2 Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved.
4 The Weather Channel (TM) funded Tungsten Graphics to develop the
5 initial release of the Radeon 8500 driver under the XFree86 license.
6 This notice must be preserved.
8 Permission is hereby granted, free of charge, to any person obtaining
9 a copy of this software and associated documentation files (the
10 "Software"), to deal in the Software without restriction, including
11 without limitation the rights to use, copy, modify, merge, publish,
12 distribute, sublicense, and/or sell copies of the Software, and to
13 permit persons to whom the Software is furnished to do so, subject to
14 the following conditions:
16 The above copyright notice and this permission notice (including the
17 next paragraph) shall be included in all copies or substantial
18 portions of the Software.
20 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
21 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
22 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
23 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
24 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
25 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
26 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28 **************************************************************************/
32 * Keith Whitwell <keith@tungstengraphics.com>
35 #include "main/glheader.h"
36 #include "main/imports.h"
37 #include "main/context.h"
38 #include "main/macros.h"
39 #include "main/texformat.h"
40 #include "main/texobj.h"
41 #include "main/enums.h"
43 #include "radeon_common.h"
44 #include "radeon_mipmap_tree.h"
45 #include "r200_context.h"
46 #include "r200_state.h"
47 #include "r200_ioctl.h"
48 #include "r200_swtcl.h"
53 #define R200_TXFORMAT_A8 R200_TXFORMAT_I8
54 #define R200_TXFORMAT_L8 R200_TXFORMAT_I8
55 #define R200_TXFORMAT_AL88 R200_TXFORMAT_AI88
56 #define R200_TXFORMAT_YCBCR R200_TXFORMAT_YVYU422
57 #define R200_TXFORMAT_YCBCR_REV R200_TXFORMAT_VYUY422
58 #define R200_TXFORMAT_RGB_DXT1 R200_TXFORMAT_DXT1
59 #define R200_TXFORMAT_RGBA_DXT1 R200_TXFORMAT_DXT1
60 #define R200_TXFORMAT_RGBA_DXT3 R200_TXFORMAT_DXT23
61 #define R200_TXFORMAT_RGBA_DXT5 R200_TXFORMAT_DXT45
64 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f, 0 }
65 #define _COLOR_REV(f) \
66 [ MESA_FORMAT_ ## f ## _REV ] = { R200_TXFORMAT_ ## f, 0 }
68 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f | R200_TXFORMAT_ALPHA_IN_MAP, 0 }
69 #define _ALPHA_REV(f) \
70 [ MESA_FORMAT_ ## f ## _REV ] = { R200_TXFORMAT_ ## f | R200_TXFORMAT_ALPHA_IN_MAP, 0 }
72 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f, R200_YUV_TO_RGB }
74 [ MESA_FORMAT_ ## f ] = { 0xffffffff, 0 }
75 #define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5) \
76 && (tx_table_be[f].format != 0xffffffff) )
79 GLuint format
, filter
;
82 static const struct tx_table tx_table_be
[] =
84 [ MESA_FORMAT_RGBA8888
] = { R200_TXFORMAT_ABGR8888
| R200_TXFORMAT_ALPHA_IN_MAP
, 0 },
111 static const struct tx_table tx_table_le
[] =
114 [ MESA_FORMAT_RGBA8888_REV
] = { R200_TXFORMAT_ABGR8888
| R200_TXFORMAT_ALPHA_IN_MAP
, 0 },
116 _ALPHA_REV(ARGB8888
),
117 [ MESA_FORMAT_RGB888
] = { R200_TXFORMAT_ARGB8888
, 0 },
121 _ALPHA_REV(ARGB4444
),
123 _ALPHA_REV(ARGB1555
),
144 /* ================================================================
145 * Texture combine functions
148 /* GL_ARB_texture_env_combine support
151 /* The color tables have combine functions for GL_SRC_COLOR,
152 * GL_ONE_MINUS_SRC_COLOR, GL_SRC_ALPHA and GL_ONE_MINUS_SRC_ALPHA.
154 static GLuint r200_register_color
[][R200_MAX_TEXTURE_UNITS
] =
157 R200_TXC_ARG_A_R0_COLOR
,
158 R200_TXC_ARG_A_R1_COLOR
,
159 R200_TXC_ARG_A_R2_COLOR
,
160 R200_TXC_ARG_A_R3_COLOR
,
161 R200_TXC_ARG_A_R4_COLOR
,
162 R200_TXC_ARG_A_R5_COLOR
165 R200_TXC_ARG_A_R0_COLOR
| R200_TXC_COMP_ARG_A
,
166 R200_TXC_ARG_A_R1_COLOR
| R200_TXC_COMP_ARG_A
,
167 R200_TXC_ARG_A_R2_COLOR
| R200_TXC_COMP_ARG_A
,
168 R200_TXC_ARG_A_R3_COLOR
| R200_TXC_COMP_ARG_A
,
169 R200_TXC_ARG_A_R4_COLOR
| R200_TXC_COMP_ARG_A
,
170 R200_TXC_ARG_A_R5_COLOR
| R200_TXC_COMP_ARG_A
173 R200_TXC_ARG_A_R0_ALPHA
,
174 R200_TXC_ARG_A_R1_ALPHA
,
175 R200_TXC_ARG_A_R2_ALPHA
,
176 R200_TXC_ARG_A_R3_ALPHA
,
177 R200_TXC_ARG_A_R4_ALPHA
,
178 R200_TXC_ARG_A_R5_ALPHA
181 R200_TXC_ARG_A_R0_ALPHA
| R200_TXC_COMP_ARG_A
,
182 R200_TXC_ARG_A_R1_ALPHA
| R200_TXC_COMP_ARG_A
,
183 R200_TXC_ARG_A_R2_ALPHA
| R200_TXC_COMP_ARG_A
,
184 R200_TXC_ARG_A_R3_ALPHA
| R200_TXC_COMP_ARG_A
,
185 R200_TXC_ARG_A_R4_ALPHA
| R200_TXC_COMP_ARG_A
,
186 R200_TXC_ARG_A_R5_ALPHA
| R200_TXC_COMP_ARG_A
190 static GLuint r200_tfactor_color
[] =
192 R200_TXC_ARG_A_TFACTOR_COLOR
,
193 R200_TXC_ARG_A_TFACTOR_COLOR
| R200_TXC_COMP_ARG_A
,
194 R200_TXC_ARG_A_TFACTOR_ALPHA
,
195 R200_TXC_ARG_A_TFACTOR_ALPHA
| R200_TXC_COMP_ARG_A
198 static GLuint r200_tfactor1_color
[] =
200 R200_TXC_ARG_A_TFACTOR1_COLOR
,
201 R200_TXC_ARG_A_TFACTOR1_COLOR
| R200_TXC_COMP_ARG_A
,
202 R200_TXC_ARG_A_TFACTOR1_ALPHA
,
203 R200_TXC_ARG_A_TFACTOR1_ALPHA
| R200_TXC_COMP_ARG_A
206 static GLuint r200_primary_color
[] =
208 R200_TXC_ARG_A_DIFFUSE_COLOR
,
209 R200_TXC_ARG_A_DIFFUSE_COLOR
| R200_TXC_COMP_ARG_A
,
210 R200_TXC_ARG_A_DIFFUSE_ALPHA
,
211 R200_TXC_ARG_A_DIFFUSE_ALPHA
| R200_TXC_COMP_ARG_A
214 /* GL_ZERO table - indices 0-3
215 * GL_ONE table - indices 1-4
217 static GLuint r200_zero_color
[] =
220 R200_TXC_ARG_A_ZERO
| R200_TXC_COMP_ARG_A
,
222 R200_TXC_ARG_A_ZERO
| R200_TXC_COMP_ARG_A
,
226 /* The alpha tables only have GL_SRC_ALPHA and GL_ONE_MINUS_SRC_ALPHA.
228 static GLuint r200_register_alpha
[][R200_MAX_TEXTURE_UNITS
] =
231 R200_TXA_ARG_A_R0_ALPHA
,
232 R200_TXA_ARG_A_R1_ALPHA
,
233 R200_TXA_ARG_A_R2_ALPHA
,
234 R200_TXA_ARG_A_R3_ALPHA
,
235 R200_TXA_ARG_A_R4_ALPHA
,
236 R200_TXA_ARG_A_R5_ALPHA
239 R200_TXA_ARG_A_R0_ALPHA
| R200_TXA_COMP_ARG_A
,
240 R200_TXA_ARG_A_R1_ALPHA
| R200_TXA_COMP_ARG_A
,
241 R200_TXA_ARG_A_R2_ALPHA
| R200_TXA_COMP_ARG_A
,
242 R200_TXA_ARG_A_R3_ALPHA
| R200_TXA_COMP_ARG_A
,
243 R200_TXA_ARG_A_R4_ALPHA
| R200_TXA_COMP_ARG_A
,
244 R200_TXA_ARG_A_R5_ALPHA
| R200_TXA_COMP_ARG_A
248 static GLuint r200_tfactor_alpha
[] =
250 R200_TXA_ARG_A_TFACTOR_ALPHA
,
251 R200_TXA_ARG_A_TFACTOR_ALPHA
| R200_TXA_COMP_ARG_A
254 static GLuint r200_tfactor1_alpha
[] =
256 R200_TXA_ARG_A_TFACTOR1_ALPHA
,
257 R200_TXA_ARG_A_TFACTOR1_ALPHA
| R200_TXA_COMP_ARG_A
260 static GLuint r200_primary_alpha
[] =
262 R200_TXA_ARG_A_DIFFUSE_ALPHA
,
263 R200_TXA_ARG_A_DIFFUSE_ALPHA
| R200_TXA_COMP_ARG_A
266 /* GL_ZERO table - indices 0-1
267 * GL_ONE table - indices 1-2
269 static GLuint r200_zero_alpha
[] =
272 R200_TXA_ARG_A_ZERO
| R200_TXA_COMP_ARG_A
,
277 /* Extract the arg from slot A, shift it into the correct argument slot
278 * and set the corresponding complement bit.
280 #define R200_COLOR_ARG( n, arg ) \
283 ((color_arg[n] & R200_TXC_ARG_A_MASK) \
284 << R200_TXC_ARG_##arg##_SHIFT); \
286 ((color_arg[n] >> R200_TXC_COMP_ARG_A_SHIFT) \
287 << R200_TXC_COMP_ARG_##arg##_SHIFT); \
290 #define R200_ALPHA_ARG( n, arg ) \
293 ((alpha_arg[n] & R200_TXA_ARG_A_MASK) \
294 << R200_TXA_ARG_##arg##_SHIFT); \
296 ((alpha_arg[n] >> R200_TXA_COMP_ARG_A_SHIFT) \
297 << R200_TXA_COMP_ARG_##arg##_SHIFT); \
301 /* ================================================================
302 * Texture unit state management
305 static GLboolean
r200UpdateTextureEnv( GLcontext
*ctx
, int unit
, int slot
, GLuint replaceargs
)
307 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
308 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[unit
];
309 GLuint color_combine
, alpha_combine
;
310 GLuint color_scale
= rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] &
311 ~(R200_TXC_SCALE_MASK
| R200_TXC_OUTPUT_REG_MASK
| R200_TXC_TFACTOR_SEL_MASK
|
312 R200_TXC_TFACTOR1_SEL_MASK
);
313 GLuint alpha_scale
= rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] &
314 ~(R200_TXA_DOT_ALPHA
| R200_TXA_SCALE_MASK
| R200_TXA_OUTPUT_REG_MASK
|
315 R200_TXA_TFACTOR_SEL_MASK
| R200_TXA_TFACTOR1_SEL_MASK
);
317 /* texUnit->_Current can be NULL if and only if the texture unit is
318 * not actually enabled.
320 assert( (texUnit
->_ReallyEnabled
== 0)
321 || (texUnit
->_Current
!= NULL
) );
323 if ( R200_DEBUG
& DEBUG_TEXTURE
) {
324 fprintf( stderr
, "%s( %p, %d )\n", __FUNCTION__
, (void *)ctx
, unit
);
327 /* Set the texture environment state. Isn't this nice and clean?
328 * The chip will automagically set the texture alpha to 0xff when
329 * the texture format does not include an alpha component. This
330 * reduces the amount of special-casing we have to do, alpha-only
331 * textures being a notable exception.
334 color_scale
|= ((rmesa
->state
.texture
.unit
[unit
].outputreg
+ 1) << R200_TXC_OUTPUT_REG_SHIFT
) |
335 (unit
<< R200_TXC_TFACTOR_SEL_SHIFT
) |
336 (replaceargs
<< R200_TXC_TFACTOR1_SEL_SHIFT
);
337 alpha_scale
|= ((rmesa
->state
.texture
.unit
[unit
].outputreg
+ 1) << R200_TXA_OUTPUT_REG_SHIFT
) |
338 (unit
<< R200_TXA_TFACTOR_SEL_SHIFT
) |
339 (replaceargs
<< R200_TXA_TFACTOR1_SEL_SHIFT
);
341 if ( !texUnit
->_ReallyEnabled
) {
343 color_combine
= R200_TXC_ARG_A_ZERO
| R200_TXC_ARG_B_ZERO
344 | R200_TXC_ARG_C_DIFFUSE_COLOR
| R200_TXC_OP_MADD
;
345 alpha_combine
= R200_TXA_ARG_A_ZERO
| R200_TXA_ARG_B_ZERO
346 | R200_TXA_ARG_C_DIFFUSE_ALPHA
| R200_TXA_OP_MADD
;
349 GLuint color_arg
[3], alpha_arg
[3];
351 const GLuint numColorArgs
= texUnit
->_CurrentCombine
->_NumArgsRGB
;
352 const GLuint numAlphaArgs
= texUnit
->_CurrentCombine
->_NumArgsA
;
353 GLuint RGBshift
= texUnit
->_CurrentCombine
->ScaleShiftRGB
;
354 GLuint Ashift
= texUnit
->_CurrentCombine
->ScaleShiftA
;
357 const GLint replaceoprgb
=
358 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->OperandRGB
[0] - GL_SRC_COLOR
;
359 const GLint replaceopa
=
360 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->OperandA
[0] - GL_SRC_ALPHA
;
363 * Extract the color and alpha combine function arguments.
365 for ( i
= 0 ; i
< numColorArgs
; i
++ ) {
366 GLint op
= texUnit
->_CurrentCombine
->OperandRGB
[i
] - GL_SRC_COLOR
;
367 const GLint srcRGBi
= texUnit
->_CurrentCombine
->SourceRGB
[i
];
372 color_arg
[i
] = r200_register_color
[op
][unit
];
375 color_arg
[i
] = r200_tfactor_color
[op
];
377 case GL_PRIMARY_COLOR
:
378 color_arg
[i
] = r200_primary_color
[op
];
381 if (replaceargs
!= unit
) {
382 const GLint srcRGBreplace
=
383 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->SourceRGB
[0];
385 op
= op
^ replaceopa
;
388 op
= op
^ replaceoprgb
;
390 switch (srcRGBreplace
) {
392 color_arg
[i
] = r200_register_color
[op
][replaceargs
];
395 color_arg
[i
] = r200_tfactor1_color
[op
];
397 case GL_PRIMARY_COLOR
:
398 color_arg
[i
] = r200_primary_color
[op
];
402 color_arg
[i
] = r200_primary_color
[op
];
404 color_arg
[i
] = r200_register_color
[op
]
405 [rmesa
->state
.texture
.unit
[replaceargs
- 1].outputreg
];
408 color_arg
[i
] = r200_zero_color
[op
];
411 color_arg
[i
] = r200_zero_color
[op
+1];
419 color_arg
[i
] = r200_register_color
[op
][srcRGBreplace
- GL_TEXTURE0
];
427 color_arg
[i
] = r200_primary_color
[op
];
429 color_arg
[i
] = r200_register_color
[op
]
430 [rmesa
->state
.texture
.unit
[unit
- 1].outputreg
];
434 color_arg
[i
] = r200_zero_color
[op
];
437 color_arg
[i
] = r200_zero_color
[op
+1];
445 color_arg
[i
] = r200_register_color
[op
][srcRGBi
- GL_TEXTURE0
];
452 for ( i
= 0 ; i
< numAlphaArgs
; i
++ ) {
453 GLint op
= texUnit
->_CurrentCombine
->OperandA
[i
] - GL_SRC_ALPHA
;
454 const GLint srcAi
= texUnit
->_CurrentCombine
->SourceA
[i
];
459 alpha_arg
[i
] = r200_register_alpha
[op
][unit
];
462 alpha_arg
[i
] = r200_tfactor_alpha
[op
];
464 case GL_PRIMARY_COLOR
:
465 alpha_arg
[i
] = r200_primary_alpha
[op
];
468 if (replaceargs
!= unit
) {
469 const GLint srcAreplace
=
470 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->SourceA
[0];
471 op
= op
^ replaceopa
;
472 switch (srcAreplace
) {
474 alpha_arg
[i
] = r200_register_alpha
[op
][replaceargs
];
477 alpha_arg
[i
] = r200_tfactor1_alpha
[op
];
479 case GL_PRIMARY_COLOR
:
480 alpha_arg
[i
] = r200_primary_alpha
[op
];
484 alpha_arg
[i
] = r200_primary_alpha
[op
];
486 alpha_arg
[i
] = r200_register_alpha
[op
]
487 [rmesa
->state
.texture
.unit
[replaceargs
- 1].outputreg
];
490 alpha_arg
[i
] = r200_zero_alpha
[op
];
493 alpha_arg
[i
] = r200_zero_alpha
[op
+1];
501 alpha_arg
[i
] = r200_register_alpha
[op
][srcAreplace
- GL_TEXTURE0
];
509 alpha_arg
[i
] = r200_primary_alpha
[op
];
511 alpha_arg
[i
] = r200_register_alpha
[op
]
512 [rmesa
->state
.texture
.unit
[unit
- 1].outputreg
];
516 alpha_arg
[i
] = r200_zero_alpha
[op
];
519 alpha_arg
[i
] = r200_zero_alpha
[op
+1];
527 alpha_arg
[i
] = r200_register_alpha
[op
][srcAi
- GL_TEXTURE0
];
535 * Build up the color and alpha combine functions.
537 switch ( texUnit
->_CurrentCombine
->ModeRGB
) {
539 color_combine
= (R200_TXC_ARG_A_ZERO
|
540 R200_TXC_ARG_B_ZERO
|
542 R200_COLOR_ARG( 0, C
);
545 color_combine
= (R200_TXC_ARG_C_ZERO
|
547 R200_COLOR_ARG( 0, A
);
548 R200_COLOR_ARG( 1, B
);
551 color_combine
= (R200_TXC_ARG_B_ZERO
|
552 R200_TXC_COMP_ARG_B
|
554 R200_COLOR_ARG( 0, A
);
555 R200_COLOR_ARG( 1, C
);
558 color_combine
= (R200_TXC_ARG_B_ZERO
|
559 R200_TXC_COMP_ARG_B
|
560 R200_TXC_BIAS_ARG_C
| /* new */
561 R200_TXC_OP_MADD
); /* was ADDSIGNED */
562 R200_COLOR_ARG( 0, A
);
563 R200_COLOR_ARG( 1, C
);
566 color_combine
= (R200_TXC_ARG_B_ZERO
|
567 R200_TXC_COMP_ARG_B
|
570 R200_COLOR_ARG( 0, A
);
571 R200_COLOR_ARG( 1, C
);
574 color_combine
= (R200_TXC_OP_LERP
);
575 R200_COLOR_ARG( 0, B
);
576 R200_COLOR_ARG( 1, A
);
577 R200_COLOR_ARG( 2, C
);
580 case GL_DOT3_RGB_EXT
:
581 case GL_DOT3_RGBA_EXT
:
582 /* The EXT version of the DOT3 extension does not support the
583 * scale factor, but the ARB version (and the version in OpenGL
591 /* DOT3 works differently on R200 than on R100. On R100, just
592 * setting the DOT3 mode did everything for you. On R200, the
593 * driver has to enable the biasing and scale in the inputs to
594 * put them in the proper [-1,1] range. This is what the 4x and
595 * the -0.5 in the DOT3 spec do. The post-scale is then set
599 color_combine
= (R200_TXC_ARG_C_ZERO
|
601 R200_TXC_BIAS_ARG_A
|
602 R200_TXC_BIAS_ARG_B
|
603 R200_TXC_SCALE_ARG_A
|
604 R200_TXC_SCALE_ARG_B
);
605 R200_COLOR_ARG( 0, A
);
606 R200_COLOR_ARG( 1, B
);
609 case GL_MODULATE_ADD_ATI
:
610 color_combine
= (R200_TXC_OP_MADD
);
611 R200_COLOR_ARG( 0, A
);
612 R200_COLOR_ARG( 1, C
);
613 R200_COLOR_ARG( 2, B
);
615 case GL_MODULATE_SIGNED_ADD_ATI
:
616 color_combine
= (R200_TXC_BIAS_ARG_C
| /* new */
617 R200_TXC_OP_MADD
); /* was ADDSIGNED */
618 R200_COLOR_ARG( 0, A
);
619 R200_COLOR_ARG( 1, C
);
620 R200_COLOR_ARG( 2, B
);
622 case GL_MODULATE_SUBTRACT_ATI
:
623 color_combine
= (R200_TXC_NEG_ARG_C
|
625 R200_COLOR_ARG( 0, A
);
626 R200_COLOR_ARG( 1, C
);
627 R200_COLOR_ARG( 2, B
);
633 switch ( texUnit
->_CurrentCombine
->ModeA
) {
635 alpha_combine
= (R200_TXA_ARG_A_ZERO
|
636 R200_TXA_ARG_B_ZERO
|
638 R200_ALPHA_ARG( 0, C
);
641 alpha_combine
= (R200_TXA_ARG_C_ZERO
|
643 R200_ALPHA_ARG( 0, A
);
644 R200_ALPHA_ARG( 1, B
);
647 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
648 R200_TXA_COMP_ARG_B
|
650 R200_ALPHA_ARG( 0, A
);
651 R200_ALPHA_ARG( 1, C
);
654 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
655 R200_TXA_COMP_ARG_B
|
656 R200_TXA_BIAS_ARG_C
| /* new */
657 R200_TXA_OP_MADD
); /* was ADDSIGNED */
658 R200_ALPHA_ARG( 0, A
);
659 R200_ALPHA_ARG( 1, C
);
662 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
663 R200_TXA_COMP_ARG_B
|
666 R200_ALPHA_ARG( 0, A
);
667 R200_ALPHA_ARG( 1, C
);
670 alpha_combine
= (R200_TXA_OP_LERP
);
671 R200_ALPHA_ARG( 0, B
);
672 R200_ALPHA_ARG( 1, A
);
673 R200_ALPHA_ARG( 2, C
);
676 case GL_MODULATE_ADD_ATI
:
677 alpha_combine
= (R200_TXA_OP_MADD
);
678 R200_ALPHA_ARG( 0, A
);
679 R200_ALPHA_ARG( 1, C
);
680 R200_ALPHA_ARG( 2, B
);
682 case GL_MODULATE_SIGNED_ADD_ATI
:
683 alpha_combine
= (R200_TXA_BIAS_ARG_C
| /* new */
684 R200_TXA_OP_MADD
); /* was ADDSIGNED */
685 R200_ALPHA_ARG( 0, A
);
686 R200_ALPHA_ARG( 1, C
);
687 R200_ALPHA_ARG( 2, B
);
689 case GL_MODULATE_SUBTRACT_ATI
:
690 alpha_combine
= (R200_TXA_NEG_ARG_C
|
692 R200_ALPHA_ARG( 0, A
);
693 R200_ALPHA_ARG( 1, C
);
694 R200_ALPHA_ARG( 2, B
);
700 if ( (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA_EXT
)
701 || (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA
) ) {
702 alpha_scale
|= R200_TXA_DOT_ALPHA
;
707 * Apply the scale factor.
709 color_scale
|= (RGBshift
<< R200_TXC_SCALE_SHIFT
);
710 alpha_scale
|= (Ashift
<< R200_TXA_SCALE_SHIFT
);
716 if ( rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND
] != color_combine
||
717 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND
] != alpha_combine
||
718 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] != color_scale
||
719 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] != alpha_scale
) {
720 R200_STATECHANGE( rmesa
, pix
[slot
] );
721 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND
] = color_combine
;
722 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND
] = alpha_combine
;
723 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] = color_scale
;
724 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] = alpha_scale
;
730 void r200SetTexOffset(__DRIcontext
* pDRICtx
, GLint texname
,
731 unsigned long long offset
, GLint depth
, GLuint pitch
)
733 r200ContextPtr rmesa
= pDRICtx
->driverPrivate
;
734 struct gl_texture_object
*tObj
=
735 _mesa_lookup_texture(rmesa
->radeon
.glCtx
, texname
);
741 t
= (radeonTexObjPtr
) tObj
->DriverData
;
743 t
->image_override
= GL_TRUE
;
748 t
->pp_txoffset
= offset
;
749 t
->pp_txpitch
= pitch
- 32;
753 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_ARGB8888
].format
;
754 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_ARGB8888
].filter
;
758 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB888
].format
;
759 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB888
].filter
;
762 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB565
].format
;
763 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB565
].filter
;
771 static GLboolean
r200UpdateAllTexEnv( GLcontext
*ctx
)
773 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
774 GLint i
, j
, currslot
;
775 GLint maxunitused
= -1;
776 GLboolean texregfree
[6] = {GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
};
777 GLubyte stageref
[7] = {0, 0, 0, 0, 0, 0, 0};
778 GLint nextunit
[R200_MAX_TEXTURE_UNITS
] = {0, 0, 0, 0, 0, 0};
779 GLint currentnext
= -1;
782 /* find highest used unit */
783 for ( j
= 0; j
< R200_MAX_TEXTURE_UNITS
; j
++) {
784 if (ctx
->Texture
.Unit
[j
]._ReallyEnabled
) {
788 stageref
[maxunitused
+ 1] = REF_COLOR
| REF_ALPHA
;
790 for ( j
= maxunitused
; j
>= 0; j
-- ) {
791 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[j
];
793 rmesa
->state
.texture
.unit
[j
].outputreg
= -1;
795 if (stageref
[j
+ 1]) {
797 /* use the lowest available reg. That gets us automatically reg0 for the last stage.
798 need this even for disabled units, as it may get referenced due to the replace
800 for ( i
= 0 ; i
< R200_MAX_TEXTURE_UNITS
; i
++ ) {
802 rmesa
->state
.texture
.unit
[j
].outputreg
= i
;
806 if (rmesa
->state
.texture
.unit
[j
].outputreg
== -1) {
807 /* no more free regs we can use. Need a fallback :-( */
811 nextunit
[j
] = currentnext
;
813 if (!texUnit
->_ReallyEnabled
) {
814 /* the not enabled stages are referenced "indirectly",
815 must not cut off the lower stages */
816 stageref
[j
] = REF_COLOR
| REF_ALPHA
;
821 const GLuint numColorArgs
= texUnit
->_CurrentCombine
->_NumArgsRGB
;
822 const GLuint numAlphaArgs
= texUnit
->_CurrentCombine
->_NumArgsA
;
823 const GLboolean isdot3rgba
= (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA
) ||
824 (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA_EXT
);
827 /* check if we need the color part, special case for dot3_rgba
828 as if only the alpha part is referenced later on it still is using the color part */
829 if ((stageref
[j
+ 1] & REF_COLOR
) || isdot3rgba
) {
830 for ( i
= 0 ; i
< numColorArgs
; i
++ ) {
831 const GLuint srcRGBi
= texUnit
->_CurrentCombine
->SourceRGB
[i
];
832 const GLuint op
= texUnit
->_CurrentCombine
->OperandRGB
[i
];
835 /* op 0/1 are referencing color, op 2/3 alpha */
836 stageref
[j
] |= (op
>> 1) + 1;
839 texregfree
[j
] = GL_FALSE
;
847 texregfree
[srcRGBi
- GL_TEXTURE0
] = GL_FALSE
;
849 default: /* don't care about other sources here */
855 /* alpha args are ignored for dot3_rgba */
856 if ((stageref
[j
+ 1] & REF_ALPHA
) && !isdot3rgba
) {
858 for ( i
= 0 ; i
< numAlphaArgs
; i
++ ) {
859 const GLuint srcAi
= texUnit
->_CurrentCombine
->SourceA
[i
];
862 stageref
[j
] |= REF_ALPHA
;
865 texregfree
[j
] = GL_FALSE
;
873 texregfree
[srcAi
- GL_TEXTURE0
] = GL_FALSE
;
875 default: /* don't care about other sources here */
883 /* don't enable texture sampling for units if the result is not used */
884 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
885 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& !texregfree
[i
])
886 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
887 else rmesa
->state
.texture
.unit
[i
].unitneeded
= 0;
892 rmesa
->state
.envneeded
= 1;
895 while ((i
<= maxunitused
) && (i
>= 0)) {
896 /* only output instruction if the results are referenced */
897 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& stageref
[i
+1]) {
898 GLuint replaceunit
= i
;
899 /* try to optimize GL_REPLACE away (only one level deep though) */
900 if ( (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeRGB
== GL_REPLACE
) &&
901 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeA
== GL_REPLACE
) &&
902 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftRGB
== 0) &&
903 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftA
== 0) &&
904 (nextunit
[i
] > 0) ) {
905 /* yippie! can optimize it away! */
910 /* need env instruction slot */
911 rmesa
->state
.envneeded
|= 1 << currslot
;
912 ok
= r200UpdateTextureEnv( ctx
, i
, currslot
, replaceunit
);
913 if (!ok
) return GL_FALSE
;
920 /* need one stage at least */
921 rmesa
->state
.texture
.unit
[0].outputreg
= 0;
922 ok
= r200UpdateTextureEnv( ctx
, 0, 0, 0 );
925 R200_STATECHANGE( rmesa
, ctx
);
926 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] &= ~(R200_TEX_BLEND_ENABLE_MASK
| R200_MULTI_PASS_ENABLE
);
927 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= rmesa
->state
.envneeded
<< R200_TEX_BLEND_0_ENABLE_SHIFT
;
936 #define TEXOBJ_TXFILTER_MASK (R200_MAX_MIP_LEVEL_MASK | \
937 R200_MIN_FILTER_MASK | \
938 R200_MAG_FILTER_MASK | \
939 R200_MAX_ANISO_MASK | \
941 R200_YUV_TEMPERATURE_MASK | \
942 R200_CLAMP_S_MASK | \
943 R200_CLAMP_T_MASK | \
944 R200_BORDER_MODE_D3D )
946 #define TEXOBJ_TXFORMAT_MASK (R200_TXFORMAT_WIDTH_MASK | \
947 R200_TXFORMAT_HEIGHT_MASK | \
948 R200_TXFORMAT_FORMAT_MASK | \
949 R200_TXFORMAT_F5_WIDTH_MASK | \
950 R200_TXFORMAT_F5_HEIGHT_MASK | \
951 R200_TXFORMAT_ALPHA_IN_MAP | \
952 R200_TXFORMAT_CUBIC_MAP_ENABLE | \
953 R200_TXFORMAT_NON_POWER2)
955 #define TEXOBJ_TXFORMAT_X_MASK (R200_DEPTH_LOG2_MASK | \
956 R200_TEXCOORD_MASK | \
957 R200_CLAMP_Q_MASK | \
958 R200_VOLUME_FILTER_MASK)
961 static void import_tex_obj_state( r200ContextPtr rmesa
,
963 radeonTexObjPtr texobj
)
965 /* do not use RADEON_DB_STATE to avoid stale texture caches */
966 GLuint
*cmd
= &rmesa
->hw
.tex
[unit
].cmd
[TEX_CMD_0
];
968 R200_STATECHANGE( rmesa
, tex
[unit
] );
970 cmd
[TEX_PP_TXFILTER
] &= ~TEXOBJ_TXFILTER_MASK
;
971 cmd
[TEX_PP_TXFILTER
] |= texobj
->pp_txfilter
& TEXOBJ_TXFILTER_MASK
;
972 cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
973 cmd
[TEX_PP_TXFORMAT
] |= texobj
->pp_txformat
& TEXOBJ_TXFORMAT_MASK
;
974 cmd
[TEX_PP_TXFORMAT_X
] &= ~TEXOBJ_TXFORMAT_X_MASK
;
975 cmd
[TEX_PP_TXFORMAT_X
] |= texobj
->pp_txformat_x
& TEXOBJ_TXFORMAT_X_MASK
;
976 cmd
[TEX_PP_TXSIZE
] = texobj
->pp_txsize
; /* NPOT only! */
977 cmd
[TEX_PP_TXPITCH
] = texobj
->pp_txpitch
; /* NPOT only! */
978 cmd
[TEX_PP_BORDER_COLOR
] = texobj
->pp_border_color
;
979 if (rmesa
->radeon
.radeonScreen
->drmSupportsFragShader
) {
980 cmd
[TEX_PP_TXOFFSET_NEWDRM
] = texobj
->pp_txoffset
;
983 cmd
[TEX_PP_TXOFFSET_OLDDRM
] = texobj
->pp_txoffset
;
986 if (texobj
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
987 GLuint
*cube_cmd
= &rmesa
->hw
.cube
[unit
].cmd
[CUBE_CMD_0
];
988 // GLuint bytesPerFace = texobj->base.totalSize / 6;
989 // ASSERT(texobj->base.totalSize % 6 == 0);
990 GLuint bytesPerFace
= 1; // TODO
992 R200_STATECHANGE( rmesa
, cube
[unit
] );
993 cube_cmd
[CUBE_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
994 if (rmesa
->radeon
.radeonScreen
->drmSupportsFragShader
) {
995 /* that value is submitted twice. could change cube atom
996 to not include that command when new drm is used */
997 cmd
[TEX_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
999 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F1
] = texobj
->pp_txoffset
+ 1 * bytesPerFace
;
1000 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F2
] = texobj
->pp_txoffset
+ 2 * bytesPerFace
;
1001 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F3
] = texobj
->pp_txoffset
+ 3 * bytesPerFace
;
1002 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F4
] = texobj
->pp_txoffset
+ 4 * bytesPerFace
;
1003 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F5
] = texobj
->pp_txoffset
+ 5 * bytesPerFace
;
1006 texobj
->dirty_state
&= ~(1<<unit
);
1009 static void set_texgen_matrix( r200ContextPtr rmesa
,
1011 const GLfloat
*s_plane
,
1012 const GLfloat
*t_plane
,
1013 const GLfloat
*r_plane
,
1014 const GLfloat
*q_plane
)
1038 _math_matrix_loadf( &(rmesa
->TexGenMatrix
[unit
]), m
);
1039 _math_matrix_analyse( &(rmesa
->TexGenMatrix
[unit
]) );
1040 rmesa
->TexGenEnabled
|= R200_TEXMAT_0_ENABLE
<<unit
;
1044 static GLuint
r200_need_dis_texgen(const GLbitfield texGenEnabled
,
1045 const GLfloat
*planeS
,
1046 const GLfloat
*planeT
,
1047 const GLfloat
*planeR
,
1048 const GLfloat
*planeQ
)
1050 GLuint needtgenable
= 0;
1052 if (!(texGenEnabled
& S_BIT
)) {
1053 if (((texGenEnabled
& T_BIT
) && planeT
[0] != 0.0) ||
1054 ((texGenEnabled
& R_BIT
) && planeR
[0] != 0.0) ||
1055 ((texGenEnabled
& Q_BIT
) && planeQ
[0] != 0.0)) {
1056 needtgenable
|= S_BIT
;
1059 if (!(texGenEnabled
& T_BIT
)) {
1060 if (((texGenEnabled
& S_BIT
) && planeS
[1] != 0.0) ||
1061 ((texGenEnabled
& R_BIT
) && planeR
[1] != 0.0) ||
1062 ((texGenEnabled
& Q_BIT
) && planeQ
[1] != 0.0)) {
1063 needtgenable
|= T_BIT
;
1066 if (!(texGenEnabled
& R_BIT
)) {
1067 if (((texGenEnabled
& S_BIT
) && planeS
[2] != 0.0) ||
1068 ((texGenEnabled
& T_BIT
) && planeT
[2] != 0.0) ||
1069 ((texGenEnabled
& Q_BIT
) && planeQ
[2] != 0.0)) {
1070 needtgenable
|= R_BIT
;
1073 if (!(texGenEnabled
& Q_BIT
)) {
1074 if (((texGenEnabled
& S_BIT
) && planeS
[3] != 0.0) ||
1075 ((texGenEnabled
& T_BIT
) && planeT
[3] != 0.0) ||
1076 ((texGenEnabled
& R_BIT
) && planeR
[3] != 0.0)) {
1077 needtgenable
|= Q_BIT
;
1081 return needtgenable
;
1086 * Returns GL_FALSE if fallback required.
1088 static GLboolean
r200_validate_texgen( GLcontext
*ctx
, GLuint unit
)
1090 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1091 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[unit
];
1092 GLuint inputshift
= R200_TEXGEN_0_INPUT_SHIFT
+ unit
*4;
1095 GLboolean mixed_fallback
= GL_FALSE
;
1096 static const GLfloat I
[16] = {
1101 static const GLfloat reflect
[16] = {
1107 rmesa
->TexGenCompSel
&= ~(R200_OUTPUT_TEX_0
<< unit
);
1108 rmesa
->TexGenEnabled
&= ~(R200_TEXGEN_TEXMAT_0_ENABLE
<<unit
);
1109 rmesa
->TexGenEnabled
&= ~(R200_TEXMAT_0_ENABLE
<<unit
);
1110 rmesa
->TexGenNeedNormals
[unit
] = GL_FALSE
;
1111 tgi
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] & ~(R200_TEXGEN_INPUT_MASK
<<
1113 tgcm
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] & ~(R200_TEXGEN_COMP_MASK
<<
1117 fprintf(stderr
, "%s unit %d\n", __FUNCTION__
, unit
);
1119 if (texUnit
->TexGenEnabled
& S_BIT
) {
1120 mode
= texUnit
->GenModeS
;
1122 tgcm
|= R200_TEXGEN_COMP_S
<< (unit
* 4);
1125 if (texUnit
->TexGenEnabled
& T_BIT
) {
1126 if (texUnit
->GenModeT
!= mode
)
1127 mixed_fallback
= GL_TRUE
;
1129 tgcm
|= R200_TEXGEN_COMP_T
<< (unit
* 4);
1131 if (texUnit
->TexGenEnabled
& R_BIT
) {
1132 if (texUnit
->GenModeR
!= mode
)
1133 mixed_fallback
= GL_TRUE
;
1135 tgcm
|= R200_TEXGEN_COMP_R
<< (unit
* 4);
1138 if (texUnit
->TexGenEnabled
& Q_BIT
) {
1139 if (texUnit
->GenModeQ
!= mode
)
1140 mixed_fallback
= GL_TRUE
;
1142 tgcm
|= R200_TEXGEN_COMP_Q
<< (unit
* 4);
1145 if (mixed_fallback
) {
1146 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1147 fprintf(stderr
, "fallback mixed texgen, 0x%x (0x%x 0x%x 0x%x 0x%x)\n",
1148 texUnit
->TexGenEnabled
, texUnit
->GenModeS
, texUnit
->GenModeT
,
1149 texUnit
->GenModeR
, texUnit
->GenModeQ
);
1153 /* we CANNOT do mixed mode if the texgen mode requires a plane where the input
1154 is not enabled for texgen, since the planes are concatenated into texmat,
1155 and thus the input will come from texcoord rather than tex gen equation!
1156 Either fallback or just hope that those texcoords aren't really needed...
1157 Assuming the former will cause lots of unnecessary fallbacks, the latter will
1158 generate bogus results sometimes - it's pretty much impossible to really know
1159 when a fallback is needed, depends on texmat and what sort of texture is bound
1160 etc, - for now fallback if we're missing either S or T bits, there's a high
1161 probability we need the texcoords in that case.
1162 That's a lot of work for some obscure texgen mixed mode fixup - why oh why
1163 doesn't the chip just directly accept the plane parameters :-(. */
1165 case GL_OBJECT_LINEAR
: {
1166 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1167 texUnit
->ObjectPlaneS
, texUnit
->ObjectPlaneT
,
1168 texUnit
->ObjectPlaneR
, texUnit
->ObjectPlaneQ
);
1169 if (needtgenable
& (S_BIT
| T_BIT
)) {
1170 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1171 fprintf(stderr
, "fallback mixed texgen / obj plane, 0x%x\n",
1172 texUnit
->TexGenEnabled
);
1175 if (needtgenable
& (R_BIT
)) {
1176 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1178 if (needtgenable
& (Q_BIT
)) {
1179 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1182 tgi
|= R200_TEXGEN_INPUT_OBJ
<< inputshift
;
1183 set_texgen_matrix( rmesa
, unit
,
1184 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->ObjectPlaneS
: I
,
1185 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->ObjectPlaneT
: I
+ 4,
1186 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->ObjectPlaneR
: I
+ 8,
1187 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->ObjectPlaneQ
: I
+ 12);
1191 case GL_EYE_LINEAR
: {
1192 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1193 texUnit
->EyePlaneS
, texUnit
->EyePlaneT
,
1194 texUnit
->EyePlaneR
, texUnit
->EyePlaneQ
);
1195 if (needtgenable
& (S_BIT
| T_BIT
)) {
1196 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1197 fprintf(stderr
, "fallback mixed texgen / eye plane, 0x%x\n",
1198 texUnit
->TexGenEnabled
);
1201 if (needtgenable
& (R_BIT
)) {
1202 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1204 if (needtgenable
& (Q_BIT
)) {
1205 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1207 tgi
|= R200_TEXGEN_INPUT_EYE
<< inputshift
;
1208 set_texgen_matrix( rmesa
, unit
,
1209 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->EyePlaneS
: I
,
1210 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->EyePlaneT
: I
+ 4,
1211 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->EyePlaneR
: I
+ 8,
1212 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->EyePlaneQ
: I
+ 12);
1216 case GL_REFLECTION_MAP_NV
:
1217 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1218 tgi
|= R200_TEXGEN_INPUT_EYE_REFLECT
<< inputshift
;
1219 /* pretty weird, must only negate when lighting is enabled? */
1220 if (ctx
->Light
.Enabled
)
1221 set_texgen_matrix( rmesa
, unit
,
1222 (texUnit
->TexGenEnabled
& S_BIT
) ? reflect
: I
,
1223 (texUnit
->TexGenEnabled
& T_BIT
) ? reflect
+ 4 : I
+ 4,
1224 (texUnit
->TexGenEnabled
& R_BIT
) ? reflect
+ 8 : I
+ 8,
1228 case GL_NORMAL_MAP_NV
:
1229 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1230 tgi
|= R200_TEXGEN_INPUT_EYE_NORMAL
<<inputshift
;
1234 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1235 tgi
|= R200_TEXGEN_INPUT_SPHERE
<<inputshift
;
1239 /* All texgen units were disabled, so just pass coords through. */
1240 tgi
|= unit
<< inputshift
;
1244 /* Unsupported mode, fallback:
1246 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1247 fprintf(stderr
, "fallback unsupported texgen, %d\n",
1252 rmesa
->TexGenEnabled
|= R200_TEXGEN_TEXMAT_0_ENABLE
<< unit
;
1253 rmesa
->TexGenCompSel
|= R200_OUTPUT_TEX_0
<< unit
;
1255 if (tgi
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] ||
1256 tgcm
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
])
1258 R200_STATECHANGE(rmesa
, tcg
);
1259 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] = tgi
;
1260 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] = tgcm
;
1266 void set_re_cntl_d3d( GLcontext
*ctx
, int unit
, GLboolean use_d3d
)
1268 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1272 re_cntl
= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] & ~(R200_VTX_STQ0_D3D
<< (2 * unit
));
1274 re_cntl
|= R200_VTX_STQ0_D3D
<< (2 * unit
);
1276 if ( re_cntl
!= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] ) {
1277 R200_STATECHANGE( rmesa
, set
);
1278 rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] = re_cntl
;
1283 * Compute the cached hardware register values for the given texture object.
1285 * \param rmesa Context pointer
1286 * \param t the r300 texture object
1288 static void setup_hardware_state(r200ContextPtr rmesa
, radeonTexObj
*t
)
1290 const struct gl_texture_image
*firstImage
=
1291 t
->base
.Image
[0][t
->mt
->firstLevel
];
1292 GLint log2Width
, log2Height
, log2Depth
, texelBytes
;
1294 log2Width
= firstImage
->WidthLog2
;
1295 log2Height
= firstImage
->HeightLog2
;
1296 log2Depth
= firstImage
->DepthLog2
;
1297 texelBytes
= firstImage
->TexFormat
->TexelBytes
;
1300 if (!t
->image_override
) {
1301 if (VALID_FORMAT(firstImage
->TexFormat
->MesaFormat
)) {
1302 const struct tx_table
*table
= _mesa_little_endian() ? tx_table_le
:
1305 t
->pp_txformat
&= ~(R200_TXFORMAT_FORMAT_MASK
|
1306 R200_TXFORMAT_ALPHA_IN_MAP
);
1307 t
->pp_txfilter
&= ~R200_YUV_TO_RGB
;
1309 t
->pp_txformat
|= table
[ firstImage
->TexFormat
->MesaFormat
].format
;
1310 t
->pp_txfilter
|= table
[ firstImage
->TexFormat
->MesaFormat
].filter
;
1312 _mesa_problem(NULL
, "unexpected texture format in %s",
1318 t
->pp_txfilter
&= ~R200_MAX_MIP_LEVEL_MASK
;
1319 t
->pp_txfilter
|= (t
->mt
->lastLevel
- t
->mt
->firstLevel
) << R200_MAX_MIP_LEVEL_SHIFT
;
1321 t
->pp_txformat
&= ~(R200_TXFORMAT_WIDTH_MASK
|
1322 R200_TXFORMAT_HEIGHT_MASK
|
1323 R200_TXFORMAT_CUBIC_MAP_ENABLE
|
1324 R200_TXFORMAT_F5_WIDTH_MASK
|
1325 R200_TXFORMAT_F5_HEIGHT_MASK
);
1326 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_WIDTH_SHIFT
) |
1327 (log2Height
<< R200_TXFORMAT_HEIGHT_SHIFT
));
1331 t
->pp_txformat_x
&= ~(R200_DEPTH_LOG2_MASK
| R200_TEXCOORD_MASK
);
1332 if (t
->base
.Target
== GL_TEXTURE_3D
) {
1333 t
->pp_txformat_x
|= (log2Depth
<< R200_DEPTH_LOG2_SHIFT
);
1334 t
->pp_txformat_x
|= R200_TEXCOORD_VOLUME
;
1337 else if (t
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
1338 ASSERT(log2Width
== log2Height
);
1339 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_F5_WIDTH_SHIFT
) |
1340 (log2Height
<< R200_TXFORMAT_F5_HEIGHT_SHIFT
) |
1341 /* don't think we need this bit, if it exists at all - fglrx does not set it */
1342 (R200_TXFORMAT_CUBIC_MAP_ENABLE
));
1343 t
->pp_txformat_x
|= R200_TEXCOORD_CUBIC_ENV
;
1344 t
->pp_cubic_faces
= ((log2Width
<< R200_FACE_WIDTH_1_SHIFT
) |
1345 (log2Height
<< R200_FACE_HEIGHT_1_SHIFT
) |
1346 (log2Width
<< R200_FACE_WIDTH_2_SHIFT
) |
1347 (log2Height
<< R200_FACE_HEIGHT_2_SHIFT
) |
1348 (log2Width
<< R200_FACE_WIDTH_3_SHIFT
) |
1349 (log2Height
<< R200_FACE_HEIGHT_3_SHIFT
) |
1350 (log2Width
<< R200_FACE_WIDTH_4_SHIFT
) |
1351 (log2Height
<< R200_FACE_HEIGHT_4_SHIFT
));
1354 /* If we don't in fact send enough texture coordinates, q will be 1,
1355 * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
1357 t
->pp_txformat_x
|= R200_TEXCOORD_PROJ
;
1360 t
->pp_txsize
= (((firstImage
->Width
- 1) << R200_PP_TX_WIDTHMASK_SHIFT
)
1361 | ((firstImage
->Height
- 1) << R200_PP_TX_HEIGHTMASK_SHIFT
));
1363 if ( !t
->image_override
) {
1364 if (firstImage
->IsCompressed
)
1365 t
->pp_txpitch
= (firstImage
->Width
+ 63) & ~(63);
1367 t
->pp_txpitch
= ((firstImage
->Width
* texelBytes
) + 63) & ~(63);
1368 t
->pp_txpitch
-= 32;
1371 if (t
->base
.Target
== GL_TEXTURE_RECTANGLE_NV
) {
1372 t
->pp_txformat
|= R200_TXFORMAT_NON_POWER2
;
1377 static GLboolean
r200_validate_texture(GLcontext
*ctx
, struct gl_texture_object
*texObj
, int unit
)
1379 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1380 radeonTexObj
*t
= radeon_tex_obj(texObj
);
1382 if (!radeon_validate_texture_miptree(ctx
, texObj
))
1385 r200_validate_texgen(ctx
, unit
);
1386 /* Configure the hardware registers (more precisely, the cached version
1387 * of the hardware registers). */
1388 setup_hardware_state(rmesa
, t
);
1390 if (texObj
->Target
== GL_TEXTURE_RECTANGLE_NV
||
1391 texObj
->Target
== GL_TEXTURE_2D
||
1392 texObj
->Target
== GL_TEXTURE_1D
)
1393 set_re_cntl_d3d( ctx
, unit
, GL_FALSE
);
1395 set_re_cntl_d3d( ctx
, unit
, GL_TRUE
);
1396 R200_STATECHANGE( rmesa
, ctx
);
1397 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_0_ENABLE
<< unit
;
1399 R200_STATECHANGE( rmesa
, vtx
);
1400 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] &= ~(7 << (unit
* 3));
1401 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] |= 4 << (unit
* 3);
1403 rmesa
->recheck_texgen
[unit
] = GL_TRUE
;
1404 if (t
->dirty_state
& (1<<unit
)) {
1405 import_tex_obj_state( rmesa
, unit
, t
);
1408 t
->dirty_state
= R200_TEX_ALL
;
1410 t
->validated
= GL_TRUE
;
1414 static GLboolean
r200UpdateTextureUnit(GLcontext
*ctx
, int unit
)
1416 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1417 GLuint unitneeded
= rmesa
->state
.texture
.unit
[unit
].unitneeded
;
1422 if (!r200_validate_texture(ctx
, ctx
->Texture
.Unit
[unit
]._Current
, unit
)) {
1424 "failed to validate texture for unit %d.\n",
1426 rmesa
->state
.texture
.unit
[unit
].texobj
= NULL
;
1430 rmesa
->state
.texture
.unit
[unit
].texobj
= radeon_tex_obj(ctx
->Texture
.Unit
[unit
]._Current
);
1435 void r200UpdateTextureState( GLcontext
*ctx
)
1437 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1441 /* NOTE: must not manipulate rmesa->state.texture.unit[].unitneeded or
1442 rmesa->state.envneeded before a R200_STATECHANGE (or R200_NEWPRIM) since
1443 we use these to determine if we want to emit the corresponding state
1445 R200_NEWPRIM( rmesa
);
1447 if (ctx
->ATIFragmentShader
._Enabled
) {
1449 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
1450 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
1455 ok
= r200UpdateAllTexEnv( ctx
);
1458 ok
= (r200UpdateTextureUnit( ctx
, 0 ) &&
1459 r200UpdateTextureUnit( ctx
, 1 ) &&
1460 r200UpdateTextureUnit( ctx
, 2 ) &&
1461 r200UpdateTextureUnit( ctx
, 3 ) &&
1462 r200UpdateTextureUnit( ctx
, 4 ) &&
1463 r200UpdateTextureUnit( ctx
, 5 ));
1466 if (ok
&& ctx
->ATIFragmentShader
._Enabled
) {
1467 r200UpdateFragmentShader(ctx
);
1470 FALLBACK( rmesa
, R200_FALLBACK_TEXTURE
, !ok
);
1472 if (rmesa
->radeon
.TclFallback
)
1473 r200ChooseVertexState( ctx
);
1476 if (rmesa
->radeon
.radeonScreen
->chip_family
== CHIP_FAMILY_R200
) {
1479 * T0 hang workaround -------------
1480 * not needed for r200 derivatives
1482 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_ENABLE_MASK
) == R200_TEX_0_ENABLE
&&
1483 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1485 R200_STATECHANGE(rmesa
, ctx
);
1486 R200_STATECHANGE(rmesa
, tex
[1]);
1487 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_1_ENABLE
;
1488 if (!(rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_1_ENABLE
))
1489 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1490 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] |= R200_TXFORMAT_LOOKUP_DISABLE
;
1492 else if (!ctx
->ATIFragmentShader
._Enabled
) {
1493 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
) &&
1494 (rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] & R200_TXFORMAT_LOOKUP_DISABLE
)) {
1495 R200_STATECHANGE(rmesa
, tex
[1]);
1496 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~R200_TXFORMAT_LOOKUP_DISABLE
;
1499 /* do the same workaround for the first pass of a fragment shader.
1500 * completely unknown if necessary / sufficient.
1502 if ((rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_ENABLE_MASK
) == R200_PPX_TEX_0_ENABLE
&&
1503 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1505 R200_STATECHANGE(rmesa
, cst
);
1506 R200_STATECHANGE(rmesa
, tex
[1]);
1507 rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] |= R200_PPX_TEX_1_ENABLE
;
1508 if (!(rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
))
1509 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1510 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXMULTI_CTL
] |= R200_PASS1_TXFORMAT_LOOKUP_DISABLE
;
1513 /* maybe needs to be done pairwise due to 2 parallel (physical) tex units ?
1514 looks like that's not the case, if 8500/9100 owners don't complain remove this...
1515 for ( i = 0; i < ctx->Const.MaxTextureUnits; i += 2) {
1516 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & ((R200_TEX_0_ENABLE |
1517 R200_TEX_1_ENABLE ) << i)) == (R200_TEX_0_ENABLE << i)) &&
1518 ((rmesa->hw.tex[i].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK) >
1519 R200_MIN_FILTER_LINEAR)) {
1520 R200_STATECHANGE(rmesa, ctx);
1521 R200_STATECHANGE(rmesa, tex[i+1]);
1522 rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= (R200_TEX_1_ENABLE << i);
1523 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
1524 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] |= 0x08000000;
1527 if ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE << i)) &&
1528 (rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] & 0x08000000)) {
1529 R200_STATECHANGE(rmesa, tex[i+1]);
1530 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~0x08000000;
1536 * Texture cache LRU hang workaround -------------
1537 * not needed for r200 derivatives
1538 * hopefully this covers first pass of a shader as well
1541 /* While the cases below attempt to only enable the workaround in the
1542 * specific cases necessary, they were insufficient. See bugzilla #1519,
1543 * #729, #814. Tests with quake3 showed no impact on performance.
1548 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE )) &&
1549 ((((rmesa->hw.tex[0].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1551 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_2_ENABLE) &&
1552 ((((rmesa->hw.tex[2].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1554 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_4_ENABLE) &&
1555 ((((rmesa->hw.tex[4].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1561 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE )) &&
1562 ((((rmesa->hw.tex[1].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1564 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_3_ENABLE) &&
1565 ((((rmesa->hw.tex[3].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1567 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_5_ENABLE) &&
1568 ((((rmesa->hw.tex[5].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1574 if (dbg
!= rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
]) {
1575 R200_STATECHANGE( rmesa
, tam
);
1576 rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
] = dbg
;
1577 if (0) printf("TEXCACHE LRU HANG WORKAROUND %x\n", dbg
);