300 gr_def(ctx, 0x401978, 0xffff0000);
301 gr_def(ctx, 0x40197c, 0x00000001);
302 gr_def(ctx, 0x401990, 0x46400000);
303 if (dev_priv->chipset == 0x40) {
304 cp_ctx(ctx, 0x4019a0, 2);
305 cp_ctx(ctx, 0x4019ac, 5);
306 } else {
307 cp_ctx(ctx, 0x4019a0, 1);
308 cp_ctx(ctx, 0x4019b4, 3);
309 }
310 gr_def(ctx, 0x4019bc, 0xffff0000);
311 switch (dev_priv->chipset) {
312 case 0x46:
313 case 0x47:
314 case 0x49:
315 case 0x4b:
316 cp_ctx(ctx, 0x4019c0, 18);
317 for (i = 0; i < 16; i++)
318 gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888);
319 break;
320 }
321 cp_ctx(ctx, 0x401a08, 8);
322 gr_def(ctx, 0x401a10, 0x0fff0000);
323 gr_def(ctx, 0x401a14, 0x0fff0000);
324 gr_def(ctx, 0x401a1c, 0x00011100);
325 cp_ctx(ctx, 0x401a2c, 4);
326 cp_ctx(ctx, 0x401a44, 26);
327 for (i = 0; i < 16; i++)
328 gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000);
329 gr_def(ctx, 0x401a8c, 0x4b7fffff);
330 if (dev_priv->chipset == 0x40) {
331 cp_ctx(ctx, 0x401ab8, 3);
332 } else {
333 cp_ctx(ctx, 0x401ab8, 1);
334 cp_ctx(ctx, 0x401ac0, 1);
335 }
336 cp_ctx(ctx, 0x401ad0, 8);
337 gr_def(ctx, 0x401ad0, 0x30201000);
338 gr_def(ctx, 0x401ad4, 0x70605040);
339 gr_def(ctx, 0x401ad8, 0xb8a89888);
340 gr_def(ctx, 0x401adc, 0xf8e8d8c8);
341 cp_ctx(ctx, 0x401b10, dev_priv->chipset == 0x40 ? 2 : 1);
342 gr_def(ctx, 0x401b10, 0x40100000);
343 cp_ctx(ctx, 0x401b18, dev_priv->chipset == 0x40 ? 6 : 5);
344 gr_def(ctx, 0x401b28, dev_priv->chipset == 0x40 ?
345 0x00000004 : 0x00000000);
346 cp_ctx(ctx, 0x401b30, 25);
347 gr_def(ctx, 0x401b34, 0x0000ffff);
348 gr_def(ctx, 0x401b68, 0x435185d6);
349 gr_def(ctx, 0x401b6c, 0x2155b699);
350 gr_def(ctx, 0x401b70, 0xfedcba98);
351 gr_def(ctx, 0x401b74, 0x00000098);
352 gr_def(ctx, 0x401b84, 0xffffffff);
353 gr_def(ctx, 0x401b88, 0x00ff7000);
354 gr_def(ctx, 0x401b8c, 0x0000ffff);
355 if (dev_priv->chipset != 0x44 && dev_priv->chipset != 0x4a &&
356 dev_priv->chipset != 0x4e)
357 cp_ctx(ctx, 0x401b94, 1);
358 cp_ctx(ctx, 0x401b98, 8);
359 gr_def(ctx, 0x401b9c, 0x00ff0000);
360 cp_ctx(ctx, 0x401bc0, 9);
361 gr_def(ctx, 0x401be0, 0x00ffff00);
362 cp_ctx(ctx, 0x401c00, 192);
363 for (i = 0; i < 16; i++) { /* fragment texture units */
364 gr_def(ctx, 0x401c40 + (i * 4), 0x00018488);
365 gr_def(ctx, 0x401c80 + (i * 4), 0x00028202);
366 gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4);
367 gr_def(ctx, 0x401d40 + (i * 4), 0x01012000);
368 gr_def(ctx, 0x401d80 + (i * 4), 0x00080008);
369 gr_def(ctx, 0x401e00 + (i * 4), 0x00100008);
370 }
371 for (i = 0; i < 4; i++) { /* vertex texture units */
372 gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80);
373 gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202);
374 gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008);
375 gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008);
376 }
377 cp_ctx(ctx, 0x400f5c, 3);
378 gr_def(ctx, 0x400f5c, 0x00000002);
379 cp_ctx(ctx, 0x400f84, 1);
380}
381
382static void
383nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx)
384{
385 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
386 int i;
387
388 cp_ctx(ctx, 0x402000, 1);
389 cp_ctx(ctx, 0x402404, dev_priv->chipset == 0x40 ? 1 : 2);
390 switch (dev_priv->chipset) {
391 case 0x40:
392 gr_def(ctx, 0x402404, 0x00000001);
393 break;
394 case 0x4c:
395 case 0x4e:
396 case 0x67:
397 gr_def(ctx, 0x402404, 0x00000020);
398 break;
399 case 0x46:
400 case 0x49:
401 case 0x4b:
402 gr_def(ctx, 0x402404, 0x00000421);
403 break;
404 default:
405 gr_def(ctx, 0x402404, 0x00000021);
406 }
407 if (dev_priv->chipset != 0x40)
408 gr_def(ctx, 0x402408, 0x030c30c3);
409 switch (dev_priv->chipset) {
410 case 0x44:
411 case 0x46:
412 case 0x4a:
413 case 0x4c:
414 case 0x4e:
415 case 0x67:
416 cp_ctx(ctx, 0x402440, 1);
417 gr_def(ctx, 0x402440, 0x00011001);
418 break;
419 default:
420 break;
421 }
422 cp_ctx(ctx, 0x402480, dev_priv->chipset == 0x40 ? 8 : 9);
423 gr_def(ctx, 0x402488, 0x3e020200);
424 gr_def(ctx, 0x40248c, 0x00ffffff);
425 switch (dev_priv->chipset) {
426 case 0x40:
427 gr_def(ctx, 0x402490, 0x60103f00);
428 break;
429 case 0x47:
430 gr_def(ctx, 0x402490, 0x40103f00);
431 break;
432 case 0x41:
433 case 0x42:
434 case 0x49:
435 case 0x4b:
436 gr_def(ctx, 0x402490, 0x20103f00);
437 break;
438 default:
439 gr_def(ctx, 0x402490, 0x0c103f00);
440 break;
441 }
442 gr_def(ctx, 0x40249c, dev_priv->chipset <= 0x43 ?
443 0x00020000 : 0x00040000);
444 cp_ctx(ctx, 0x402500, 31);
445 gr_def(ctx, 0x402530, 0x00008100);
446 if (dev_priv->chipset == 0x40)
447 cp_ctx(ctx, 0x40257c, 6);
448 cp_ctx(ctx, 0x402594, 16);
449 cp_ctx(ctx, 0x402800, 17);
450 gr_def(ctx, 0x402800, 0x00000001);
451 switch (dev_priv->chipset) {
452 case 0x47:
453 case 0x49:
454 case 0x4b:
455 cp_ctx(ctx, 0x402864, 1);
456 gr_def(ctx, 0x402864, 0x00001001);
457 cp_ctx(ctx, 0x402870, 3);
458 gr_def(ctx, 0x402878, 0x00000003);
459 if (dev_priv->chipset != 0x47) { /* belong at end!! */
460 cp_ctx(ctx, 0x402900, 1);
461 cp_ctx(ctx, 0x402940, 1);
462 cp_ctx(ctx, 0x402980, 1);
463 cp_ctx(ctx, 0x4029c0, 1);
464 cp_ctx(ctx, 0x402a00, 1);
465 cp_ctx(ctx, 0x402a40, 1);
466 cp_ctx(ctx, 0x402a80, 1);
467 cp_ctx(ctx, 0x402ac0, 1);
468 }
469 break;
470 case 0x40:
471 cp_ctx(ctx, 0x402844, 1);
472 gr_def(ctx, 0x402844, 0x00000001);
473 cp_ctx(ctx, 0x402850, 1);
474 break;
475 default:
476 cp_ctx(ctx, 0x402844, 1);
477 gr_def(ctx, 0x402844, 0x00001001);
478 cp_ctx(ctx, 0x402850, 2);
479 gr_def(ctx, 0x402854, 0x00000003);
480 break;
481 }
482
483 cp_ctx(ctx, 0x402c00, 4);
484 gr_def(ctx, 0x402c00, dev_priv->chipset == 0x40 ?
485 0x80800001 : 0x00888001);
486 switch (dev_priv->chipset) {
487 case 0x47:
488 case 0x49:
489 case 0x4b:
490 cp_ctx(ctx, 0x402c20, 40);
491 for (i = 0; i < 32; i++)
492 gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff);
493 cp_ctx(ctx, 0x4030b8, 13);
494 gr_def(ctx, 0x4030dc, 0x00000005);
495 gr_def(ctx, 0x4030e8, 0x0000ffff);
496 break;
497 default:
498 cp_ctx(ctx, 0x402c10, 4);
499 if (dev_priv->chipset == 0x40)
500 cp_ctx(ctx, 0x402c20, 36);
501 else
502 if (dev_priv->chipset <= 0x42)
503 cp_ctx(ctx, 0x402c20, 24);
504 else
505 if (dev_priv->chipset <= 0x4a)
506 cp_ctx(ctx, 0x402c20, 16);
507 else
508 cp_ctx(ctx, 0x402c20, 8);
509 cp_ctx(ctx, 0x402cb0, dev_priv->chipset == 0x40 ? 12 : 13);
510 gr_def(ctx, 0x402cd4, 0x00000005);
511 if (dev_priv->chipset != 0x40)
512 gr_def(ctx, 0x402ce0, 0x0000ffff);
513 break;
514 }
515
516 cp_ctx(ctx, 0x403400, dev_priv->chipset == 0x40 ? 4 : 3);
517 cp_ctx(ctx, 0x403410, dev_priv->chipset == 0x40 ? 4 : 3);
518 cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->dev));
519 for (i = 0; i < nv40_graph_vs_count(ctx->dev); i++)
520 gr_def(ctx, 0x403420 + (i * 4), 0x00005555);
521
522 if (dev_priv->chipset != 0x40) {
523 cp_ctx(ctx, 0x403600, 1);
524 gr_def(ctx, 0x403600, 0x00000001);
525 }
526 cp_ctx(ctx, 0x403800, 1);
527
528 cp_ctx(ctx, 0x403c18, 1);
529 gr_def(ctx, 0x403c18, 0x00000001);
530 switch (dev_priv->chipset) {
531 case 0x46:
532 case 0x47:
533 case 0x49:
534 case 0x4b:
535 cp_ctx(ctx, 0x405018, 1);
536 gr_def(ctx, 0x405018, 0x08e00001);
537 cp_ctx(ctx, 0x405c24, 1);
538 gr_def(ctx, 0x405c24, 0x000e3000);
539 break;
540 }
541 if (dev_priv->chipset != 0x4e)
542 cp_ctx(ctx, 0x405800, 11);
543 cp_ctx(ctx, 0x407000, 1);
544}
545
546static void
547nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx)
548{
549 int len = nv40_graph_4097(ctx->dev) ? 0x0684 : 0x0084;
550
551 cp_out (ctx, 0x300000);
552 cp_lsr (ctx, len - 4);
553 cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save);
554 cp_lsr (ctx, len);
555 cp_name(ctx, cp_swap_state3d_3_is_save);
556 cp_out (ctx, 0x800001);
557
558 ctx->ctxvals_pos += len;
559}
560
561static void
562nv40_graph_construct_shader(struct nouveau_grctx *ctx)
563{
564 struct drm_device *dev = ctx->dev;
565 struct drm_nouveau_private *dev_priv = dev->dev_private;
566 struct nouveau_gpuobj *obj = ctx->data;
567 int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset;
568 int offset, i;
569
570 vs_nr = nv40_graph_vs_count(ctx->dev);
571 vs_nr_b0 = 363;
572 vs_nr_b1 = dev_priv->chipset == 0x40 ? 128 : 64;
573 if (dev_priv->chipset == 0x40) {
574 b0_offset = 0x2200/4; /* 33a0 */
575 b1_offset = 0x55a0/4; /* 1500 */
576 vs_len = 0x6aa0/4;
577 } else
578 if (dev_priv->chipset == 0x41 || dev_priv->chipset == 0x42) {
579 b0_offset = 0x2200/4; /* 2200 */
580 b1_offset = 0x4400/4; /* 0b00 */
581 vs_len = 0x4f00/4;
582 } else {
583 b0_offset = 0x1d40/4; /* 2200 */
584 b1_offset = 0x3f40/4; /* 0b00 : 0a40 */
585 vs_len = nv40_graph_4097(dev) ? 0x4a40/4 : 0x4980/4;
586 }
587
588 cp_lsr(ctx, vs_len * vs_nr + 0x300/4);
589 cp_out(ctx, nv40_graph_4097(dev) ? 0x800041 : 0x800029);
590
591 offset = ctx->ctxvals_pos;
592 ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len));
593
594 if (ctx->mode != NOUVEAU_GRCTX_VALS)
595 return;
596
597 offset += 0x0280/4;
598 for (i = 0; i < 16; i++, offset += 2)
599 nv_wo32(dev, obj, offset, 0x3f800000);
600
601 for (vs = 0; vs < vs_nr; vs++, offset += vs_len) {
602 for (i = 0; i < vs_nr_b0 * 6; i += 6)
603 nv_wo32(dev, obj, offset + b0_offset + i, 0x00000001);
604 for (i = 0; i < vs_nr_b1 * 4; i += 4)
605 nv_wo32(dev, obj, offset + b1_offset + i, 0x3f800000);
606 }
607}
608
609void
610nv40_grctx_init(struct nouveau_grctx *ctx)
611{
612 /* decide whether we're loading/unloading the context */
613 cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
614 cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
615
616 cp_name(ctx, cp_check_load);
617 cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
618 cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
619 cp_bra (ctx, ALWAYS, TRUE, cp_exit);
620
621 /* setup for context load */
622 cp_name(ctx, cp_setup_auto_load);
623 cp_wait(ctx, STATUS, IDLE);
624 cp_out (ctx, CP_NEXT_TO_SWAP);
625 cp_name(ctx, cp_setup_load);
626 cp_wait(ctx, STATUS, IDLE);
627 cp_set (ctx, SWAP_DIRECTION, LOAD);
628 cp_out (ctx, 0x00910880); /* ?? */
629 cp_out (ctx, 0x00901ffe); /* ?? */
630 cp_out (ctx, 0x01940000); /* ?? */
631 cp_lsr (ctx, 0x20);
632 cp_out (ctx, 0x0060000b); /* ?? */
633 cp_wait(ctx, UNK57, CLEAR);
634 cp_out (ctx, 0x0060000c); /* ?? */
635 cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
636
637 /* setup for context save */
638 cp_name(ctx, cp_setup_save);
639 cp_set (ctx, SWAP_DIRECTION, SAVE);
640
641 /* general PGRAPH state */
642 cp_name(ctx, cp_swap_state);
643 cp_pos (ctx, 0x00020/4);
644 nv40_graph_construct_general(ctx);
645 cp_wait(ctx, STATUS, IDLE);
646
647 /* 3D state, block 1 */
648 cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit);
649 nv40_graph_construct_state3d(ctx);
650 cp_wait(ctx, STATUS, IDLE);
651
652 /* 3D state, block 2 */
653 nv40_graph_construct_state3d_2(ctx);
654
655 /* Some other block of "random" state */
656 nv40_graph_construct_state3d_3(ctx);
657
658 /* Per-vertex shader state */
659 cp_pos (ctx, ctx->ctxvals_pos);
660 nv40_graph_construct_shader(ctx);
661
662 /* pre-exit state updates */
663 cp_name(ctx, cp_prepare_exit);
664 cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
665 cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
666 cp_out (ctx, CP_NEXT_TO_CURRENT);
667
668 cp_name(ctx, cp_exit);
669 cp_set (ctx, USER_SAVE, NOT_PENDING);
670 cp_set (ctx, USER_LOAD, NOT_PENDING);
671 cp_out (ctx, CP_END);
672}
673