nv50_grctx.c revision e69b4418825c2e4c6563ae1d69bd75377826e263
1/*
2 * Copyright 2009 Marcin Kościelnicki
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
23#define CP_FLAG_CLEAR                 0
24#define CP_FLAG_SET                   1
25#define CP_FLAG_SWAP_DIRECTION        ((0 * 32) + 0)
26#define CP_FLAG_SWAP_DIRECTION_LOAD   0
27#define CP_FLAG_SWAP_DIRECTION_SAVE   1
28#define CP_FLAG_UNK01                 ((0 * 32) + 1)
29#define CP_FLAG_UNK01_CLEAR           0
30#define CP_FLAG_UNK01_SET             1
31#define CP_FLAG_UNK03                 ((0 * 32) + 3)
32#define CP_FLAG_UNK03_CLEAR           0
33#define CP_FLAG_UNK03_SET             1
34#define CP_FLAG_USER_SAVE             ((0 * 32) + 5)
35#define CP_FLAG_USER_SAVE_NOT_PENDING 0
36#define CP_FLAG_USER_SAVE_PENDING     1
37#define CP_FLAG_USER_LOAD             ((0 * 32) + 6)
38#define CP_FLAG_USER_LOAD_NOT_PENDING 0
39#define CP_FLAG_USER_LOAD_PENDING     1
40#define CP_FLAG_UNK0B                 ((0 * 32) + 0xb)
41#define CP_FLAG_UNK0B_CLEAR           0
42#define CP_FLAG_UNK0B_SET             1
43#define CP_FLAG_UNK1D                 ((0 * 32) + 0x1d)
44#define CP_FLAG_UNK1D_CLEAR           0
45#define CP_FLAG_UNK1D_SET             1
46#define CP_FLAG_UNK20                 ((1 * 32) + 0)
47#define CP_FLAG_UNK20_CLEAR           0
48#define CP_FLAG_UNK20_SET             1
49#define CP_FLAG_STATUS                ((2 * 32) + 0)
50#define CP_FLAG_STATUS_BUSY           0
51#define CP_FLAG_STATUS_IDLE           1
52#define CP_FLAG_AUTO_SAVE             ((2 * 32) + 4)
53#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
54#define CP_FLAG_AUTO_SAVE_PENDING     1
55#define CP_FLAG_AUTO_LOAD             ((2 * 32) + 5)
56#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
57#define CP_FLAG_AUTO_LOAD_PENDING     1
58#define CP_FLAG_NEWCTX                ((2 * 32) + 10)
59#define CP_FLAG_NEWCTX_BUSY           0
60#define CP_FLAG_NEWCTX_DONE           1
61#define CP_FLAG_XFER                  ((2 * 32) + 11)
62#define CP_FLAG_XFER_IDLE             0
63#define CP_FLAG_XFER_BUSY             1
64#define CP_FLAG_ALWAYS                ((2 * 32) + 13)
65#define CP_FLAG_ALWAYS_FALSE          0
66#define CP_FLAG_ALWAYS_TRUE           1
67#define CP_FLAG_INTR                  ((2 * 32) + 15)
68#define CP_FLAG_INTR_NOT_PENDING      0
69#define CP_FLAG_INTR_PENDING          1
70
71#define CP_CTX                   0x00100000
72#define CP_CTX_COUNT             0x000f0000
73#define CP_CTX_COUNT_SHIFT               16
74#define CP_CTX_REG               0x00003fff
75#define CP_LOAD_SR               0x00200000
76#define CP_LOAD_SR_VALUE         0x000fffff
77#define CP_BRA                   0x00400000
78#define CP_BRA_IP                0x0001ff00
79#define CP_BRA_IP_SHIFT                   8
80#define CP_BRA_IF_CLEAR          0x00000080
81#define CP_BRA_FLAG              0x0000007f
82#define CP_WAIT                  0x00500000
83#define CP_WAIT_SET              0x00000080
84#define CP_WAIT_FLAG             0x0000007f
85#define CP_SET                   0x00700000
86#define CP_SET_1                 0x00000080
87#define CP_SET_FLAG              0x0000007f
88#define CP_NEWCTX                0x00600004
89#define CP_NEXT_TO_SWAP          0x00600005
90#define CP_SET_CONTEXT_POINTER   0x00600006
91#define CP_SET_XFER_POINTER      0x00600007
92#define CP_ENABLE                0x00600009
93#define CP_END                   0x0060000c
94#define CP_NEXT_TO_CURRENT       0x0060000d
95#define CP_DISABLE1              0x0090ffff
96#define CP_DISABLE2              0x0091ffff
97#define CP_XFER_1      0x008000ff
98#define CP_XFER_2      0x008800ff
99#define CP_SEEK_1      0x00c000ff
100#define CP_SEEK_2      0x00c800ff
101
102#include "drmP.h"
103#include "nouveau_drv.h"
104#include "nouveau_grctx.h"
105
106#define IS_NVA3F(x) (((x) > 0xa0 && (x) < 0xaa) || (x) == 0xaf)
107#define IS_NVAAF(x) ((x) >= 0xaa && (x) <= 0xac)
108
109/*
110 * This code deals with PGRAPH contexts on NV50 family cards. Like NV40, it's
111 * the GPU itself that does context-switching, but it needs a special
112 * microcode to do it. And it's the driver's task to supply this microcode,
113 * further known as ctxprog, as well as the initial context values, known
114 * as ctxvals.
115 *
116 * Without ctxprog, you cannot switch contexts. Not even in software, since
117 * the majority of context [xfer strands] isn't accessible directly. You're
118 * stuck with a single channel, and you also suffer all the problems resulting
119 * from missing ctxvals, since you cannot load them.
120 *
121 * Without ctxvals, you're stuck with PGRAPH's default context. It's enough to
122 * run 2d operations, but trying to utilise 3d or CUDA will just lock you up,
123 * since you don't have... some sort of needed setup.
124 *
125 * Nouveau will just disable acceleration if not given ctxprog + ctxvals, since
126 * it's too much hassle to handle no-ctxprog as a special case.
127 */
128
129/*
130 * How ctxprogs work.
131 *
132 * The ctxprog is written in its own kind of microcode, with very small and
133 * crappy set of available commands. You upload it to a small [512 insns]
134 * area of memory on PGRAPH, and it'll be run when PFIFO wants PGRAPH to
135 * switch channel. or when the driver explicitely requests it. Stuff visible
136 * to ctxprog consists of: PGRAPH MMIO registers, PGRAPH context strands,
137 * the per-channel context save area in VRAM [known as ctxvals or grctx],
138 * 4 flags registers, a scratch register, two grctx pointers, plus many
139 * random poorly-understood details.
140 *
141 * When ctxprog runs, it's supposed to check what operations are asked of it,
142 * save old context if requested, optionally reset PGRAPH and switch to the
143 * new channel, and load the new context. Context consists of three major
144 * parts: subset of MMIO registers and two "xfer areas".
145 */
146
147/* TODO:
148 *  - document unimplemented bits compared to nvidia
149 *  - NVAx: make a TP subroutine, use it.
150 *  - use 0x4008fc instead of 0x1540?
151 */
152
153enum cp_label {
154	cp_check_load = 1,
155	cp_setup_auto_load,
156	cp_setup_load,
157	cp_setup_save,
158	cp_swap_state,
159	cp_prepare_exit,
160	cp_exit,
161};
162
163static void nv50_graph_construct_mmio(struct nouveau_grctx *ctx);
164static void nv50_graph_construct_xfer1(struct nouveau_grctx *ctx);
165static void nv50_graph_construct_xfer2(struct nouveau_grctx *ctx);
166
167/* Main function: construct the ctxprog skeleton, call the other functions. */
168
169int
170nv50_grctx_init(struct nouveau_grctx *ctx)
171{
172	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
173
174	switch (dev_priv->chipset) {
175	case 0x50:
176	case 0x84:
177	case 0x86:
178	case 0x92:
179	case 0x94:
180	case 0x96:
181	case 0x98:
182	case 0xa0:
183	case 0xa3:
184	case 0xa5:
185	case 0xa8:
186	case 0xaa:
187	case 0xac:
188	case 0xaf:
189		break;
190	default:
191		NV_ERROR(ctx->dev, "I don't know how to make a ctxprog for "
192				   "your NV%x card.\n", dev_priv->chipset);
193		NV_ERROR(ctx->dev, "Disabling acceleration. Please contact "
194				   "the devs.\n");
195		return -ENOSYS;
196	}
197	/* decide whether we're loading/unloading the context */
198	cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
199	cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
200
201	cp_name(ctx, cp_check_load);
202	cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
203	cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
204	cp_bra (ctx, ALWAYS, TRUE, cp_exit);
205
206	/* setup for context load */
207	cp_name(ctx, cp_setup_auto_load);
208	cp_out (ctx, CP_DISABLE1);
209	cp_out (ctx, CP_DISABLE2);
210	cp_out (ctx, CP_ENABLE);
211	cp_out (ctx, CP_NEXT_TO_SWAP);
212	cp_set (ctx, UNK01, SET);
213	cp_name(ctx, cp_setup_load);
214	cp_out (ctx, CP_NEWCTX);
215	cp_wait(ctx, NEWCTX, BUSY);
216	cp_set (ctx, UNK1D, CLEAR);
217	cp_set (ctx, SWAP_DIRECTION, LOAD);
218	cp_bra (ctx, UNK0B, SET, cp_prepare_exit);
219	cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
220
221	/* setup for context save */
222	cp_name(ctx, cp_setup_save);
223	cp_set (ctx, UNK1D, SET);
224	cp_wait(ctx, STATUS, BUSY);
225	cp_wait(ctx, INTR, PENDING);
226	cp_bra (ctx, STATUS, BUSY, cp_setup_save);
227	cp_set (ctx, UNK01, SET);
228	cp_set (ctx, SWAP_DIRECTION, SAVE);
229
230	/* general PGRAPH state */
231	cp_name(ctx, cp_swap_state);
232	cp_set (ctx, UNK03, SET);
233	cp_pos (ctx, 0x00004/4);
234	cp_ctx (ctx, 0x400828, 1); /* needed. otherwise, flickering happens. */
235	cp_pos (ctx, 0x00100/4);
236	nv50_graph_construct_mmio(ctx);
237	nv50_graph_construct_xfer1(ctx);
238	nv50_graph_construct_xfer2(ctx);
239
240	cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
241
242	cp_set (ctx, UNK20, SET);
243	cp_set (ctx, SWAP_DIRECTION, SAVE); /* no idea why this is needed, but fixes at least one lockup. */
244	cp_lsr (ctx, ctx->ctxvals_base);
245	cp_out (ctx, CP_SET_XFER_POINTER);
246	cp_lsr (ctx, 4);
247	cp_out (ctx, CP_SEEK_1);
248	cp_out (ctx, CP_XFER_1);
249	cp_wait(ctx, XFER, BUSY);
250
251	/* pre-exit state updates */
252	cp_name(ctx, cp_prepare_exit);
253	cp_set (ctx, UNK01, CLEAR);
254	cp_set (ctx, UNK03, CLEAR);
255	cp_set (ctx, UNK1D, CLEAR);
256
257	cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
258	cp_out (ctx, CP_NEXT_TO_CURRENT);
259
260	cp_name(ctx, cp_exit);
261	cp_set (ctx, USER_SAVE, NOT_PENDING);
262	cp_set (ctx, USER_LOAD, NOT_PENDING);
263	cp_out (ctx, CP_END);
264	ctx->ctxvals_pos += 0x400; /* padding... no idea why you need it */
265
266	return 0;
267}
268
269/*
270 * Constructs MMIO part of ctxprog and ctxvals. Just a matter of knowing which
271 * registers to save/restore and the default values for them.
272 */
273
274static void
275nv50_graph_construct_mmio_ddata(struct nouveau_grctx *ctx);
276
277static void
278nv50_graph_construct_mmio(struct nouveau_grctx *ctx)
279{
280	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
281	int i, j;
282	int offset, base;
283	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
284
285	/* 0800: DISPATCH */
286	cp_ctx(ctx, 0x400808, 7);
287	gr_def(ctx, 0x400814, 0x00000030);
288	cp_ctx(ctx, 0x400834, 0x32);
289	if (dev_priv->chipset == 0x50) {
290		gr_def(ctx, 0x400834, 0xff400040);
291		gr_def(ctx, 0x400838, 0xfff00080);
292		gr_def(ctx, 0x40083c, 0xfff70090);
293		gr_def(ctx, 0x400840, 0xffe806a8);
294	}
295	gr_def(ctx, 0x400844, 0x00000002);
296	if (IS_NVA3F(dev_priv->chipset))
297		gr_def(ctx, 0x400894, 0x00001000);
298	gr_def(ctx, 0x4008e8, 0x00000003);
299	gr_def(ctx, 0x4008ec, 0x00001000);
300	if (dev_priv->chipset == 0x50)
301		cp_ctx(ctx, 0x400908, 0xb);
302	else if (dev_priv->chipset < 0xa0)
303		cp_ctx(ctx, 0x400908, 0xc);
304	else
305		cp_ctx(ctx, 0x400908, 0xe);
306
307	if (dev_priv->chipset >= 0xa0)
308		cp_ctx(ctx, 0x400b00, 0x1);
309	if (IS_NVA3F(dev_priv->chipset)) {
310		cp_ctx(ctx, 0x400b10, 0x1);
311		gr_def(ctx, 0x400b10, 0x0001629d);
312		cp_ctx(ctx, 0x400b20, 0x1);
313		gr_def(ctx, 0x400b20, 0x0001629d);
314	}
315
316	nv50_graph_construct_mmio_ddata(ctx);
317
318	/* 0C00: VFETCH */
319	cp_ctx(ctx, 0x400c08, 0x2);
320	gr_def(ctx, 0x400c08, 0x0000fe0c);
321
322	/* 1000 */
323	if (dev_priv->chipset < 0xa0) {
324		cp_ctx(ctx, 0x401008, 0x4);
325		gr_def(ctx, 0x401014, 0x00001000);
326	} else if (!IS_NVA3F(dev_priv->chipset)) {
327		cp_ctx(ctx, 0x401008, 0x5);
328		gr_def(ctx, 0x401018, 0x00001000);
329	} else {
330		cp_ctx(ctx, 0x401008, 0x5);
331		gr_def(ctx, 0x401018, 0x00004000);
332	}
333
334	/* 1400 */
335	cp_ctx(ctx, 0x401400, 0x8);
336	cp_ctx(ctx, 0x401424, 0x3);
337	if (dev_priv->chipset == 0x50)
338		gr_def(ctx, 0x40142c, 0x0001fd87);
339	else
340		gr_def(ctx, 0x40142c, 0x00000187);
341	cp_ctx(ctx, 0x401540, 0x5);
342	gr_def(ctx, 0x401550, 0x00001018);
343
344	/* 1800: STREAMOUT */
345	cp_ctx(ctx, 0x401814, 0x1);
346	gr_def(ctx, 0x401814, 0x000000ff);
347	if (dev_priv->chipset == 0x50) {
348		cp_ctx(ctx, 0x40181c, 0xe);
349		gr_def(ctx, 0x401850, 0x00000004);
350	} else if (dev_priv->chipset < 0xa0) {
351		cp_ctx(ctx, 0x40181c, 0xf);
352		gr_def(ctx, 0x401854, 0x00000004);
353	} else {
354		cp_ctx(ctx, 0x40181c, 0x13);
355		gr_def(ctx, 0x401864, 0x00000004);
356	}
357
358	/* 1C00 */
359	cp_ctx(ctx, 0x401c00, 0x1);
360	switch (dev_priv->chipset) {
361	case 0x50:
362		gr_def(ctx, 0x401c00, 0x0001005f);
363		break;
364	case 0x84:
365	case 0x86:
366	case 0x94:
367		gr_def(ctx, 0x401c00, 0x044d00df);
368		break;
369	case 0x92:
370	case 0x96:
371	case 0x98:
372	case 0xa0:
373	case 0xaa:
374	case 0xac:
375		gr_def(ctx, 0x401c00, 0x042500df);
376		break;
377	case 0xa3:
378	case 0xa5:
379	case 0xa8:
380	case 0xaf:
381		gr_def(ctx, 0x401c00, 0x142500df);
382		break;
383	}
384
385	/* 2000 */
386
387	/* 2400 */
388	cp_ctx(ctx, 0x402400, 0x1);
389	if (dev_priv->chipset == 0x50)
390		cp_ctx(ctx, 0x402408, 0x1);
391	else
392		cp_ctx(ctx, 0x402408, 0x2);
393	gr_def(ctx, 0x402408, 0x00000600);
394
395	/* 2800: CSCHED */
396	cp_ctx(ctx, 0x402800, 0x1);
397	if (dev_priv->chipset == 0x50)
398		gr_def(ctx, 0x402800, 0x00000006);
399
400	/* 2C00: ZCULL */
401	cp_ctx(ctx, 0x402c08, 0x6);
402	if (dev_priv->chipset != 0x50)
403		gr_def(ctx, 0x402c14, 0x01000000);
404	gr_def(ctx, 0x402c18, 0x000000ff);
405	if (dev_priv->chipset == 0x50)
406		cp_ctx(ctx, 0x402ca0, 0x1);
407	else
408		cp_ctx(ctx, 0x402ca0, 0x2);
409	if (dev_priv->chipset < 0xa0)
410		gr_def(ctx, 0x402ca0, 0x00000400);
411	else if (!IS_NVA3F(dev_priv->chipset))
412		gr_def(ctx, 0x402ca0, 0x00000800);
413	else
414		gr_def(ctx, 0x402ca0, 0x00000400);
415	cp_ctx(ctx, 0x402cac, 0x4);
416
417	/* 3000: ENG2D */
418	cp_ctx(ctx, 0x403004, 0x1);
419	gr_def(ctx, 0x403004, 0x00000001);
420
421	/* 3400 */
422	if (dev_priv->chipset >= 0xa0) {
423		cp_ctx(ctx, 0x403404, 0x1);
424		gr_def(ctx, 0x403404, 0x00000001);
425	}
426
427	/* 5000: CCACHE */
428	cp_ctx(ctx, 0x405000, 0x1);
429	switch (dev_priv->chipset) {
430	case 0x50:
431		gr_def(ctx, 0x405000, 0x00300080);
432		break;
433	case 0x84:
434	case 0xa0:
435	case 0xa3:
436	case 0xa5:
437	case 0xa8:
438	case 0xaa:
439	case 0xac:
440	case 0xaf:
441		gr_def(ctx, 0x405000, 0x000e0080);
442		break;
443	case 0x86:
444	case 0x92:
445	case 0x94:
446	case 0x96:
447	case 0x98:
448		gr_def(ctx, 0x405000, 0x00000080);
449		break;
450	}
451	cp_ctx(ctx, 0x405014, 0x1);
452	gr_def(ctx, 0x405014, 0x00000004);
453	cp_ctx(ctx, 0x40501c, 0x1);
454	cp_ctx(ctx, 0x405024, 0x1);
455	cp_ctx(ctx, 0x40502c, 0x1);
456
457	/* 6000? */
458	if (dev_priv->chipset == 0x50)
459		cp_ctx(ctx, 0x4063e0, 0x1);
460
461	/* 6800: M2MF */
462	if (dev_priv->chipset < 0x90) {
463		cp_ctx(ctx, 0x406814, 0x2b);
464		gr_def(ctx, 0x406818, 0x00000f80);
465		gr_def(ctx, 0x406860, 0x007f0080);
466		gr_def(ctx, 0x40689c, 0x007f0080);
467	} else {
468		cp_ctx(ctx, 0x406814, 0x4);
469		if (dev_priv->chipset == 0x98)
470			gr_def(ctx, 0x406818, 0x00000f80);
471		else
472			gr_def(ctx, 0x406818, 0x00001f80);
473		if (IS_NVA3F(dev_priv->chipset))
474			gr_def(ctx, 0x40681c, 0x00000030);
475		cp_ctx(ctx, 0x406830, 0x3);
476	}
477
478	/* 7000: per-ROP group state */
479	for (i = 0; i < 8; i++) {
480		if (units & (1<<(i+16))) {
481			cp_ctx(ctx, 0x407000 + (i<<8), 3);
482			if (dev_priv->chipset == 0x50)
483				gr_def(ctx, 0x407000 + (i<<8), 0x1b74f820);
484			else if (dev_priv->chipset != 0xa5)
485				gr_def(ctx, 0x407000 + (i<<8), 0x3b74f821);
486			else
487				gr_def(ctx, 0x407000 + (i<<8), 0x7b74f821);
488			gr_def(ctx, 0x407004 + (i<<8), 0x89058001);
489
490			if (dev_priv->chipset == 0x50) {
491				cp_ctx(ctx, 0x407010 + (i<<8), 1);
492			} else if (dev_priv->chipset < 0xa0) {
493				cp_ctx(ctx, 0x407010 + (i<<8), 2);
494				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
495				gr_def(ctx, 0x407014 + (i<<8), 0x0000001f);
496			} else {
497				cp_ctx(ctx, 0x407010 + (i<<8), 3);
498				gr_def(ctx, 0x407010 + (i<<8), 0x00001000);
499				if (dev_priv->chipset != 0xa5)
500					gr_def(ctx, 0x407014 + (i<<8), 0x000000ff);
501				else
502					gr_def(ctx, 0x407014 + (i<<8), 0x000001ff);
503			}
504
505			cp_ctx(ctx, 0x407080 + (i<<8), 4);
506			if (dev_priv->chipset != 0xa5)
507				gr_def(ctx, 0x407080 + (i<<8), 0x027c10fa);
508			else
509				gr_def(ctx, 0x407080 + (i<<8), 0x827c10fa);
510			if (dev_priv->chipset == 0x50)
511				gr_def(ctx, 0x407084 + (i<<8), 0x000000c0);
512			else
513				gr_def(ctx, 0x407084 + (i<<8), 0x400000c0);
514			gr_def(ctx, 0x407088 + (i<<8), 0xb7892080);
515
516			if (dev_priv->chipset < 0xa0)
517				cp_ctx(ctx, 0x407094 + (i<<8), 1);
518			else if (!IS_NVA3F(dev_priv->chipset))
519				cp_ctx(ctx, 0x407094 + (i<<8), 3);
520			else {
521				cp_ctx(ctx, 0x407094 + (i<<8), 4);
522				gr_def(ctx, 0x4070a0 + (i<<8), 1);
523			}
524		}
525	}
526
527	cp_ctx(ctx, 0x407c00, 0x3);
528	if (dev_priv->chipset < 0x90)
529		gr_def(ctx, 0x407c00, 0x00010040);
530	else if (dev_priv->chipset < 0xa0)
531		gr_def(ctx, 0x407c00, 0x00390040);
532	else
533		gr_def(ctx, 0x407c00, 0x003d0040);
534	gr_def(ctx, 0x407c08, 0x00000022);
535	if (dev_priv->chipset >= 0xa0) {
536		cp_ctx(ctx, 0x407c10, 0x3);
537		cp_ctx(ctx, 0x407c20, 0x1);
538		cp_ctx(ctx, 0x407c2c, 0x1);
539	}
540
541	if (dev_priv->chipset < 0xa0) {
542		cp_ctx(ctx, 0x407d00, 0x9);
543	} else {
544		cp_ctx(ctx, 0x407d00, 0x15);
545	}
546	if (dev_priv->chipset == 0x98)
547		gr_def(ctx, 0x407d08, 0x00380040);
548	else {
549		if (dev_priv->chipset < 0x90)
550			gr_def(ctx, 0x407d08, 0x00010040);
551		else if (dev_priv->chipset < 0xa0)
552			gr_def(ctx, 0x407d08, 0x00390040);
553		else
554			gr_def(ctx, 0x407d08, 0x003d0040);
555		gr_def(ctx, 0x407d0c, 0x00000022);
556	}
557
558	/* 8000+: per-TP state */
559	for (i = 0; i < 10; i++) {
560		if (units & (1<<i)) {
561			if (dev_priv->chipset < 0xa0)
562				base = 0x408000 + (i<<12);
563			else
564				base = 0x408000 + (i<<11);
565			if (dev_priv->chipset < 0xa0)
566				offset = base + 0xc00;
567			else
568				offset = base + 0x80;
569			cp_ctx(ctx, offset + 0x00, 1);
570			gr_def(ctx, offset + 0x00, 0x0000ff0a);
571			cp_ctx(ctx, offset + 0x08, 1);
572
573			/* per-MP state */
574			for (j = 0; j < (dev_priv->chipset < 0xa0 ? 2 : 4); j++) {
575				if (!(units & (1 << (j+24)))) continue;
576				if (dev_priv->chipset < 0xa0)
577					offset = base + 0x200 + (j<<7);
578				else
579					offset = base + 0x100 + (j<<7);
580				cp_ctx(ctx, offset, 0x20);
581				gr_def(ctx, offset + 0x00, 0x01800000);
582				gr_def(ctx, offset + 0x04, 0x00160000);
583				gr_def(ctx, offset + 0x08, 0x01800000);
584				gr_def(ctx, offset + 0x18, 0x0003ffff);
585				switch (dev_priv->chipset) {
586				case 0x50:
587					gr_def(ctx, offset + 0x1c, 0x00080000);
588					break;
589				case 0x84:
590					gr_def(ctx, offset + 0x1c, 0x00880000);
591					break;
592				case 0x86:
593					gr_def(ctx, offset + 0x1c, 0x008c0000);
594					break;
595				case 0x92:
596				case 0x96:
597				case 0x98:
598					gr_def(ctx, offset + 0x1c, 0x118c0000);
599					break;
600				case 0x94:
601					gr_def(ctx, offset + 0x1c, 0x10880000);
602					break;
603				case 0xa0:
604				case 0xa5:
605					gr_def(ctx, offset + 0x1c, 0x310c0000);
606					break;
607				case 0xa3:
608				case 0xa8:
609				case 0xaa:
610				case 0xac:
611				case 0xaf:
612					gr_def(ctx, offset + 0x1c, 0x300c0000);
613					break;
614				}
615				gr_def(ctx, offset + 0x40, 0x00010401);
616				if (dev_priv->chipset == 0x50)
617					gr_def(ctx, offset + 0x48, 0x00000040);
618				else
619					gr_def(ctx, offset + 0x48, 0x00000078);
620				gr_def(ctx, offset + 0x50, 0x000000bf);
621				gr_def(ctx, offset + 0x58, 0x00001210);
622				if (dev_priv->chipset == 0x50)
623					gr_def(ctx, offset + 0x5c, 0x00000080);
624				else
625					gr_def(ctx, offset + 0x5c, 0x08000080);
626				if (dev_priv->chipset >= 0xa0)
627					gr_def(ctx, offset + 0x68, 0x0000003e);
628			}
629
630			if (dev_priv->chipset < 0xa0)
631				cp_ctx(ctx, base + 0x300, 0x4);
632			else
633				cp_ctx(ctx, base + 0x300, 0x5);
634			if (dev_priv->chipset == 0x50)
635				gr_def(ctx, base + 0x304, 0x00007070);
636			else if (dev_priv->chipset < 0xa0)
637				gr_def(ctx, base + 0x304, 0x00027070);
638			else if (!IS_NVA3F(dev_priv->chipset))
639				gr_def(ctx, base + 0x304, 0x01127070);
640			else
641				gr_def(ctx, base + 0x304, 0x05127070);
642
643			if (dev_priv->chipset < 0xa0)
644				cp_ctx(ctx, base + 0x318, 1);
645			else
646				cp_ctx(ctx, base + 0x320, 1);
647			if (dev_priv->chipset == 0x50)
648				gr_def(ctx, base + 0x318, 0x0003ffff);
649			else if (dev_priv->chipset < 0xa0)
650				gr_def(ctx, base + 0x318, 0x03ffffff);
651			else
652				gr_def(ctx, base + 0x320, 0x07ffffff);
653
654			if (dev_priv->chipset < 0xa0)
655				cp_ctx(ctx, base + 0x324, 5);
656			else
657				cp_ctx(ctx, base + 0x328, 4);
658
659			if (dev_priv->chipset < 0xa0) {
660				cp_ctx(ctx, base + 0x340, 9);
661				offset = base + 0x340;
662			} else if (!IS_NVA3F(dev_priv->chipset)) {
663				cp_ctx(ctx, base + 0x33c, 0xb);
664				offset = base + 0x344;
665			} else {
666				cp_ctx(ctx, base + 0x33c, 0xd);
667				offset = base + 0x344;
668			}
669			gr_def(ctx, offset + 0x0, 0x00120407);
670			gr_def(ctx, offset + 0x4, 0x05091507);
671			if (dev_priv->chipset == 0x84)
672				gr_def(ctx, offset + 0x8, 0x05100202);
673			else
674				gr_def(ctx, offset + 0x8, 0x05010202);
675			gr_def(ctx, offset + 0xc, 0x00030201);
676			if (dev_priv->chipset == 0xa3)
677				cp_ctx(ctx, base + 0x36c, 1);
678
679			cp_ctx(ctx, base + 0x400, 2);
680			gr_def(ctx, base + 0x404, 0x00000040);
681			cp_ctx(ctx, base + 0x40c, 2);
682			gr_def(ctx, base + 0x40c, 0x0d0c0b0a);
683			gr_def(ctx, base + 0x410, 0x00141210);
684
685			if (dev_priv->chipset < 0xa0)
686				offset = base + 0x800;
687			else
688				offset = base + 0x500;
689			cp_ctx(ctx, offset, 6);
690			gr_def(ctx, offset + 0x0, 0x000001f0);
691			gr_def(ctx, offset + 0x4, 0x00000001);
692			gr_def(ctx, offset + 0x8, 0x00000003);
693			if (dev_priv->chipset == 0x50 || IS_NVAAF(dev_priv->chipset))
694				gr_def(ctx, offset + 0xc, 0x00008000);
695			gr_def(ctx, offset + 0x14, 0x00039e00);
696			cp_ctx(ctx, offset + 0x1c, 2);
697			if (dev_priv->chipset == 0x50)
698				gr_def(ctx, offset + 0x1c, 0x00000040);
699			else
700				gr_def(ctx, offset + 0x1c, 0x00000100);
701			gr_def(ctx, offset + 0x20, 0x00003800);
702
703			if (dev_priv->chipset >= 0xa0) {
704				cp_ctx(ctx, base + 0x54c, 2);
705				if (!IS_NVA3F(dev_priv->chipset))
706					gr_def(ctx, base + 0x54c, 0x003fe006);
707				else
708					gr_def(ctx, base + 0x54c, 0x003fe007);
709				gr_def(ctx, base + 0x550, 0x003fe000);
710			}
711
712			if (dev_priv->chipset < 0xa0)
713				offset = base + 0xa00;
714			else
715				offset = base + 0x680;
716			cp_ctx(ctx, offset, 1);
717			gr_def(ctx, offset, 0x00404040);
718
719			if (dev_priv->chipset < 0xa0)
720				offset = base + 0xe00;
721			else
722				offset = base + 0x700;
723			cp_ctx(ctx, offset, 2);
724			if (dev_priv->chipset < 0xa0)
725				gr_def(ctx, offset, 0x0077f005);
726			else if (dev_priv->chipset == 0xa5)
727				gr_def(ctx, offset, 0x6cf7f007);
728			else if (dev_priv->chipset == 0xa8)
729				gr_def(ctx, offset, 0x6cfff007);
730			else if (dev_priv->chipset == 0xac)
731				gr_def(ctx, offset, 0x0cfff007);
732			else
733				gr_def(ctx, offset, 0x0cf7f007);
734			if (dev_priv->chipset == 0x50)
735				gr_def(ctx, offset + 0x4, 0x00007fff);
736			else if (dev_priv->chipset < 0xa0)
737				gr_def(ctx, offset + 0x4, 0x003f7fff);
738			else
739				gr_def(ctx, offset + 0x4, 0x02bf7fff);
740			cp_ctx(ctx, offset + 0x2c, 1);
741			if (dev_priv->chipset == 0x50) {
742				cp_ctx(ctx, offset + 0x50, 9);
743				gr_def(ctx, offset + 0x54, 0x000003ff);
744				gr_def(ctx, offset + 0x58, 0x00000003);
745				gr_def(ctx, offset + 0x5c, 0x00000003);
746				gr_def(ctx, offset + 0x60, 0x000001ff);
747				gr_def(ctx, offset + 0x64, 0x0000001f);
748				gr_def(ctx, offset + 0x68, 0x0000000f);
749				gr_def(ctx, offset + 0x6c, 0x0000000f);
750			} else if(dev_priv->chipset < 0xa0) {
751				cp_ctx(ctx, offset + 0x50, 1);
752				cp_ctx(ctx, offset + 0x70, 1);
753			} else {
754				cp_ctx(ctx, offset + 0x50, 1);
755				cp_ctx(ctx, offset + 0x60, 5);
756			}
757		}
758	}
759}
760
761static void
762dd_emit(struct nouveau_grctx *ctx, int num, uint32_t val) {
763	int i;
764	if (val && ctx->mode == NOUVEAU_GRCTX_VALS)
765		for (i = 0; i < num; i++)
766			nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + i), val);
767	ctx->ctxvals_pos += num;
768}
769
770static void
771nv50_graph_construct_mmio_ddata(struct nouveau_grctx *ctx)
772{
773	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
774	int base, num;
775	base = ctx->ctxvals_pos;
776
777	/* tesla state */
778	dd_emit(ctx, 1, 0);	/* 00000001 UNK0F90 */
779	dd_emit(ctx, 1, 0);	/* 00000001 UNK135C */
780
781	/* SRC_TIC state */
782	dd_emit(ctx, 1, 0);	/* 00000007 SRC_TILE_MODE_Z */
783	dd_emit(ctx, 1, 2);	/* 00000007 SRC_TILE_MODE_Y */
784	dd_emit(ctx, 1, 1);	/* 00000001 SRC_LINEAR #1 */
785	dd_emit(ctx, 1, 0);	/* 000000ff SRC_ADDRESS_HIGH */
786	dd_emit(ctx, 1, 0);	/* 00000001 SRC_SRGB */
787	if (dev_priv->chipset >= 0x94)
788		dd_emit(ctx, 1, 0);	/* 00000003 eng2d UNK0258 */
789	dd_emit(ctx, 1, 1);	/* 00000fff SRC_DEPTH */
790	dd_emit(ctx, 1, 0x100);	/* 0000ffff SRC_HEIGHT */
791
792	/* turing state */
793	dd_emit(ctx, 1, 0);		/* 0000000f TEXTURES_LOG2 */
794	dd_emit(ctx, 1, 0);		/* 0000000f SAMPLERS_LOG2 */
795	dd_emit(ctx, 1, 0);		/* 000000ff CB_DEF_ADDRESS_HIGH */
796	dd_emit(ctx, 1, 0);		/* ffffffff CB_DEF_ADDRESS_LOW */
797	dd_emit(ctx, 1, 0);		/* ffffffff SHARED_SIZE */
798	dd_emit(ctx, 1, 2);		/* ffffffff REG_MODE */
799	dd_emit(ctx, 1, 1);		/* 0000ffff BLOCK_ALLOC_THREADS */
800	dd_emit(ctx, 1, 1);		/* 00000001 LANES32 */
801	dd_emit(ctx, 1, 0);		/* 000000ff UNK370 */
802	dd_emit(ctx, 1, 0);		/* 000000ff USER_PARAM_UNK */
803	dd_emit(ctx, 1, 0);		/* 000000ff USER_PARAM_COUNT */
804	dd_emit(ctx, 1, 1);		/* 000000ff UNK384 bits 8-15 */
805	dd_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
806	dd_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
807	dd_emit(ctx, 1, 0);		/* 0000ffff CB_ADDR_INDEX */
808	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_X */
809	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_XMY */
810	dd_emit(ctx, 1, 0);		/* 00000001 BLOCKDIM_XMY_OVERFLOW */
811	dd_emit(ctx, 1, 1);		/* 0003ffff BLOCKDIM_XMYMZ */
812	dd_emit(ctx, 1, 1);		/* 000007ff BLOCKDIM_Y */
813	dd_emit(ctx, 1, 1);		/* 0000007f BLOCKDIM_Z */
814	dd_emit(ctx, 1, 4);		/* 000000ff CP_REG_ALLOC_TEMP */
815	dd_emit(ctx, 1, 1);		/* 00000001 BLOCKDIM_DIRTY */
816	if (IS_NVA3F(dev_priv->chipset))
817		dd_emit(ctx, 1, 0);	/* 00000003 UNK03E8 */
818	dd_emit(ctx, 1, 1);		/* 0000007f BLOCK_ALLOC_HALFWARPS */
819	dd_emit(ctx, 1, 1);		/* 00000007 LOCAL_WARPS_NO_CLAMP */
820	dd_emit(ctx, 1, 7);		/* 00000007 LOCAL_WARPS_LOG_ALLOC */
821	dd_emit(ctx, 1, 1);		/* 00000007 STACK_WARPS_NO_CLAMP */
822	dd_emit(ctx, 1, 7);		/* 00000007 STACK_WARPS_LOG_ALLOC */
823	dd_emit(ctx, 1, 1);		/* 00001fff BLOCK_ALLOC_REGSLOTS_PACKED */
824	dd_emit(ctx, 1, 1);		/* 00001fff BLOCK_ALLOC_REGSLOTS_STRIDED */
825	dd_emit(ctx, 1, 1);		/* 000007ff BLOCK_ALLOC_THREADS */
826
827	/* compat 2d state */
828	if (dev_priv->chipset == 0x50) {
829		dd_emit(ctx, 4, 0);		/* 0000ffff clip X, Y, W, H */
830
831		dd_emit(ctx, 1, 1);		/* ffffffff chroma COLOR_FORMAT */
832
833		dd_emit(ctx, 1, 1);		/* ffffffff pattern COLOR_FORMAT */
834		dd_emit(ctx, 1, 0);		/* ffffffff pattern SHAPE */
835		dd_emit(ctx, 1, 1);		/* ffffffff pattern PATTERN_SELECT */
836
837		dd_emit(ctx, 1, 0xa);		/* ffffffff surf2d SRC_FORMAT */
838		dd_emit(ctx, 1, 0);		/* ffffffff surf2d DMA_SRC */
839		dd_emit(ctx, 1, 0);		/* 000000ff surf2d SRC_ADDRESS_HIGH */
840		dd_emit(ctx, 1, 0);		/* ffffffff surf2d SRC_ADDRESS_LOW */
841		dd_emit(ctx, 1, 0x40);		/* 0000ffff surf2d SRC_PITCH */
842		dd_emit(ctx, 1, 0);		/* 0000000f surf2d SRC_TILE_MODE_Z */
843		dd_emit(ctx, 1, 2);		/* 0000000f surf2d SRC_TILE_MODE_Y */
844		dd_emit(ctx, 1, 0x100);		/* ffffffff surf2d SRC_HEIGHT */
845		dd_emit(ctx, 1, 1);		/* 00000001 surf2d SRC_LINEAR */
846		dd_emit(ctx, 1, 0x100);		/* ffffffff surf2d SRC_WIDTH */
847
848		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_B_X */
849		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_B_Y */
850		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_C_X */
851		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_C_Y */
852		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_D_X */
853		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect CLIP_D_Y */
854		dd_emit(ctx, 1, 1);		/* ffffffff gdirect COLOR_FORMAT */
855		dd_emit(ctx, 1, 0);		/* ffffffff gdirect OPERATION */
856		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect POINT_X */
857		dd_emit(ctx, 1, 0);		/* 0000ffff gdirect POINT_Y */
858
859		dd_emit(ctx, 1, 0);		/* 0000ffff blit SRC_Y */
860		dd_emit(ctx, 1, 0);		/* ffffffff blit OPERATION */
861
862		dd_emit(ctx, 1, 0);		/* ffffffff ifc OPERATION */
863
864		dd_emit(ctx, 1, 0);		/* ffffffff iifc INDEX_FORMAT */
865		dd_emit(ctx, 1, 0);		/* ffffffff iifc LUT_OFFSET */
866		dd_emit(ctx, 1, 4);		/* ffffffff iifc COLOR_FORMAT */
867		dd_emit(ctx, 1, 0);		/* ffffffff iifc OPERATION */
868	}
869
870	/* m2mf state */
871	dd_emit(ctx, 1, 0);		/* ffffffff m2mf LINE_COUNT */
872	dd_emit(ctx, 1, 0);		/* ffffffff m2mf LINE_LENGTH_IN */
873	dd_emit(ctx, 2, 0);		/* ffffffff m2mf OFFSET_IN, OFFSET_OUT */
874	dd_emit(ctx, 1, 1);		/* ffffffff m2mf TILING_DEPTH_OUT */
875	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_HEIGHT_OUT */
876	dd_emit(ctx, 1, 0);		/* ffffffff m2mf TILING_POSITION_OUT_Z */
877	dd_emit(ctx, 1, 1);		/* 00000001 m2mf LINEAR_OUT */
878	dd_emit(ctx, 2, 0);		/* 0000ffff m2mf TILING_POSITION_OUT_X, Y */
879	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_PITCH_OUT */
880	dd_emit(ctx, 1, 1);		/* ffffffff m2mf TILING_DEPTH_IN */
881	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_HEIGHT_IN */
882	dd_emit(ctx, 1, 0);		/* ffffffff m2mf TILING_POSITION_IN_Z */
883	dd_emit(ctx, 1, 1);		/* 00000001 m2mf LINEAR_IN */
884	dd_emit(ctx, 2, 0);		/* 0000ffff m2mf TILING_POSITION_IN_X, Y */
885	dd_emit(ctx, 1, 0x100);		/* ffffffff m2mf TILING_PITCH_IN */
886
887	/* more compat 2d state */
888	if (dev_priv->chipset == 0x50) {
889		dd_emit(ctx, 1, 1);		/* ffffffff line COLOR_FORMAT */
890		dd_emit(ctx, 1, 0);		/* ffffffff line OPERATION */
891
892		dd_emit(ctx, 1, 1);		/* ffffffff triangle COLOR_FORMAT */
893		dd_emit(ctx, 1, 0);		/* ffffffff triangle OPERATION */
894
895		dd_emit(ctx, 1, 0);		/* 0000000f sifm TILE_MODE_Z */
896		dd_emit(ctx, 1, 2);		/* 0000000f sifm TILE_MODE_Y */
897		dd_emit(ctx, 1, 0);		/* 000000ff sifm FORMAT_FILTER */
898		dd_emit(ctx, 1, 1);		/* 000000ff sifm FORMAT_ORIGIN */
899		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_PITCH */
900		dd_emit(ctx, 1, 1);		/* 00000001 sifm SRC_LINEAR */
901		dd_emit(ctx, 1, 0);		/* 000000ff sifm SRC_OFFSET_HIGH */
902		dd_emit(ctx, 1, 0);		/* ffffffff sifm SRC_OFFSET */
903		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_HEIGHT */
904		dd_emit(ctx, 1, 0);		/* 0000ffff sifm SRC_WIDTH */
905		dd_emit(ctx, 1, 3);		/* ffffffff sifm COLOR_FORMAT */
906		dd_emit(ctx, 1, 0);		/* ffffffff sifm OPERATION */
907
908		dd_emit(ctx, 1, 0);		/* ffffffff sifc OPERATION */
909	}
910
911	/* tesla state */
912	dd_emit(ctx, 1, 0);		/* 0000000f GP_TEXTURES_LOG2 */
913	dd_emit(ctx, 1, 0);		/* 0000000f GP_SAMPLERS_LOG2 */
914	dd_emit(ctx, 1, 0);		/* 000000ff */
915	dd_emit(ctx, 1, 0);		/* ffffffff */
916	dd_emit(ctx, 1, 4);		/* 000000ff UNK12B0_0 */
917	dd_emit(ctx, 1, 0x70);		/* 000000ff UNK12B0_1 */
918	dd_emit(ctx, 1, 0x80);		/* 000000ff UNK12B0_3 */
919	dd_emit(ctx, 1, 0);		/* 000000ff UNK12B0_2 */
920	dd_emit(ctx, 1, 0);		/* 0000000f FP_TEXTURES_LOG2 */
921	dd_emit(ctx, 1, 0);		/* 0000000f FP_SAMPLERS_LOG2 */
922	if (IS_NVA3F(dev_priv->chipset)) {
923		dd_emit(ctx, 1, 0);	/* ffffffff */
924		dd_emit(ctx, 1, 0);	/* 0000007f MULTISAMPLE_SAMPLES_LOG2 */
925	} else {
926		dd_emit(ctx, 1, 0);	/* 0000000f MULTISAMPLE_SAMPLES_LOG2 */
927	}
928	dd_emit(ctx, 1, 0xc);		/* 000000ff SEMANTIC_COLOR.BFC0_ID */
929	if (dev_priv->chipset != 0x50)
930		dd_emit(ctx, 1, 0);	/* 00000001 SEMANTIC_COLOR.CLMP_EN */
931	dd_emit(ctx, 1, 8);		/* 000000ff SEMANTIC_COLOR.COLR_NR */
932	dd_emit(ctx, 1, 0x14);		/* 000000ff SEMANTIC_COLOR.FFC0_ID */
933	if (dev_priv->chipset == 0x50) {
934		dd_emit(ctx, 1, 0);	/* 000000ff SEMANTIC_LAYER */
935		dd_emit(ctx, 1, 0);	/* 00000001 */
936	} else {
937		dd_emit(ctx, 1, 0);	/* 00000001 SEMANTIC_PTSZ.ENABLE */
938		dd_emit(ctx, 1, 0x29);	/* 000000ff SEMANTIC_PTSZ.PTSZ_ID */
939		dd_emit(ctx, 1, 0x27);	/* 000000ff SEMANTIC_PRIM */
940		dd_emit(ctx, 1, 0x26);	/* 000000ff SEMANTIC_LAYER */
941		dd_emit(ctx, 1, 8);	/* 0000000f SMENATIC_CLIP.CLIP_HIGH */
942		dd_emit(ctx, 1, 4);	/* 000000ff SEMANTIC_CLIP.CLIP_LO */
943		dd_emit(ctx, 1, 0x27);	/* 000000ff UNK0FD4 */
944		dd_emit(ctx, 1, 0);	/* 00000001 UNK1900 */
945	}
946	dd_emit(ctx, 1, 0);		/* 00000007 RT_CONTROL_MAP0 */
947	dd_emit(ctx, 1, 1);		/* 00000007 RT_CONTROL_MAP1 */
948	dd_emit(ctx, 1, 2);		/* 00000007 RT_CONTROL_MAP2 */
949	dd_emit(ctx, 1, 3);		/* 00000007 RT_CONTROL_MAP3 */
950	dd_emit(ctx, 1, 4);		/* 00000007 RT_CONTROL_MAP4 */
951	dd_emit(ctx, 1, 5);		/* 00000007 RT_CONTROL_MAP5 */
952	dd_emit(ctx, 1, 6);		/* 00000007 RT_CONTROL_MAP6 */
953	dd_emit(ctx, 1, 7);		/* 00000007 RT_CONTROL_MAP7 */
954	dd_emit(ctx, 1, 1);		/* 0000000f RT_CONTROL_COUNT */
955	dd_emit(ctx, 8, 0);		/* 00000001 RT_HORIZ_UNK */
956	dd_emit(ctx, 8, 0);		/* ffffffff RT_ADDRESS_LOW */
957	dd_emit(ctx, 1, 0xcf);		/* 000000ff RT_FORMAT */
958	dd_emit(ctx, 7, 0);		/* 000000ff RT_FORMAT */
959	if (dev_priv->chipset != 0x50)
960		dd_emit(ctx, 3, 0);	/* 1, 1, 1 */
961	else
962		dd_emit(ctx, 2, 0);	/* 1, 1 */
963	dd_emit(ctx, 1, 0);		/* ffffffff GP_ENABLE */
964	dd_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT*/
965	dd_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
966	dd_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
967	if (IS_NVA3F(dev_priv->chipset)) {
968		dd_emit(ctx, 1, 3);	/* 00000003 */
969		dd_emit(ctx, 1, 0);	/* 00000001 UNK1418. Alone. */
970	}
971	if (dev_priv->chipset != 0x50)
972		dd_emit(ctx, 1, 3);	/* 00000003 UNK15AC */
973	dd_emit(ctx, 1, 1);		/* ffffffff RASTERIZE_ENABLE */
974	dd_emit(ctx, 1, 0);		/* 00000001 FP_CONTROL.EXPORTS_Z */
975	if (dev_priv->chipset != 0x50)
976		dd_emit(ctx, 1, 0);	/* 00000001 FP_CONTROL.MULTIPLE_RESULTS */
977	dd_emit(ctx, 1, 0x12);		/* 000000ff FP_INTERPOLANT_CTRL.COUNT */
978	dd_emit(ctx, 1, 0x10);		/* 000000ff FP_INTERPOLANT_CTRL.COUNT_NONFLAT */
979	dd_emit(ctx, 1, 0xc);		/* 000000ff FP_INTERPOLANT_CTRL.OFFSET */
980	dd_emit(ctx, 1, 1);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.W */
981	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.X */
982	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.Y */
983	dd_emit(ctx, 1, 0);		/* 00000001 FP_INTERPOLANT_CTRL.UMASK.Z */
984	dd_emit(ctx, 1, 4);		/* 000000ff FP_RESULT_COUNT */
985	dd_emit(ctx, 1, 2);		/* ffffffff REG_MODE */
986	dd_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
987	if (dev_priv->chipset >= 0xa0)
988		dd_emit(ctx, 1, 0);	/* ffffffff */
989	dd_emit(ctx, 1, 0);		/* 00000001 GP_BUILTIN_RESULT_EN.LAYER_IDX */
990	dd_emit(ctx, 1, 0);		/* ffffffff STRMOUT_ENABLE */
991	dd_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
992	dd_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
993	dd_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE*/
994	if (dev_priv->chipset != 0x50)
995		dd_emit(ctx, 8, 0);	/* 00000001 */
996	if (dev_priv->chipset >= 0xa0) {
997		dd_emit(ctx, 1, 1);	/* 00000007 VTX_ATTR_DEFINE.COMP */
998		dd_emit(ctx, 1, 1);	/* 00000007 VTX_ATTR_DEFINE.SIZE */
999		dd_emit(ctx, 1, 2);	/* 00000007 VTX_ATTR_DEFINE.TYPE */
1000		dd_emit(ctx, 1, 0);	/* 000000ff VTX_ATTR_DEFINE.ATTR */
1001	}
1002	dd_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
1003	dd_emit(ctx, 1, 0x14);		/* 0000001f ZETA_FORMAT */
1004	dd_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
1005	dd_emit(ctx, 1, 0);		/* 0000000f VP_TEXTURES_LOG2 */
1006	dd_emit(ctx, 1, 0);		/* 0000000f VP_SAMPLERS_LOG2 */
1007	if (IS_NVA3F(dev_priv->chipset))
1008		dd_emit(ctx, 1, 0);	/* 00000001 */
1009	dd_emit(ctx, 1, 2);		/* 00000003 POLYGON_MODE_BACK */
1010	if (dev_priv->chipset >= 0xa0)
1011		dd_emit(ctx, 1, 0);	/* 00000003 VTX_ATTR_DEFINE.SIZE - 1 */
1012	dd_emit(ctx, 1, 0);		/* 0000ffff CB_ADDR_INDEX */
1013	if (dev_priv->chipset >= 0xa0)
1014		dd_emit(ctx, 1, 0);	/* 00000003 */
1015	dd_emit(ctx, 1, 0);		/* 00000001 CULL_FACE_ENABLE */
1016	dd_emit(ctx, 1, 1);		/* 00000003 CULL_FACE */
1017	dd_emit(ctx, 1, 0);		/* 00000001 FRONT_FACE */
1018	dd_emit(ctx, 1, 2);		/* 00000003 POLYGON_MODE_FRONT */
1019	dd_emit(ctx, 1, 0x1000);	/* 00007fff UNK141C */
1020	if (dev_priv->chipset != 0x50) {
1021		dd_emit(ctx, 1, 0xe00);		/* 7fff */
1022		dd_emit(ctx, 1, 0x1000);	/* 7fff */
1023		dd_emit(ctx, 1, 0x1e00);	/* 7fff */
1024	}
1025	dd_emit(ctx, 1, 0);		/* 00000001 BEGIN_END_ACTIVE */
1026	dd_emit(ctx, 1, 1);		/* 00000001 POLYGON_MODE_??? */
1027	dd_emit(ctx, 1, 1);		/* 000000ff GP_REG_ALLOC_TEMP / 4 rounded up */
1028	dd_emit(ctx, 1, 1);		/* 000000ff FP_REG_ALLOC_TEMP... without /4? */
1029	dd_emit(ctx, 1, 1);		/* 000000ff VP_REG_ALLOC_TEMP / 4 rounded up */
1030	dd_emit(ctx, 1, 1);		/* 00000001 */
1031	dd_emit(ctx, 1, 0);		/* 00000001 */
1032	dd_emit(ctx, 1, 0);		/* 00000001 VTX_ATTR_MASK_UNK0 nonempty */
1033	dd_emit(ctx, 1, 0);		/* 00000001 VTX_ATTR_MASK_UNK1 nonempty */
1034	dd_emit(ctx, 1, 0x200);		/* 0003ffff GP_VERTEX_OUTPUT_COUNT*GP_REG_ALLOC_RESULT */
1035	if (IS_NVA3F(dev_priv->chipset))
1036		dd_emit(ctx, 1, 0x200);
1037	dd_emit(ctx, 1, 0);		/* 00000001 */
1038	if (dev_priv->chipset < 0xa0) {
1039		dd_emit(ctx, 1, 1);	/* 00000001 */
1040		dd_emit(ctx, 1, 0x70);	/* 000000ff */
1041		dd_emit(ctx, 1, 0x80);	/* 000000ff */
1042		dd_emit(ctx, 1, 0);	/* 000000ff */
1043		dd_emit(ctx, 1, 0);	/* 00000001 */
1044		dd_emit(ctx, 1, 1);	/* 00000001 */
1045		dd_emit(ctx, 1, 0x70);	/* 000000ff */
1046		dd_emit(ctx, 1, 0x80);	/* 000000ff */
1047		dd_emit(ctx, 1, 0);	/* 000000ff */
1048	} else {
1049		dd_emit(ctx, 1, 1);	/* 00000001 */
1050		dd_emit(ctx, 1, 0xf0);	/* 000000ff */
1051		dd_emit(ctx, 1, 0xff);	/* 000000ff */
1052		dd_emit(ctx, 1, 0);	/* 000000ff */
1053		dd_emit(ctx, 1, 0);	/* 00000001 */
1054		dd_emit(ctx, 1, 1);	/* 00000001 */
1055		dd_emit(ctx, 1, 0xf0);	/* 000000ff */
1056		dd_emit(ctx, 1, 0xff);	/* 000000ff */
1057		dd_emit(ctx, 1, 0);	/* 000000ff */
1058		dd_emit(ctx, 1, 9);	/* 0000003f UNK114C.COMP,SIZE */
1059	}
1060
1061	/* eng2d state */
1062	dd_emit(ctx, 1, 0);		/* 00000001 eng2d COLOR_KEY_ENABLE */
1063	dd_emit(ctx, 1, 0);		/* 00000007 eng2d COLOR_KEY_FORMAT */
1064	dd_emit(ctx, 1, 1);		/* ffffffff eng2d DST_DEPTH */
1065	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d DST_FORMAT */
1066	dd_emit(ctx, 1, 0);		/* ffffffff eng2d DST_LAYER */
1067	dd_emit(ctx, 1, 1);		/* 00000001 eng2d DST_LINEAR */
1068	dd_emit(ctx, 1, 0);		/* 00000007 eng2d PATTERN_COLOR_FORMAT */
1069	dd_emit(ctx, 1, 0);		/* 00000007 eng2d OPERATION */
1070	dd_emit(ctx, 1, 0);		/* 00000003 eng2d PATTERN_SELECT */
1071	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d SIFC_FORMAT */
1072	dd_emit(ctx, 1, 0);		/* 00000001 eng2d SIFC_BITMAP_ENABLE */
1073	dd_emit(ctx, 1, 2);		/* 00000003 eng2d SIFC_BITMAP_UNK808 */
1074	dd_emit(ctx, 1, 0);		/* ffffffff eng2d BLIT_DU_DX_FRACT */
1075	dd_emit(ctx, 1, 1);		/* ffffffff eng2d BLIT_DU_DX_INT */
1076	dd_emit(ctx, 1, 0);		/* ffffffff eng2d BLIT_DV_DY_FRACT */
1077	dd_emit(ctx, 1, 1);		/* ffffffff eng2d BLIT_DV_DY_INT */
1078	dd_emit(ctx, 1, 0);		/* 00000001 eng2d BLIT_CONTROL_FILTER */
1079	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d DRAW_COLOR_FORMAT */
1080	dd_emit(ctx, 1, 0xcf);		/* 000000ff eng2d SRC_FORMAT */
1081	dd_emit(ctx, 1, 1);		/* 00000001 eng2d SRC_LINEAR #2 */
1082
1083	num = ctx->ctxvals_pos - base;
1084	ctx->ctxvals_pos = base;
1085	if (IS_NVA3F(dev_priv->chipset))
1086		cp_ctx(ctx, 0x404800, num);
1087	else
1088		cp_ctx(ctx, 0x405400, num);
1089}
1090
1091/*
1092 * xfer areas. These are a pain.
1093 *
1094 * There are 2 xfer areas: the first one is big and contains all sorts of
1095 * stuff, the second is small and contains some per-TP context.
1096 *
1097 * Each area is split into 8 "strands". The areas, when saved to grctx,
1098 * are made of 8-word blocks. Each block contains a single word from
1099 * each strand. The strands are independent of each other, their
1100 * addresses are unrelated to each other, and data in them is closely
1101 * packed together. The strand layout varies a bit between cards: here
1102 * and there, a single word is thrown out in the middle and the whole
1103 * strand is offset by a bit from corresponding one on another chipset.
1104 * For this reason, addresses of stuff in strands are almost useless.
1105 * Knowing sequence of stuff and size of gaps between them is much more
1106 * useful, and that's how we build the strands in our generator.
1107 *
1108 * NVA0 takes this mess to a whole new level by cutting the old strands
1109 * into a few dozen pieces [known as genes], rearranging them randomly,
1110 * and putting them back together to make new strands. Hopefully these
1111 * genes correspond more or less directly to the same PGRAPH subunits
1112 * as in 400040 register.
1113 *
1114 * The most common value in default context is 0, and when the genes
1115 * are separated by 0's, gene bounduaries are quite speculative...
1116 * some of them can be clearly deduced, others can be guessed, and yet
1117 * others won't be resolved without figuring out the real meaning of
1118 * given ctxval. For the same reason, ending point of each strand
1119 * is unknown. Except for strand 0, which is the longest strand and
1120 * its end corresponds to end of the whole xfer.
1121 *
1122 * An unsolved mystery is the seek instruction: it takes an argument
1123 * in bits 8-18, and that argument is clearly the place in strands to
1124 * seek to... but the offsets don't seem to correspond to offsets as
1125 * seen in grctx. Perhaps there's another, real, not randomly-changing
1126 * addressing in strands, and the xfer insn just happens to skip over
1127 * the unused bits? NV10-NV30 PIPE comes to mind...
1128 *
1129 * As far as I know, there's no way to access the xfer areas directly
1130 * without the help of ctxprog.
1131 */
1132
1133static void
1134xf_emit(struct nouveau_grctx *ctx, int num, uint32_t val) {
1135	int i;
1136	if (val && ctx->mode == NOUVEAU_GRCTX_VALS)
1137		for (i = 0; i < num; i++)
1138			nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + (i << 3)), val);
1139	ctx->ctxvals_pos += num << 3;
1140}
1141
1142/* Gene declarations... */
1143
1144static void nv50_graph_construct_gene_dispatch(struct nouveau_grctx *ctx);
1145static void nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx);
1146static void nv50_graph_construct_gene_ccache(struct nouveau_grctx *ctx);
1147static void nv50_graph_construct_gene_unk10xx(struct nouveau_grctx *ctx);
1148static void nv50_graph_construct_gene_unk14xx(struct nouveau_grctx *ctx);
1149static void nv50_graph_construct_gene_zcull(struct nouveau_grctx *ctx);
1150static void nv50_graph_construct_gene_clipid(struct nouveau_grctx *ctx);
1151static void nv50_graph_construct_gene_unk24xx(struct nouveau_grctx *ctx);
1152static void nv50_graph_construct_gene_vfetch(struct nouveau_grctx *ctx);
1153static void nv50_graph_construct_gene_eng2d(struct nouveau_grctx *ctx);
1154static void nv50_graph_construct_gene_csched(struct nouveau_grctx *ctx);
1155static void nv50_graph_construct_gene_unk1cxx(struct nouveau_grctx *ctx);
1156static void nv50_graph_construct_gene_strmout(struct nouveau_grctx *ctx);
1157static void nv50_graph_construct_gene_unk34xx(struct nouveau_grctx *ctx);
1158static void nv50_graph_construct_gene_ropm1(struct nouveau_grctx *ctx);
1159static void nv50_graph_construct_gene_ropm2(struct nouveau_grctx *ctx);
1160static void nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx);
1161static void nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx);
1162
1163static void
1164nv50_graph_construct_xfer1(struct nouveau_grctx *ctx)
1165{
1166	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1167	int i;
1168	int offset;
1169	int size = 0;
1170	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
1171
1172	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
1173	ctx->ctxvals_base = offset;
1174
1175	if (dev_priv->chipset < 0xa0) {
1176		/* Strand 0 */
1177		ctx->ctxvals_pos = offset;
1178		nv50_graph_construct_gene_dispatch(ctx);
1179		nv50_graph_construct_gene_m2mf(ctx);
1180		nv50_graph_construct_gene_unk24xx(ctx);
1181		nv50_graph_construct_gene_clipid(ctx);
1182		nv50_graph_construct_gene_zcull(ctx);
1183		if ((ctx->ctxvals_pos-offset)/8 > size)
1184			size = (ctx->ctxvals_pos-offset)/8;
1185
1186		/* Strand 1 */
1187		ctx->ctxvals_pos = offset + 0x1;
1188		nv50_graph_construct_gene_vfetch(ctx);
1189		nv50_graph_construct_gene_eng2d(ctx);
1190		nv50_graph_construct_gene_csched(ctx);
1191		nv50_graph_construct_gene_ropm1(ctx);
1192		nv50_graph_construct_gene_ropm2(ctx);
1193		if ((ctx->ctxvals_pos-offset)/8 > size)
1194			size = (ctx->ctxvals_pos-offset)/8;
1195
1196		/* Strand 2 */
1197		ctx->ctxvals_pos = offset + 0x2;
1198		nv50_graph_construct_gene_ccache(ctx);
1199		nv50_graph_construct_gene_unk1cxx(ctx);
1200		nv50_graph_construct_gene_strmout(ctx);
1201		nv50_graph_construct_gene_unk14xx(ctx);
1202		nv50_graph_construct_gene_unk10xx(ctx);
1203		nv50_graph_construct_gene_unk34xx(ctx);
1204		if ((ctx->ctxvals_pos-offset)/8 > size)
1205			size = (ctx->ctxvals_pos-offset)/8;
1206
1207		/* Strand 3: per-ROP group state */
1208		ctx->ctxvals_pos = offset + 3;
1209		for (i = 0; i < 6; i++)
1210			if (units & (1 << (i + 16)))
1211				nv50_graph_construct_gene_ropc(ctx);
1212		if ((ctx->ctxvals_pos-offset)/8 > size)
1213			size = (ctx->ctxvals_pos-offset)/8;
1214
1215		/* Strands 4-7: per-TP state */
1216		for (i = 0; i < 4; i++) {
1217			ctx->ctxvals_pos = offset + 4 + i;
1218			if (units & (1 << (2 * i)))
1219				nv50_graph_construct_xfer_tp(ctx);
1220			if (units & (1 << (2 * i + 1)))
1221				nv50_graph_construct_xfer_tp(ctx);
1222			if ((ctx->ctxvals_pos-offset)/8 > size)
1223				size = (ctx->ctxvals_pos-offset)/8;
1224		}
1225	} else {
1226		/* Strand 0 */
1227		ctx->ctxvals_pos = offset;
1228		nv50_graph_construct_gene_dispatch(ctx);
1229		nv50_graph_construct_gene_m2mf(ctx);
1230		nv50_graph_construct_gene_unk34xx(ctx);
1231		nv50_graph_construct_gene_csched(ctx);
1232		nv50_graph_construct_gene_unk1cxx(ctx);
1233		nv50_graph_construct_gene_strmout(ctx);
1234		if ((ctx->ctxvals_pos-offset)/8 > size)
1235			size = (ctx->ctxvals_pos-offset)/8;
1236
1237		/* Strand 1 */
1238		ctx->ctxvals_pos = offset + 1;
1239		nv50_graph_construct_gene_unk10xx(ctx);
1240		if ((ctx->ctxvals_pos-offset)/8 > size)
1241			size = (ctx->ctxvals_pos-offset)/8;
1242
1243		/* Strand 2 */
1244		ctx->ctxvals_pos = offset + 2;
1245		if (dev_priv->chipset == 0xa0)
1246			nv50_graph_construct_gene_unk14xx(ctx);
1247		nv50_graph_construct_gene_unk24xx(ctx);
1248		if ((ctx->ctxvals_pos-offset)/8 > size)
1249			size = (ctx->ctxvals_pos-offset)/8;
1250
1251		/* Strand 3 */
1252		ctx->ctxvals_pos = offset + 3;
1253		nv50_graph_construct_gene_vfetch(ctx);
1254		if ((ctx->ctxvals_pos-offset)/8 > size)
1255			size = (ctx->ctxvals_pos-offset)/8;
1256
1257		/* Strand 4 */
1258		ctx->ctxvals_pos = offset + 4;
1259		nv50_graph_construct_gene_ccache(ctx);
1260		if ((ctx->ctxvals_pos-offset)/8 > size)
1261			size = (ctx->ctxvals_pos-offset)/8;
1262
1263		/* Strand 5 */
1264		ctx->ctxvals_pos = offset + 5;
1265		nv50_graph_construct_gene_ropm2(ctx);
1266		nv50_graph_construct_gene_ropm1(ctx);
1267		/* per-ROP context */
1268		for (i = 0; i < 8; i++)
1269			if (units & (1<<(i+16)))
1270				nv50_graph_construct_gene_ropc(ctx);
1271		if ((ctx->ctxvals_pos-offset)/8 > size)
1272			size = (ctx->ctxvals_pos-offset)/8;
1273
1274		/* Strand 6 */
1275		ctx->ctxvals_pos = offset + 6;
1276		nv50_graph_construct_gene_zcull(ctx);
1277		nv50_graph_construct_gene_clipid(ctx);
1278		nv50_graph_construct_gene_eng2d(ctx);
1279		if (units & (1 << 0))
1280			nv50_graph_construct_xfer_tp(ctx);
1281		if (units & (1 << 1))
1282			nv50_graph_construct_xfer_tp(ctx);
1283		if (units & (1 << 2))
1284			nv50_graph_construct_xfer_tp(ctx);
1285		if (units & (1 << 3))
1286			nv50_graph_construct_xfer_tp(ctx);
1287		if ((ctx->ctxvals_pos-offset)/8 > size)
1288			size = (ctx->ctxvals_pos-offset)/8;
1289
1290		/* Strand 7 */
1291		ctx->ctxvals_pos = offset + 7;
1292		if (dev_priv->chipset == 0xa0) {
1293			if (units & (1 << 4))
1294				nv50_graph_construct_xfer_tp(ctx);
1295			if (units & (1 << 5))
1296				nv50_graph_construct_xfer_tp(ctx);
1297			if (units & (1 << 6))
1298				nv50_graph_construct_xfer_tp(ctx);
1299			if (units & (1 << 7))
1300				nv50_graph_construct_xfer_tp(ctx);
1301			if (units & (1 << 8))
1302				nv50_graph_construct_xfer_tp(ctx);
1303			if (units & (1 << 9))
1304				nv50_graph_construct_xfer_tp(ctx);
1305		} else {
1306			nv50_graph_construct_gene_unk14xx(ctx);
1307		}
1308		if ((ctx->ctxvals_pos-offset)/8 > size)
1309			size = (ctx->ctxvals_pos-offset)/8;
1310	}
1311
1312	ctx->ctxvals_pos = offset + size * 8;
1313	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
1314	cp_lsr (ctx, offset);
1315	cp_out (ctx, CP_SET_XFER_POINTER);
1316	cp_lsr (ctx, size);
1317	cp_out (ctx, CP_SEEK_1);
1318	cp_out (ctx, CP_XFER_1);
1319	cp_wait(ctx, XFER, BUSY);
1320}
1321
1322/*
1323 * non-trivial demagiced parts of ctx init go here
1324 */
1325
1326static void
1327nv50_graph_construct_gene_dispatch(struct nouveau_grctx *ctx)
1328{
1329	/* start of strand 0 */
1330	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1331	/* SEEK */
1332	if (dev_priv->chipset == 0x50)
1333		xf_emit(ctx, 5, 0);
1334	else if (!IS_NVA3F(dev_priv->chipset))
1335		xf_emit(ctx, 6, 0);
1336	else
1337		xf_emit(ctx, 4, 0);
1338	/* SEEK */
1339	/* the PGRAPH's internal FIFO */
1340	if (dev_priv->chipset == 0x50)
1341		xf_emit(ctx, 8*3, 0);
1342	else
1343		xf_emit(ctx, 0x100*3, 0);
1344	/* and another bonus slot?!? */
1345	xf_emit(ctx, 3, 0);
1346	/* and YET ANOTHER bonus slot? */
1347	if (IS_NVA3F(dev_priv->chipset))
1348		xf_emit(ctx, 3, 0);
1349	/* SEEK */
1350	/* CTX_SWITCH: caches of gr objects bound to subchannels. 8 values, last used index */
1351	xf_emit(ctx, 9, 0);
1352	/* SEEK */
1353	xf_emit(ctx, 9, 0);
1354	/* SEEK */
1355	xf_emit(ctx, 9, 0);
1356	/* SEEK */
1357	xf_emit(ctx, 9, 0);
1358	/* SEEK */
1359	if (dev_priv->chipset < 0x90)
1360		xf_emit(ctx, 4, 0);
1361	/* SEEK */
1362	xf_emit(ctx, 2, 0);
1363	/* SEEK */
1364	xf_emit(ctx, 6*2, 0);
1365	xf_emit(ctx, 2, 0);
1366	/* SEEK */
1367	xf_emit(ctx, 2, 0);
1368	/* SEEK */
1369	xf_emit(ctx, 6*2, 0);
1370	xf_emit(ctx, 2, 0);
1371	/* SEEK */
1372	if (dev_priv->chipset == 0x50)
1373		xf_emit(ctx, 0x1c, 0);
1374	else if (dev_priv->chipset < 0xa0)
1375		xf_emit(ctx, 0x1e, 0);
1376	else
1377		xf_emit(ctx, 0x22, 0);
1378	/* SEEK */
1379	xf_emit(ctx, 0x15, 0);
1380}
1381
1382static void
1383nv50_graph_construct_gene_m2mf(struct nouveau_grctx *ctx)
1384{
1385	/* Strand 0, right after dispatch */
1386	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1387	int smallm2mf = 0;
1388	if (dev_priv->chipset < 0x92 || dev_priv->chipset == 0x98)
1389		smallm2mf = 1;
1390	/* SEEK */
1391	xf_emit (ctx, 1, 0);		/* DMA_NOTIFY instance >> 4 */
1392	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_IN instance >> 4 */
1393	xf_emit (ctx, 1, 0);		/* DMA_BUFFER_OUT instance >> 4 */
1394	xf_emit (ctx, 1, 0);		/* OFFSET_IN */
1395	xf_emit (ctx, 1, 0);		/* OFFSET_OUT */
1396	xf_emit (ctx, 1, 0);		/* PITCH_IN */
1397	xf_emit (ctx, 1, 0);		/* PITCH_OUT */
1398	xf_emit (ctx, 1, 0);		/* LINE_LENGTH */
1399	xf_emit (ctx, 1, 0);		/* LINE_COUNT */
1400	xf_emit (ctx, 1, 0x21);		/* FORMAT: bits 0-4 INPUT_INC, bits 5-9 OUTPUT_INC */
1401	xf_emit (ctx, 1, 1);		/* LINEAR_IN */
1402	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_IN: bits 0-2 y tiling, bits 3-5 z tiling */
1403	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_IN */
1404	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_IN */
1405	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_IN */
1406	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN_Z */
1407	xf_emit (ctx, 1, 0);		/* TILING_POSITION_IN */
1408	xf_emit (ctx, 1, 1);		/* LINEAR_OUT */
1409	xf_emit (ctx, 1, 0x2);		/* TILING_MODE_OUT: bits 0-2 y tiling, bits 3-5 z tiling */
1410	xf_emit (ctx, 1, 0x100);	/* TILING_PITCH_OUT */
1411	xf_emit (ctx, 1, 0x100);	/* TILING_HEIGHT_OUT */
1412	xf_emit (ctx, 1, 1);		/* TILING_DEPTH_OUT */
1413	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT_Z */
1414	xf_emit (ctx, 1, 0);		/* TILING_POSITION_OUT */
1415	xf_emit (ctx, 1, 0);		/* OFFSET_IN_HIGH */
1416	xf_emit (ctx, 1, 0);		/* OFFSET_OUT_HIGH */
1417	/* SEEK */
1418	if (smallm2mf)
1419		xf_emit(ctx, 0x40, 0);	/* 20 * ffffffff, 3ffff */
1420	else
1421		xf_emit(ctx, 0x100, 0);	/* 80 * ffffffff, 3ffff */
1422	xf_emit(ctx, 4, 0);		/* 1f/7f, 0, 1f/7f, 0 [1f for smallm2mf, 7f otherwise] */
1423	/* SEEK */
1424	if (smallm2mf)
1425		xf_emit(ctx, 0x400, 0);	/* ffffffff */
1426	else
1427		xf_emit(ctx, 0x800, 0);	/* ffffffff */
1428	xf_emit(ctx, 4, 0);		/* ff/1ff, 0, 0, 0 [ff for smallm2mf, 1ff otherwise] */
1429	/* SEEK */
1430	xf_emit(ctx, 0x40, 0);		/* 20 * bits ffffffff, 3ffff */
1431	xf_emit(ctx, 0x6, 0);		/* 1f, 0, 1f, 0, 1f, 0 */
1432}
1433
1434static void
1435nv50_graph_construct_gene_ccache(struct nouveau_grctx *ctx)
1436{
1437	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1438	xf_emit(ctx, 2, 0);		/* RO */
1439	xf_emit(ctx, 0x800, 0);		/* ffffffff */
1440	switch (dev_priv->chipset) {
1441	case 0x50:
1442	case 0x92:
1443	case 0xa0:
1444		xf_emit(ctx, 0x2b, 0);
1445		break;
1446	case 0x84:
1447		xf_emit(ctx, 0x29, 0);
1448		break;
1449	case 0x94:
1450	case 0x96:
1451	case 0xa3:
1452		xf_emit(ctx, 0x27, 0);
1453		break;
1454	case 0x86:
1455	case 0x98:
1456	case 0xa5:
1457	case 0xa8:
1458	case 0xaa:
1459	case 0xac:
1460	case 0xaf:
1461		xf_emit(ctx, 0x25, 0);
1462		break;
1463	}
1464	/* CB bindings, 0x80 of them. first word is address >> 8, second is
1465	 * size >> 4 | valid << 24 */
1466	xf_emit(ctx, 0x100, 0);		/* ffffffff CB_DEF */
1467	xf_emit(ctx, 1, 0);		/* 0000007f CB_ADDR_BUFFER */
1468	xf_emit(ctx, 1, 0);		/* 0 */
1469	xf_emit(ctx, 0x30, 0);		/* ff SET_PROGRAM_CB */
1470	xf_emit(ctx, 1, 0);		/* 3f last SET_PROGRAM_CB */
1471	xf_emit(ctx, 4, 0);		/* RO */
1472	xf_emit(ctx, 0x100, 0);		/* ffffffff */
1473	xf_emit(ctx, 8, 0);		/* 1f, 0, 0, ... */
1474	xf_emit(ctx, 8, 0);		/* ffffffff */
1475	xf_emit(ctx, 4, 0);		/* ffffffff */
1476	xf_emit(ctx, 1, 0);		/* 3 */
1477	xf_emit(ctx, 1, 0);		/* ffffffff */
1478	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_CODE_CB */
1479	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TIC */
1480	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TSC */
1481	xf_emit(ctx, 1, 0);		/* 00000001 LINKED_TSC */
1482	xf_emit(ctx, 1, 0);		/* 000000ff TIC_ADDRESS_HIGH */
1483	xf_emit(ctx, 1, 0);		/* ffffffff TIC_ADDRESS_LOW */
1484	xf_emit(ctx, 1, 0x3fffff);	/* 003fffff TIC_LIMIT */
1485	xf_emit(ctx, 1, 0);		/* 000000ff TSC_ADDRESS_HIGH */
1486	xf_emit(ctx, 1, 0);		/* ffffffff TSC_ADDRESS_LOW */
1487	xf_emit(ctx, 1, 0x1fff);	/* 000fffff TSC_LIMIT */
1488	xf_emit(ctx, 1, 0);		/* 000000ff VP_ADDRESS_HIGH */
1489	xf_emit(ctx, 1, 0);		/* ffffffff VP_ADDRESS_LOW */
1490	xf_emit(ctx, 1, 0);		/* 00ffffff VP_START_ID */
1491	xf_emit(ctx, 1, 0);		/* 000000ff CB_DEF_ADDRESS_HIGH */
1492	xf_emit(ctx, 1, 0);		/* ffffffff CB_DEF_ADDRESS_LOW */
1493	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1494	xf_emit(ctx, 1, 0);		/* 000000ff GP_ADDRESS_HIGH */
1495	xf_emit(ctx, 1, 0);		/* ffffffff GP_ADDRESS_LOW */
1496	xf_emit(ctx, 1, 0);		/* 00ffffff GP_START_ID */
1497	xf_emit(ctx, 1, 0);		/* 000000ff FP_ADDRESS_HIGH */
1498	xf_emit(ctx, 1, 0);		/* ffffffff FP_ADDRESS_LOW */
1499	xf_emit(ctx, 1, 0);		/* 00ffffff FP_START_ID */
1500}
1501
1502static void
1503nv50_graph_construct_gene_unk10xx(struct nouveau_grctx *ctx)
1504{
1505	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1506	int i;
1507	/* end of area 2 on pre-NVA0, area 1 on NVAx */
1508	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
1509	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
1510	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1511	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
1512	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
1513	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
1514	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
1515	if (dev_priv->chipset == 0x50)
1516		xf_emit(ctx, 1, 0x3ff);
1517	else
1518		xf_emit(ctx, 1, 0x7ff);	/* 000007ff */
1519	xf_emit(ctx, 1, 0);		/* 111/113 */
1520	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
1521	for (i = 0; i < 8; i++) {
1522		switch (dev_priv->chipset) {
1523		case 0x50:
1524		case 0x86:
1525		case 0x98:
1526		case 0xaa:
1527		case 0xac:
1528			xf_emit(ctx, 0xa0, 0);	/* ffffffff */
1529			break;
1530		case 0x84:
1531		case 0x92:
1532		case 0x94:
1533		case 0x96:
1534			xf_emit(ctx, 0x120, 0);
1535			break;
1536		case 0xa5:
1537		case 0xa8:
1538			xf_emit(ctx, 0x100, 0);	/* ffffffff */
1539			break;
1540		case 0xa0:
1541		case 0xa3:
1542		case 0xaf:
1543			xf_emit(ctx, 0x400, 0);	/* ffffffff */
1544			break;
1545		}
1546		xf_emit(ctx, 4, 0);	/* 3f, 0, 0, 0 */
1547		xf_emit(ctx, 4, 0);	/* ffffffff */
1548	}
1549	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
1550	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
1551	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1552	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
1553	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_TEMP */
1554	xf_emit(ctx, 1, 1);		/* 00000001 RASTERIZE_ENABLE */
1555	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
1556	xf_emit(ctx, 1, 0x27);		/* 000000ff UNK0FD4 */
1557	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
1558	xf_emit(ctx, 1, 0x26);		/* 000000ff SEMANTIC_LAYER */
1559	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
1560}
1561
1562static void
1563nv50_graph_construct_gene_unk34xx(struct nouveau_grctx *ctx)
1564{
1565	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1566	/* end of area 2 on pre-NVA0, area 1 on NVAx */
1567	xf_emit(ctx, 1, 0);		/* 00000001 VIEWPORT_CLIP_RECTS_EN */
1568	xf_emit(ctx, 1, 0);		/* 00000003 VIEWPORT_CLIP_MODE */
1569	xf_emit(ctx, 0x10, 0x04000000);	/* 07ffffff VIEWPORT_CLIP_HORIZ*8, VIEWPORT_CLIP_VERT*8 */
1570	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_STIPPLE_ENABLE */
1571	xf_emit(ctx, 0x20, 0);		/* ffffffff POLYGON_STIPPLE */
1572	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
1573	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
1574	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0D64 */
1575	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0DF4 */
1576	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
1577	xf_emit(ctx, 1, 0);		/* 00000007 */
1578	xf_emit(ctx, 1, 0x1fe21);	/* 0001ffff tesla UNK0FAC */
1579	if (dev_priv->chipset >= 0xa0)
1580		xf_emit(ctx, 1, 0x0fac6881);
1581	if (IS_NVA3F(dev_priv->chipset)) {
1582		xf_emit(ctx, 1, 1);
1583		xf_emit(ctx, 3, 0);
1584	}
1585}
1586
1587static void
1588nv50_graph_construct_gene_unk14xx(struct nouveau_grctx *ctx)
1589{
1590	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1591	/* middle of area 2 on pre-NVA0, beginning of area 2 on NVA0, area 7 on >NVA0 */
1592	if (dev_priv->chipset != 0x50) {
1593		xf_emit(ctx, 5, 0);		/* ffffffff */
1594		xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
1595		xf_emit(ctx, 1, 0);		/* 00000001 */
1596		xf_emit(ctx, 1, 0);		/* 000003ff */
1597		xf_emit(ctx, 1, 0x804);		/* 00000fff SEMANTIC_CLIP */
1598		xf_emit(ctx, 1, 0);		/* 00000001 */
1599		xf_emit(ctx, 2, 4);		/* 7f, ff */
1600		xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
1601	}
1602	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1603	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
1604	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
1605	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
1606	xf_emit(ctx, 1, 0x10);			/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
1607	xf_emit(ctx, 1, 0);			/* 000000ff VP_CLIP_DISTANCE_ENABLE */
1608	if (dev_priv->chipset != 0x50)
1609		xf_emit(ctx, 1, 0);		/* 3ff */
1610	xf_emit(ctx, 1, 0);			/* 000000ff tesla UNK1940 */
1611	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0D7C */
1612	xf_emit(ctx, 1, 0x804);			/* 00000fff SEMANTIC_CLIP */
1613	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
1614	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
1615	if (dev_priv->chipset != 0x50)
1616		xf_emit(ctx, 1, 0x7f);		/* 000000ff tesla UNK0FFC */
1617	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1618	xf_emit(ctx, 1, 1);			/* 00000001 SHADE_MODEL */
1619	xf_emit(ctx, 1, 0x80c14);		/* 01ffffff SEMANTIC_COLOR */
1620	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
1621	xf_emit(ctx, 1, 0x8100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
1622	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
1623	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
1624	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
1625	xf_emit(ctx, 1, 0x10);			/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
1626	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0D7C */
1627	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK0F8C */
1628	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1629	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
1630	xf_emit(ctx, 1, 0x8100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
1631	xf_emit(ctx, 4, 0);			/* ffffffff NOPERSPECTIVE_BITMAP */
1632	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
1633	xf_emit(ctx, 1, 0);			/* 0000000f */
1634	if (dev_priv->chipset == 0x50)
1635		xf_emit(ctx, 1, 0x3ff);		/* 000003ff tesla UNK0D68 */
1636	else
1637		xf_emit(ctx, 1, 0x7ff);		/* 000007ff tesla UNK0D68 */
1638	xf_emit(ctx, 1, 0x80c14);		/* 01ffffff SEMANTIC_COLOR */
1639	xf_emit(ctx, 1, 0);			/* 00000001 VERTEX_TWO_SIDE_ENABLE */
1640	xf_emit(ctx, 0x30, 0);			/* ffffffff VIEWPORT_SCALE: X0, Y0, Z0, X1, Y1, ... */
1641	xf_emit(ctx, 3, 0);			/* f, 0, 0 */
1642	xf_emit(ctx, 3, 0);			/* ffffffff last VIEWPORT_SCALE? */
1643	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1644	xf_emit(ctx, 1, 1);			/* 00000001 VIEWPORT_TRANSFORM_EN */
1645	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
1646	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1924 */
1647	xf_emit(ctx, 1, 0x10);			/* 000000ff VIEW_VOLUME_CLIP_CTRL */
1648	xf_emit(ctx, 1, 0);			/* 00000001 */
1649	xf_emit(ctx, 0x30, 0);			/* ffffffff VIEWPORT_TRANSLATE */
1650	xf_emit(ctx, 3, 0);			/* f, 0, 0 */
1651	xf_emit(ctx, 3, 0);			/* ffffffff */
1652	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1653	xf_emit(ctx, 2, 0x88);			/* 000001ff tesla UNK19D8 */
1654	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1924 */
1655	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1656	xf_emit(ctx, 1, 4);			/* 0000000f CULL_MODE */
1657	xf_emit(ctx, 2, 0);			/* 07ffffff SCREEN_SCISSOR */
1658	xf_emit(ctx, 2, 0);			/* 00007fff WINDOW_OFFSET_XY */
1659	xf_emit(ctx, 1, 0);			/* 00000003 WINDOW_ORIGIN */
1660	xf_emit(ctx, 0x10, 0);			/* 00000001 SCISSOR_ENABLE */
1661	xf_emit(ctx, 1, 0);			/* 0001ffff GP_BUILTIN_RESULT_EN */
1662	xf_emit(ctx, 1, 0x26);			/* 000000ff SEMANTIC_LAYER */
1663	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
1664	xf_emit(ctx, 1, 0);			/* 0000000f */
1665	xf_emit(ctx, 1, 0x3f800000);		/* ffffffff LINE_WIDTH */
1666	xf_emit(ctx, 1, 0);			/* 00000001 LINE_STIPPLE_ENABLE */
1667	xf_emit(ctx, 1, 0);			/* 00000001 LINE_SMOOTH_ENABLE */
1668	xf_emit(ctx, 1, 0);			/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
1669	if (IS_NVA3F(dev_priv->chipset))
1670		xf_emit(ctx, 1, 0);		/* 00000001 */
1671	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
1672	xf_emit(ctx, 1, 0x10);			/* 000000ff VIEW_VOLUME_CLIP_CTRL */
1673	if (dev_priv->chipset != 0x50) {
1674		xf_emit(ctx, 1, 0);		/* ffffffff */
1675		xf_emit(ctx, 1, 0);		/* 00000001 */
1676		xf_emit(ctx, 1, 0);		/* 000003ff */
1677	}
1678	xf_emit(ctx, 0x20, 0);			/* 10xbits ffffffff, 3fffff. SCISSOR_* */
1679	xf_emit(ctx, 1, 0);			/* f */
1680	xf_emit(ctx, 1, 0);			/* 0? */
1681	xf_emit(ctx, 1, 0);			/* ffffffff */
1682	xf_emit(ctx, 1, 0);			/* 003fffff */
1683	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A30 */
1684	xf_emit(ctx, 1, 0x52);			/* 000001ff SEMANTIC_PTSZ */
1685	xf_emit(ctx, 1, 0);			/* 0001ffff GP_BUILTIN_RESULT_EN */
1686	xf_emit(ctx, 1, 0x26);			/* 000000ff SEMANTIC_LAYER */
1687	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1900 */
1688	xf_emit(ctx, 1, 4);			/* 0000007f VP_RESULT_MAP_SIZE */
1689	xf_emit(ctx, 1, 4);			/* 000000ff GP_RESULT_MAP_SIZE */
1690	xf_emit(ctx, 1, 0);			/* 00000001 GP_ENABLE */
1691	xf_emit(ctx, 1, 0x1a);			/* 0000001f POLYGON_MODE */
1692	xf_emit(ctx, 1, 0);			/* 00000001 LINE_SMOOTH_ENABLE */
1693	xf_emit(ctx, 1, 0);			/* 00000001 LINE_STIPPLE_ENABLE */
1694	xf_emit(ctx, 1, 0x00ffff00);		/* 00ffffff LINE_STIPPLE_PATTERN */
1695	xf_emit(ctx, 1, 0);			/* 0000000f */
1696}
1697
1698static void
1699nv50_graph_construct_gene_zcull(struct nouveau_grctx *ctx)
1700{
1701	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1702	/* end of strand 0 on pre-NVA0, beginning of strand 6 on NVAx */
1703	/* SEEK */
1704	xf_emit(ctx, 1, 0x3f);		/* 0000003f UNK1590 */
1705	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
1706	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
1707	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
1708	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
1709	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
1710	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
1711	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
1712	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
1713	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
1714	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff tesla UNK0D6C */
1715	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
1716	xf_emit(ctx, 1, 0);		/* 00000001 CLIPID_ENABLE */
1717	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
1718	xf_emit(ctx, 1, 0);		/* 00000001 */
1719	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
1720	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
1721	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
1722	xf_emit(ctx, 1, 4);		/* 0000000f CULL_MODE */
1723	xf_emit(ctx, 1, 0);		/* 0000ffff */
1724	xf_emit(ctx, 1, 0);		/* 00000001 UNK0FB0 */
1725	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_STIPPLE_ENABLE */
1726	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
1727	xf_emit(ctx, 1, 0);		/* ffffffff */
1728	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
1729	xf_emit(ctx, 1, 0);		/* 000000ff CLEAR_STENCIL */
1730	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
1731	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
1732	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
1733	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
1734	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
1735	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
1736	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
1737	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
1738	xf_emit(ctx, 1, 0);		/* 00000007 */
1739	if (dev_priv->chipset != 0x50)
1740		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1108 */
1741	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
1742	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
1743	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
1744	xf_emit(ctx, 1, 0x1001);	/* 00001fff ZETA_ARRAY_MODE */
1745	/* SEEK */
1746	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
1747	xf_emit(ctx, 0x10, 0);		/* 00000001 SCISSOR_ENABLE */
1748	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
1749	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
1750	xf_emit(ctx, 1, 0x10);		/* 7f/ff/3ff VIEW_VOLUME_CLIP_CTRL */
1751	xf_emit(ctx, 1, 0);		/* 00000001 VIEWPORT_CLIP_RECTS_EN */
1752	xf_emit(ctx, 1, 3);		/* 00000003 FP_CTRL_UNK196C */
1753	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1968 */
1754	if (dev_priv->chipset != 0x50)
1755		xf_emit(ctx, 1, 0);	/* 0fffffff tesla UNK1104 */
1756	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK151C */
1757}
1758
1759static void
1760nv50_graph_construct_gene_clipid(struct nouveau_grctx *ctx)
1761{
1762	/* middle of strand 0 on pre-NVA0 [after 24xx], middle of area 6 on NVAx */
1763	/* SEEK */
1764	xf_emit(ctx, 1, 0);		/* 00000007 UNK0FB4 */
1765	/* SEEK */
1766	xf_emit(ctx, 4, 0);		/* 07ffffff CLIPID_REGION_HORIZ */
1767	xf_emit(ctx, 4, 0);		/* 07ffffff CLIPID_REGION_VERT */
1768	xf_emit(ctx, 2, 0);		/* 07ffffff SCREEN_SCISSOR */
1769	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff UNK1508 */
1770	xf_emit(ctx, 1, 0);		/* 00000001 CLIPID_ENABLE */
1771	xf_emit(ctx, 1, 0x80);		/* 00003fff CLIPID_WIDTH */
1772	xf_emit(ctx, 1, 0);		/* 000000ff CLIPID_ID */
1773	xf_emit(ctx, 1, 0);		/* 000000ff CLIPID_ADDRESS_HIGH */
1774	xf_emit(ctx, 1, 0);		/* ffffffff CLIPID_ADDRESS_LOW */
1775	xf_emit(ctx, 1, 0x80);		/* 00003fff CLIPID_HEIGHT */
1776	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_CLIPID */
1777}
1778
1779static void
1780nv50_graph_construct_gene_unk24xx(struct nouveau_grctx *ctx)
1781{
1782	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1783	int i;
1784	/* middle of strand 0 on pre-NVA0 [after m2mf], end of strand 2 on NVAx */
1785	/* SEEK */
1786	xf_emit(ctx, 0x33, 0);
1787	/* SEEK */
1788	xf_emit(ctx, 2, 0);
1789	/* SEEK */
1790	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1791	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
1792	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
1793	/* SEEK */
1794	if (IS_NVA3F(dev_priv->chipset)) {
1795		xf_emit(ctx, 4, 0);	/* RO */
1796		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
1797		xf_emit(ctx, 1, 0);	/* 1ff */
1798		xf_emit(ctx, 8, 0);	/* 0? */
1799		xf_emit(ctx, 9, 0);	/* ffffffff, 7ff */
1800
1801		xf_emit(ctx, 4, 0);	/* RO */
1802		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
1803		xf_emit(ctx, 1, 0);	/* 1ff */
1804		xf_emit(ctx, 8, 0);	/* 0? */
1805		xf_emit(ctx, 9, 0);	/* ffffffff, 7ff */
1806	}
1807	else
1808	{
1809		xf_emit(ctx, 0xc, 0);	/* RO */
1810		/* SEEK */
1811		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
1812		xf_emit(ctx, 1, 0);	/* 1ff */
1813		xf_emit(ctx, 8, 0);	/* 0? */
1814
1815		/* SEEK */
1816		xf_emit(ctx, 0xc, 0);	/* RO */
1817		/* SEEK */
1818		xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */
1819		xf_emit(ctx, 1, 0);	/* 1ff */
1820		xf_emit(ctx, 8, 0);	/* 0? */
1821	}
1822	/* SEEK */
1823	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1824	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
1825	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
1826	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
1827	if (dev_priv->chipset != 0x50)
1828		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
1829	/* SEEK */
1830	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1831	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
1832	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
1833	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
1834	xf_emit(ctx, 1, 1);		/* 00000001 */
1835	/* SEEK */
1836	if (dev_priv->chipset >= 0xa0)
1837		xf_emit(ctx, 2, 4);	/* 000000ff */
1838	xf_emit(ctx, 1, 0x80c14);	/* 01ffffff SEMANTIC_COLOR */
1839	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
1840	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
1841	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
1842	xf_emit(ctx, 1, 0x27);		/* 000000ff SEMANTIC_PRIM_ID */
1843	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1844	xf_emit(ctx, 1, 0);		/* 0000000f */
1845	xf_emit(ctx, 1, 1);		/* 00000001 */
1846	for (i = 0; i < 10; i++) {
1847		/* SEEK */
1848		xf_emit(ctx, 0x40, 0);		/* ffffffff */
1849		xf_emit(ctx, 0x10, 0);		/* 3, 0, 0.... */
1850		xf_emit(ctx, 0x10, 0);		/* ffffffff */
1851	}
1852	/* SEEK */
1853	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_CTRL */
1854	xf_emit(ctx, 1, 1);		/* 00000001 */
1855	xf_emit(ctx, 1, 0);		/* ffffffff */
1856	xf_emit(ctx, 4, 0);		/* ffffffff NOPERSPECTIVE_BITMAP */
1857	xf_emit(ctx, 0x10, 0);		/* 00ffffff POINT_COORD_REPLACE_MAP */
1858	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
1859	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
1860	if (dev_priv->chipset != 0x50)
1861		xf_emit(ctx, 1, 0);	/* 000003ff */
1862}
1863
1864static void
1865nv50_graph_construct_gene_vfetch(struct nouveau_grctx *ctx)
1866{
1867	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
1868	int acnt = 0x10, rep, i;
1869	/* beginning of strand 1 on pre-NVA0, strand 3 on NVAx */
1870	if (IS_NVA3F(dev_priv->chipset))
1871		acnt = 0x20;
1872	/* SEEK */
1873	if (dev_priv->chipset >= 0xa0) {
1874		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK13A4 */
1875		xf_emit(ctx, 1, 1);	/* 00000fff tesla UNK1318 */
1876	}
1877	xf_emit(ctx, 1, 0);		/* ffffffff VERTEX_BUFFER_FIRST */
1878	xf_emit(ctx, 1, 0);		/* 00000001 PRIMITIVE_RESTART_ENABLE */
1879	xf_emit(ctx, 1, 0);		/* 00000001 UNK0DE8 */
1880	xf_emit(ctx, 1, 0);		/* ffffffff PRIMITIVE_RESTART_INDEX */
1881	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
1882	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
1883	xf_emit(ctx, acnt/8, 0);	/* ffffffff VTX_ATR_MASK_UNK0DD0 */
1884	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
1885	xf_emit(ctx, 1, 0x20);		/* 0000ffff tesla UNK129C */
1886	xf_emit(ctx, 1, 0);		/* 000000ff turing UNK370??? */
1887	xf_emit(ctx, 1, 0);		/* 0000ffff turing USER_PARAM_COUNT */
1888	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
1889	/* SEEK */
1890	if (IS_NVA3F(dev_priv->chipset))
1891		xf_emit(ctx, 0xb, 0);	/* RO */
1892	else if (dev_priv->chipset >= 0xa0)
1893		xf_emit(ctx, 0x9, 0);	/* RO */
1894	else
1895		xf_emit(ctx, 0x8, 0);	/* RO */
1896	/* SEEK */
1897	xf_emit(ctx, 1, 0);		/* 00000001 EDGE_FLAG */
1898	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
1899	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1900	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
1901	/* SEEK */
1902	xf_emit(ctx, 0xc, 0);		/* RO */
1903	/* SEEK */
1904	xf_emit(ctx, 1, 0);		/* 7f/ff */
1905	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
1906	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
1907	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
1908	xf_emit(ctx, 1, 4);		/* 000001ff UNK1A28 */
1909	xf_emit(ctx, 1, 8);		/* 000001ff UNK0DF0 */
1910	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
1911	if (dev_priv->chipset == 0x50)
1912		xf_emit(ctx, 1, 0x3ff);	/* 3ff tesla UNK0D68 */
1913	else
1914		xf_emit(ctx, 1, 0x7ff);	/* 7ff tesla UNK0D68 */
1915	if (dev_priv->chipset == 0xa8)
1916		xf_emit(ctx, 1, 0x1e00);	/* 7fff */
1917	/* SEEK */
1918	xf_emit(ctx, 0xc, 0);		/* RO or close */
1919	/* SEEK */
1920	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
1921	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
1922	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
1923	if (dev_priv->chipset > 0x50 && dev_priv->chipset < 0xa0)
1924		xf_emit(ctx, 2, 0);	/* ffffffff */
1925	else
1926		xf_emit(ctx, 1, 0);	/* ffffffff */
1927	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK0FD8 */
1928	/* SEEK */
1929	if (IS_NVA3F(dev_priv->chipset)) {
1930		xf_emit(ctx, 0x10, 0);	/* 0? */
1931		xf_emit(ctx, 2, 0);	/* weird... */
1932		xf_emit(ctx, 2, 0);	/* RO */
1933	} else {
1934		xf_emit(ctx, 8, 0);	/* 0? */
1935		xf_emit(ctx, 1, 0);	/* weird... */
1936		xf_emit(ctx, 2, 0);	/* RO */
1937	}
1938	/* SEEK */
1939	xf_emit(ctx, 1, 0);		/* ffffffff VB_ELEMENT_BASE */
1940	xf_emit(ctx, 1, 0);		/* ffffffff UNK1438 */
1941	xf_emit(ctx, acnt, 0);		/* 1 tesla UNK1000 */
1942	if (dev_priv->chipset >= 0xa0)
1943		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1118? */
1944	/* SEEK */
1945	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_UNK90C */
1946	xf_emit(ctx, 1, 0);		/* f/1f */
1947	/* SEEK */
1948	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_UNK90C */
1949	xf_emit(ctx, 1, 0);		/* f/1f */
1950	/* SEEK */
1951	xf_emit(ctx, acnt, 0);		/* RO */
1952	xf_emit(ctx, 2, 0);		/* RO */
1953	/* SEEK */
1954	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK111C? */
1955	xf_emit(ctx, 1, 0);		/* RO */
1956	/* SEEK */
1957	xf_emit(ctx, 1, 0);		/* 000000ff UNK15F4_ADDRESS_HIGH */
1958	xf_emit(ctx, 1, 0);		/* ffffffff UNK15F4_ADDRESS_LOW */
1959	xf_emit(ctx, 1, 0);		/* 000000ff UNK0F84_ADDRESS_HIGH */
1960	xf_emit(ctx, 1, 0);		/* ffffffff UNK0F84_ADDRESS_LOW */
1961	/* SEEK */
1962	xf_emit(ctx, acnt, 0);		/* 00003fff VERTEX_ARRAY_ATTRIB_OFFSET */
1963	xf_emit(ctx, 3, 0);		/* f/1f */
1964	/* SEEK */
1965	xf_emit(ctx, acnt, 0);		/* 00000fff VERTEX_ARRAY_STRIDE */
1966	xf_emit(ctx, 3, 0);		/* f/1f */
1967	/* SEEK */
1968	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_ARRAY_LOW */
1969	xf_emit(ctx, 3, 0);		/* f/1f */
1970	/* SEEK */
1971	xf_emit(ctx, acnt, 0);		/* 000000ff VERTEX_ARRAY_HIGH */
1972	xf_emit(ctx, 3, 0);		/* f/1f */
1973	/* SEEK */
1974	xf_emit(ctx, acnt, 0);		/* ffffffff VERTEX_LIMIT_LOW */
1975	xf_emit(ctx, 3, 0);		/* f/1f */
1976	/* SEEK */
1977	xf_emit(ctx, acnt, 0);		/* 000000ff VERTEX_LIMIT_HIGH */
1978	xf_emit(ctx, 3, 0);		/* f/1f */
1979	/* SEEK */
1980	if (IS_NVA3F(dev_priv->chipset)) {
1981		xf_emit(ctx, acnt, 0);		/* f */
1982		xf_emit(ctx, 3, 0);		/* f/1f */
1983	}
1984	/* SEEK */
1985	if (IS_NVA3F(dev_priv->chipset))
1986		xf_emit(ctx, 2, 0);	/* RO */
1987	else
1988		xf_emit(ctx, 5, 0);	/* RO */
1989	/* SEEK */
1990	xf_emit(ctx, 1, 0);		/* ffff DMA_VTXBUF */
1991	/* SEEK */
1992	if (dev_priv->chipset < 0xa0) {
1993		xf_emit(ctx, 0x41, 0);	/* RO */
1994		/* SEEK */
1995		xf_emit(ctx, 0x11, 0);	/* RO */
1996	} else if (!IS_NVA3F(dev_priv->chipset))
1997		xf_emit(ctx, 0x50, 0);	/* RO */
1998	else
1999		xf_emit(ctx, 0x58, 0);	/* RO */
2000	/* SEEK */
2001	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
2002	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
2003	xf_emit(ctx, 1, 1);		/* 1 UNK0DEC */
2004	/* SEEK */
2005	xf_emit(ctx, acnt*4, 0);	/* ffffffff VTX_ATTR */
2006	xf_emit(ctx, 4, 0);		/* f/1f, 0, 0, 0 */
2007	/* SEEK */
2008	if (IS_NVA3F(dev_priv->chipset))
2009		xf_emit(ctx, 0x1d, 0);	/* RO */
2010	else
2011		xf_emit(ctx, 0x16, 0);	/* RO */
2012	/* SEEK */
2013	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
2014	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
2015	/* SEEK */
2016	if (dev_priv->chipset < 0xa0)
2017		xf_emit(ctx, 8, 0);	/* RO */
2018	else if (IS_NVA3F(dev_priv->chipset))
2019		xf_emit(ctx, 0xc, 0);	/* RO */
2020	else
2021		xf_emit(ctx, 7, 0);	/* RO */
2022	/* SEEK */
2023	xf_emit(ctx, 0xa, 0);		/* RO */
2024	if (dev_priv->chipset == 0xa0)
2025		rep = 0xc;
2026	else
2027		rep = 4;
2028	for (i = 0; i < rep; i++) {
2029		/* SEEK */
2030		if (IS_NVA3F(dev_priv->chipset))
2031			xf_emit(ctx, 0x20, 0);	/* ffffffff */
2032		xf_emit(ctx, 0x200, 0);	/* ffffffff */
2033		xf_emit(ctx, 4, 0);	/* 7f/ff, 0, 0, 0 */
2034		xf_emit(ctx, 4, 0);	/* ffffffff */
2035	}
2036	/* SEEK */
2037	xf_emit(ctx, 1, 0);		/* 113/111 */
2038	xf_emit(ctx, 1, 0xf);		/* ffffffff VP_ATTR_EN */
2039	xf_emit(ctx, (acnt/8)-1, 0);	/* ffffffff VP_ATTR_EN */
2040	xf_emit(ctx, acnt/8, 0);	/* ffffffff VTX_ATTR_MASK_UNK0DD0 */
2041	xf_emit(ctx, 1, 0);		/* 0000000f VP_GP_BUILTIN_ATTR_EN */
2042	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2043	/* SEEK */
2044	if (IS_NVA3F(dev_priv->chipset))
2045		xf_emit(ctx, 7, 0);	/* weird... */
2046	else
2047		xf_emit(ctx, 5, 0);	/* weird... */
2048}
2049
2050static void
2051nv50_graph_construct_gene_eng2d(struct nouveau_grctx *ctx)
2052{
2053	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2054	/* middle of strand 1 on pre-NVA0 [after vfetch], middle of strand 6 on NVAx */
2055	/* SEEK */
2056	xf_emit(ctx, 2, 0);		/* 0001ffff CLIP_X, CLIP_Y */
2057	xf_emit(ctx, 2, 0);		/* 0000ffff CLIP_W, CLIP_H */
2058	xf_emit(ctx, 1, 0);		/* 00000001 CLIP_ENABLE */
2059	if (dev_priv->chipset < 0xa0) {
2060		/* this is useless on everything but the original NV50,
2061		 * guess they forgot to nuke it. Or just didn't bother. */
2062		xf_emit(ctx, 2, 0);	/* 0000ffff IFC_CLIP_X, Y */
2063		xf_emit(ctx, 2, 1);	/* 0000ffff IFC_CLIP_W, H */
2064		xf_emit(ctx, 1, 0);	/* 00000001 IFC_CLIP_ENABLE */
2065	}
2066	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2067	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_WIDTH */
2068	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_HEIGHT */
2069	xf_emit(ctx, 1, 0x11);		/* 3f[NV50]/7f[NV84+] DST_FORMAT */
2070	xf_emit(ctx, 1, 0);		/* 0001ffff DRAW_POINT_X */
2071	xf_emit(ctx, 1, 8);		/* 0000000f DRAW_UNK58C */
2072	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DST_X_FRACT */
2073	xf_emit(ctx, 1, 0);		/* 0001ffff SIFC_DST_X_INT */
2074	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DST_Y_FRACT */
2075	xf_emit(ctx, 1, 0);		/* 0001ffff SIFC_DST_Y_INT */
2076	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DX_DU_FRACT */
2077	xf_emit(ctx, 1, 1);		/* 0001ffff SIFC_DX_DU_INT */
2078	xf_emit(ctx, 1, 0);		/* 000fffff SIFC_DY_DV_FRACT */
2079	xf_emit(ctx, 1, 1);		/* 0001ffff SIFC_DY_DV_INT */
2080	xf_emit(ctx, 1, 1);		/* 0000ffff SIFC_WIDTH */
2081	xf_emit(ctx, 1, 1);		/* 0000ffff SIFC_HEIGHT */
2082	xf_emit(ctx, 1, 0xcf);		/* 000000ff SIFC_FORMAT */
2083	xf_emit(ctx, 1, 2);		/* 00000003 SIFC_BITMAP_UNK808 */
2084	xf_emit(ctx, 1, 0);		/* 00000003 SIFC_BITMAP_LINE_PACK_MODE */
2085	xf_emit(ctx, 1, 0);		/* 00000001 SIFC_BITMAP_LSB_FIRST */
2086	xf_emit(ctx, 1, 0);		/* 00000001 SIFC_BITMAP_ENABLE */
2087	xf_emit(ctx, 1, 0);		/* 0000ffff BLIT_DST_X */
2088	xf_emit(ctx, 1, 0);		/* 0000ffff BLIT_DST_Y */
2089	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DU_DX_FRACT */
2090	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DU_DX_INT */
2091	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DV_DY_FRACT */
2092	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DV_DY_INT */
2093	xf_emit(ctx, 1, 1);		/* 0000ffff BLIT_DST_W */
2094	xf_emit(ctx, 1, 1);		/* 0000ffff BLIT_DST_H */
2095	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_SRC_X_FRACT */
2096	xf_emit(ctx, 1, 0);		/* 0001ffff BLIT_SRC_X_INT */
2097	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_SRC_Y_FRACT */
2098	xf_emit(ctx, 1, 0);		/* 00000001 UNK888 */
2099	xf_emit(ctx, 1, 4);		/* 0000003f UNK884 */
2100	xf_emit(ctx, 1, 0);		/* 00000007 UNK880 */
2101	xf_emit(ctx, 1, 1);		/* 0000001f tesla UNK0FB8 */
2102	xf_emit(ctx, 1, 0x15);		/* 000000ff tesla UNK128C */
2103	xf_emit(ctx, 2, 0);		/* 00000007, ffff0ff3 */
2104	xf_emit(ctx, 1, 0);		/* 00000001 UNK260 */
2105	xf_emit(ctx, 1, 0x4444480);	/* 1fffffff UNK870 */
2106	/* SEEK */
2107	xf_emit(ctx, 0x10, 0);
2108	/* SEEK */
2109	xf_emit(ctx, 0x27, 0);
2110}
2111
2112static void
2113nv50_graph_construct_gene_csched(struct nouveau_grctx *ctx)
2114{
2115	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2116	/* middle of strand 1 on pre-NVA0 [after eng2d], middle of strand 0 on NVAx */
2117	/* SEEK */
2118	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY... what is it doing here??? */
2119	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1924 */
2120	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
2121	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
2122	xf_emit(ctx, 1, 0);		/* 000003ff */
2123	/* SEEK */
2124	xf_emit(ctx, 1, 0);		/* ffffffff turing UNK364 */
2125	xf_emit(ctx, 1, 0);		/* 0000000f turing UNK36C */
2126	xf_emit(ctx, 1, 0);		/* 0000ffff USER_PARAM_COUNT */
2127	xf_emit(ctx, 1, 0x100);		/* 00ffffff turing UNK384 */
2128	xf_emit(ctx, 1, 0);		/* 0000000f turing UNK2A0 */
2129	xf_emit(ctx, 1, 0);		/* 0000ffff GRIDID */
2130	xf_emit(ctx, 1, 0x10001);	/* ffffffff GRIDDIM_XY */
2131	xf_emit(ctx, 1, 0);		/* ffffffff */
2132	xf_emit(ctx, 1, 0x10001);	/* ffffffff BLOCKDIM_XY */
2133	xf_emit(ctx, 1, 1);		/* 0000ffff BLOCKDIM_Z */
2134	xf_emit(ctx, 1, 0x10001);	/* 00ffffff BLOCK_ALLOC */
2135	xf_emit(ctx, 1, 1);		/* 00000001 LANES32 */
2136	xf_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
2137	xf_emit(ctx, 1, 2);		/* 00000003 REG_MODE */
2138	/* SEEK */
2139	xf_emit(ctx, 0x40, 0);		/* ffffffff USER_PARAM */
2140	switch (dev_priv->chipset) {
2141	case 0x50:
2142	case 0x92:
2143		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2144		xf_emit(ctx, 0x80, 0);	/* fff */
2145		xf_emit(ctx, 2, 0);	/* ff, fff */
2146		xf_emit(ctx, 0x10*2, 0);	/* ffffffff, 1f */
2147		break;
2148	case 0x84:
2149		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2150		xf_emit(ctx, 0x60, 0);	/* fff */
2151		xf_emit(ctx, 2, 0);	/* ff, fff */
2152		xf_emit(ctx, 0xc*2, 0);	/* ffffffff, 1f */
2153		break;
2154	case 0x94:
2155	case 0x96:
2156		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2157		xf_emit(ctx, 0x40, 0);	/* fff */
2158		xf_emit(ctx, 2, 0);	/* ff, fff */
2159		xf_emit(ctx, 8*2, 0);	/* ffffffff, 1f */
2160		break;
2161	case 0x86:
2162	case 0x98:
2163		xf_emit(ctx, 4, 0);	/* f, 0, 0, 0 */
2164		xf_emit(ctx, 0x10, 0);	/* fff */
2165		xf_emit(ctx, 2, 0);	/* ff, fff */
2166		xf_emit(ctx, 2*2, 0);	/* ffffffff, 1f */
2167		break;
2168	case 0xa0:
2169		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2170		xf_emit(ctx, 0xf0, 0);	/* fff */
2171		xf_emit(ctx, 2, 0);	/* ff, fff */
2172		xf_emit(ctx, 0x1e*2, 0);	/* ffffffff, 1f */
2173		break;
2174	case 0xa3:
2175		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2176		xf_emit(ctx, 0x60, 0);	/* fff */
2177		xf_emit(ctx, 2, 0);	/* ff, fff */
2178		xf_emit(ctx, 0xc*2, 0);	/* ffffffff, 1f */
2179		break;
2180	case 0xa5:
2181	case 0xaf:
2182		xf_emit(ctx, 8, 0);	/* 7, 0, 0, 0, ... */
2183		xf_emit(ctx, 0x30, 0);	/* fff */
2184		xf_emit(ctx, 2, 0);	/* ff, fff */
2185		xf_emit(ctx, 6*2, 0);	/* ffffffff, 1f */
2186		break;
2187	case 0xaa:
2188		xf_emit(ctx, 0x12, 0);
2189		break;
2190	case 0xa8:
2191	case 0xac:
2192		xf_emit(ctx, 4, 0);	/* f, 0, 0, 0 */
2193		xf_emit(ctx, 0x10, 0);	/* fff */
2194		xf_emit(ctx, 2, 0);	/* ff, fff */
2195		xf_emit(ctx, 2*2, 0);	/* ffffffff, 1f */
2196		break;
2197	}
2198	xf_emit(ctx, 1, 0);		/* 0000000f */
2199	xf_emit(ctx, 1, 0);		/* 00000000 */
2200	xf_emit(ctx, 1, 0);		/* ffffffff */
2201	xf_emit(ctx, 1, 0);		/* 0000001f */
2202	xf_emit(ctx, 4, 0);		/* ffffffff */
2203	xf_emit(ctx, 1, 0);		/* 00000003 turing UNK35C */
2204	xf_emit(ctx, 1, 0);		/* ffffffff */
2205	xf_emit(ctx, 4, 0);		/* ffffffff */
2206	xf_emit(ctx, 1, 0);		/* 00000003 turing UNK35C */
2207	xf_emit(ctx, 1, 0);		/* ffffffff */
2208	xf_emit(ctx, 1, 0);		/* 000000ff */
2209}
2210
2211static void
2212nv50_graph_construct_gene_unk1cxx(struct nouveau_grctx *ctx)
2213{
2214	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2215	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
2216	xf_emit(ctx, 1, 0x3f800000);	/* ffffffff LINE_WIDTH */
2217	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
2218	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1658 */
2219	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_SMOOTH_ENABLE */
2220	xf_emit(ctx, 3, 0);		/* 00000001 POLYGON_OFFSET_*_ENABLE */
2221	xf_emit(ctx, 1, 4);		/* 0000000f CULL_MODE */
2222	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
2223	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2224	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
2225	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK165C */
2226	xf_emit(ctx, 0x10, 0);		/* 00000001 SCISSOR_ENABLE */
2227	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2228	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
2229	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
2230	xf_emit(ctx, 1, 0);		/* ffffffff POLYGON_OFFSET_UNITS */
2231	xf_emit(ctx, 1, 0);		/* ffffffff POLYGON_OFFSET_FACTOR */
2232	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1668 */
2233	xf_emit(ctx, 2, 0);		/* 07ffffff SCREEN_SCISSOR */
2234	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
2235	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2236	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2237	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2238	xf_emit(ctx, 1, 0x11);		/* 0000007f RT_FORMAT */
2239	xf_emit(ctx, 7, 0);		/* 0000007f RT_FORMAT */
2240	xf_emit(ctx, 8, 0);		/* 00000001 RT_HORIZ_LINEAR */
2241	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
2242	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
2243	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
2244	if (IS_NVA3F(dev_priv->chipset))
2245		xf_emit(ctx, 1, 3);	/* 00000003 UNK16B4 */
2246	else if (dev_priv->chipset >= 0xa0)
2247		xf_emit(ctx, 1, 1);	/* 00000001 UNK16B4 */
2248	xf_emit(ctx, 1, 0);		/* 00000003 MULTISAMPLE_CTRL */
2249	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK0F90 */
2250	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
2251	xf_emit(ctx, 2, 0x04000000);	/* 07ffffff tesla UNK0D6C */
2252	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
2253	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2254	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
2255	xf_emit(ctx, 1, 5);		/* 0000000f UNK1408 */
2256	xf_emit(ctx, 1, 0x52);		/* 000001ff SEMANTIC_PTSZ */
2257	xf_emit(ctx, 1, 0);		/* ffffffff POINT_SIZE */
2258	xf_emit(ctx, 1, 0);		/* 00000001 */
2259	xf_emit(ctx, 1, 0);		/* 00000007 tesla UNK0FB4 */
2260	if (dev_priv->chipset != 0x50) {
2261		xf_emit(ctx, 1, 0);	/* 3ff */
2262		xf_emit(ctx, 1, 1);	/* 00000001 tesla UNK1110 */
2263	}
2264	if (IS_NVA3F(dev_priv->chipset))
2265		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1928 */
2266	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
2267	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
2268	xf_emit(ctx, 1, 0x10);		/* 000000ff VIEW_VOLUME_CLIP_CTRL */
2269	xf_emit(ctx, 0x20, 0);		/* 07ffffff VIEWPORT_HORIZ, then VIEWPORT_VERT. (W&0x3fff)<<13 | (X&0x1fff). */
2270	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK187C */
2271	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
2272	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2273	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2274	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2275	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
2276	xf_emit(ctx, 1, 0x8100c12);	/* 1fffffff FP_INTERPOLANT_CTRL */
2277	xf_emit(ctx, 1, 5);		/* 0000000f tesla UNK1220 */
2278	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2279	xf_emit(ctx, 1, 0);		/* 000000ff tesla UNK1A20 */
2280	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2281	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
2282	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
2283	if (dev_priv->chipset != 0x50)
2284		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
2285	if (dev_priv->chipset < 0xa0)
2286		xf_emit(ctx, 0x1c, 0);	/* RO */
2287	else if (IS_NVA3F(dev_priv->chipset))
2288		xf_emit(ctx, 0x9, 0);
2289	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
2290	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
2291	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
2292	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
2293	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
2294	xf_emit(ctx, 1, 0);		/* 00000003 WINDOW_ORIGIN */
2295	if (dev_priv->chipset != 0x50) {
2296		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK1100 */
2297		xf_emit(ctx, 1, 0);	/* 3ff */
2298	}
2299	/* XXX: the following block could belong either to unk1cxx, or
2300	 * to STRMOUT. Rather hard to tell. */
2301	if (dev_priv->chipset < 0xa0)
2302		xf_emit(ctx, 0x25, 0);
2303	else
2304		xf_emit(ctx, 0x3b, 0);
2305}
2306
2307static void
2308nv50_graph_construct_gene_strmout(struct nouveau_grctx *ctx)
2309{
2310	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2311	xf_emit(ctx, 1, 0x102);		/* 0000ffff STRMOUT_BUFFER_CTRL */
2312	xf_emit(ctx, 1, 0);		/* ffffffff STRMOUT_PRIMITIVE_COUNT */
2313	xf_emit(ctx, 4, 4);		/* 000000ff STRMOUT_NUM_ATTRIBS */
2314	if (dev_priv->chipset >= 0xa0) {
2315		xf_emit(ctx, 4, 0);	/* ffffffff UNK1A8C */
2316		xf_emit(ctx, 4, 0);	/* ffffffff UNK1780 */
2317	}
2318	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
2319	xf_emit(ctx, 1, 4);		/* 0000007f VP_RESULT_MAP_SIZE */
2320	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2321	if (dev_priv->chipset == 0x50)
2322		xf_emit(ctx, 1, 0x3ff);	/* 000003ff tesla UNK0D68 */
2323	else
2324		xf_emit(ctx, 1, 0x7ff);	/* 000007ff tesla UNK0D68 */
2325	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2326	/* SEEK */
2327	xf_emit(ctx, 1, 0x102);		/* 0000ffff STRMOUT_BUFFER_CTRL */
2328	xf_emit(ctx, 1, 0);		/* ffffffff STRMOUT_PRIMITIVE_COUNT */
2329	xf_emit(ctx, 4, 0);		/* 000000ff STRMOUT_ADDRESS_HIGH */
2330	xf_emit(ctx, 4, 0);		/* ffffffff STRMOUT_ADDRESS_LOW */
2331	xf_emit(ctx, 4, 4);		/* 000000ff STRMOUT_NUM_ATTRIBS */
2332	if (dev_priv->chipset >= 0xa0) {
2333		xf_emit(ctx, 4, 0);	/* ffffffff UNK1A8C */
2334		xf_emit(ctx, 4, 0);	/* ffffffff UNK1780 */
2335	}
2336	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_STRMOUT */
2337	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
2338	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
2339	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW QUERY_COUNTER */
2340	xf_emit(ctx, 2, 0);		/* ffffffff */
2341	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2342	/* SEEK */
2343	xf_emit(ctx, 0x20, 0);		/* ffffffff STRMOUT_MAP */
2344	xf_emit(ctx, 1, 0);		/* 0000000f */
2345	xf_emit(ctx, 1, 0);		/* 00000000? */
2346	xf_emit(ctx, 2, 0);		/* ffffffff */
2347}
2348
2349static void
2350nv50_graph_construct_gene_ropm1(struct nouveau_grctx *ctx)
2351{
2352	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2353	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0D64 */
2354	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0DF4 */
2355	xf_emit(ctx, 1, 0);		/* 00000007 */
2356	xf_emit(ctx, 1, 0);		/* 000003ff */
2357	if (IS_NVA3F(dev_priv->chipset))
2358		xf_emit(ctx, 1, 0x11);	/* 000000ff tesla UNK1968 */
2359	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2360}
2361
2362static void
2363nv50_graph_construct_gene_ropm2(struct nouveau_grctx *ctx)
2364{
2365	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2366	/* SEEK */
2367	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
2368	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2369	xf_emit(ctx, 2, 0);		/* ffffffff */
2370	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
2371	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW, COUNTER */
2372	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
2373	xf_emit(ctx, 1, 0);		/* 7 */
2374	/* SEEK */
2375	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_QUERY */
2376	xf_emit(ctx, 1, 0);		/* 000000ff QUERY_ADDRESS_HIGH */
2377	xf_emit(ctx, 2, 0);		/* ffffffff QUERY_ADDRESS_LOW, COUNTER */
2378	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0D64 */
2379	xf_emit(ctx, 1, 0x4e3bfdf);	/* ffffffff UNK0DF4 */
2380	xf_emit(ctx, 1, 0);		/* 00000001 eng2d UNK260 */
2381	xf_emit(ctx, 1, 0);		/* ff/3ff */
2382	xf_emit(ctx, 1, 0);		/* 00000007 */
2383	if (IS_NVA3F(dev_priv->chipset))
2384		xf_emit(ctx, 1, 0x11);	/* 000000ff tesla UNK1968 */
2385	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2386}
2387
2388static void
2389nv50_graph_construct_gene_ropc(struct nouveau_grctx *ctx)
2390{
2391	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2392	int magic2;
2393	if (dev_priv->chipset == 0x50) {
2394		magic2 = 0x00003e60;
2395	} else if (!IS_NVA3F(dev_priv->chipset)) {
2396		magic2 = 0x001ffe67;
2397	} else {
2398		magic2 = 0x00087e67;
2399	}
2400	xf_emit(ctx, 1, 0);		/* f/7 MUTISAMPLE_SAMPLES_LOG2 */
2401	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2402	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
2403	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
2404	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
2405	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
2406	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
2407	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2408	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
2409	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2410	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
2411	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2412	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2413	if (IS_NVA3F(dev_priv->chipset))
2414		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2415	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
2416	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
2417	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
2418	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
2419	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2420	if (dev_priv->chipset >= 0xa0 && !IS_NVAAF(dev_priv->chipset))
2421		xf_emit(ctx, 1, 0x15);	/* 000000ff */
2422	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2423	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2424	xf_emit(ctx, 1, 0x10);		/* 3ff/ff VIEW_VOLUME_CLIP_CTRL */
2425	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
2426	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2427	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2428	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2429	if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x92 || dev_priv->chipset == 0x98 || dev_priv->chipset >= 0xa0) {
2430		xf_emit(ctx, 3, 0);	/* ff, ffffffff, ffffffff */
2431		xf_emit(ctx, 1, 4);	/* 7 */
2432		xf_emit(ctx, 1, 0x400);	/* fffffff */
2433		xf_emit(ctx, 1, 0x300);	/* ffff */
2434		xf_emit(ctx, 1, 0x1001);	/* 1fff */
2435		if (dev_priv->chipset != 0xa0) {
2436			if (IS_NVA3F(dev_priv->chipset))
2437				xf_emit(ctx, 1, 0);	/* 0000000f UNK15C8 */
2438			else
2439				xf_emit(ctx, 1, 0x15);	/* ff */
2440		}
2441	}
2442	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2443	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2444	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
2445	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
2446	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2447	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
2448	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2449	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
2450	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2451	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2452	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
2453	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
2454	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2455	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2456	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2457	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
2458	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2459	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2460	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2461	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2462	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1900 */
2463	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
2464	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
2465	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
2466	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
2467	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2468	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
2469	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2470	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2471	xf_emit(ctx, 1, 0);		/* 0000000f */
2472	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
2473	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
2474	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
2475	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
2476	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2477	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2478	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
2479	xf_emit(ctx, 0x10, 0);		/* ffffffff DEPTH_RANGE_NEAR */
2480	xf_emit(ctx, 0x10, 0x3f800000);	/* ffffffff DEPTH_RANGE_FAR */
2481	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2482	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2483	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_BACK_FUNC_FUNC */
2484	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_MASK */
2485	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_FUNC_REF */
2486	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
2487	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
2488	xf_emit(ctx, 2, 0);		/* ffffffff DEPTH_BOUNDS */
2489	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2490	xf_emit(ctx, 1, 0);		/* 00000007 DEPTH_TEST_FUNC */
2491	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2492	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2493	xf_emit(ctx, 1, 0);		/* 000000ff CLEAR_STENCIL */
2494	xf_emit(ctx, 1, 0);		/* 00000007 STENCIL_FRONT_FUNC_FUNC */
2495	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_MASK */
2496	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_FUNC_REF */
2497	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
2498	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
2499	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2500	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2501	xf_emit(ctx, 1, 0x10);		/* 7f/ff VIEW_VOLUME_CLIP_CTRL */
2502	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2503	xf_emit(ctx, 1, 0x3f);		/* 0000003f UNK1590 */
2504	xf_emit(ctx, 1, 0);		/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2505	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2506	xf_emit(ctx, 2, 0);		/* ffff0ff3, ffff */
2507	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
2508	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
2509	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2510	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2511	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2512	xf_emit(ctx, 1, 0);		/* ffffffff CLEAR_DEPTH */
2513	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
2514	if (dev_priv->chipset >= 0xa0) {
2515		xf_emit(ctx, 2, 0);
2516		xf_emit(ctx, 1, 0x1001);
2517		xf_emit(ctx, 0xb, 0);
2518	} else {
2519		xf_emit(ctx, 1, 0);	/* 00000007 */
2520		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1534 */
2521		xf_emit(ctx, 1, 0);	/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2522		xf_emit(ctx, 8, 0);	/* 00000001 BLEND_ENABLE */
2523		xf_emit(ctx, 1, 0);	/* ffff0ff3 */
2524	}
2525	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2526	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2527	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2528	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2529	xf_emit(ctx, 1, 0x11);		/* 3f/7f */
2530	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
2531	if (dev_priv->chipset != 0x50) {
2532		xf_emit(ctx, 1, 0);	/* 0000000f LOGIC_OP */
2533		xf_emit(ctx, 1, 0);	/* 000000ff */
2534	}
2535	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
2536	xf_emit(ctx, 1, 0);		/* ff/3ff */
2537	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
2538	xf_emit(ctx, 2, 1);		/* 00000007 BLEND_EQUATION_RGB, ALPHA */
2539	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
2540	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
2541	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
2542	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
2543	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
2544	xf_emit(ctx, 1, 0);		/* 00000001 */
2545	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
2546	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2547	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2548	if (IS_NVA3F(dev_priv->chipset)) {
2549		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK12E4 */
2550		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
2551		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
2552		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
2553		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_RGB */
2554		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_RGB */
2555		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_ALPHA */
2556		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_ALPHA */
2557		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1140 */
2558		xf_emit(ctx, 2, 0);	/* 00000001 */
2559		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2560		xf_emit(ctx, 1, 0);	/* 0000000f */
2561		xf_emit(ctx, 1, 0);	/* 00000003 */
2562		xf_emit(ctx, 1, 0);	/* ffffffff */
2563		xf_emit(ctx, 2, 0);	/* 00000001 */
2564		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2565		xf_emit(ctx, 1, 0);	/* 00000001 */
2566		xf_emit(ctx, 1, 0);	/* 000003ff */
2567	} else if (dev_priv->chipset >= 0xa0) {
2568		xf_emit(ctx, 2, 0);	/* 00000001 */
2569		xf_emit(ctx, 1, 0);	/* 00000007 */
2570		xf_emit(ctx, 1, 0);	/* 00000003 */
2571		xf_emit(ctx, 1, 0);	/* ffffffff */
2572		xf_emit(ctx, 2, 0);	/* 00000001 */
2573	} else {
2574		xf_emit(ctx, 1, 0);	/* 00000007 MULTISAMPLE_SAMPLES_LOG2 */
2575		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1430 */
2576		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A3C */
2577	}
2578	xf_emit(ctx, 4, 0);		/* ffffffff CLEAR_COLOR */
2579	xf_emit(ctx, 4, 0);		/* ffffffff BLEND_COLOR A R G B */
2580	xf_emit(ctx, 1, 0);		/* 00000fff eng2d UNK2B0 */
2581	if (dev_priv->chipset >= 0xa0)
2582		xf_emit(ctx, 2, 0);	/* 00000001 */
2583	xf_emit(ctx, 1, 0);		/* 000003ff */
2584	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
2585	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
2586	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
2587	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
2588	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
2589	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
2590	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
2591	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
2592	xf_emit(ctx, 1, 0);		/* 00000001 UNK19C0 */
2593	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
2594	xf_emit(ctx, 1, 0);		/* 0000000f LOGIC_OP */
2595	if (dev_priv->chipset >= 0xa0)
2596		xf_emit(ctx, 1, 0);	/* 00000001 UNK12E4? NVA3+ only? */
2597	if (IS_NVA3F(dev_priv->chipset)) {
2598		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
2599		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
2600		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_RGB */
2601		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_RGB */
2602		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
2603		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_FUNC_SRC_ALPHA */
2604		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_FUNC_DST_ALPHA */
2605		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK15C4 */
2606		xf_emit(ctx, 1, 0);	/* 00000001 */
2607		xf_emit(ctx, 1, 0);	/* 00000001 tesla UNK1140 */
2608	}
2609	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2610	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2611	xf_emit(ctx, 1, 0);		/* 00000007 PATTERN_COLOR_FORMAT */
2612	xf_emit(ctx, 2, 0);		/* ffffffff PATTERN_MONO_COLOR */
2613	xf_emit(ctx, 1, 0);		/* 00000001 PATTERN_MONO_FORMAT */
2614	xf_emit(ctx, 2, 0);		/* ffffffff PATTERN_MONO_BITMAP */
2615	xf_emit(ctx, 1, 0);		/* 00000003 PATTERN_SELECT */
2616	xf_emit(ctx, 1, 0);		/* 000000ff ROP */
2617	xf_emit(ctx, 1, 0);		/* ffffffff BETA1 */
2618	xf_emit(ctx, 1, 0);		/* ffffffff BETA4 */
2619	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
2620	xf_emit(ctx, 0x50, 0);		/* 10x ffffff, ffffff, ffffff, ffffff, 3 PATTERN */
2621}
2622
2623static void
2624nv50_graph_construct_xfer_unk84xx(struct nouveau_grctx *ctx)
2625{
2626	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2627	int magic3;
2628	switch (dev_priv->chipset) {
2629	case 0x50:
2630		magic3 = 0x1000;
2631		break;
2632	case 0x86:
2633	case 0x98:
2634	case 0xa8:
2635	case 0xaa:
2636	case 0xac:
2637	case 0xaf:
2638		magic3 = 0x1e00;
2639		break;
2640	default:
2641		magic3 = 0;
2642	}
2643	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2644	xf_emit(ctx, 1, 4);		/* 7f/ff[NVA0+] VP_REG_ALLOC_RESULT */
2645	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2646	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2647	xf_emit(ctx, 1, 0);		/* 111/113[NVA0+] */
2648	if (IS_NVA3F(dev_priv->chipset))
2649		xf_emit(ctx, 0x1f, 0);	/* ffffffff */
2650	else if (dev_priv->chipset >= 0xa0)
2651		xf_emit(ctx, 0x0f, 0);	/* ffffffff */
2652	else
2653		xf_emit(ctx, 0x10, 0);	/* fffffff VP_RESULT_MAP_1 up */
2654	xf_emit(ctx, 2, 0);		/* f/1f[NVA3], fffffff/ffffffff[NVA0+] */
2655	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
2656	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
2657	if (dev_priv->chipset >= 0xa0)
2658		xf_emit(ctx, 1, 0x03020100);	/* ffffffff */
2659	else
2660		xf_emit(ctx, 1, 0x00608080);	/* fffffff VP_RESULT_MAP_0 */
2661	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2662	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2663	xf_emit(ctx, 2, 0);		/* 111/113, 7f/ff */
2664	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
2665	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2666	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2667	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
2668	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
2669	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
2670	if (magic3)
2671		xf_emit(ctx, 1, magic3);	/* 00007fff tesla UNK141C */
2672	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
2673	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2674	xf_emit(ctx, 1, 0);		/* 111/113 */
2675	xf_emit(ctx, 0x1f, 0);		/* ffffffff GP_RESULT_MAP_1 up */
2676	xf_emit(ctx, 1, 0);		/* 0000001f */
2677	xf_emit(ctx, 1, 0);		/* ffffffff */
2678	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2679	xf_emit(ctx, 1, 4);		/* 000000ff GP_REG_ALLOC_RESULT */
2680	xf_emit(ctx, 1, 0x80);		/* 0000ffff GP_VERTEX_OUTPUT_COUNT */
2681	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
2682	xf_emit(ctx, 1, 0x03020100);	/* ffffffff GP_RESULT_MAP_0 */
2683	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
2684	if (magic3)
2685		xf_emit(ctx, 1, magic3);	/* 7fff tesla UNK141C */
2686	xf_emit(ctx, 1, 4);		/* 7f/ff VP_RESULT_MAP_SIZE */
2687	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
2688	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2689	xf_emit(ctx, 1, 0);		/* 111/113 */
2690	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2691	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
2692	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
2693	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
2694	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2695	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK13A0 */
2696	xf_emit(ctx, 1, 4);		/* 7f/ff VP_REG_ALLOC_RESULT */
2697	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2698	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2699	xf_emit(ctx, 1, 0);		/* 111/113 */
2700	if (dev_priv->chipset == 0x94 || dev_priv->chipset == 0x96)
2701		xf_emit(ctx, 0x1020, 0);	/* 4 x (0x400 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */
2702	else if (dev_priv->chipset < 0xa0)
2703		xf_emit(ctx, 0xa20, 0);	/* 4 x (0x280 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */
2704	else if (!IS_NVA3F(dev_priv->chipset))
2705		xf_emit(ctx, 0x210, 0);	/* ffffffff */
2706	else
2707		xf_emit(ctx, 0x410, 0);	/* ffffffff */
2708	xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
2709	xf_emit(ctx, 1, 4);		/* 000000ff GP_RESULT_MAP_SIZE */
2710	xf_emit(ctx, 1, 3);		/* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */
2711	xf_emit(ctx, 1, 0);		/* 00000001 PROVOKING_VERTEX_LAST */
2712	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
2713}
2714
2715static void
2716nv50_graph_construct_xfer_tprop(struct nouveau_grctx *ctx)
2717{
2718	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
2719	int magic1, magic2;
2720	if (dev_priv->chipset == 0x50) {
2721		magic1 = 0x3ff;
2722		magic2 = 0x00003e60;
2723	} else if (!IS_NVA3F(dev_priv->chipset)) {
2724		magic1 = 0x7ff;
2725		magic2 = 0x001ffe67;
2726	} else {
2727		magic1 = 0x7ff;
2728		magic2 = 0x00087e67;
2729	}
2730	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
2731	xf_emit(ctx, 1, 0);		/* ffffffff ALPHA_TEST_REF */
2732	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
2733	if (IS_NVA3F(dev_priv->chipset))
2734		xf_emit(ctx, 1, 1);	/* 0000000f UNK16A0 */
2735	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2736	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2737	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_BACK_MASK */
2738	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */
2739	xf_emit(ctx, 4, 0);		/* ffffffff BLEND_COLOR */
2740	xf_emit(ctx, 1, 0);		/* 00000001 UNK19C0 */
2741	xf_emit(ctx, 1, 0);		/* 00000001 UNK0FDC */
2742	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2743	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2744	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2745	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2746	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
2747	xf_emit(ctx, 1, 0);		/* ff[NV50]/3ff[NV84+] */
2748	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
2749	xf_emit(ctx, 4, 0xffff);	/* 0000ffff MSAA_MASK */
2750	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
2751	xf_emit(ctx, 3, 0);		/* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */
2752	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2753	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_BACK_ENABLE */
2754	xf_emit(ctx, 2, 0);		/* 00007fff WINDOW_OFFSET_XY */
2755	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
2756	xf_emit(ctx, 1, 0);		/* 7 */
2757	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
2758	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2759	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2760	xf_emit(ctx, 1, 0);		/* ffffffff COLOR_KEY */
2761	xf_emit(ctx, 1, 0);		/* 00000001 COLOR_KEY_ENABLE */
2762	xf_emit(ctx, 1, 0);		/* 00000007 COLOR_KEY_FORMAT */
2763	xf_emit(ctx, 2, 0);		/* ffffffff SIFC_BITMAP_COLOR */
2764	xf_emit(ctx, 1, 1);		/* 00000001 SIFC_BITMAP_WRITE_BIT0_ENABLE */
2765	xf_emit(ctx, 1, 0);		/* 00000007 ALPHA_TEST_FUNC */
2766	xf_emit(ctx, 1, 0);		/* 00000001 ALPHA_TEST_ENABLE */
2767	if (IS_NVA3F(dev_priv->chipset)) {
2768		xf_emit(ctx, 1, 3);	/* 00000003 tesla UNK16B4 */
2769		xf_emit(ctx, 1, 0);	/* 00000003 */
2770		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1298 */
2771	} else if (dev_priv->chipset >= 0xa0) {
2772		xf_emit(ctx, 1, 1);	/* 00000001 tesla UNK16B4 */
2773		xf_emit(ctx, 1, 0);	/* 00000003 */
2774	} else {
2775		xf_emit(ctx, 1, 0);	/* 00000003 MULTISAMPLE_CTRL */
2776	}
2777	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2778	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
2779	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
2780	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
2781	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
2782	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
2783	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
2784	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
2785	if (IS_NVA3F(dev_priv->chipset)) {
2786		xf_emit(ctx, 1, 0);	/* 00000001 UNK12E4 */
2787		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_RGB */
2788		xf_emit(ctx, 8, 1);	/* 00000007 IBLEND_EQUATION_ALPHA */
2789		xf_emit(ctx, 8, 1);	/* 00000001 IBLEND_UNK00 */
2790		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_SRC_RGB */
2791		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_DST_RGB */
2792		xf_emit(ctx, 8, 2);	/* 0000001f IBLEND_SRC_ALPHA */
2793		xf_emit(ctx, 8, 1);	/* 0000001f IBLEND_DST_ALPHA */
2794		xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
2795	}
2796	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
2797	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2798	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2799	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2800	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2801	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
2802	xf_emit(ctx, 1, 0);		/* ff/3ff */
2803	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
2804	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
2805	xf_emit(ctx, 1, 0);		/* 00000001 FRAMEBUFFER_SRGB */
2806	xf_emit(ctx, 1, 0);		/* 7 */
2807	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2808	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2809	xf_emit(ctx, 1, 0);		/* 00000007 OPERATION */
2810	xf_emit(ctx, 1, 0xcf);		/* 000000ff SIFC_FORMAT */
2811	xf_emit(ctx, 1, 0xcf);		/* 000000ff DRAW_COLOR_FORMAT */
2812	xf_emit(ctx, 1, 0xcf);		/* 000000ff SRC_FORMAT */
2813	if (IS_NVA3F(dev_priv->chipset))
2814		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2815	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2816	xf_emit(ctx, 1, 0);		/* 7/f[NVA3] MULTISAMPLE_SAMPLES_LOG2 */
2817	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
2818	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_ALPHA */
2819	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_ALPHA */
2820	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_ALPHA */
2821	xf_emit(ctx, 1, 1);		/* 0000001f BLEND_FUNC_DST_RGB */
2822	xf_emit(ctx, 1, 1);		/* 00000007 BLEND_EQUATION_RGB */
2823	xf_emit(ctx, 1, 2);		/* 0000001f BLEND_FUNC_SRC_RGB */
2824	xf_emit(ctx, 1, 1);		/* 00000001 UNK133C */
2825	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2826	xf_emit(ctx, 8, 1);		/* 00000001 UNK19E0 */
2827	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2828	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2829	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2830	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2831	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2832	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
2833	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2834	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2835	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2836	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2837	if (IS_NVA3F(dev_priv->chipset))
2838		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2839	if(dev_priv->chipset == 0x50)
2840		xf_emit(ctx, 1, 0);	/* ff */
2841	else
2842		xf_emit(ctx, 3, 0);	/* 1, 7, 3ff */
2843	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
2844	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
2845	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2846	xf_emit(ctx, 1, 0);		/* 00000007 */
2847	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
2848	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2849	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2850	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2851	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2852	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2853	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2854	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2855	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2856	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2857	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2858	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2859	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2860	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2861	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2862	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DU_DX_FRACT */
2863	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DU_DX_INT */
2864	xf_emit(ctx, 1, 0);		/* 000fffff BLIT_DV_DY_FRACT */
2865	xf_emit(ctx, 1, 1);		/* 0001ffff BLIT_DV_DY_INT */
2866	xf_emit(ctx, 1, 0);		/* ff/3ff */
2867	xf_emit(ctx, 1, magic1);	/* 3ff/7ff tesla UNK0D68 */
2868	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2869	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2870	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2871	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2872	xf_emit(ctx, 1, 0);		/* 00000007 */
2873	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2874	if (IS_NVA3F(dev_priv->chipset))
2875		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2876	xf_emit(ctx, 8, 0);		/* 0000ffff DMA_COLOR */
2877	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_GLOBAL */
2878	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_LOCAL */
2879	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_STACK */
2880	xf_emit(ctx, 1, 0);		/* ff/3ff */
2881	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_DST */
2882	xf_emit(ctx, 1, 0);		/* 7 */
2883	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2884	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2885	xf_emit(ctx, 8, 0);		/* 000000ff RT_ADDRESS_HIGH */
2886	xf_emit(ctx, 8, 0);		/* ffffffff RT_LAYER_STRIDE */
2887	xf_emit(ctx, 8, 0);		/* ffffffff RT_ADDRESS_LOW */
2888	xf_emit(ctx, 8, 8);		/* 0000007f RT_TILE_MODE */
2889	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2890	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2891	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2892	xf_emit(ctx, 8, 0x400);		/* 0fffffff RT_HORIZ */
2893	xf_emit(ctx, 8, 0x300);		/* 0000ffff RT_VERT */
2894	xf_emit(ctx, 1, 1);		/* 00001fff RT_ARRAY_MODE */
2895	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2896	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2897	xf_emit(ctx, 1, 0x20);		/* 00000fff DST_TILE_MODE */
2898	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2899	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_HEIGHT */
2900	xf_emit(ctx, 1, 0);		/* 000007ff DST_LAYER */
2901	xf_emit(ctx, 1, 1);		/* 00000001 DST_LINEAR */
2902	xf_emit(ctx, 1, 0);		/* ffffffff DST_ADDRESS_LOW */
2903	xf_emit(ctx, 1, 0);		/* 000000ff DST_ADDRESS_HIGH */
2904	xf_emit(ctx, 1, 0x40);		/* 0007ffff DST_PITCH */
2905	xf_emit(ctx, 1, 0x100);		/* 0001ffff DST_WIDTH */
2906	xf_emit(ctx, 1, 0);		/* 0000ffff */
2907	xf_emit(ctx, 1, 3);		/* 00000003 tesla UNK15AC */
2908	xf_emit(ctx, 1, 0);		/* ff/3ff */
2909	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
2910	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
2911	xf_emit(ctx, 1, 0);		/* 00000007 */
2912	if (IS_NVA3F(dev_priv->chipset))
2913		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2914	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
2915	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2916	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1534 */
2917	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2918	xf_emit(ctx, 1, 2);		/* 00000003 tesla UNK143C */
2919	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2920	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_ZETA */
2921	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2922	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2923	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2924	xf_emit(ctx, 2, 0);		/* ffff, ff/3ff */
2925	xf_emit(ctx, 1, 0);		/* 0001ffff GP_BUILTIN_RESULT_EN */
2926	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2927	xf_emit(ctx, 1, 0);		/* 000000ff STENCIL_FRONT_MASK */
2928	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2929	xf_emit(ctx, 1, 0);		/* 00000007 */
2930	xf_emit(ctx, 1, 0);		/* ffffffff ZETA_LAYER_STRIDE */
2931	xf_emit(ctx, 1, 0);		/* 000000ff ZETA_ADDRESS_HIGH */
2932	xf_emit(ctx, 1, 0);		/* ffffffff ZETA_ADDRESS_LOW */
2933	xf_emit(ctx, 1, 4);		/* 00000007 ZETA_TILE_MODE */
2934	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2935	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2936	xf_emit(ctx, 1, 0x400);		/* 0fffffff ZETA_HORIZ */
2937	xf_emit(ctx, 1, 0x300);		/* 0000ffff ZETA_VERT */
2938	xf_emit(ctx, 1, 0x1001);	/* 00001fff ZETA_ARRAY_MODE */
2939	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2940	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2941	if (IS_NVA3F(dev_priv->chipset))
2942		xf_emit(ctx, 1, 0);	/* 00000001 */
2943	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2944	xf_emit(ctx, 1, 0x11);		/* 3f/7f RT_FORMAT */
2945	xf_emit(ctx, 7, 0);		/* 3f/7f RT_FORMAT */
2946	xf_emit(ctx, 1, 0x0fac6881);	/* 0fffffff RT_CONTROL */
2947	xf_emit(ctx, 1, 0xf);		/* 0000000f COLOR_MASK */
2948	xf_emit(ctx, 7, 0);		/* 0000000f COLOR_MASK */
2949	xf_emit(ctx, 1, 0);		/* ff/3ff */
2950	xf_emit(ctx, 8, 0);		/* 00000001 BLEND_ENABLE */
2951	xf_emit(ctx, 1, 0);		/* 00000003 UNK0F90 */
2952	xf_emit(ctx, 1, 0);		/* 00000001 FRAMEBUFFER_SRGB */
2953	xf_emit(ctx, 1, 0);		/* 7 */
2954	xf_emit(ctx, 1, 0);		/* 00000001 LOGIC_OP_ENABLE */
2955	if (IS_NVA3F(dev_priv->chipset)) {
2956		xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
2957		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2958	}
2959	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
2960	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
2961	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2962	if (dev_priv->chipset >= 0xa0)
2963		xf_emit(ctx, 1, 0x0fac6881);	/* fffffff */
2964	xf_emit(ctx, 1, magic2);	/* 001fffff tesla UNK0F78 */
2965	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_BOUNDS_EN */
2966	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
2967	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE_ENABLE */
2968	xf_emit(ctx, 1, 0x11);		/* 3f/7f DST_FORMAT */
2969	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK0FB0 */
2970	xf_emit(ctx, 1, 0);		/* ff/3ff */
2971	xf_emit(ctx, 1, 4);		/* 00000007 FP_CONTROL */
2972	xf_emit(ctx, 1, 0);		/* 00000001 STENCIL_FRONT_ENABLE */
2973	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK15B4 */
2974	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK19CC */
2975	xf_emit(ctx, 1, 0);		/* 00000007 */
2976	xf_emit(ctx, 1, 0);		/* 00000001 SAMPLECNT_ENABLE */
2977	xf_emit(ctx, 1, 0);		/* 0000000f ZETA_FORMAT */
2978	xf_emit(ctx, 1, 1);		/* 00000001 ZETA_ENABLE */
2979	if (IS_NVA3F(dev_priv->chipset)) {
2980		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
2981		xf_emit(ctx, 1, 0);	/* 0000000f tesla UNK15C8 */
2982	}
2983	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A3C */
2984	if (dev_priv->chipset >= 0xa0) {
2985		xf_emit(ctx, 3, 0);		/* 7/f, 1, ffff0ff3 */
2986		xf_emit(ctx, 1, 0xfac6881);	/* fffffff */
2987		xf_emit(ctx, 4, 0);		/* 1, 1, 1, 3ff */
2988		xf_emit(ctx, 1, 4);		/* 7 */
2989		xf_emit(ctx, 1, 0);		/* 1 */
2990		xf_emit(ctx, 2, 1);		/* 1 */
2991		xf_emit(ctx, 2, 0);		/* 7, f */
2992		xf_emit(ctx, 1, 1);		/* 1 */
2993		xf_emit(ctx, 1, 0);		/* 7/f */
2994		if (IS_NVA3F(dev_priv->chipset))
2995			xf_emit(ctx, 0x9, 0);	/* 1 */
2996		else
2997			xf_emit(ctx, 0x8, 0);	/* 1 */
2998		xf_emit(ctx, 1, 0);		/* ffff0ff3 */
2999		xf_emit(ctx, 8, 1);		/* 1 */
3000		xf_emit(ctx, 1, 0x11);		/* 7f */
3001		xf_emit(ctx, 7, 0);		/* 7f */
3002		xf_emit(ctx, 1, 0xfac6881);	/* fffffff */
3003		xf_emit(ctx, 1, 0xf);		/* f */
3004		xf_emit(ctx, 7, 0);		/* f */
3005		xf_emit(ctx, 1, 0x11);		/* 7f */
3006		xf_emit(ctx, 1, 1);		/* 1 */
3007		xf_emit(ctx, 5, 0);		/* 1, 7, 3ff, 3, 7 */
3008		if (IS_NVA3F(dev_priv->chipset)) {
3009			xf_emit(ctx, 1, 0);	/* 00000001 UNK1140 */
3010			xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
3011		}
3012	}
3013}
3014
3015static void
3016nv50_graph_construct_xfer_tex(struct nouveau_grctx *ctx)
3017{
3018	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
3019	xf_emit(ctx, 2, 0);		/* 1 LINKED_TSC. yes, 2. */
3020	if (dev_priv->chipset != 0x50)
3021		xf_emit(ctx, 1, 0);	/* 3 */
3022	xf_emit(ctx, 1, 1);		/* 1ffff BLIT_DU_DX_INT */
3023	xf_emit(ctx, 1, 0);		/* fffff BLIT_DU_DX_FRACT */
3024	xf_emit(ctx, 1, 1);		/* 1ffff BLIT_DV_DY_INT */
3025	xf_emit(ctx, 1, 0);		/* fffff BLIT_DV_DY_FRACT */
3026	if (dev_priv->chipset == 0x50)
3027		xf_emit(ctx, 1, 0);	/* 3 BLIT_CONTROL */
3028	else
3029		xf_emit(ctx, 2, 0);	/* 3ff, 1 */
3030	xf_emit(ctx, 1, 0x2a712488);	/* ffffffff SRC_TIC_0 */
3031	xf_emit(ctx, 1, 0);		/* ffffffff SRC_TIC_1 */
3032	xf_emit(ctx, 1, 0x4085c000);	/* ffffffff SRC_TIC_2 */
3033	xf_emit(ctx, 1, 0x40);		/* ffffffff SRC_TIC_3 */
3034	xf_emit(ctx, 1, 0x100);		/* ffffffff SRC_TIC_4 */
3035	xf_emit(ctx, 1, 0x10100);	/* ffffffff SRC_TIC_5 */
3036	xf_emit(ctx, 1, 0x02800000);	/* ffffffff SRC_TIC_6 */
3037	xf_emit(ctx, 1, 0);		/* ffffffff SRC_TIC_7 */
3038	if (dev_priv->chipset == 0x50) {
3039		xf_emit(ctx, 1, 0);	/* 00000001 turing UNK358 */
3040		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A34? */
3041		xf_emit(ctx, 1, 0);	/* 00000003 turing UNK37C tesla UNK1690 */
3042		xf_emit(ctx, 1, 0);	/* 00000003 BLIT_CONTROL */
3043		xf_emit(ctx, 1, 0);	/* 00000001 turing UNK32C tesla UNK0F94 */
3044	} else if (!IS_NVAAF(dev_priv->chipset)) {
3045		xf_emit(ctx, 1, 0);	/* ffffffff tesla UNK1A34? */
3046		xf_emit(ctx, 1, 0);	/* 00000003 */
3047		xf_emit(ctx, 1, 0);	/* 000003ff */
3048		xf_emit(ctx, 1, 0);	/* 00000003 */
3049		xf_emit(ctx, 1, 0);	/* 000003ff */
3050		xf_emit(ctx, 1, 0);	/* 00000003 tesla UNK1664 / turing UNK03E8 */
3051		xf_emit(ctx, 1, 0);	/* 00000003 */
3052		xf_emit(ctx, 1, 0);	/* 000003ff */
3053	} else {
3054		xf_emit(ctx, 0x6, 0);
3055	}
3056	xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A34 */
3057	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_TEXTURE */
3058	xf_emit(ctx, 1, 0);		/* 0000ffff DMA_SRC */
3059}
3060
3061static void
3062nv50_graph_construct_xfer_unk8cxx(struct nouveau_grctx *ctx)
3063{
3064	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
3065	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
3066	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
3067	xf_emit(ctx, 2, 0);		/* 7, ffff0ff3 */
3068	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
3069	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE */
3070	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0D64 */
3071	xf_emit(ctx, 1, 0x04e3bfdf);	/* ffffffff UNK0DF4 */
3072	xf_emit(ctx, 1, 1);		/* 00000001 UNK15B4 */
3073	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
3074	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
3075	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK0F98 */
3076	if (IS_NVA3F(dev_priv->chipset))
3077		xf_emit(ctx, 1, 1);	/* 0000001f tesla UNK169C */
3078	xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1668 */
3079	xf_emit(ctx, 1, 0);		/* 00000001 LINE_STIPPLE_ENABLE */
3080	xf_emit(ctx, 1, 0x00ffff00);	/* 00ffffff LINE_STIPPLE_PATTERN */
3081	xf_emit(ctx, 1, 0);		/* 00000001 POLYGON_SMOOTH_ENABLE */
3082	xf_emit(ctx, 1, 0);		/* 00000001 UNK1534 */
3083	xf_emit(ctx, 1, 0);		/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
3084	xf_emit(ctx, 1, 0);		/* 00000001 tesla UNK1658 */
3085	xf_emit(ctx, 1, 0);		/* 00000001 LINE_SMOOTH_ENABLE */
3086	xf_emit(ctx, 1, 0);		/* ffff0ff3 */
3087	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_TEST_ENABLE */
3088	xf_emit(ctx, 1, 0);		/* 00000001 DEPTH_WRITE */
3089	xf_emit(ctx, 1, 1);		/* 00000001 UNK15B4 */
3090	xf_emit(ctx, 1, 0);		/* 00000001 POINT_SPRITE_ENABLE */
3091	xf_emit(ctx, 1, 1);		/* 00000001 tesla UNK165C */
3092	xf_emit(ctx, 1, 0x30201000);	/* ffffffff tesla UNK1670 */
3093	xf_emit(ctx, 1, 0x70605040);	/* ffffffff tesla UNK1670 */
3094	xf_emit(ctx, 1, 0xb8a89888);	/* ffffffff tesla UNK1670 */
3095	xf_emit(ctx, 1, 0xf8e8d8c8);	/* ffffffff tesla UNK1670 */
3096	xf_emit(ctx, 1, 0);		/* 00000001 VERTEX_TWO_SIDE_ENABLE */
3097	xf_emit(ctx, 1, 0x1a);		/* 0000001f POLYGON_MODE */
3098}
3099
3100static void
3101nv50_graph_construct_xfer_tp(struct nouveau_grctx *ctx)
3102{
3103	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
3104	if (dev_priv->chipset < 0xa0) {
3105		nv50_graph_construct_xfer_unk84xx(ctx);
3106		nv50_graph_construct_xfer_tprop(ctx);
3107		nv50_graph_construct_xfer_tex(ctx);
3108		nv50_graph_construct_xfer_unk8cxx(ctx);
3109	} else {
3110		nv50_graph_construct_xfer_tex(ctx);
3111		nv50_graph_construct_xfer_tprop(ctx);
3112		nv50_graph_construct_xfer_unk8cxx(ctx);
3113		nv50_graph_construct_xfer_unk84xx(ctx);
3114	}
3115}
3116
3117static void
3118nv50_graph_construct_xfer_mpc(struct nouveau_grctx *ctx)
3119{
3120	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
3121	int i, mpcnt = 2;
3122	switch (dev_priv->chipset) {
3123		case 0x98:
3124		case 0xaa:
3125			mpcnt = 1;
3126			break;
3127		case 0x50:
3128		case 0x84:
3129		case 0x86:
3130		case 0x92:
3131		case 0x94:
3132		case 0x96:
3133		case 0xa8:
3134		case 0xac:
3135			mpcnt = 2;
3136			break;
3137		case 0xa0:
3138		case 0xa3:
3139		case 0xa5:
3140		case 0xaf:
3141			mpcnt = 3;
3142			break;
3143	}
3144	for (i = 0; i < mpcnt; i++) {
3145		xf_emit(ctx, 1, 0);		/* ff */
3146		xf_emit(ctx, 1, 0x80);		/* ffffffff tesla UNK1404 */
3147		xf_emit(ctx, 1, 0x80007004);	/* ffffffff tesla UNK12B0 */
3148		xf_emit(ctx, 1, 0x04000400);	/* ffffffff */
3149		if (dev_priv->chipset >= 0xa0)
3150			xf_emit(ctx, 1, 0xc0);	/* 00007fff tesla UNK152C */
3151		xf_emit(ctx, 1, 0x1000);	/* 0000ffff tesla UNK0D60 */
3152		xf_emit(ctx, 1, 0);		/* ff/3ff */
3153		xf_emit(ctx, 1, 0);		/* ffffffff tesla UNK1A30 */
3154		if (dev_priv->chipset == 0x86 || dev_priv->chipset == 0x98 || dev_priv->chipset == 0xa8 || IS_NVAAF(dev_priv->chipset)) {
3155			xf_emit(ctx, 1, 0xe00);		/* 7fff */
3156			xf_emit(ctx, 1, 0x1e00);	/* 7fff */
3157		}
3158		xf_emit(ctx, 1, 1);		/* 000000ff VP_REG_ALLOC_TEMP */
3159		xf_emit(ctx, 1, 0);		/* 00000001 LINKED_TSC */
3160		xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
3161		if (dev_priv->chipset == 0x50)
3162			xf_emit(ctx, 2, 0x1000);	/* 7fff tesla UNK141C */
3163		xf_emit(ctx, 1, 1);		/* 000000ff GP_REG_ALLOC_TEMP */
3164		xf_emit(ctx, 1, 0);		/* 00000001 GP_ENABLE */
3165		xf_emit(ctx, 1, 4);		/* 000000ff FP_REG_ALLOC_TEMP */
3166		xf_emit(ctx, 1, 2);		/* 00000003 REG_MODE */
3167		if (IS_NVAAF(dev_priv->chipset))
3168			xf_emit(ctx, 0xb, 0);	/* RO */
3169		else if (dev_priv->chipset >= 0xa0)
3170			xf_emit(ctx, 0xc, 0);	/* RO */
3171		else
3172			xf_emit(ctx, 0xa, 0);	/* RO */
3173	}
3174	xf_emit(ctx, 1, 0x08100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
3175	xf_emit(ctx, 1, 0);			/* ff/3ff */
3176	if (dev_priv->chipset >= 0xa0) {
3177		xf_emit(ctx, 1, 0x1fe21);	/* 0003ffff tesla UNK0FAC */
3178	}
3179	xf_emit(ctx, 3, 0);			/* 7fff, 0, 0 */
3180	xf_emit(ctx, 1, 0);			/* 00000001 tesla UNK1534 */
3181	xf_emit(ctx, 1, 0);			/* 7/f MULTISAMPLE_SAMPLES_LOG2 */
3182	xf_emit(ctx, 4, 0xffff);		/* 0000ffff MSAA_MASK */
3183	xf_emit(ctx, 1, 1);			/* 00000001 LANES32 */
3184	xf_emit(ctx, 1, 0x10001);		/* 00ffffff BLOCK_ALLOC */
3185	xf_emit(ctx, 1, 0x10001);		/* ffffffff BLOCKDIM_XY */
3186	xf_emit(ctx, 1, 1);			/* 0000ffff BLOCKDIM_Z */
3187	xf_emit(ctx, 1, 0);			/* ffffffff SHARED_SIZE */
3188	xf_emit(ctx, 1, 0x1fe21);		/* 1ffff/3ffff[NVA0+] tesla UNk0FAC */
3189	xf_emit(ctx, 1, 0);			/* ffffffff tesla UNK1A34 */
3190	if (IS_NVA3F(dev_priv->chipset))
3191		xf_emit(ctx, 1, 1);		/* 0000001f tesla UNK169C */
3192	xf_emit(ctx, 1, 0);			/* ff/3ff */
3193	xf_emit(ctx, 1, 0);			/* 1 LINKED_TSC */
3194	xf_emit(ctx, 1, 0);			/* ff FP_ADDRESS_HIGH */
3195	xf_emit(ctx, 1, 0);			/* ffffffff FP_ADDRESS_LOW */
3196	xf_emit(ctx, 1, 0x08100c12);		/* 1fffffff FP_INTERPOLANT_CTRL */
3197	xf_emit(ctx, 1, 4);			/* 00000007 FP_CONTROL */
3198	xf_emit(ctx, 1, 0);			/* 000000ff FRAG_COLOR_CLAMP_EN */
3199	xf_emit(ctx, 1, 2);			/* 00000003 REG_MODE */
3200	xf_emit(ctx, 1, 0x11);			/* 0000007f RT_FORMAT */
3201	xf_emit(ctx, 7, 0);			/* 0000007f RT_FORMAT */
3202	xf_emit(ctx, 1, 0);			/* 00000007 */
3203	xf_emit(ctx, 1, 0xfac6881);		/* 0fffffff RT_CONTROL */
3204	xf_emit(ctx, 1, 0);			/* 00000003 MULTISAMPLE_CTRL */
3205	if (IS_NVA3F(dev_priv->chipset))
3206		xf_emit(ctx, 1, 3);		/* 00000003 tesla UNK16B4 */
3207	xf_emit(ctx, 1, 0);			/* 00000001 ALPHA_TEST_ENABLE */
3208	xf_emit(ctx, 1, 0);			/* 00000007 ALPHA_TEST_FUNC */
3209	xf_emit(ctx, 1, 0);			/* 00000001 FRAMEBUFFER_SRGB */
3210	xf_emit(ctx, 1, 4);			/* ffffffff tesla UNK1400 */
3211	xf_emit(ctx, 8, 0);			/* 00000001 BLEND_ENABLE */
3212	xf_emit(ctx, 1, 0);			/* 00000001 LOGIC_OP_ENABLE */
3213	xf_emit(ctx, 1, 2);			/* 0000001f BLEND_FUNC_SRC_RGB */
3214	xf_emit(ctx, 1, 1);			/* 0000001f BLEND_FUNC_DST_RGB */
3215	xf_emit(ctx, 1, 1);			/* 00000007 BLEND_EQUATION_RGB */
3216	xf_emit(ctx, 1, 2);			/* 0000001f BLEND_FUNC_SRC_ALPHA */
3217	xf_emit(ctx, 1, 1);			/* 0000001f BLEND_FUNC_DST_ALPHA */
3218	xf_emit(ctx, 1, 1);			/* 00000007 BLEND_EQUATION_ALPHA */
3219	xf_emit(ctx, 1, 1);			/* 00000001 UNK133C */
3220	if (IS_NVA3F(dev_priv->chipset)) {
3221		xf_emit(ctx, 1, 0);		/* 00000001 UNK12E4 */
3222		xf_emit(ctx, 8, 2);		/* 0000001f IBLEND_FUNC_SRC_RGB */
3223		xf_emit(ctx, 8, 1);		/* 0000001f IBLEND_FUNC_DST_RGB */
3224		xf_emit(ctx, 8, 1);		/* 00000007 IBLEND_EQUATION_RGB */
3225		xf_emit(ctx, 8, 2);		/* 0000001f IBLEND_FUNC_SRC_ALPHA */
3226		xf_emit(ctx, 8, 1);		/* 0000001f IBLEND_FUNC_DST_ALPHA */
3227		xf_emit(ctx, 8, 1);		/* 00000007 IBLEND_EQUATION_ALPHA */
3228		xf_emit(ctx, 8, 1);		/* 00000001 IBLEND_UNK00 */
3229		xf_emit(ctx, 1, 0);		/* 00000003 tesla UNK1928 */
3230		xf_emit(ctx, 1, 0);		/* 00000001 UNK1140 */
3231	}
3232	xf_emit(ctx, 1, 0);			/* 00000003 tesla UNK0F90 */
3233	xf_emit(ctx, 1, 4);			/* 000000ff FP_RESULT_COUNT */
3234	/* XXX: demagic this part some day */
3235	if (dev_priv->chipset == 0x50)
3236		xf_emit(ctx, 0x3a0, 0);
3237	else if (dev_priv->chipset < 0x94)
3238		xf_emit(ctx, 0x3a2, 0);
3239	else if (dev_priv->chipset == 0x98 || dev_priv->chipset == 0xaa)
3240		xf_emit(ctx, 0x39f, 0);
3241	else
3242		xf_emit(ctx, 0x3a3, 0);
3243	xf_emit(ctx, 1, 0x11);			/* 3f/7f DST_FORMAT */
3244	xf_emit(ctx, 1, 0);			/* 7 OPERATION */
3245	xf_emit(ctx, 1, 1);			/* 1 DST_LINEAR */
3246	xf_emit(ctx, 0x2d, 0);
3247}
3248
3249static void
3250nv50_graph_construct_xfer2(struct nouveau_grctx *ctx)
3251{
3252	struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
3253	int i;
3254	uint32_t offset;
3255	uint32_t units = nv_rd32 (ctx->dev, 0x1540);
3256	int size = 0;
3257
3258	offset = (ctx->ctxvals_pos+0x3f)&~0x3f;
3259
3260	if (dev_priv->chipset < 0xa0) {
3261		for (i = 0; i < 8; i++) {
3262			ctx->ctxvals_pos = offset + i;
3263			/* that little bugger belongs to csched. No idea
3264			 * what it's doing here. */
3265			if (i == 0)
3266				xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */
3267			if (units & (1 << i))
3268				nv50_graph_construct_xfer_mpc(ctx);
3269			if ((ctx->ctxvals_pos-offset)/8 > size)
3270				size = (ctx->ctxvals_pos-offset)/8;
3271		}
3272	} else {
3273		/* Strand 0: TPs 0, 1 */
3274		ctx->ctxvals_pos = offset;
3275		/* that little bugger belongs to csched. No idea
3276		 * what it's doing here. */
3277		xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */
3278		if (units & (1 << 0))
3279			nv50_graph_construct_xfer_mpc(ctx);
3280		if (units & (1 << 1))
3281			nv50_graph_construct_xfer_mpc(ctx);
3282		if ((ctx->ctxvals_pos-offset)/8 > size)
3283			size = (ctx->ctxvals_pos-offset)/8;
3284
3285		/* Strand 1: TPs 2, 3 */
3286		ctx->ctxvals_pos = offset + 1;
3287		if (units & (1 << 2))
3288			nv50_graph_construct_xfer_mpc(ctx);
3289		if (units & (1 << 3))
3290			nv50_graph_construct_xfer_mpc(ctx);
3291		if ((ctx->ctxvals_pos-offset)/8 > size)
3292			size = (ctx->ctxvals_pos-offset)/8;
3293
3294		/* Strand 2: TPs 4, 5, 6 */
3295		ctx->ctxvals_pos = offset + 2;
3296		if (units & (1 << 4))
3297			nv50_graph_construct_xfer_mpc(ctx);
3298		if (units & (1 << 5))
3299			nv50_graph_construct_xfer_mpc(ctx);
3300		if (units & (1 << 6))
3301			nv50_graph_construct_xfer_mpc(ctx);
3302		if ((ctx->ctxvals_pos-offset)/8 > size)
3303			size = (ctx->ctxvals_pos-offset)/8;
3304
3305		/* Strand 3: TPs 7, 8, 9 */
3306		ctx->ctxvals_pos = offset + 3;
3307		if (units & (1 << 7))
3308			nv50_graph_construct_xfer_mpc(ctx);
3309		if (units & (1 << 8))
3310			nv50_graph_construct_xfer_mpc(ctx);
3311		if (units & (1 << 9))
3312			nv50_graph_construct_xfer_mpc(ctx);
3313		if ((ctx->ctxvals_pos-offset)/8 > size)
3314			size = (ctx->ctxvals_pos-offset)/8;
3315	}
3316	ctx->ctxvals_pos = offset + size * 8;
3317	ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f;
3318	cp_lsr (ctx, offset);
3319	cp_out (ctx, CP_SET_XFER_POINTER);
3320	cp_lsr (ctx, size);
3321	cp_out (ctx, CP_SEEK_2);
3322	cp_out (ctx, CP_XFER_2);
3323	cp_wait(ctx, XFER, BUSY);
3324}
3325