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
);
736 radeonTexObjPtr t
= radeon_tex_obj(tObj
);
741 t
->image_override
= GL_TRUE
;
746 t
->pp_txoffset
= offset
;
747 t
->pp_txpitch
= pitch
- 32;
751 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_ARGB8888
].format
;
752 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_ARGB8888
].filter
;
756 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB888
].format
;
757 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB888
].filter
;
760 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB565
].format
;
761 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB565
].filter
;
769 static GLboolean
r200UpdateAllTexEnv( GLcontext
*ctx
)
771 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
772 GLint i
, j
, currslot
;
773 GLint maxunitused
= -1;
774 GLboolean texregfree
[6] = {GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
};
775 GLubyte stageref
[7] = {0, 0, 0, 0, 0, 0, 0};
776 GLint nextunit
[R200_MAX_TEXTURE_UNITS
] = {0, 0, 0, 0, 0, 0};
777 GLint currentnext
= -1;
780 /* find highest used unit */
781 for ( j
= 0; j
< R200_MAX_TEXTURE_UNITS
; j
++) {
782 if (ctx
->Texture
.Unit
[j
]._ReallyEnabled
) {
786 stageref
[maxunitused
+ 1] = REF_COLOR
| REF_ALPHA
;
788 for ( j
= maxunitused
; j
>= 0; j
-- ) {
789 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[j
];
791 rmesa
->state
.texture
.unit
[j
].outputreg
= -1;
793 if (stageref
[j
+ 1]) {
795 /* use the lowest available reg. That gets us automatically reg0 for the last stage.
796 need this even for disabled units, as it may get referenced due to the replace
798 for ( i
= 0 ; i
< R200_MAX_TEXTURE_UNITS
; i
++ ) {
800 rmesa
->state
.texture
.unit
[j
].outputreg
= i
;
804 if (rmesa
->state
.texture
.unit
[j
].outputreg
== -1) {
805 /* no more free regs we can use. Need a fallback :-( */
809 nextunit
[j
] = currentnext
;
811 if (!texUnit
->_ReallyEnabled
) {
812 /* the not enabled stages are referenced "indirectly",
813 must not cut off the lower stages */
814 stageref
[j
] = REF_COLOR
| REF_ALPHA
;
819 const GLuint numColorArgs
= texUnit
->_CurrentCombine
->_NumArgsRGB
;
820 const GLuint numAlphaArgs
= texUnit
->_CurrentCombine
->_NumArgsA
;
821 const GLboolean isdot3rgba
= (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA
) ||
822 (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA_EXT
);
825 /* check if we need the color part, special case for dot3_rgba
826 as if only the alpha part is referenced later on it still is using the color part */
827 if ((stageref
[j
+ 1] & REF_COLOR
) || isdot3rgba
) {
828 for ( i
= 0 ; i
< numColorArgs
; i
++ ) {
829 const GLuint srcRGBi
= texUnit
->_CurrentCombine
->SourceRGB
[i
];
830 const GLuint op
= texUnit
->_CurrentCombine
->OperandRGB
[i
];
833 /* op 0/1 are referencing color, op 2/3 alpha */
834 stageref
[j
] |= (op
>> 1) + 1;
837 texregfree
[j
] = GL_FALSE
;
845 texregfree
[srcRGBi
- GL_TEXTURE0
] = GL_FALSE
;
847 default: /* don't care about other sources here */
853 /* alpha args are ignored for dot3_rgba */
854 if ((stageref
[j
+ 1] & REF_ALPHA
) && !isdot3rgba
) {
856 for ( i
= 0 ; i
< numAlphaArgs
; i
++ ) {
857 const GLuint srcAi
= texUnit
->_CurrentCombine
->SourceA
[i
];
860 stageref
[j
] |= REF_ALPHA
;
863 texregfree
[j
] = GL_FALSE
;
871 texregfree
[srcAi
- GL_TEXTURE0
] = GL_FALSE
;
873 default: /* don't care about other sources here */
881 /* don't enable texture sampling for units if the result is not used */
882 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
883 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& !texregfree
[i
])
884 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
885 else rmesa
->state
.texture
.unit
[i
].unitneeded
= 0;
890 rmesa
->state
.envneeded
= 1;
893 while ((i
<= maxunitused
) && (i
>= 0)) {
894 /* only output instruction if the results are referenced */
895 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& stageref
[i
+1]) {
896 GLuint replaceunit
= i
;
897 /* try to optimize GL_REPLACE away (only one level deep though) */
898 if ( (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeRGB
== GL_REPLACE
) &&
899 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeA
== GL_REPLACE
) &&
900 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftRGB
== 0) &&
901 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftA
== 0) &&
902 (nextunit
[i
] > 0) ) {
903 /* yippie! can optimize it away! */
908 /* need env instruction slot */
909 rmesa
->state
.envneeded
|= 1 << currslot
;
910 ok
= r200UpdateTextureEnv( ctx
, i
, currslot
, replaceunit
);
911 if (!ok
) return GL_FALSE
;
918 /* need one stage at least */
919 rmesa
->state
.texture
.unit
[0].outputreg
= 0;
920 ok
= r200UpdateTextureEnv( ctx
, 0, 0, 0 );
923 R200_STATECHANGE( rmesa
, ctx
);
924 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] &= ~(R200_TEX_BLEND_ENABLE_MASK
| R200_MULTI_PASS_ENABLE
);
925 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= rmesa
->state
.envneeded
<< R200_TEX_BLEND_0_ENABLE_SHIFT
;
934 #define TEXOBJ_TXFILTER_MASK (R200_MAX_MIP_LEVEL_MASK | \
935 R200_MIN_FILTER_MASK | \
936 R200_MAG_FILTER_MASK | \
937 R200_MAX_ANISO_MASK | \
939 R200_YUV_TEMPERATURE_MASK | \
940 R200_CLAMP_S_MASK | \
941 R200_CLAMP_T_MASK | \
942 R200_BORDER_MODE_D3D )
944 #define TEXOBJ_TXFORMAT_MASK (R200_TXFORMAT_WIDTH_MASK | \
945 R200_TXFORMAT_HEIGHT_MASK | \
946 R200_TXFORMAT_FORMAT_MASK | \
947 R200_TXFORMAT_F5_WIDTH_MASK | \
948 R200_TXFORMAT_F5_HEIGHT_MASK | \
949 R200_TXFORMAT_ALPHA_IN_MAP | \
950 R200_TXFORMAT_CUBIC_MAP_ENABLE | \
951 R200_TXFORMAT_NON_POWER2)
953 #define TEXOBJ_TXFORMAT_X_MASK (R200_DEPTH_LOG2_MASK | \
954 R200_TEXCOORD_MASK | \
955 R200_CLAMP_Q_MASK | \
956 R200_VOLUME_FILTER_MASK)
959 static void disable_tex_obj_state( r200ContextPtr rmesa
,
963 R200_STATECHANGE( rmesa
, vtx
);
964 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] &= ~(7 << (unit
* 3));
966 if (rmesa
->radeon
.TclFallback
& (R200_TCL_FALLBACK_TEXGEN_0
<<unit
)) {
967 TCL_FALLBACK( rmesa
->radeon
.glCtx
, (R200_TCL_FALLBACK_TEXGEN_0
<<unit
), GL_FALSE
);
970 /* Actually want to keep all units less than max active texture
971 * enabled, right? Fix this for >2 texunits.
975 GLuint tmp
= rmesa
->TexGenEnabled
;
977 rmesa
->TexGenEnabled
&= ~(R200_TEXGEN_TEXMAT_0_ENABLE
<<unit
);
978 rmesa
->TexGenEnabled
&= ~(R200_TEXMAT_0_ENABLE
<<unit
);
979 rmesa
->TexGenNeedNormals
[unit
] = GL_FALSE
;
980 rmesa
->TexGenCompSel
&= ~(R200_OUTPUT_TEX_0
<< unit
);
982 if (tmp
!= rmesa
->TexGenEnabled
) {
983 rmesa
->recheck_texgen
[unit
] = GL_TRUE
;
984 rmesa
->radeon
.NewGLState
|= _NEW_TEXTURE_MATRIX
;
988 static void import_tex_obj_state( r200ContextPtr rmesa
,
990 radeonTexObjPtr texobj
)
992 /* do not use RADEON_DB_STATE to avoid stale texture caches */
993 GLuint
*cmd
= &rmesa
->hw
.tex
[unit
].cmd
[TEX_CMD_0
];
995 R200_STATECHANGE( rmesa
, tex
[unit
] );
997 cmd
[TEX_PP_TXFILTER
] &= ~TEXOBJ_TXFILTER_MASK
;
998 cmd
[TEX_PP_TXFILTER
] |= texobj
->pp_txfilter
& TEXOBJ_TXFILTER_MASK
;
999 cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1000 cmd
[TEX_PP_TXFORMAT
] |= texobj
->pp_txformat
& TEXOBJ_TXFORMAT_MASK
;
1001 cmd
[TEX_PP_TXFORMAT_X
] &= ~TEXOBJ_TXFORMAT_X_MASK
;
1002 cmd
[TEX_PP_TXFORMAT_X
] |= texobj
->pp_txformat_x
& TEXOBJ_TXFORMAT_X_MASK
;
1003 cmd
[TEX_PP_TXSIZE
] = texobj
->pp_txsize
; /* NPOT only! */
1004 cmd
[TEX_PP_TXPITCH
] = texobj
->pp_txpitch
; /* NPOT only! */
1005 cmd
[TEX_PP_BORDER_COLOR
] = texobj
->pp_border_color
;
1006 if (rmesa
->radeon
.radeonScreen
->drmSupportsFragShader
) {
1007 cmd
[TEX_PP_TXOFFSET_NEWDRM
] = texobj
->pp_txoffset
;
1010 cmd
[TEX_PP_TXOFFSET_OLDDRM
] = texobj
->pp_txoffset
;
1013 if (texobj
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
1014 GLuint
*cube_cmd
= &rmesa
->hw
.cube
[unit
].cmd
[CUBE_CMD_0
];
1015 // GLuint bytesPerFace = texobj->base.totalSize / 6;
1016 // ASSERT(texobj->base.totalSize % 6 == 0);
1017 GLuint bytesPerFace
= 1; // TODO
1019 R200_STATECHANGE( rmesa
, cube
[unit
] );
1020 cube_cmd
[CUBE_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
1021 if (rmesa
->radeon
.radeonScreen
->drmSupportsFragShader
) {
1022 /* that value is submitted twice. could change cube atom
1023 to not include that command when new drm is used */
1024 cmd
[TEX_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
1026 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F1
] = texobj
->pp_txoffset
+ 1 * bytesPerFace
;
1027 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F2
] = texobj
->pp_txoffset
+ 2 * bytesPerFace
;
1028 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F3
] = texobj
->pp_txoffset
+ 3 * bytesPerFace
;
1029 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F4
] = texobj
->pp_txoffset
+ 4 * bytesPerFace
;
1030 cube_cmd
[CUBE_PP_CUBIC_OFFSET_F5
] = texobj
->pp_txoffset
+ 5 * bytesPerFace
;
1033 texobj
->dirty_state
&= ~(1<<unit
);
1036 static void set_texgen_matrix( r200ContextPtr rmesa
,
1038 const GLfloat
*s_plane
,
1039 const GLfloat
*t_plane
,
1040 const GLfloat
*r_plane
,
1041 const GLfloat
*q_plane
)
1065 _math_matrix_loadf( &(rmesa
->TexGenMatrix
[unit
]), m
);
1066 _math_matrix_analyse( &(rmesa
->TexGenMatrix
[unit
]) );
1067 rmesa
->TexGenEnabled
|= R200_TEXMAT_0_ENABLE
<<unit
;
1071 static GLuint
r200_need_dis_texgen(const GLbitfield texGenEnabled
,
1072 const GLfloat
*planeS
,
1073 const GLfloat
*planeT
,
1074 const GLfloat
*planeR
,
1075 const GLfloat
*planeQ
)
1077 GLuint needtgenable
= 0;
1079 if (!(texGenEnabled
& S_BIT
)) {
1080 if (((texGenEnabled
& T_BIT
) && planeT
[0] != 0.0) ||
1081 ((texGenEnabled
& R_BIT
) && planeR
[0] != 0.0) ||
1082 ((texGenEnabled
& Q_BIT
) && planeQ
[0] != 0.0)) {
1083 needtgenable
|= S_BIT
;
1086 if (!(texGenEnabled
& T_BIT
)) {
1087 if (((texGenEnabled
& S_BIT
) && planeS
[1] != 0.0) ||
1088 ((texGenEnabled
& R_BIT
) && planeR
[1] != 0.0) ||
1089 ((texGenEnabled
& Q_BIT
) && planeQ
[1] != 0.0)) {
1090 needtgenable
|= T_BIT
;
1093 if (!(texGenEnabled
& R_BIT
)) {
1094 if (((texGenEnabled
& S_BIT
) && planeS
[2] != 0.0) ||
1095 ((texGenEnabled
& T_BIT
) && planeT
[2] != 0.0) ||
1096 ((texGenEnabled
& Q_BIT
) && planeQ
[2] != 0.0)) {
1097 needtgenable
|= R_BIT
;
1100 if (!(texGenEnabled
& Q_BIT
)) {
1101 if (((texGenEnabled
& S_BIT
) && planeS
[3] != 0.0) ||
1102 ((texGenEnabled
& T_BIT
) && planeT
[3] != 0.0) ||
1103 ((texGenEnabled
& R_BIT
) && planeR
[3] != 0.0)) {
1104 needtgenable
|= Q_BIT
;
1108 return needtgenable
;
1113 * Returns GL_FALSE if fallback required.
1115 static GLboolean
r200_validate_texgen( GLcontext
*ctx
, GLuint unit
)
1117 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1118 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[unit
];
1119 GLuint inputshift
= R200_TEXGEN_0_INPUT_SHIFT
+ unit
*4;
1122 GLboolean mixed_fallback
= GL_FALSE
;
1123 static const GLfloat I
[16] = {
1128 static const GLfloat reflect
[16] = {
1134 rmesa
->TexGenCompSel
&= ~(R200_OUTPUT_TEX_0
<< unit
);
1135 rmesa
->TexGenEnabled
&= ~(R200_TEXGEN_TEXMAT_0_ENABLE
<<unit
);
1136 rmesa
->TexGenEnabled
&= ~(R200_TEXMAT_0_ENABLE
<<unit
);
1137 rmesa
->TexGenNeedNormals
[unit
] = GL_FALSE
;
1138 tgi
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] & ~(R200_TEXGEN_INPUT_MASK
<<
1140 tgcm
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] & ~(R200_TEXGEN_COMP_MASK
<<
1144 fprintf(stderr
, "%s unit %d\n", __FUNCTION__
, unit
);
1146 if (texUnit
->TexGenEnabled
& S_BIT
) {
1147 mode
= texUnit
->GenModeS
;
1149 tgcm
|= R200_TEXGEN_COMP_S
<< (unit
* 4);
1152 if (texUnit
->TexGenEnabled
& T_BIT
) {
1153 if (texUnit
->GenModeT
!= mode
)
1154 mixed_fallback
= GL_TRUE
;
1156 tgcm
|= R200_TEXGEN_COMP_T
<< (unit
* 4);
1158 if (texUnit
->TexGenEnabled
& R_BIT
) {
1159 if (texUnit
->GenModeR
!= mode
)
1160 mixed_fallback
= GL_TRUE
;
1162 tgcm
|= R200_TEXGEN_COMP_R
<< (unit
* 4);
1165 if (texUnit
->TexGenEnabled
& Q_BIT
) {
1166 if (texUnit
->GenModeQ
!= mode
)
1167 mixed_fallback
= GL_TRUE
;
1169 tgcm
|= R200_TEXGEN_COMP_Q
<< (unit
* 4);
1172 if (mixed_fallback
) {
1173 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1174 fprintf(stderr
, "fallback mixed texgen, 0x%x (0x%x 0x%x 0x%x 0x%x)\n",
1175 texUnit
->TexGenEnabled
, texUnit
->GenModeS
, texUnit
->GenModeT
,
1176 texUnit
->GenModeR
, texUnit
->GenModeQ
);
1180 /* we CANNOT do mixed mode if the texgen mode requires a plane where the input
1181 is not enabled for texgen, since the planes are concatenated into texmat,
1182 and thus the input will come from texcoord rather than tex gen equation!
1183 Either fallback or just hope that those texcoords aren't really needed...
1184 Assuming the former will cause lots of unnecessary fallbacks, the latter will
1185 generate bogus results sometimes - it's pretty much impossible to really know
1186 when a fallback is needed, depends on texmat and what sort of texture is bound
1187 etc, - for now fallback if we're missing either S or T bits, there's a high
1188 probability we need the texcoords in that case.
1189 That's a lot of work for some obscure texgen mixed mode fixup - why oh why
1190 doesn't the chip just directly accept the plane parameters :-(. */
1192 case GL_OBJECT_LINEAR
: {
1193 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1194 texUnit
->ObjectPlaneS
, texUnit
->ObjectPlaneT
,
1195 texUnit
->ObjectPlaneR
, texUnit
->ObjectPlaneQ
);
1196 if (needtgenable
& (S_BIT
| T_BIT
)) {
1197 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1198 fprintf(stderr
, "fallback mixed texgen / obj plane, 0x%x\n",
1199 texUnit
->TexGenEnabled
);
1202 if (needtgenable
& (R_BIT
)) {
1203 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1205 if (needtgenable
& (Q_BIT
)) {
1206 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1209 tgi
|= R200_TEXGEN_INPUT_OBJ
<< inputshift
;
1210 set_texgen_matrix( rmesa
, unit
,
1211 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->ObjectPlaneS
: I
,
1212 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->ObjectPlaneT
: I
+ 4,
1213 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->ObjectPlaneR
: I
+ 8,
1214 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->ObjectPlaneQ
: I
+ 12);
1218 case GL_EYE_LINEAR
: {
1219 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1220 texUnit
->EyePlaneS
, texUnit
->EyePlaneT
,
1221 texUnit
->EyePlaneR
, texUnit
->EyePlaneQ
);
1222 if (needtgenable
& (S_BIT
| T_BIT
)) {
1223 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1224 fprintf(stderr
, "fallback mixed texgen / eye plane, 0x%x\n",
1225 texUnit
->TexGenEnabled
);
1228 if (needtgenable
& (R_BIT
)) {
1229 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1231 if (needtgenable
& (Q_BIT
)) {
1232 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1234 tgi
|= R200_TEXGEN_INPUT_EYE
<< inputshift
;
1235 set_texgen_matrix( rmesa
, unit
,
1236 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->EyePlaneS
: I
,
1237 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->EyePlaneT
: I
+ 4,
1238 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->EyePlaneR
: I
+ 8,
1239 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->EyePlaneQ
: I
+ 12);
1243 case GL_REFLECTION_MAP_NV
:
1244 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1245 tgi
|= R200_TEXGEN_INPUT_EYE_REFLECT
<< inputshift
;
1246 /* pretty weird, must only negate when lighting is enabled? */
1247 if (ctx
->Light
.Enabled
)
1248 set_texgen_matrix( rmesa
, unit
,
1249 (texUnit
->TexGenEnabled
& S_BIT
) ? reflect
: I
,
1250 (texUnit
->TexGenEnabled
& T_BIT
) ? reflect
+ 4 : I
+ 4,
1251 (texUnit
->TexGenEnabled
& R_BIT
) ? reflect
+ 8 : I
+ 8,
1255 case GL_NORMAL_MAP_NV
:
1256 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1257 tgi
|= R200_TEXGEN_INPUT_EYE_NORMAL
<<inputshift
;
1261 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1262 tgi
|= R200_TEXGEN_INPUT_SPHERE
<<inputshift
;
1266 /* All texgen units were disabled, so just pass coords through. */
1267 tgi
|= unit
<< inputshift
;
1271 /* Unsupported mode, fallback:
1273 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1274 fprintf(stderr
, "fallback unsupported texgen, %d\n",
1279 rmesa
->TexGenEnabled
|= R200_TEXGEN_TEXMAT_0_ENABLE
<< unit
;
1280 rmesa
->TexGenCompSel
|= R200_OUTPUT_TEX_0
<< unit
;
1282 if (tgi
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] ||
1283 tgcm
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
])
1285 R200_STATECHANGE(rmesa
, tcg
);
1286 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] = tgi
;
1287 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] = tgcm
;
1293 void set_re_cntl_d3d( GLcontext
*ctx
, int unit
, GLboolean use_d3d
)
1295 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1299 re_cntl
= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] & ~(R200_VTX_STQ0_D3D
<< (2 * unit
));
1301 re_cntl
|= R200_VTX_STQ0_D3D
<< (2 * unit
);
1303 if ( re_cntl
!= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] ) {
1304 R200_STATECHANGE( rmesa
, set
);
1305 rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] = re_cntl
;
1310 * Compute the cached hardware register values for the given texture object.
1312 * \param rmesa Context pointer
1313 * \param t the r300 texture object
1315 static void setup_hardware_state(r200ContextPtr rmesa
, radeonTexObj
*t
)
1317 const struct gl_texture_image
*firstImage
=
1318 t
->base
.Image
[0][t
->mt
->firstLevel
];
1319 GLint log2Width
, log2Height
, log2Depth
, texelBytes
;
1321 log2Width
= firstImage
->WidthLog2
;
1322 log2Height
= firstImage
->HeightLog2
;
1323 log2Depth
= firstImage
->DepthLog2
;
1324 texelBytes
= firstImage
->TexFormat
->TexelBytes
;
1327 if (!t
->image_override
) {
1328 if (VALID_FORMAT(firstImage
->TexFormat
->MesaFormat
)) {
1329 const struct tx_table
*table
= _mesa_little_endian() ? tx_table_le
:
1332 t
->pp_txformat
&= ~(R200_TXFORMAT_FORMAT_MASK
|
1333 R200_TXFORMAT_ALPHA_IN_MAP
);
1334 t
->pp_txfilter
&= ~R200_YUV_TO_RGB
;
1336 t
->pp_txformat
|= table
[ firstImage
->TexFormat
->MesaFormat
].format
;
1337 t
->pp_txfilter
|= table
[ firstImage
->TexFormat
->MesaFormat
].filter
;
1339 _mesa_problem(NULL
, "unexpected texture format in %s",
1345 t
->pp_txfilter
&= ~R200_MAX_MIP_LEVEL_MASK
;
1346 t
->pp_txfilter
|= (t
->mt
->lastLevel
- t
->mt
->firstLevel
) << R200_MAX_MIP_LEVEL_SHIFT
;
1348 t
->pp_txformat
&= ~(R200_TXFORMAT_WIDTH_MASK
|
1349 R200_TXFORMAT_HEIGHT_MASK
|
1350 R200_TXFORMAT_CUBIC_MAP_ENABLE
|
1351 R200_TXFORMAT_F5_WIDTH_MASK
|
1352 R200_TXFORMAT_F5_HEIGHT_MASK
);
1353 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_WIDTH_SHIFT
) |
1354 (log2Height
<< R200_TXFORMAT_HEIGHT_SHIFT
));
1358 t
->pp_txformat_x
&= ~(R200_DEPTH_LOG2_MASK
| R200_TEXCOORD_MASK
);
1359 if (t
->base
.Target
== GL_TEXTURE_3D
) {
1360 t
->pp_txformat_x
|= (log2Depth
<< R200_DEPTH_LOG2_SHIFT
);
1361 t
->pp_txformat_x
|= R200_TEXCOORD_VOLUME
;
1364 else if (t
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
1365 ASSERT(log2Width
== log2Height
);
1366 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_F5_WIDTH_SHIFT
) |
1367 (log2Height
<< R200_TXFORMAT_F5_HEIGHT_SHIFT
) |
1368 /* don't think we need this bit, if it exists at all - fglrx does not set it */
1369 (R200_TXFORMAT_CUBIC_MAP_ENABLE
));
1370 t
->pp_txformat_x
|= R200_TEXCOORD_CUBIC_ENV
;
1371 t
->pp_cubic_faces
= ((log2Width
<< R200_FACE_WIDTH_1_SHIFT
) |
1372 (log2Height
<< R200_FACE_HEIGHT_1_SHIFT
) |
1373 (log2Width
<< R200_FACE_WIDTH_2_SHIFT
) |
1374 (log2Height
<< R200_FACE_HEIGHT_2_SHIFT
) |
1375 (log2Width
<< R200_FACE_WIDTH_3_SHIFT
) |
1376 (log2Height
<< R200_FACE_HEIGHT_3_SHIFT
) |
1377 (log2Width
<< R200_FACE_WIDTH_4_SHIFT
) |
1378 (log2Height
<< R200_FACE_HEIGHT_4_SHIFT
));
1381 /* If we don't in fact send enough texture coordinates, q will be 1,
1382 * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
1384 t
->pp_txformat_x
|= R200_TEXCOORD_PROJ
;
1387 t
->pp_txsize
= (((firstImage
->Width
- 1) << R200_PP_TX_WIDTHMASK_SHIFT
)
1388 | ((firstImage
->Height
- 1) << R200_PP_TX_HEIGHTMASK_SHIFT
));
1390 if ( !t
->image_override
) {
1391 if (firstImage
->IsCompressed
)
1392 t
->pp_txpitch
= (firstImage
->Width
+ 63) & ~(63);
1394 t
->pp_txpitch
= ((firstImage
->Width
* texelBytes
) + 63) & ~(63);
1395 t
->pp_txpitch
-= 32;
1398 if (t
->base
.Target
== GL_TEXTURE_RECTANGLE_NV
) {
1399 t
->pp_txformat
|= R200_TXFORMAT_NON_POWER2
;
1404 static GLboolean
r200_validate_texture(GLcontext
*ctx
, struct gl_texture_object
*texObj
, int unit
)
1406 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1407 radeonTexObj
*t
= radeon_tex_obj(texObj
);
1409 if (!radeon_validate_texture_miptree(ctx
, texObj
))
1412 r200_validate_texgen(ctx
, unit
);
1413 /* Configure the hardware registers (more precisely, the cached version
1414 * of the hardware registers). */
1415 setup_hardware_state(rmesa
, t
);
1417 if (texObj
->Target
== GL_TEXTURE_RECTANGLE_NV
||
1418 texObj
->Target
== GL_TEXTURE_2D
||
1419 texObj
->Target
== GL_TEXTURE_1D
)
1420 set_re_cntl_d3d( ctx
, unit
, GL_FALSE
);
1422 set_re_cntl_d3d( ctx
, unit
, GL_TRUE
);
1423 R200_STATECHANGE( rmesa
, ctx
);
1424 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_0_ENABLE
<< unit
;
1426 R200_STATECHANGE( rmesa
, vtx
);
1427 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] &= ~(7 << (unit
* 3));
1428 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] |= 4 << (unit
* 3);
1430 rmesa
->recheck_texgen
[unit
] = GL_TRUE
;
1431 if (t
->dirty_state
& (1<<unit
)) {
1432 import_tex_obj_state( rmesa
, unit
, t
);
1435 if (rmesa
->recheck_texgen
[unit
]) {
1436 GLboolean fallback
= !r200_validate_texgen( ctx
, unit
);
1437 TCL_FALLBACK( ctx
, (R200_TCL_FALLBACK_TEXGEN_0
<<unit
), fallback
);
1438 rmesa
->recheck_texgen
[unit
] = 0;
1439 rmesa
->radeon
.NewGLState
|= _NEW_TEXTURE_MATRIX
;
1442 t
->dirty_state
= R200_TEX_ALL
;
1444 t
->validated
= GL_TRUE
;
1446 FALLBACK( rmesa
, RADEON_FALLBACK_BORDER_MODE
, t
->border_fallback
);
1448 return !t
->border_fallback
;
1451 static GLboolean
r200UpdateTextureUnit(GLcontext
*ctx
, int unit
)
1453 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1454 GLuint unitneeded
= rmesa
->state
.texture
.unit
[unit
].unitneeded
;
1457 /* disable the unit */
1458 disable_tex_obj_state(rmesa
, unit
);
1462 if (!r200_validate_texture(ctx
, ctx
->Texture
.Unit
[unit
]._Current
, unit
)) {
1464 "failed to validate texture for unit %d.\n",
1466 rmesa
->state
.texture
.unit
[unit
].texobj
= NULL
;
1470 rmesa
->state
.texture
.unit
[unit
].texobj
= radeon_tex_obj(ctx
->Texture
.Unit
[unit
]._Current
);
1475 void r200UpdateTextureState( GLcontext
*ctx
)
1477 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1481 /* NOTE: must not manipulate rmesa->state.texture.unit[].unitneeded or
1482 rmesa->state.envneeded before a R200_STATECHANGE (or R200_NEWPRIM) since
1483 we use these to determine if we want to emit the corresponding state
1485 R200_NEWPRIM( rmesa
);
1487 if (ctx
->ATIFragmentShader
._Enabled
) {
1489 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
1490 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
1495 ok
= r200UpdateAllTexEnv( ctx
);
1498 ok
= (r200UpdateTextureUnit( ctx
, 0 ) &&
1499 r200UpdateTextureUnit( ctx
, 1 ) &&
1500 r200UpdateTextureUnit( ctx
, 2 ) &&
1501 r200UpdateTextureUnit( ctx
, 3 ) &&
1502 r200UpdateTextureUnit( ctx
, 4 ) &&
1503 r200UpdateTextureUnit( ctx
, 5 ));
1506 if (ok
&& ctx
->ATIFragmentShader
._Enabled
) {
1507 r200UpdateFragmentShader(ctx
);
1510 FALLBACK( rmesa
, R200_FALLBACK_TEXTURE
, !ok
);
1512 if (rmesa
->radeon
.TclFallback
)
1513 r200ChooseVertexState( ctx
);
1516 if (rmesa
->radeon
.radeonScreen
->chip_family
== CHIP_FAMILY_R200
) {
1519 * T0 hang workaround -------------
1520 * not needed for r200 derivatives
1522 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_ENABLE_MASK
) == R200_TEX_0_ENABLE
&&
1523 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1525 R200_STATECHANGE(rmesa
, ctx
);
1526 R200_STATECHANGE(rmesa
, tex
[1]);
1527 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_1_ENABLE
;
1528 if (!(rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_1_ENABLE
))
1529 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1530 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] |= R200_TXFORMAT_LOOKUP_DISABLE
;
1532 else if (!ctx
->ATIFragmentShader
._Enabled
) {
1533 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
) &&
1534 (rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] & R200_TXFORMAT_LOOKUP_DISABLE
)) {
1535 R200_STATECHANGE(rmesa
, tex
[1]);
1536 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~R200_TXFORMAT_LOOKUP_DISABLE
;
1539 /* do the same workaround for the first pass of a fragment shader.
1540 * completely unknown if necessary / sufficient.
1542 if ((rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_ENABLE_MASK
) == R200_PPX_TEX_0_ENABLE
&&
1543 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1545 R200_STATECHANGE(rmesa
, cst
);
1546 R200_STATECHANGE(rmesa
, tex
[1]);
1547 rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] |= R200_PPX_TEX_1_ENABLE
;
1548 if (!(rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
))
1549 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1550 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXMULTI_CTL
] |= R200_PASS1_TXFORMAT_LOOKUP_DISABLE
;
1553 /* maybe needs to be done pairwise due to 2 parallel (physical) tex units ?
1554 looks like that's not the case, if 8500/9100 owners don't complain remove this...
1555 for ( i = 0; i < ctx->Const.MaxTextureUnits; i += 2) {
1556 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & ((R200_TEX_0_ENABLE |
1557 R200_TEX_1_ENABLE ) << i)) == (R200_TEX_0_ENABLE << i)) &&
1558 ((rmesa->hw.tex[i].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK) >
1559 R200_MIN_FILTER_LINEAR)) {
1560 R200_STATECHANGE(rmesa, ctx);
1561 R200_STATECHANGE(rmesa, tex[i+1]);
1562 rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= (R200_TEX_1_ENABLE << i);
1563 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
1564 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] |= 0x08000000;
1567 if ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE << i)) &&
1568 (rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] & 0x08000000)) {
1569 R200_STATECHANGE(rmesa, tex[i+1]);
1570 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~0x08000000;
1576 * Texture cache LRU hang workaround -------------
1577 * not needed for r200 derivatives
1578 * hopefully this covers first pass of a shader as well
1581 /* While the cases below attempt to only enable the workaround in the
1582 * specific cases necessary, they were insufficient. See bugzilla #1519,
1583 * #729, #814. Tests with quake3 showed no impact on performance.
1588 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE )) &&
1589 ((((rmesa->hw.tex[0].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1591 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_2_ENABLE) &&
1592 ((((rmesa->hw.tex[2].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1594 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_4_ENABLE) &&
1595 ((((rmesa->hw.tex[4].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1601 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE )) &&
1602 ((((rmesa->hw.tex[1].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1604 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_3_ENABLE) &&
1605 ((((rmesa->hw.tex[3].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1607 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_5_ENABLE) &&
1608 ((((rmesa->hw.tex[5].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1614 if (dbg
!= rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
]) {
1615 R200_STATECHANGE( rmesa
, tam
);
1616 rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
] = dbg
;
1617 if (0) printf("TEXCACHE LRU HANG WORKAROUND %x\n", dbg
);