2 * Copyright 2008 Ben Skeggs
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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 AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
18 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
19 * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
23 #include "util/u_format.h"
25 #include "nv50_context.h"
26 #include "nv50_resource.h"
27 #include "nouveau/nouveau_stateobj.h"
29 static struct nouveau_stateobj
*
30 validate_fb(struct nv50_context
*nv50
)
32 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
33 struct nouveau_stateobj
*so
= so_new(32, 79, 18);
34 struct pipe_framebuffer_state
*fb
= &nv50
->framebuffer
;
35 unsigned i
, w
= 0, h
= 0, gw
= 0;
37 /* Set nr of active RTs and select RT for each colour output.
38 * FP result 0 always goes to RT[0], bits 4 - 6 are ignored.
39 * Ambiguous assignment results in no rendering (no DATA_ERROR).
41 so_method(so
, tesla
, NV50TCL_RT_CONTROL
, 1);
42 so_data (so
, fb
->nr_cbufs
|
43 (0 << 4) | (1 << 7) | (2 << 10) | (3 << 13) |
44 (4 << 16) | (5 << 19) | (6 << 22) | (7 << 25));
46 for (i
= 0; i
< fb
->nr_cbufs
; i
++) {
47 struct pipe_resource
*pt
= fb
->cbufs
[i
]->texture
;
48 struct nouveau_bo
*bo
= nv50_miptree(pt
)->base
.bo
;
51 w
= fb
->cbufs
[i
]->width
;
52 h
= fb
->cbufs
[i
]->height
;
55 assert(w
== fb
->cbufs
[i
]->width
);
56 assert(h
== fb
->cbufs
[i
]->height
);
59 so_method(so
, tesla
, NV50TCL_RT_HORIZ(i
), 2);
60 so_data (so
, fb
->cbufs
[i
]->width
);
61 so_data (so
, fb
->cbufs
[i
]->height
);
63 so_method(so
, tesla
, NV50TCL_RT_ADDRESS_HIGH(i
), 5);
64 so_reloc (so
, bo
, fb
->cbufs
[i
]->offset
, NOUVEAU_BO_VRAM
|
65 NOUVEAU_BO_HIGH
| NOUVEAU_BO_RDWR
, 0, 0);
66 so_reloc (so
, bo
, fb
->cbufs
[i
]->offset
, NOUVEAU_BO_VRAM
|
67 NOUVEAU_BO_LOW
| NOUVEAU_BO_RDWR
, 0, 0);
68 switch (fb
->cbufs
[i
]->format
) {
69 case PIPE_FORMAT_B8G8R8A8_UNORM
:
70 so_data(so
, NV50TCL_RT_FORMAT_A8R8G8B8_UNORM
);
72 case PIPE_FORMAT_B8G8R8X8_UNORM
:
73 so_data(so
, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM
);
75 case PIPE_FORMAT_B5G6R5_UNORM
:
76 so_data(so
, NV50TCL_RT_FORMAT_R5G6B5_UNORM
);
78 case PIPE_FORMAT_R16G16B16A16_SNORM
:
79 so_data(so
, NV50TCL_RT_FORMAT_R16G16B16A16_SNORM
);
81 case PIPE_FORMAT_R16G16B16A16_UNORM
:
82 so_data(so
, NV50TCL_RT_FORMAT_R16G16B16A16_UNORM
);
84 case PIPE_FORMAT_R32G32B32A32_FLOAT
:
85 so_data(so
, NV50TCL_RT_FORMAT_R32G32B32A32_FLOAT
);
87 case PIPE_FORMAT_R16G16_SNORM
:
88 so_data(so
, NV50TCL_RT_FORMAT_R16G16_SNORM
);
90 case PIPE_FORMAT_R16G16_UNORM
:
91 so_data(so
, NV50TCL_RT_FORMAT_R16G16_UNORM
);
94 NOUVEAU_ERR("AIIII unknown format %s\n",
95 util_format_name(fb
->cbufs
[i
]->format
));
96 so_data(so
, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM
);
99 so_data(so
, nv50_miptree(pt
)->
100 level
[fb
->cbufs
[i
]->level
].tile_mode
<< 4);
101 so_data(so
, 0x00000000);
103 so_method(so
, tesla
, NV50TCL_RT_ARRAY_MODE
, 1);
108 struct pipe_resource
*pt
= fb
->zsbuf
->texture
;
109 struct nouveau_bo
*bo
= nv50_miptree(pt
)->base
.bo
;
112 w
= fb
->zsbuf
->width
;
113 h
= fb
->zsbuf
->height
;
116 assert(w
== fb
->zsbuf
->width
);
117 assert(h
== fb
->zsbuf
->height
);
120 so_method(so
, tesla
, NV50TCL_ZETA_ADDRESS_HIGH
, 5);
121 so_reloc (so
, bo
, fb
->zsbuf
->offset
, NOUVEAU_BO_VRAM
|
122 NOUVEAU_BO_HIGH
| NOUVEAU_BO_RDWR
, 0, 0);
123 so_reloc (so
, bo
, fb
->zsbuf
->offset
, NOUVEAU_BO_VRAM
|
124 NOUVEAU_BO_LOW
| NOUVEAU_BO_RDWR
, 0, 0);
125 switch (fb
->zsbuf
->format
) {
126 case PIPE_FORMAT_Z24_UNORM_S8_USCALED
:
127 so_data(so
, NV50TCL_ZETA_FORMAT_S8Z24_UNORM
);
129 case PIPE_FORMAT_Z24X8_UNORM
:
130 so_data(so
, NV50TCL_ZETA_FORMAT_X8Z24_UNORM
);
132 case PIPE_FORMAT_S8_USCALED_Z24_UNORM
:
133 so_data(so
, NV50TCL_ZETA_FORMAT_Z24S8_UNORM
);
135 case PIPE_FORMAT_Z32_FLOAT
:
136 so_data(so
, NV50TCL_ZETA_FORMAT_Z32_FLOAT
);
139 NOUVEAU_ERR("AIIII unknown format %s\n",
140 util_format_name(fb
->zsbuf
->format
));
141 so_data(so
, NV50TCL_ZETA_FORMAT_S8Z24_UNORM
);
144 so_data(so
, nv50_miptree(pt
)->
145 level
[fb
->zsbuf
->level
].tile_mode
<< 4);
146 so_data(so
, 0x00000000);
148 so_method(so
, tesla
, NV50TCL_ZETA_ENABLE
, 1);
150 so_method(so
, tesla
, NV50TCL_ZETA_HORIZ
, 3);
151 so_data (so
, fb
->zsbuf
->width
);
152 so_data (so
, fb
->zsbuf
->height
);
153 so_data (so
, 0x00010001);
155 so_method(so
, tesla
, NV50TCL_ZETA_ENABLE
, 1);
159 so_method(so
, tesla
, NV50TCL_VIEWPORT_HORIZ(0), 2);
160 so_data (so
, w
<< 16);
161 so_data (so
, h
<< 16);
162 /* set window lower left corner */
163 so_method(so
, tesla
, NV50TCL_WINDOW_OFFSET_X
, 2);
166 /* set screen scissor rectangle */
167 so_method(so
, tesla
, NV50TCL_SCREEN_SCISSOR_HORIZ
, 2);
168 so_data (so
, w
<< 16);
169 so_data (so
, h
<< 16);
175 nv50_validate_samplers(struct nv50_context
*nv50
, struct nouveau_stateobj
*so
,
178 struct nouveau_grobj
*eng2d
= nv50
->screen
->eng2d
;
179 unsigned i
, j
, dw
= nv50
->sampler_nr
[p
] * 8;
183 nv50_so_init_sifc(nv50
, so
, nv50
->screen
->tsc
, NOUVEAU_BO_VRAM
,
184 p
* (32 * 8 * 4), dw
* 4);
186 so_method(so
, eng2d
, NV50_2D_SIFC_DATA
| (2 << 29), dw
);
188 for (i
= 0; i
< nv50
->sampler_nr
[p
]; ++i
) {
189 if (nv50
->sampler
[p
][i
])
190 so_datap(so
, nv50
->sampler
[p
][i
]->tsc
, 8);
192 for (j
= 0; j
< 8; ++j
) /* you get punished */
193 so_data(so
, 0); /* ... for leaving holes */
198 static struct nouveau_stateobj
*
199 validate_blend(struct nv50_context
*nv50
)
201 struct nouveau_stateobj
*so
= NULL
;
202 so_ref(nv50
->blend
->so
, &so
);
206 static struct nouveau_stateobj
*
207 validate_zsa(struct nv50_context
*nv50
)
209 struct nouveau_stateobj
*so
= NULL
;
210 so_ref(nv50
->zsa
->so
, &so
);
214 static struct nouveau_stateobj
*
215 validate_rast(struct nv50_context
*nv50
)
217 struct nouveau_stateobj
*so
= NULL
;
218 so_ref(nv50
->rasterizer
->so
, &so
);
222 static struct nouveau_stateobj
*
223 validate_blend_colour(struct nv50_context
*nv50
)
225 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
226 struct nouveau_stateobj
*so
= so_new(1, 4, 0);
228 so_method(so
, tesla
, NV50TCL_BLEND_COLOR(0), 4);
229 so_data (so
, fui(nv50
->blend_colour
.color
[0]));
230 so_data (so
, fui(nv50
->blend_colour
.color
[1]));
231 so_data (so
, fui(nv50
->blend_colour
.color
[2]));
232 so_data (so
, fui(nv50
->blend_colour
.color
[3]));
236 static struct nouveau_stateobj
*
237 validate_stencil_ref(struct nv50_context
*nv50
)
239 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
240 struct nouveau_stateobj
*so
= so_new(2, 2, 0);
242 so_method(so
, tesla
, NV50TCL_STENCIL_FRONT_FUNC_REF
, 1);
243 so_data (so
, nv50
->stencil_ref
.ref_value
[0]);
244 so_method(so
, tesla
, NV50TCL_STENCIL_BACK_FUNC_REF
, 1);
245 so_data (so
, nv50
->stencil_ref
.ref_value
[1]);
249 static struct nouveau_stateobj
*
250 validate_stipple(struct nv50_context
*nv50
)
252 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
253 struct nouveau_stateobj
*so
= so_new(1, 32, 0);
256 so_method(so
, tesla
, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32);
257 for (i
= 0; i
< 32; i
++)
258 so_data(so
, util_bswap32(nv50
->stipple
.stipple
[i
]));
262 static struct nouveau_stateobj
*
263 validate_scissor(struct nv50_context
*nv50
)
265 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
266 struct pipe_scissor_state
*s
= &nv50
->scissor
;
267 struct nouveau_stateobj
*so
;
269 so
= so_new(1, 2, 0);
270 so_method(so
, tesla
, NV50TCL_SCISSOR_HORIZ(0), 2);
271 so_data (so
, (s
->maxx
<< 16) | s
->minx
);
272 so_data (so
, (s
->maxy
<< 16) | s
->miny
);
276 static struct nouveau_stateobj
*
277 validate_viewport(struct nv50_context
*nv50
)
279 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
280 struct nouveau_stateobj
*so
= so_new(5, 9, 0);
282 so_method(so
, tesla
, NV50TCL_VIEWPORT_TRANSLATE_X(0), 3);
283 so_data (so
, fui(nv50
->viewport
.translate
[0]));
284 so_data (so
, fui(nv50
->viewport
.translate
[1]));
285 so_data (so
, fui(nv50
->viewport
.translate
[2]));
286 so_method(so
, tesla
, NV50TCL_VIEWPORT_SCALE_X(0), 3);
287 so_data (so
, fui(nv50
->viewport
.scale
[0]));
288 so_data (so
, fui(nv50
->viewport
.scale
[1]));
289 so_data (so
, fui(nv50
->viewport
.scale
[2]));
291 so_method(so
, tesla
, NV50TCL_VIEWPORT_TRANSFORM_EN
, 1);
293 /* 0x0000 = remove whole primitive only (xyz)
294 * 0x1018 = remove whole primitive only (xy), clamp z
295 * 0x1080 = clip primitive (xyz)
296 * 0x1098 = clip primitive (xy), clamp z
298 so_method(so
, tesla
, NV50TCL_VIEW_VOLUME_CLIP_CTRL
, 1);
299 so_data (so
, 0x1080);
300 /* no idea what 0f90 does */
301 so_method(so
, tesla
, 0x0f90, 1);
307 static struct nouveau_stateobj
*
308 validate_sampler(struct nv50_context
*nv50
)
310 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
311 struct nouveau_stateobj
*so
;
314 for (i
= 0; i
< 3; ++i
)
315 nr
+= nv50
->sampler_nr
[i
];
317 so
= so_new(1 + 5 * 3, 1 + 19 * 3 + nr
* 8, 3 * 2);
319 nv50_validate_samplers(nv50
, so
, 0); /* VP */
320 nv50_validate_samplers(nv50
, so
, 2); /* FP */
322 so_method(so
, tesla
, 0x1334, 1); /* flush TSC */
328 static struct nouveau_stateobj
*
329 validate_vtxbuf(struct nv50_context
*nv50
)
331 struct nouveau_stateobj
*so
= NULL
;
332 so_ref(nv50
->state
.vtxbuf
, &so
);
336 static struct nouveau_stateobj
*
337 validate_vtxattr(struct nv50_context
*nv50
)
339 struct nouveau_stateobj
*so
= NULL
;
340 so_ref(nv50
->state
.vtxattr
, &so
);
344 struct state_validate
{
345 struct nouveau_stateobj
*(*func
)(struct nv50_context
*nv50
);
347 } validate_list
[] = {
348 { validate_fb
, NV50_NEW_FRAMEBUFFER
},
349 { validate_blend
, NV50_NEW_BLEND
},
350 { validate_zsa
, NV50_NEW_ZSA
},
351 { nv50_vertprog_validate
, NV50_NEW_VERTPROG
| NV50_NEW_VERTPROG_CB
},
352 { nv50_fragprog_validate
, NV50_NEW_FRAGPROG
| NV50_NEW_FRAGPROG_CB
},
353 { nv50_geomprog_validate
, NV50_NEW_GEOMPROG
| NV50_NEW_GEOMPROG_CB
},
354 { nv50_fp_linkage_validate
, NV50_NEW_VERTPROG
| NV50_NEW_GEOMPROG
|
355 NV50_NEW_FRAGPROG
| NV50_NEW_RASTERIZER
},
356 { nv50_gp_linkage_validate
, NV50_NEW_VERTPROG
| NV50_NEW_GEOMPROG
},
357 { validate_rast
, NV50_NEW_RASTERIZER
},
358 { validate_blend_colour
, NV50_NEW_BLEND_COLOUR
},
359 { validate_stencil_ref
, NV50_NEW_STENCIL_REF
},
360 { validate_stipple
, NV50_NEW_STIPPLE
},
361 { validate_scissor
, NV50_NEW_SCISSOR
},
362 { validate_viewport
, NV50_NEW_VIEWPORT
},
363 { validate_sampler
, NV50_NEW_SAMPLER
},
364 { nv50_tex_validate
, NV50_NEW_TEXTURE
| NV50_NEW_SAMPLER
},
365 { nv50_vbo_validate
, NV50_NEW_ARRAYS
},
366 { validate_vtxbuf
, NV50_NEW_ARRAYS
},
367 { validate_vtxattr
, NV50_NEW_ARRAYS
},
370 #define validate_list_len (sizeof(validate_list) / sizeof(validate_list[0]))
373 nv50_state_validate(struct nv50_context
*nv50
, unsigned wait_dwords
)
375 struct nouveau_channel
*chan
= nv50
->screen
->base
.channel
;
376 struct nouveau_grobj
*tesla
= nv50
->screen
->tesla
;
377 unsigned nr_relocs
= 128, nr_dwords
= wait_dwords
+ 128 + 4;
380 for (i
= 0; i
< validate_list_len
; i
++) {
381 struct state_validate
*validate
= &validate_list
[i
];
382 struct nouveau_stateobj
*so
;
384 if (!(nv50
->dirty
& validate
->states
))
387 so
= validate
->func(nv50
);
391 nr_dwords
+= (so
->total
+ so
->cur
);
392 nr_relocs
+= so
->cur_reloc
;
394 so_ref(so
, &nv50
->state
.hw
[i
]);
396 nv50
->state
.hw_dirty
|= (1 << i
);
400 if (nv50
->screen
->cur_ctx
!= nv50
) {
401 for (i
= 0; i
< validate_list_len
; i
++) {
402 if (!nv50
->state
.hw
[i
] ||
403 (nv50
->state
.hw_dirty
& (1 << i
)))
406 nr_dwords
+= (nv50
->state
.hw
[i
]->total
+
407 nv50
->state
.hw
[i
]->cur
);
408 nr_relocs
+= nv50
->state
.hw
[i
]->cur_reloc
;
409 nv50
->state
.hw_dirty
|= (1 << i
);
412 nv50
->screen
->cur_ctx
= nv50
;
415 ret
= MARK_RING(chan
, nr_dwords
, nr_relocs
);
417 debug_printf("MARK_RING(%d, %d) failed: %d\n",
418 nr_dwords
, nr_relocs
, ret
);
422 while (nv50
->state
.hw_dirty
) {
423 i
= ffs(nv50
->state
.hw_dirty
) - 1;
424 nv50
->state
.hw_dirty
&= ~(1 << i
);
426 so_emit(chan
, nv50
->state
.hw
[i
]);
429 /* Yes, really, we need to do this. If a buffer that is referenced
430 * on the hardware isn't part of changed state above, without doing
431 * this the kernel is given no clue that the buffer is being used
432 * still. This can cause all sorts of fun issues.
434 nv50_tex_relocs(nv50
);
435 so_emit_reloc_markers(chan
, nv50
->state
.hw
[0]); /* fb */
436 so_emit_reloc_markers(chan
, nv50
->state
.hw
[3]); /* vp */
437 so_emit_reloc_markers(chan
, nv50
->state
.hw
[4]); /* fp */
438 so_emit_reloc_markers(chan
, nv50
->state
.hw
[17]); /* vb */
439 nv50_screen_relocs(nv50
->screen
);
442 BEGIN_RING(chan
, tesla
, 0x142c, 1);
444 BEGIN_RING(chan
, tesla
, 0x142c, 1);
449 void nv50_so_init_sifc(struct nv50_context
*nv50
,
450 struct nouveau_stateobj
*so
,
451 struct nouveau_bo
*bo
, unsigned reloc
,
452 unsigned offset
, unsigned size
)
454 struct nouveau_grobj
*eng2d
= nv50
->screen
->eng2d
;
456 reloc
|= NOUVEAU_BO_WR
;
458 so_method(so
, eng2d
, NV50_2D_DST_FORMAT
, 2);
459 so_data (so
, NV50_2D_DST_FORMAT_R8_UNORM
);
461 so_method(so
, eng2d
, NV50_2D_DST_PITCH
, 5);
462 so_data (so
, 262144);
465 so_reloc (so
, bo
, offset
, reloc
| NOUVEAU_BO_HIGH
, 0, 0);
466 so_reloc (so
, bo
, offset
, reloc
| NOUVEAU_BO_LOW
, 0, 0);
467 so_method(so
, eng2d
, NV50_2D_SIFC_BITMAP_ENABLE
, 2);
469 so_data (so
, NV50_2D_SIFC_FORMAT_R8_UNORM
);
470 so_method(so
, eng2d
, NV50_2D_SIFC_WIDTH
, 10);