aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/nouveau/nv40_grctx.c
diff options
context:
space:
mode:
authorBen Skeggs <bskeggs@redhat.com>2009-12-15 07:02:47 -0500
committerBen Skeggs <bskeggs@redhat.com>2009-12-16 02:05:39 -0500
commit054b93e444550a72aef17115363cdef253b9ee7c (patch)
tree667d10b686c62d64e0b998115209c65884bcc6de /drivers/gpu/drm/nouveau/nv40_grctx.c
parent15bee69ee1532a29e13124b298027ee6ef54bac8 (diff)
drm/nv40: implement ctxprog/state generation
The context programs are *very* simple compared to the ones used by the binary driver. There's notes in nv40_grctx.c explaining most of the things we don't implement. If we discover if/why any of it is required further down the track, we'll handle it then. The PGRAPH state generated for each chipset should match what NVIDIA do almost exactly (there's a couple of exceptions). If someone has a lot of time on their hands, they could figure out the mapping of object/method to PGRAPH register and demagic the initial state a little, it's not terribly important however. At time of commit, confirmed to be working at least well enough for accelerated X (and where tested, for 3D apps) on NV40, NV43, NV44, NV46, NV49, NV4A, NV4B and NV4E. A module option has been added to force the use of external firmware blobs if it becomes required. Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Diffstat (limited to 'drivers/gpu/drm/nouveau/nv40_grctx.c')
-rw-r--r--drivers/gpu/drm/nouveau/nv40_grctx.c678
1 files changed, 678 insertions, 0 deletions
diff --git a/drivers/gpu/drm/nouveau/nv40_grctx.c b/drivers/gpu/drm/nouveau/nv40_grctx.c
new file mode 100644
index 000000000000..11b11c31f543
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nv40_grctx.c
@@ -0,0 +1,678 @@
1/*
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
118 * - document unimplemented bits compared to nvidia
119 * - nsource handling
120 * - R0 & 0x0200 handling
121 * - single-vs handling
122 * - 400314 bit 0
123 */
124
125static int
126nv40_graph_4097(struct drm_device *dev)
127{
128 struct drm_nouveau_private *dev_priv = dev->dev_private;
129
130 if ((dev_priv->chipset & 0xf0) == 0x60)
131 return 0;
132
133 return !!(0x0baf & (1 << dev_priv->chipset));
134}
135
136static int
137nv40_graph_vs_count(struct drm_device *dev)
138{
139 struct drm_nouveau_private *dev_priv = dev->dev_private;
140
141 switch (dev_priv->chipset) {
142 case 0x47:
143 case 0x49:
144 case 0x4b:
145 return 8;
146 case 0x40:
147 return 6;
148 case 0x41:
149 case 0x42:
150 return 5;
151 case 0x43:
152 case 0x44:
153 case 0x46:
154 case 0x4a:
155 return 3;
156 case 0x4c:
157 case 0x4e:
158 case 0x67:
159 default:
160 return 1;
161 }
162}
163
164
165enum cp_label {
166 cp_check_load = 1,
167 cp_setup_auto_load,
168 cp_setup_load,
169 cp_setup_save,
170 cp_swap_state,
171 cp_swap_state3d_3_is_save,
172 cp_prepare_exit,
173 cp_exit,
174};
175
176static void
177nv40_graph_construct_general(struct nouveau_grctx *ctx)
178{
179 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
180 int i;
181
182 cp_ctx(ctx, 0x4000a4, 1);
183 gr_def(ctx, 0x4000a4, 0x00000008);
184 cp_ctx(ctx, 0x400144, 58);
185 gr_def(ctx, 0x400144, 0x00000001);
186 cp_ctx(ctx, 0x400314, 1);
187 gr_def(ctx, 0x400314, 0x00000000);
188 cp_ctx(ctx, 0x400400, 10);
189 cp_ctx(ctx, 0x400480, 10);
190 cp_ctx(ctx, 0x400500, 19);
191 gr_def(ctx, 0x400514, 0x00040000);
192 gr_def(ctx, 0x400524, 0x55555555);
193 gr_def(ctx, 0x400528, 0x55555555);
194 gr_def(ctx, 0x40052c, 0x55555555);
195 gr_def(ctx, 0x400530, 0x55555555);
196 cp_ctx(ctx, 0x400560, 6);
197 gr_def(ctx, 0x400568, 0x0000ffff);
198 gr_def(ctx, 0x40056c, 0x0000ffff);
199 cp_ctx(ctx, 0x40057c, 5);
200 cp_ctx(ctx, 0x400710, 3);
201 gr_def(ctx, 0x400710, 0x20010001);
202 gr_def(ctx, 0x400714, 0x0f73ef00);
203 cp_ctx(ctx, 0x400724, 1);
204 gr_def(ctx, 0x400724, 0x02008821);
205 cp_ctx(ctx, 0x400770, 3);
206 if (dev_priv->chipset == 0x40) {
207 cp_ctx(ctx, 0x400814, 4);
208 cp_ctx(ctx, 0x400828, 5);
209 cp_ctx(ctx, 0x400840, 5);
210 gr_def(ctx, 0x400850, 0x00000040);
211 cp_ctx(ctx, 0x400858, 4);
212 gr_def(ctx, 0x400858, 0x00000040);
213 gr_def(ctx, 0x40085c, 0x00000040);
214 gr_def(ctx, 0x400864, 0x80000000);
215 cp_ctx(ctx, 0x40086c, 9);
216 gr_def(ctx, 0x40086c, 0x80000000);
217 gr_def(ctx, 0x400870, 0x80000000);
218 gr_def(ctx, 0x400874, 0x80000000);
219 gr_def(ctx, 0x400878, 0x80000000);
220 gr_def(ctx, 0x400888, 0x00000040);
221 gr_def(ctx, 0x40088c, 0x80000000);
222 cp_ctx(ctx, 0x4009c0, 8);
223 gr_def(ctx, 0x4009cc, 0x80000000);
224 gr_def(ctx, 0x4009dc, 0x80000000);
225 } else {
226 cp_ctx(ctx, 0x400840, 20);
227 if (!nv40_graph_4097(ctx->dev)) {
228 for (i = 0; i < 8; i++)
229 gr_def(ctx, 0x400860 + (i * 4), 0x00000001);
230 }
231 gr_def(ctx, 0x400880, 0x00000040);
232 gr_def(ctx, 0x400884, 0x00000040);
233 gr_def(ctx, 0x400888, 0x00000040);
234 cp_ctx(ctx, 0x400894, 11);
235 gr_def(ctx, 0x400894, 0x00000040);
236 if (nv40_graph_4097(ctx->dev)) {
237 for (i = 0; i < 8; i++)
238 gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000);
239 }
240 cp_ctx(ctx, 0x4008e0, 2);
241 cp_ctx(ctx, 0x4008f8, 2);
242 if (dev_priv->chipset == 0x4c ||
243 (dev_priv->chipset & 0xf0) == 0x60)
244 cp_ctx(ctx, 0x4009f8, 1);
245 }
246 cp_ctx(ctx, 0x400a00, 73);
247 gr_def(ctx, 0x400b0c, 0x0b0b0b0c);
248 cp_ctx(ctx, 0x401000, 4);
249 cp_ctx(ctx, 0x405004, 1);
250 switch (dev_priv->chipset) {
251 case 0x47:
252 case 0x49:
253 case 0x4b:
254 cp_ctx(ctx, 0x403448, 1);
255 gr_def(ctx, 0x403448, 0x00001010);
256 break;
257 default:
258 cp_ctx(ctx, 0x403440, 1);
259 switch (dev_priv->chipset) {
260 case 0x40:
261 gr_def(ctx, 0x403440, 0x00000010);
262 break;
263 case 0x44:
264 case 0x46:
265 case 0x4a:
266 gr_def(ctx, 0x403440, 0x00003010);
267 break;
268 case 0x41:
269 case 0x42:
270 case 0x43:
271 case 0x4c:
272 case 0x4e:
273 case 0x67:
274 default:
275 gr_def(ctx, 0x403440, 0x00001010);
276 break;
277 }
278 break;
279 }
280}
281
282static void
283nv40_graph_construct_state3d(struct nouveau_grctx *ctx)
284{
285 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
286 int i;
287
288 if (dev_priv->chipset == 0x40) {
289 cp_ctx(ctx, 0x401880, 51);
290 gr_def(ctx, 0x401940, 0x00000100);
291 } else
292 if (dev_priv->chipset == 0x46 || dev_priv->chipset == 0x47 ||
293 dev_priv->chipset == 0x49 || dev_priv->chipset == 0x4b) {
294 cp_ctx(ctx, 0x401880, 32);
295 for (i = 0; i < 16; i++)
296 gr_def(ctx, 0x401880 + (i * 4), 0x00000111);
297 if (dev_priv->chipset == 0x46)
298 cp_ctx(ctx, 0x401900, 16);
299 cp_ctx(ctx, 0x401940, 3);
300 }
301 cp_ctx(ctx, 0x40194c, 18);
302 gr_def(ctx, 0x401954, 0x00000111);
303 gr_def(ctx, 0x401958, 0x00080060);
304 gr_def(ctx, 0x401974, 0x00000080);
305 gr_def(ctx, 0x401978, 0xffff0000);
306 gr_def(ctx, 0x40197c, 0x00000001);
307 gr_def(ctx, 0x401990, 0x46400000);
308 if (dev_priv->chipset == 0x40) {
309 cp_ctx(ctx, 0x4019a0, 2);
310 cp_ctx(ctx, 0x4019ac, 5);
311 } else {
312 cp_ctx(ctx, 0x4019a0, 1);
313 cp_ctx(ctx, 0x4019b4, 3);
314 }
315 gr_def(ctx, 0x4019bc, 0xffff0000);
316 switch (dev_priv->chipset) {
317 case 0x46:
318 case 0x47:
319 case 0x49:
320 case 0x4b:
321 cp_ctx(ctx, 0x4019c0, 18);
322 for (i = 0; i < 16; i++)
323 gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888);
324 break;
325 }
326 cp_ctx(ctx, 0x401a08, 8);
327 gr_def(ctx, 0x401a10, 0x0fff0000);
328 gr_def(ctx, 0x401a14, 0x0fff0000);
329 gr_def(ctx, 0x401a1c, 0x00011100);
330 cp_ctx(ctx, 0x401a2c, 4);
331 cp_ctx(ctx, 0x401a44, 26);
332 for (i = 0; i < 16; i++)
333 gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000);
334 gr_def(ctx, 0x401a8c, 0x4b7fffff);
335 if (dev_priv->chipset == 0x40) {
336 cp_ctx(ctx, 0x401ab8, 3);
337 } else {
338 cp_ctx(ctx, 0x401ab8, 1);
339 cp_ctx(ctx, 0x401ac0, 1);
340 }
341 cp_ctx(ctx, 0x401ad0, 8);
342 gr_def(ctx, 0x401ad0, 0x30201000);
343 gr_def(ctx, 0x401ad4, 0x70605040);
344 gr_def(ctx, 0x401ad8, 0xb8a89888);
345 gr_def(ctx, 0x401adc, 0xf8e8d8c8);
346 cp_ctx(ctx, 0x401b10, dev_priv->chipset == 0x40 ? 2 : 1);
347 gr_def(ctx, 0x401b10, 0x40100000);
348 cp_ctx(ctx, 0x401b18, dev_priv->chipset == 0x40 ? 6 : 5);
349 gr_def(ctx, 0x401b28, dev_priv->chipset == 0x40 ?
350 0x00000004 : 0x00000000);
351 cp_ctx(ctx, 0x401b30, 25);
352 gr_def(ctx, 0x401b34, 0x0000ffff);
353 gr_def(ctx, 0x401b68, 0x435185d6);
354 gr_def(ctx, 0x401b6c, 0x2155b699);
355 gr_def(ctx, 0x401b70, 0xfedcba98);
356 gr_def(ctx, 0x401b74, 0x00000098);
357 gr_def(ctx, 0x401b84, 0xffffffff);
358 gr_def(ctx, 0x401b88, 0x00ff7000);
359 gr_def(ctx, 0x401b8c, 0x0000ffff);
360 if (dev_priv->chipset != 0x44 && dev_priv->chipset != 0x4a &&
361 dev_priv->chipset != 0x4e)
362 cp_ctx(ctx, 0x401b94, 1);
363 cp_ctx(ctx, 0x401b98, 8);
364 gr_def(ctx, 0x401b9c, 0x00ff0000);
365 cp_ctx(ctx, 0x401bc0, 9);
366 gr_def(ctx, 0x401be0, 0x00ffff00);
367 cp_ctx(ctx, 0x401c00, 192);
368 for (i = 0; i < 16; i++) { /* fragment texture units */
369 gr_def(ctx, 0x401c40 + (i * 4), 0x00018488);
370 gr_def(ctx, 0x401c80 + (i * 4), 0x00028202);
371 gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4);
372 gr_def(ctx, 0x401d40 + (i * 4), 0x01012000);
373 gr_def(ctx, 0x401d80 + (i * 4), 0x00080008);
374 gr_def(ctx, 0x401e00 + (i * 4), 0x00100008);
375 }
376 for (i = 0; i < 4; i++) { /* vertex texture units */
377 gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80);
378 gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202);
379 gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008);
380 gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008);
381 }
382 cp_ctx(ctx, 0x400f5c, 3);
383 gr_def(ctx, 0x400f5c, 0x00000002);
384 cp_ctx(ctx, 0x400f84, 1);
385}
386
387static void
388nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx)
389{
390 struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
391 int i;
392
393 cp_ctx(ctx, 0x402000, 1);
394 cp_ctx(ctx, 0x402404, dev_priv->chipset == 0x40 ? 1 : 2);
395 switch (dev_priv->chipset) {
396 case 0x40:
397 gr_def(ctx, 0x402404, 0x00000001);
398 break;
399 case 0x4c:
400 case 0x4e:
401 case 0x67:
402 gr_def(ctx, 0x402404, 0x00000020);
403 break;
404 case 0x46:
405 case 0x49:
406 case 0x4b:
407 gr_def(ctx, 0x402404, 0x00000421);
408 break;
409 default:
410 gr_def(ctx, 0x402404, 0x00000021);
411 }
412 if (dev_priv->chipset != 0x40)
413 gr_def(ctx, 0x402408, 0x030c30c3);
414 switch (dev_priv->chipset) {
415 case 0x44:
416 case 0x46:
417 case 0x4a:
418 case 0x4c:
419 case 0x4e:
420 case 0x67:
421 cp_ctx(ctx, 0x402440, 1);
422 gr_def(ctx, 0x402440, 0x00011001);
423 break;
424 default:
425 break;
426 }
427 cp_ctx(ctx, 0x402480, dev_priv->chipset == 0x40 ? 8 : 9);
428 gr_def(ctx, 0x402488, 0x3e020200);
429 gr_def(ctx, 0x40248c, 0x00ffffff);
430 switch (dev_priv->chipset) {
431 case 0x40:
432 gr_def(ctx, 0x402490, 0x60103f00);
433 break;
434 case 0x47:
435 gr_def(ctx, 0x402490, 0x40103f00);
436 break;
437 case 0x41:
438 case 0x42:
439 case 0x49:
440 case 0x4b:
441 gr_def(ctx, 0x402490, 0x20103f00);
442 break;
443 default:
444 gr_def(ctx, 0x402490, 0x0c103f00);
445 break;
446 }
447 gr_def(ctx, 0x40249c, dev_priv->chipset <= 0x43 ?
448 0x00020000 : 0x00040000);
449 cp_ctx(ctx, 0x402500, 31);
450 gr_def(ctx, 0x402530, 0x00008100);
451 if (dev_priv->chipset == 0x40)
452 cp_ctx(ctx, 0x40257c, 6);
453 cp_ctx(ctx, 0x402594, 16);
454 cp_ctx(ctx, 0x402800, 17);
455 gr_def(ctx, 0x402800, 0x00000001);
456 switch (dev_priv->chipset) {
457 case 0x47:
458 case 0x49:
459 case 0x4b:
460 cp_ctx(ctx, 0x402864, 1);
461 gr_def(ctx, 0x402864, 0x00001001);
462 cp_ctx(ctx, 0x402870, 3);
463 gr_def(ctx, 0x402878, 0x00000003);
464 if (dev_priv->chipset != 0x47) { /* belong at end!! */
465 cp_ctx(ctx, 0x402900, 1);
466 cp_ctx(ctx, 0x402940, 1);
467 cp_ctx(ctx, 0x402980, 1);
468 cp_ctx(ctx, 0x4029c0, 1);
469 cp_ctx(ctx, 0x402a00, 1);
470 cp_ctx(ctx, 0x402a40, 1);
471 cp_ctx(ctx, 0x402a80, 1);
472 cp_ctx(ctx, 0x402ac0, 1);
473 }
474 break;
475 case 0x40:
476 cp_ctx(ctx, 0x402844, 1);
477 gr_def(ctx, 0x402844, 0x00000001);
478 cp_ctx(ctx, 0x402850, 1);
479 break;
480 default:
481 cp_ctx(ctx, 0x402844, 1);
482 gr_def(ctx, 0x402844, 0x00001001);
483 cp_ctx(ctx, 0x402850, 2);
484 gr_def(ctx, 0x402854, 0x00000003);
485 break;
486 }
487
488 cp_ctx(ctx, 0x402c00, 4);
489 gr_def(ctx, 0x402c00, dev_priv->chipset == 0x40 ?
490 0x80800001 : 0x00888001);
491 switch (dev_priv->chipset) {
492 case 0x47:
493 case 0x49:
494 case 0x4b:
495 cp_ctx(ctx, 0x402c20, 40);
496 for (i = 0; i < 32; i++)
497 gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff);
498 cp_ctx(ctx, 0x4030b8, 13);
499 gr_def(ctx, 0x4030dc, 0x00000005);
500 gr_def(ctx, 0x4030e8, 0x0000ffff);
501 break;
502 default:
503 cp_ctx(ctx, 0x402c10, 4);
504 if (dev_priv->chipset == 0x40)
505 cp_ctx(ctx, 0x402c20, 36);
506 else
507 if (dev_priv->chipset <= 0x42)
508 cp_ctx(ctx, 0x402c20, 24);
509 else
510 if (dev_priv->chipset <= 0x4a)
511 cp_ctx(ctx, 0x402c20, 16);
512 else
513 cp_ctx(ctx, 0x402c20, 8);
514 cp_ctx(ctx, 0x402cb0, dev_priv->chipset == 0x40 ? 12 : 13);
515 gr_def(ctx, 0x402cd4, 0x00000005);
516 if (dev_priv->chipset != 0x40)
517 gr_def(ctx, 0x402ce0, 0x0000ffff);
518 break;
519 }
520
521 cp_ctx(ctx, 0x403400, dev_priv->chipset == 0x40 ? 4 : 3);
522 cp_ctx(ctx, 0x403410, dev_priv->chipset == 0x40 ? 4 : 3);
523 cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->dev));
524 for (i = 0; i < nv40_graph_vs_count(ctx->dev); i++)
525 gr_def(ctx, 0x403420 + (i * 4), 0x00005555);
526
527 if (dev_priv->chipset != 0x40) {
528 cp_ctx(ctx, 0x403600, 1);
529 gr_def(ctx, 0x403600, 0x00000001);
530 }
531 cp_ctx(ctx, 0x403800, 1);
532
533 cp_ctx(ctx, 0x403c18, 1);
534 gr_def(ctx, 0x403c18, 0x00000001);
535 switch (dev_priv->chipset) {
536 case 0x46:
537 case 0x47:
538 case 0x49:
539 case 0x4b:
540 cp_ctx(ctx, 0x405018, 1);
541 gr_def(ctx, 0x405018, 0x08e00001);
542 cp_ctx(ctx, 0x405c24, 1);
543 gr_def(ctx, 0x405c24, 0x000e3000);
544 break;
545 }
546 if (dev_priv->chipset != 0x4e)
547 cp_ctx(ctx, 0x405800, 11);
548 cp_ctx(ctx, 0x407000, 1);
549}
550
551static void
552nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx)
553{
554 int len = nv40_graph_4097(ctx->dev) ? 0x0684 : 0x0084;
555
556 cp_out (ctx, 0x300000);
557 cp_lsr (ctx, len - 4);
558 cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save);
559 cp_lsr (ctx, len);
560 cp_name(ctx, cp_swap_state3d_3_is_save);
561 cp_out (ctx, 0x800001);
562
563 ctx->ctxvals_pos += len;
564}
565
566static void
567nv40_graph_construct_shader(struct nouveau_grctx *ctx)
568{
569 struct drm_device *dev = ctx->dev;
570 struct drm_nouveau_private *dev_priv = dev->dev_private;
571 struct nouveau_gpuobj *obj = ctx->data;
572 int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset;
573 int offset, i;
574
575 vs_nr = nv40_graph_vs_count(ctx->dev);
576 vs_nr_b0 = 363;
577 vs_nr_b1 = dev_priv->chipset == 0x40 ? 128 : 64;
578 if (dev_priv->chipset == 0x40) {
579 b0_offset = 0x2200/4; /* 33a0 */
580 b1_offset = 0x55a0/4; /* 1500 */
581 vs_len = 0x6aa0/4;
582 } else
583 if (dev_priv->chipset == 0x41 || dev_priv->chipset == 0x42) {
584 b0_offset = 0x2200/4; /* 2200 */
585 b1_offset = 0x4400/4; /* 0b00 */
586 vs_len = 0x4f00/4;
587 } else {
588 b0_offset = 0x1d40/4; /* 2200 */
589 b1_offset = 0x3f40/4; /* 0b00 : 0a40 */
590 vs_len = nv40_graph_4097(dev) ? 0x4a40/4 : 0x4980/4;
591 }
592
593 cp_lsr(ctx, vs_len * vs_nr + 0x300/4);
594 cp_out(ctx, nv40_graph_4097(dev) ? 0x800041 : 0x800029);
595
596 offset = ctx->ctxvals_pos;
597 ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len));
598
599 if (ctx->mode != NOUVEAU_GRCTX_VALS)
600 return;
601
602 offset += 0x0280/4;
603 for (i = 0; i < 16; i++, offset += 2)
604 nv_wo32(dev, obj, offset, 0x3f800000);
605
606 for (vs = 0; vs < vs_nr; vs++, offset += vs_len) {
607 for (i = 0; i < vs_nr_b0 * 6; i += 6)
608 nv_wo32(dev, obj, offset + b0_offset + i, 0x00000001);
609 for (i = 0; i < vs_nr_b1 * 4; i += 4)
610 nv_wo32(dev, obj, offset + b1_offset + i, 0x3f800000);
611 }
612}
613
614void
615nv40_grctx_init(struct nouveau_grctx *ctx)
616{
617 /* decide whether we're loading/unloading the context */
618 cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
619 cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
620
621 cp_name(ctx, cp_check_load);
622 cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
623 cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
624 cp_bra (ctx, ALWAYS, TRUE, cp_exit);
625
626 /* setup for context load */
627 cp_name(ctx, cp_setup_auto_load);
628 cp_wait(ctx, STATUS, IDLE);
629 cp_out (ctx, CP_NEXT_TO_SWAP);
630 cp_name(ctx, cp_setup_load);
631 cp_wait(ctx, STATUS, IDLE);
632 cp_set (ctx, SWAP_DIRECTION, LOAD);
633 cp_out (ctx, 0x00910880); /* ?? */
634 cp_out (ctx, 0x00901ffe); /* ?? */
635 cp_out (ctx, 0x01940000); /* ?? */
636 cp_lsr (ctx, 0x20);
637 cp_out (ctx, 0x0060000b); /* ?? */
638 cp_wait(ctx, UNK57, CLEAR);
639 cp_out (ctx, 0x0060000c); /* ?? */
640 cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
641
642 /* setup for context save */
643 cp_name(ctx, cp_setup_save);
644 cp_set (ctx, SWAP_DIRECTION, SAVE);
645
646 /* general PGRAPH state */
647 cp_name(ctx, cp_swap_state);
648 cp_pos (ctx, 0x00020/4);
649 nv40_graph_construct_general(ctx);
650 cp_wait(ctx, STATUS, IDLE);
651
652 /* 3D state, block 1 */
653 cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit);
654 nv40_graph_construct_state3d(ctx);
655 cp_wait(ctx, STATUS, IDLE);
656
657 /* 3D state, block 2 */
658 nv40_graph_construct_state3d_2(ctx);
659
660 /* Some other block of "random" state */
661 nv40_graph_construct_state3d_3(ctx);
662
663 /* Per-vertex shader state */
664 cp_pos (ctx, ctx->ctxvals_pos);
665 nv40_graph_construct_shader(ctx);
666
667 /* pre-exit state updates */
668 cp_name(ctx, cp_prepare_exit);
669 cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
670 cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
671 cp_out (ctx, CP_NEXT_TO_CURRENT);
672
673 cp_name(ctx, cp_exit);
674 cp_set (ctx, USER_SAVE, NOT_PENDING);
675 cp_set (ctx, USER_LOAD, NOT_PENDING);
676 cp_out (ctx, CP_END);
677}
678
generated by cgit v1.2.2 (git 2.25.0) at 2025-01-07 19:13:35 -0500