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/teximage.h"
41 #include "main/texobj.h"
42 #include "main/enums.h"
44 #include "radeon_common.h"
45 #include "radeon_mipmap_tree.h"
46 #include "r200_context.h"
47 #include "r200_state.h"
48 #include "r200_ioctl.h"
49 #include "r200_swtcl.h"
54 #define R200_TXFORMAT_A8 R200_TXFORMAT_I8
55 #define R200_TXFORMAT_L8 R200_TXFORMAT_I8
56 #define R200_TXFORMAT_AL88 R200_TXFORMAT_AI88
57 #define R200_TXFORMAT_YCBCR R200_TXFORMAT_YVYU422
58 #define R200_TXFORMAT_YCBCR_REV R200_TXFORMAT_VYUY422
59 #define R200_TXFORMAT_RGB_DXT1 R200_TXFORMAT_DXT1
60 #define R200_TXFORMAT_RGBA_DXT1 R200_TXFORMAT_DXT1
61 #define R200_TXFORMAT_RGBA_DXT3 R200_TXFORMAT_DXT23
62 #define R200_TXFORMAT_RGBA_DXT5 R200_TXFORMAT_DXT45
65 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f, 0 }
66 #define _COLOR_REV(f) \
67 [ MESA_FORMAT_ ## f ## _REV ] = { R200_TXFORMAT_ ## f, 0 }
69 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f | R200_TXFORMAT_ALPHA_IN_MAP, 0 }
70 #define _ALPHA_REV(f) \
71 [ MESA_FORMAT_ ## f ## _REV ] = { R200_TXFORMAT_ ## f | R200_TXFORMAT_ALPHA_IN_MAP, 0 }
73 [ MESA_FORMAT_ ## f ] = { R200_TXFORMAT_ ## f, R200_YUV_TO_RGB }
75 [ MESA_FORMAT_ ## f ] = { 0xffffffff, 0 }
76 #define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5) \
77 && (tx_table_be[f].format != 0xffffffff) )
80 GLuint format
, filter
;
83 static const struct tx_table tx_table_be
[] =
85 [ MESA_FORMAT_RGBA8888
] = { R200_TXFORMAT_ABGR8888
| R200_TXFORMAT_ALPHA_IN_MAP
, 0 },
112 static const struct tx_table tx_table_le
[] =
115 [ MESA_FORMAT_RGBA8888_REV
] = { R200_TXFORMAT_ABGR8888
| R200_TXFORMAT_ALPHA_IN_MAP
, 0 },
117 _ALPHA_REV(ARGB8888
),
118 [ MESA_FORMAT_RGB888
] = { R200_TXFORMAT_ARGB8888
, 0 },
122 _ALPHA_REV(ARGB4444
),
124 _ALPHA_REV(ARGB1555
),
145 /* ================================================================
146 * Texture combine functions
149 /* GL_ARB_texture_env_combine support
152 /* The color tables have combine functions for GL_SRC_COLOR,
153 * GL_ONE_MINUS_SRC_COLOR, GL_SRC_ALPHA and GL_ONE_MINUS_SRC_ALPHA.
155 static GLuint r200_register_color
[][R200_MAX_TEXTURE_UNITS
] =
158 R200_TXC_ARG_A_R0_COLOR
,
159 R200_TXC_ARG_A_R1_COLOR
,
160 R200_TXC_ARG_A_R2_COLOR
,
161 R200_TXC_ARG_A_R3_COLOR
,
162 R200_TXC_ARG_A_R4_COLOR
,
163 R200_TXC_ARG_A_R5_COLOR
166 R200_TXC_ARG_A_R0_COLOR
| R200_TXC_COMP_ARG_A
,
167 R200_TXC_ARG_A_R1_COLOR
| R200_TXC_COMP_ARG_A
,
168 R200_TXC_ARG_A_R2_COLOR
| R200_TXC_COMP_ARG_A
,
169 R200_TXC_ARG_A_R3_COLOR
| R200_TXC_COMP_ARG_A
,
170 R200_TXC_ARG_A_R4_COLOR
| R200_TXC_COMP_ARG_A
,
171 R200_TXC_ARG_A_R5_COLOR
| R200_TXC_COMP_ARG_A
174 R200_TXC_ARG_A_R0_ALPHA
,
175 R200_TXC_ARG_A_R1_ALPHA
,
176 R200_TXC_ARG_A_R2_ALPHA
,
177 R200_TXC_ARG_A_R3_ALPHA
,
178 R200_TXC_ARG_A_R4_ALPHA
,
179 R200_TXC_ARG_A_R5_ALPHA
182 R200_TXC_ARG_A_R0_ALPHA
| R200_TXC_COMP_ARG_A
,
183 R200_TXC_ARG_A_R1_ALPHA
| R200_TXC_COMP_ARG_A
,
184 R200_TXC_ARG_A_R2_ALPHA
| R200_TXC_COMP_ARG_A
,
185 R200_TXC_ARG_A_R3_ALPHA
| R200_TXC_COMP_ARG_A
,
186 R200_TXC_ARG_A_R4_ALPHA
| R200_TXC_COMP_ARG_A
,
187 R200_TXC_ARG_A_R5_ALPHA
| R200_TXC_COMP_ARG_A
191 static GLuint r200_tfactor_color
[] =
193 R200_TXC_ARG_A_TFACTOR_COLOR
,
194 R200_TXC_ARG_A_TFACTOR_COLOR
| R200_TXC_COMP_ARG_A
,
195 R200_TXC_ARG_A_TFACTOR_ALPHA
,
196 R200_TXC_ARG_A_TFACTOR_ALPHA
| R200_TXC_COMP_ARG_A
199 static GLuint r200_tfactor1_color
[] =
201 R200_TXC_ARG_A_TFACTOR1_COLOR
,
202 R200_TXC_ARG_A_TFACTOR1_COLOR
| R200_TXC_COMP_ARG_A
,
203 R200_TXC_ARG_A_TFACTOR1_ALPHA
,
204 R200_TXC_ARG_A_TFACTOR1_ALPHA
| R200_TXC_COMP_ARG_A
207 static GLuint r200_primary_color
[] =
209 R200_TXC_ARG_A_DIFFUSE_COLOR
,
210 R200_TXC_ARG_A_DIFFUSE_COLOR
| R200_TXC_COMP_ARG_A
,
211 R200_TXC_ARG_A_DIFFUSE_ALPHA
,
212 R200_TXC_ARG_A_DIFFUSE_ALPHA
| R200_TXC_COMP_ARG_A
215 /* GL_ZERO table - indices 0-3
216 * GL_ONE table - indices 1-4
218 static GLuint r200_zero_color
[] =
221 R200_TXC_ARG_A_ZERO
| R200_TXC_COMP_ARG_A
,
223 R200_TXC_ARG_A_ZERO
| R200_TXC_COMP_ARG_A
,
227 /* The alpha tables only have GL_SRC_ALPHA and GL_ONE_MINUS_SRC_ALPHA.
229 static GLuint r200_register_alpha
[][R200_MAX_TEXTURE_UNITS
] =
232 R200_TXA_ARG_A_R0_ALPHA
,
233 R200_TXA_ARG_A_R1_ALPHA
,
234 R200_TXA_ARG_A_R2_ALPHA
,
235 R200_TXA_ARG_A_R3_ALPHA
,
236 R200_TXA_ARG_A_R4_ALPHA
,
237 R200_TXA_ARG_A_R5_ALPHA
240 R200_TXA_ARG_A_R0_ALPHA
| R200_TXA_COMP_ARG_A
,
241 R200_TXA_ARG_A_R1_ALPHA
| R200_TXA_COMP_ARG_A
,
242 R200_TXA_ARG_A_R2_ALPHA
| R200_TXA_COMP_ARG_A
,
243 R200_TXA_ARG_A_R3_ALPHA
| R200_TXA_COMP_ARG_A
,
244 R200_TXA_ARG_A_R4_ALPHA
| R200_TXA_COMP_ARG_A
,
245 R200_TXA_ARG_A_R5_ALPHA
| R200_TXA_COMP_ARG_A
249 static GLuint r200_tfactor_alpha
[] =
251 R200_TXA_ARG_A_TFACTOR_ALPHA
,
252 R200_TXA_ARG_A_TFACTOR_ALPHA
| R200_TXA_COMP_ARG_A
255 static GLuint r200_tfactor1_alpha
[] =
257 R200_TXA_ARG_A_TFACTOR1_ALPHA
,
258 R200_TXA_ARG_A_TFACTOR1_ALPHA
| R200_TXA_COMP_ARG_A
261 static GLuint r200_primary_alpha
[] =
263 R200_TXA_ARG_A_DIFFUSE_ALPHA
,
264 R200_TXA_ARG_A_DIFFUSE_ALPHA
| R200_TXA_COMP_ARG_A
267 /* GL_ZERO table - indices 0-1
268 * GL_ONE table - indices 1-2
270 static GLuint r200_zero_alpha
[] =
273 R200_TXA_ARG_A_ZERO
| R200_TXA_COMP_ARG_A
,
278 /* Extract the arg from slot A, shift it into the correct argument slot
279 * and set the corresponding complement bit.
281 #define R200_COLOR_ARG( n, arg ) \
284 ((color_arg[n] & R200_TXC_ARG_A_MASK) \
285 << R200_TXC_ARG_##arg##_SHIFT); \
287 ((color_arg[n] >> R200_TXC_COMP_ARG_A_SHIFT) \
288 << R200_TXC_COMP_ARG_##arg##_SHIFT); \
291 #define R200_ALPHA_ARG( n, arg ) \
294 ((alpha_arg[n] & R200_TXA_ARG_A_MASK) \
295 << R200_TXA_ARG_##arg##_SHIFT); \
297 ((alpha_arg[n] >> R200_TXA_COMP_ARG_A_SHIFT) \
298 << R200_TXA_COMP_ARG_##arg##_SHIFT); \
302 /* ================================================================
303 * Texture unit state management
306 static GLboolean
r200UpdateTextureEnv( GLcontext
*ctx
, int unit
, int slot
, GLuint replaceargs
)
308 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
309 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[unit
];
310 GLuint color_combine
, alpha_combine
;
311 GLuint color_scale
= rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] &
312 ~(R200_TXC_SCALE_MASK
| R200_TXC_OUTPUT_REG_MASK
| R200_TXC_TFACTOR_SEL_MASK
|
313 R200_TXC_TFACTOR1_SEL_MASK
);
314 GLuint alpha_scale
= rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] &
315 ~(R200_TXA_DOT_ALPHA
| R200_TXA_SCALE_MASK
| R200_TXA_OUTPUT_REG_MASK
|
316 R200_TXA_TFACTOR_SEL_MASK
| R200_TXA_TFACTOR1_SEL_MASK
);
318 /* texUnit->_Current can be NULL if and only if the texture unit is
319 * not actually enabled.
321 assert( (texUnit
->_ReallyEnabled
== 0)
322 || (texUnit
->_Current
!= NULL
) );
324 if ( R200_DEBUG
& DEBUG_TEXTURE
) {
325 fprintf( stderr
, "%s( %p, %d )\n", __FUNCTION__
, (void *)ctx
, unit
);
328 /* Set the texture environment state. Isn't this nice and clean?
329 * The chip will automagically set the texture alpha to 0xff when
330 * the texture format does not include an alpha component. This
331 * reduces the amount of special-casing we have to do, alpha-only
332 * textures being a notable exception.
335 color_scale
|= ((rmesa
->state
.texture
.unit
[unit
].outputreg
+ 1) << R200_TXC_OUTPUT_REG_SHIFT
) |
336 (unit
<< R200_TXC_TFACTOR_SEL_SHIFT
) |
337 (replaceargs
<< R200_TXC_TFACTOR1_SEL_SHIFT
);
338 alpha_scale
|= ((rmesa
->state
.texture
.unit
[unit
].outputreg
+ 1) << R200_TXA_OUTPUT_REG_SHIFT
) |
339 (unit
<< R200_TXA_TFACTOR_SEL_SHIFT
) |
340 (replaceargs
<< R200_TXA_TFACTOR1_SEL_SHIFT
);
342 if ( !texUnit
->_ReallyEnabled
) {
344 color_combine
= R200_TXC_ARG_A_ZERO
| R200_TXC_ARG_B_ZERO
345 | R200_TXC_ARG_C_DIFFUSE_COLOR
| R200_TXC_OP_MADD
;
346 alpha_combine
= R200_TXA_ARG_A_ZERO
| R200_TXA_ARG_B_ZERO
347 | R200_TXA_ARG_C_DIFFUSE_ALPHA
| R200_TXA_OP_MADD
;
350 GLuint color_arg
[3], alpha_arg
[3];
352 const GLuint numColorArgs
= texUnit
->_CurrentCombine
->_NumArgsRGB
;
353 const GLuint numAlphaArgs
= texUnit
->_CurrentCombine
->_NumArgsA
;
354 GLuint RGBshift
= texUnit
->_CurrentCombine
->ScaleShiftRGB
;
355 GLuint Ashift
= texUnit
->_CurrentCombine
->ScaleShiftA
;
358 const GLint replaceoprgb
=
359 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->OperandRGB
[0] - GL_SRC_COLOR
;
360 const GLint replaceopa
=
361 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->OperandA
[0] - GL_SRC_ALPHA
;
364 * Extract the color and alpha combine function arguments.
366 for ( i
= 0 ; i
< numColorArgs
; i
++ ) {
367 GLint op
= texUnit
->_CurrentCombine
->OperandRGB
[i
] - GL_SRC_COLOR
;
368 const GLint srcRGBi
= texUnit
->_CurrentCombine
->SourceRGB
[i
];
373 color_arg
[i
] = r200_register_color
[op
][unit
];
376 color_arg
[i
] = r200_tfactor_color
[op
];
378 case GL_PRIMARY_COLOR
:
379 color_arg
[i
] = r200_primary_color
[op
];
382 if (replaceargs
!= unit
) {
383 const GLint srcRGBreplace
=
384 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->SourceRGB
[0];
386 op
= op
^ replaceopa
;
389 op
= op
^ replaceoprgb
;
391 switch (srcRGBreplace
) {
393 color_arg
[i
] = r200_register_color
[op
][replaceargs
];
396 color_arg
[i
] = r200_tfactor1_color
[op
];
398 case GL_PRIMARY_COLOR
:
399 color_arg
[i
] = r200_primary_color
[op
];
403 color_arg
[i
] = r200_primary_color
[op
];
405 color_arg
[i
] = r200_register_color
[op
]
406 [rmesa
->state
.texture
.unit
[replaceargs
- 1].outputreg
];
409 color_arg
[i
] = r200_zero_color
[op
];
412 color_arg
[i
] = r200_zero_color
[op
+1];
420 color_arg
[i
] = r200_register_color
[op
][srcRGBreplace
- GL_TEXTURE0
];
428 color_arg
[i
] = r200_primary_color
[op
];
430 color_arg
[i
] = r200_register_color
[op
]
431 [rmesa
->state
.texture
.unit
[unit
- 1].outputreg
];
435 color_arg
[i
] = r200_zero_color
[op
];
438 color_arg
[i
] = r200_zero_color
[op
+1];
446 color_arg
[i
] = r200_register_color
[op
][srcRGBi
- GL_TEXTURE0
];
453 for ( i
= 0 ; i
< numAlphaArgs
; i
++ ) {
454 GLint op
= texUnit
->_CurrentCombine
->OperandA
[i
] - GL_SRC_ALPHA
;
455 const GLint srcAi
= texUnit
->_CurrentCombine
->SourceA
[i
];
460 alpha_arg
[i
] = r200_register_alpha
[op
][unit
];
463 alpha_arg
[i
] = r200_tfactor_alpha
[op
];
465 case GL_PRIMARY_COLOR
:
466 alpha_arg
[i
] = r200_primary_alpha
[op
];
469 if (replaceargs
!= unit
) {
470 const GLint srcAreplace
=
471 ctx
->Texture
.Unit
[replaceargs
]._CurrentCombine
->SourceA
[0];
472 op
= op
^ replaceopa
;
473 switch (srcAreplace
) {
475 alpha_arg
[i
] = r200_register_alpha
[op
][replaceargs
];
478 alpha_arg
[i
] = r200_tfactor1_alpha
[op
];
480 case GL_PRIMARY_COLOR
:
481 alpha_arg
[i
] = r200_primary_alpha
[op
];
485 alpha_arg
[i
] = r200_primary_alpha
[op
];
487 alpha_arg
[i
] = r200_register_alpha
[op
]
488 [rmesa
->state
.texture
.unit
[replaceargs
- 1].outputreg
];
491 alpha_arg
[i
] = r200_zero_alpha
[op
];
494 alpha_arg
[i
] = r200_zero_alpha
[op
+1];
502 alpha_arg
[i
] = r200_register_alpha
[op
][srcAreplace
- GL_TEXTURE0
];
510 alpha_arg
[i
] = r200_primary_alpha
[op
];
512 alpha_arg
[i
] = r200_register_alpha
[op
]
513 [rmesa
->state
.texture
.unit
[unit
- 1].outputreg
];
517 alpha_arg
[i
] = r200_zero_alpha
[op
];
520 alpha_arg
[i
] = r200_zero_alpha
[op
+1];
528 alpha_arg
[i
] = r200_register_alpha
[op
][srcAi
- GL_TEXTURE0
];
536 * Build up the color and alpha combine functions.
538 switch ( texUnit
->_CurrentCombine
->ModeRGB
) {
540 color_combine
= (R200_TXC_ARG_A_ZERO
|
541 R200_TXC_ARG_B_ZERO
|
543 R200_COLOR_ARG( 0, C
);
546 color_combine
= (R200_TXC_ARG_C_ZERO
|
548 R200_COLOR_ARG( 0, A
);
549 R200_COLOR_ARG( 1, B
);
552 color_combine
= (R200_TXC_ARG_B_ZERO
|
553 R200_TXC_COMP_ARG_B
|
555 R200_COLOR_ARG( 0, A
);
556 R200_COLOR_ARG( 1, C
);
559 color_combine
= (R200_TXC_ARG_B_ZERO
|
560 R200_TXC_COMP_ARG_B
|
561 R200_TXC_BIAS_ARG_C
| /* new */
562 R200_TXC_OP_MADD
); /* was ADDSIGNED */
563 R200_COLOR_ARG( 0, A
);
564 R200_COLOR_ARG( 1, C
);
567 color_combine
= (R200_TXC_ARG_B_ZERO
|
568 R200_TXC_COMP_ARG_B
|
571 R200_COLOR_ARG( 0, A
);
572 R200_COLOR_ARG( 1, C
);
575 color_combine
= (R200_TXC_OP_LERP
);
576 R200_COLOR_ARG( 0, B
);
577 R200_COLOR_ARG( 1, A
);
578 R200_COLOR_ARG( 2, C
);
581 case GL_DOT3_RGB_EXT
:
582 case GL_DOT3_RGBA_EXT
:
583 /* The EXT version of the DOT3 extension does not support the
584 * scale factor, but the ARB version (and the version in OpenGL
592 /* DOT3 works differently on R200 than on R100. On R100, just
593 * setting the DOT3 mode did everything for you. On R200, the
594 * driver has to enable the biasing and scale in the inputs to
595 * put them in the proper [-1,1] range. This is what the 4x and
596 * the -0.5 in the DOT3 spec do. The post-scale is then set
600 color_combine
= (R200_TXC_ARG_C_ZERO
|
602 R200_TXC_BIAS_ARG_A
|
603 R200_TXC_BIAS_ARG_B
|
604 R200_TXC_SCALE_ARG_A
|
605 R200_TXC_SCALE_ARG_B
);
606 R200_COLOR_ARG( 0, A
);
607 R200_COLOR_ARG( 1, B
);
610 case GL_MODULATE_ADD_ATI
:
611 color_combine
= (R200_TXC_OP_MADD
);
612 R200_COLOR_ARG( 0, A
);
613 R200_COLOR_ARG( 1, C
);
614 R200_COLOR_ARG( 2, B
);
616 case GL_MODULATE_SIGNED_ADD_ATI
:
617 color_combine
= (R200_TXC_BIAS_ARG_C
| /* new */
618 R200_TXC_OP_MADD
); /* was ADDSIGNED */
619 R200_COLOR_ARG( 0, A
);
620 R200_COLOR_ARG( 1, C
);
621 R200_COLOR_ARG( 2, B
);
623 case GL_MODULATE_SUBTRACT_ATI
:
624 color_combine
= (R200_TXC_NEG_ARG_C
|
626 R200_COLOR_ARG( 0, A
);
627 R200_COLOR_ARG( 1, C
);
628 R200_COLOR_ARG( 2, B
);
634 switch ( texUnit
->_CurrentCombine
->ModeA
) {
636 alpha_combine
= (R200_TXA_ARG_A_ZERO
|
637 R200_TXA_ARG_B_ZERO
|
639 R200_ALPHA_ARG( 0, C
);
642 alpha_combine
= (R200_TXA_ARG_C_ZERO
|
644 R200_ALPHA_ARG( 0, A
);
645 R200_ALPHA_ARG( 1, B
);
648 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
649 R200_TXA_COMP_ARG_B
|
651 R200_ALPHA_ARG( 0, A
);
652 R200_ALPHA_ARG( 1, C
);
655 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
656 R200_TXA_COMP_ARG_B
|
657 R200_TXA_BIAS_ARG_C
| /* new */
658 R200_TXA_OP_MADD
); /* was ADDSIGNED */
659 R200_ALPHA_ARG( 0, A
);
660 R200_ALPHA_ARG( 1, C
);
663 alpha_combine
= (R200_TXA_ARG_B_ZERO
|
664 R200_TXA_COMP_ARG_B
|
667 R200_ALPHA_ARG( 0, A
);
668 R200_ALPHA_ARG( 1, C
);
671 alpha_combine
= (R200_TXA_OP_LERP
);
672 R200_ALPHA_ARG( 0, B
);
673 R200_ALPHA_ARG( 1, A
);
674 R200_ALPHA_ARG( 2, C
);
677 case GL_MODULATE_ADD_ATI
:
678 alpha_combine
= (R200_TXA_OP_MADD
);
679 R200_ALPHA_ARG( 0, A
);
680 R200_ALPHA_ARG( 1, C
);
681 R200_ALPHA_ARG( 2, B
);
683 case GL_MODULATE_SIGNED_ADD_ATI
:
684 alpha_combine
= (R200_TXA_BIAS_ARG_C
| /* new */
685 R200_TXA_OP_MADD
); /* was ADDSIGNED */
686 R200_ALPHA_ARG( 0, A
);
687 R200_ALPHA_ARG( 1, C
);
688 R200_ALPHA_ARG( 2, B
);
690 case GL_MODULATE_SUBTRACT_ATI
:
691 alpha_combine
= (R200_TXA_NEG_ARG_C
|
693 R200_ALPHA_ARG( 0, A
);
694 R200_ALPHA_ARG( 1, C
);
695 R200_ALPHA_ARG( 2, B
);
701 if ( (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA_EXT
)
702 || (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA
) ) {
703 alpha_scale
|= R200_TXA_DOT_ALPHA
;
708 * Apply the scale factor.
710 color_scale
|= (RGBshift
<< R200_TXC_SCALE_SHIFT
);
711 alpha_scale
|= (Ashift
<< R200_TXA_SCALE_SHIFT
);
717 if ( rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND
] != color_combine
||
718 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND
] != alpha_combine
||
719 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] != color_scale
||
720 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] != alpha_scale
) {
721 R200_STATECHANGE( rmesa
, pix
[slot
] );
722 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND
] = color_combine
;
723 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND
] = alpha_combine
;
724 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXCBLEND2
] = color_scale
;
725 rmesa
->hw
.pix
[slot
].cmd
[PIX_PP_TXABLEND2
] = alpha_scale
;
731 void r200SetTexOffset(__DRIcontext
* pDRICtx
, GLint texname
,
732 unsigned long long offset
, GLint depth
, GLuint pitch
)
734 r200ContextPtr rmesa
= pDRICtx
->driverPrivate
;
735 struct gl_texture_object
*tObj
=
736 _mesa_lookup_texture(rmesa
->radeon
.glCtx
, texname
);
737 radeonTexObjPtr t
= radeon_tex_obj(tObj
);
742 t
->image_override
= GL_TRUE
;
748 t
->override_offset
= 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
;
768 void r200SetTexBuffer(__DRIcontext
*pDRICtx
, GLint target
, __DRIdrawable
*dPriv
)
770 struct gl_texture_unit
*texUnit
;
771 struct gl_texture_object
*texObj
;
772 struct gl_texture_image
*texImage
;
773 struct radeon_renderbuffer
*rb
;
774 radeon_texture_image
*rImage
;
775 radeonContextPtr radeon
;
776 r200ContextPtr rmesa
;
777 struct radeon_framebuffer
*rfb
;
781 target
= GL_TEXTURE_RECTANGLE_ARB
;
783 radeon
= pDRICtx
->driverPrivate
;
784 rmesa
= pDRICtx
->driverPrivate
;
786 rfb
= dPriv
->driverPrivate
;
787 texUnit
= &radeon
->glCtx
->Texture
.Unit
[radeon
->glCtx
->Texture
.CurrentUnit
];
788 texObj
= _mesa_select_tex_object(radeon
->glCtx
, texUnit
, target
);
789 texImage
= _mesa_get_tex_image(radeon
->glCtx
, texObj
, target
, 0);
791 rImage
= get_radeon_texture_image(texImage
);
792 t
= radeon_tex_obj(texObj
);
797 radeon_update_renderbuffers(pDRICtx
, dPriv
);
798 /* back & depth buffer are useless free them right away */
799 rb
= (void*)rfb
->base
.Attachment
[BUFFER_DEPTH
].Renderbuffer
;
801 radeon_bo_unref(rb
->bo
);
804 rb
= (void*)rfb
->base
.Attachment
[BUFFER_BACK_LEFT
].Renderbuffer
;
806 radeon_bo_unref(rb
->bo
);
809 rb
= rfb
->color_rb
[0];
810 if (rb
->bo
== NULL
) {
811 /* Failed to BO for the buffer */
815 _mesa_lock_texture(radeon
->glCtx
, texObj
);
817 radeon_bo_unref(t
->bo
);
821 radeon_bo_unref(rImage
->bo
);
825 radeon_miptree_unreference(t
->mt
);
829 radeon_miptree_unreference(rImage
->mt
);
832 fprintf(stderr
,"settexbuf %d %dx%d@%d\n", rb
->pitch
, rb
->width
, rb
->height
, rb
->cpp
);
833 _mesa_init_teximage_fields(radeon
->glCtx
, target
, texImage
,
834 rb
->width
, rb
->height
, 1, 0, rb
->cpp
);
835 texImage
->TexFormat
= &_mesa_texformat_rgba8888_rev
;
837 radeon_bo_ref(rImage
->bo
);
839 radeon_bo_ref(t
->bo
);
841 t
->image_override
= GL_TRUE
;
842 t
->override_offset
= 0;
843 t
->pp_txpitch
&= (1 << 13) -1;
844 pitch_val
= rb
->pitch
;
847 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_ARGB8888
].format
;
848 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_ARGB8888
].filter
;
852 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB888
].format
;
853 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB888
].filter
;
856 t
->pp_txformat
= tx_table_le
[MESA_FORMAT_RGB565
].format
;
857 t
->pp_txfilter
|= tx_table_le
[MESA_FORMAT_RGB565
].filter
;
860 t
->pp_txsize
= ((rb
->width
- 1) << RADEON_TEX_USIZE_SHIFT
)
861 | ((rb
->height
- 1) << RADEON_TEX_VSIZE_SHIFT
);
862 t
->pp_txformat
|= R200_TXFORMAT_NON_POWER2
;
863 t
->pp_txpitch
= pitch_val
;
866 t
->validated
= GL_TRUE
;
867 _mesa_unlock_texture(radeon
->glCtx
, texObj
);
874 static GLboolean
r200UpdateAllTexEnv( GLcontext
*ctx
)
876 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
877 GLint i
, j
, currslot
;
878 GLint maxunitused
= -1;
879 GLboolean texregfree
[6] = {GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
, GL_TRUE
};
880 GLubyte stageref
[7] = {0, 0, 0, 0, 0, 0, 0};
881 GLint nextunit
[R200_MAX_TEXTURE_UNITS
] = {0, 0, 0, 0, 0, 0};
882 GLint currentnext
= -1;
885 /* find highest used unit */
886 for ( j
= 0; j
< R200_MAX_TEXTURE_UNITS
; j
++) {
887 if (ctx
->Texture
.Unit
[j
]._ReallyEnabled
) {
891 stageref
[maxunitused
+ 1] = REF_COLOR
| REF_ALPHA
;
893 for ( j
= maxunitused
; j
>= 0; j
-- ) {
894 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[j
];
896 rmesa
->state
.texture
.unit
[j
].outputreg
= -1;
898 if (stageref
[j
+ 1]) {
900 /* use the lowest available reg. That gets us automatically reg0 for the last stage.
901 need this even for disabled units, as it may get referenced due to the replace
903 for ( i
= 0 ; i
< R200_MAX_TEXTURE_UNITS
; i
++ ) {
905 rmesa
->state
.texture
.unit
[j
].outputreg
= i
;
909 if (rmesa
->state
.texture
.unit
[j
].outputreg
== -1) {
910 /* no more free regs we can use. Need a fallback :-( */
914 nextunit
[j
] = currentnext
;
916 if (!texUnit
->_ReallyEnabled
) {
917 /* the not enabled stages are referenced "indirectly",
918 must not cut off the lower stages */
919 stageref
[j
] = REF_COLOR
| REF_ALPHA
;
924 const GLuint numColorArgs
= texUnit
->_CurrentCombine
->_NumArgsRGB
;
925 const GLuint numAlphaArgs
= texUnit
->_CurrentCombine
->_NumArgsA
;
926 const GLboolean isdot3rgba
= (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA
) ||
927 (texUnit
->_CurrentCombine
->ModeRGB
== GL_DOT3_RGBA_EXT
);
930 /* check if we need the color part, special case for dot3_rgba
931 as if only the alpha part is referenced later on it still is using the color part */
932 if ((stageref
[j
+ 1] & REF_COLOR
) || isdot3rgba
) {
933 for ( i
= 0 ; i
< numColorArgs
; i
++ ) {
934 const GLuint srcRGBi
= texUnit
->_CurrentCombine
->SourceRGB
[i
];
935 const GLuint op
= texUnit
->_CurrentCombine
->OperandRGB
[i
];
938 /* op 0/1 are referencing color, op 2/3 alpha */
939 stageref
[j
] |= (op
>> 1) + 1;
942 texregfree
[j
] = GL_FALSE
;
950 texregfree
[srcRGBi
- GL_TEXTURE0
] = GL_FALSE
;
952 default: /* don't care about other sources here */
958 /* alpha args are ignored for dot3_rgba */
959 if ((stageref
[j
+ 1] & REF_ALPHA
) && !isdot3rgba
) {
961 for ( i
= 0 ; i
< numAlphaArgs
; i
++ ) {
962 const GLuint srcAi
= texUnit
->_CurrentCombine
->SourceA
[i
];
965 stageref
[j
] |= REF_ALPHA
;
968 texregfree
[j
] = GL_FALSE
;
976 texregfree
[srcAi
- GL_TEXTURE0
] = GL_FALSE
;
978 default: /* don't care about other sources here */
986 /* don't enable texture sampling for units if the result is not used */
987 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
988 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& !texregfree
[i
])
989 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
990 else rmesa
->state
.texture
.unit
[i
].unitneeded
= 0;
995 rmesa
->state
.envneeded
= 1;
998 while ((i
<= maxunitused
) && (i
>= 0)) {
999 /* only output instruction if the results are referenced */
1000 if (ctx
->Texture
.Unit
[i
]._ReallyEnabled
&& stageref
[i
+1]) {
1001 GLuint replaceunit
= i
;
1002 /* try to optimize GL_REPLACE away (only one level deep though) */
1003 if ( (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeRGB
== GL_REPLACE
) &&
1004 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ModeA
== GL_REPLACE
) &&
1005 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftRGB
== 0) &&
1006 (ctx
->Texture
.Unit
[i
]._CurrentCombine
->ScaleShiftA
== 0) &&
1007 (nextunit
[i
] > 0) ) {
1008 /* yippie! can optimize it away! */
1013 /* need env instruction slot */
1014 rmesa
->state
.envneeded
|= 1 << currslot
;
1015 ok
= r200UpdateTextureEnv( ctx
, i
, currslot
, replaceunit
);
1016 if (!ok
) return GL_FALSE
;
1022 if (currslot
== 0) {
1023 /* need one stage at least */
1024 rmesa
->state
.texture
.unit
[0].outputreg
= 0;
1025 ok
= r200UpdateTextureEnv( ctx
, 0, 0, 0 );
1028 R200_STATECHANGE( rmesa
, ctx
);
1029 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] &= ~(R200_TEX_BLEND_ENABLE_MASK
| R200_MULTI_PASS_ENABLE
);
1030 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= rmesa
->state
.envneeded
<< R200_TEX_BLEND_0_ENABLE_SHIFT
;
1039 #define TEXOBJ_TXFILTER_MASK (R200_MAX_MIP_LEVEL_MASK | \
1040 R200_MIN_FILTER_MASK | \
1041 R200_MAG_FILTER_MASK | \
1042 R200_MAX_ANISO_MASK | \
1044 R200_YUV_TEMPERATURE_MASK | \
1045 R200_CLAMP_S_MASK | \
1046 R200_CLAMP_T_MASK | \
1047 R200_BORDER_MODE_D3D )
1049 #define TEXOBJ_TXFORMAT_MASK (R200_TXFORMAT_WIDTH_MASK | \
1050 R200_TXFORMAT_HEIGHT_MASK | \
1051 R200_TXFORMAT_FORMAT_MASK | \
1052 R200_TXFORMAT_F5_WIDTH_MASK | \
1053 R200_TXFORMAT_F5_HEIGHT_MASK | \
1054 R200_TXFORMAT_ALPHA_IN_MAP | \
1055 R200_TXFORMAT_CUBIC_MAP_ENABLE | \
1056 R200_TXFORMAT_NON_POWER2)
1058 #define TEXOBJ_TXFORMAT_X_MASK (R200_DEPTH_LOG2_MASK | \
1059 R200_TEXCOORD_MASK | \
1060 R200_CLAMP_Q_MASK | \
1061 R200_VOLUME_FILTER_MASK)
1064 static void disable_tex_obj_state( r200ContextPtr rmesa
,
1068 R200_STATECHANGE( rmesa
, vtx
);
1069 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] &= ~(7 << (unit
* 3));
1071 if (rmesa
->radeon
.TclFallback
& (R200_TCL_FALLBACK_TEXGEN_0
<<unit
)) {
1072 TCL_FALLBACK( rmesa
->radeon
.glCtx
, (R200_TCL_FALLBACK_TEXGEN_0
<<unit
), GL_FALSE
);
1075 /* Actually want to keep all units less than max active texture
1076 * enabled, right? Fix this for >2 texunits.
1080 GLuint tmp
= rmesa
->TexGenEnabled
;
1082 rmesa
->TexGenEnabled
&= ~(R200_TEXGEN_TEXMAT_0_ENABLE
<<unit
);
1083 rmesa
->TexGenEnabled
&= ~(R200_TEXMAT_0_ENABLE
<<unit
);
1084 rmesa
->TexGenNeedNormals
[unit
] = GL_FALSE
;
1085 rmesa
->TexGenCompSel
&= ~(R200_OUTPUT_TEX_0
<< unit
);
1087 if (tmp
!= rmesa
->TexGenEnabled
) {
1088 rmesa
->recheck_texgen
[unit
] = GL_TRUE
;
1089 rmesa
->radeon
.NewGLState
|= _NEW_TEXTURE_MATRIX
;
1093 static void import_tex_obj_state( r200ContextPtr rmesa
,
1095 radeonTexObjPtr texobj
)
1097 /* do not use RADEON_DB_STATE to avoid stale texture caches */
1098 GLuint
*cmd
= &rmesa
->hw
.tex
[unit
].cmd
[TEX_CMD_0
];
1100 R200_STATECHANGE( rmesa
, tex
[unit
] );
1102 cmd
[TEX_PP_TXFILTER
] &= ~TEXOBJ_TXFILTER_MASK
;
1103 cmd
[TEX_PP_TXFILTER
] |= texobj
->pp_txfilter
& TEXOBJ_TXFILTER_MASK
;
1104 cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1105 cmd
[TEX_PP_TXFORMAT
] |= texobj
->pp_txformat
& TEXOBJ_TXFORMAT_MASK
;
1106 cmd
[TEX_PP_TXFORMAT_X
] &= ~TEXOBJ_TXFORMAT_X_MASK
;
1107 cmd
[TEX_PP_TXFORMAT_X
] |= texobj
->pp_txformat_x
& TEXOBJ_TXFORMAT_X_MASK
;
1108 cmd
[TEX_PP_TXSIZE
] = texobj
->pp_txsize
; /* NPOT only! */
1109 cmd
[TEX_PP_TXPITCH
] = texobj
->pp_txpitch
; /* NPOT only! */
1110 cmd
[TEX_PP_BORDER_COLOR
] = texobj
->pp_border_color
;
1112 if (texobj
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
1113 GLuint
*cube_cmd
= &rmesa
->hw
.cube
[unit
].cmd
[CUBE_CMD_0
];
1115 R200_STATECHANGE( rmesa
, cube
[unit
] );
1116 cube_cmd
[CUBE_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
1117 if (rmesa
->radeon
.radeonScreen
->drmSupportsFragShader
) {
1118 /* that value is submitted twice. could change cube atom
1119 to not include that command when new drm is used */
1120 cmd
[TEX_PP_CUBIC_FACES
] = texobj
->pp_cubic_faces
;
1126 static void set_texgen_matrix( r200ContextPtr rmesa
,
1128 const GLfloat
*s_plane
,
1129 const GLfloat
*t_plane
,
1130 const GLfloat
*r_plane
,
1131 const GLfloat
*q_plane
)
1155 _math_matrix_loadf( &(rmesa
->TexGenMatrix
[unit
]), m
);
1156 _math_matrix_analyse( &(rmesa
->TexGenMatrix
[unit
]) );
1157 rmesa
->TexGenEnabled
|= R200_TEXMAT_0_ENABLE
<<unit
;
1161 static GLuint
r200_need_dis_texgen(const GLbitfield texGenEnabled
,
1162 const GLfloat
*planeS
,
1163 const GLfloat
*planeT
,
1164 const GLfloat
*planeR
,
1165 const GLfloat
*planeQ
)
1167 GLuint needtgenable
= 0;
1169 if (!(texGenEnabled
& S_BIT
)) {
1170 if (((texGenEnabled
& T_BIT
) && planeT
[0] != 0.0) ||
1171 ((texGenEnabled
& R_BIT
) && planeR
[0] != 0.0) ||
1172 ((texGenEnabled
& Q_BIT
) && planeQ
[0] != 0.0)) {
1173 needtgenable
|= S_BIT
;
1176 if (!(texGenEnabled
& T_BIT
)) {
1177 if (((texGenEnabled
& S_BIT
) && planeS
[1] != 0.0) ||
1178 ((texGenEnabled
& R_BIT
) && planeR
[1] != 0.0) ||
1179 ((texGenEnabled
& Q_BIT
) && planeQ
[1] != 0.0)) {
1180 needtgenable
|= T_BIT
;
1183 if (!(texGenEnabled
& R_BIT
)) {
1184 if (((texGenEnabled
& S_BIT
) && planeS
[2] != 0.0) ||
1185 ((texGenEnabled
& T_BIT
) && planeT
[2] != 0.0) ||
1186 ((texGenEnabled
& Q_BIT
) && planeQ
[2] != 0.0)) {
1187 needtgenable
|= R_BIT
;
1190 if (!(texGenEnabled
& Q_BIT
)) {
1191 if (((texGenEnabled
& S_BIT
) && planeS
[3] != 0.0) ||
1192 ((texGenEnabled
& T_BIT
) && planeT
[3] != 0.0) ||
1193 ((texGenEnabled
& R_BIT
) && planeR
[3] != 0.0)) {
1194 needtgenable
|= Q_BIT
;
1198 return needtgenable
;
1203 * Returns GL_FALSE if fallback required.
1205 static GLboolean
r200_validate_texgen( GLcontext
*ctx
, GLuint unit
)
1207 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1208 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[unit
];
1209 GLuint inputshift
= R200_TEXGEN_0_INPUT_SHIFT
+ unit
*4;
1212 GLboolean mixed_fallback
= GL_FALSE
;
1213 static const GLfloat I
[16] = {
1218 static const GLfloat reflect
[16] = {
1224 rmesa
->TexGenCompSel
&= ~(R200_OUTPUT_TEX_0
<< unit
);
1225 rmesa
->TexGenEnabled
&= ~(R200_TEXGEN_TEXMAT_0_ENABLE
<<unit
);
1226 rmesa
->TexGenEnabled
&= ~(R200_TEXMAT_0_ENABLE
<<unit
);
1227 rmesa
->TexGenNeedNormals
[unit
] = GL_FALSE
;
1228 tgi
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] & ~(R200_TEXGEN_INPUT_MASK
<<
1230 tgcm
= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] & ~(R200_TEXGEN_COMP_MASK
<<
1234 fprintf(stderr
, "%s unit %d\n", __FUNCTION__
, unit
);
1236 if (texUnit
->TexGenEnabled
& S_BIT
) {
1237 mode
= texUnit
->GenModeS
;
1239 tgcm
|= R200_TEXGEN_COMP_S
<< (unit
* 4);
1242 if (texUnit
->TexGenEnabled
& T_BIT
) {
1243 if (texUnit
->GenModeT
!= mode
)
1244 mixed_fallback
= GL_TRUE
;
1246 tgcm
|= R200_TEXGEN_COMP_T
<< (unit
* 4);
1248 if (texUnit
->TexGenEnabled
& R_BIT
) {
1249 if (texUnit
->GenModeR
!= mode
)
1250 mixed_fallback
= GL_TRUE
;
1252 tgcm
|= R200_TEXGEN_COMP_R
<< (unit
* 4);
1255 if (texUnit
->TexGenEnabled
& Q_BIT
) {
1256 if (texUnit
->GenModeQ
!= mode
)
1257 mixed_fallback
= GL_TRUE
;
1259 tgcm
|= R200_TEXGEN_COMP_Q
<< (unit
* 4);
1262 if (mixed_fallback
) {
1263 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1264 fprintf(stderr
, "fallback mixed texgen, 0x%x (0x%x 0x%x 0x%x 0x%x)\n",
1265 texUnit
->TexGenEnabled
, texUnit
->GenModeS
, texUnit
->GenModeT
,
1266 texUnit
->GenModeR
, texUnit
->GenModeQ
);
1270 /* we CANNOT do mixed mode if the texgen mode requires a plane where the input
1271 is not enabled for texgen, since the planes are concatenated into texmat,
1272 and thus the input will come from texcoord rather than tex gen equation!
1273 Either fallback or just hope that those texcoords aren't really needed...
1274 Assuming the former will cause lots of unnecessary fallbacks, the latter will
1275 generate bogus results sometimes - it's pretty much impossible to really know
1276 when a fallback is needed, depends on texmat and what sort of texture is bound
1277 etc, - for now fallback if we're missing either S or T bits, there's a high
1278 probability we need the texcoords in that case.
1279 That's a lot of work for some obscure texgen mixed mode fixup - why oh why
1280 doesn't the chip just directly accept the plane parameters :-(. */
1282 case GL_OBJECT_LINEAR
: {
1283 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1284 texUnit
->ObjectPlaneS
, texUnit
->ObjectPlaneT
,
1285 texUnit
->ObjectPlaneR
, texUnit
->ObjectPlaneQ
);
1286 if (needtgenable
& (S_BIT
| T_BIT
)) {
1287 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1288 fprintf(stderr
, "fallback mixed texgen / obj plane, 0x%x\n",
1289 texUnit
->TexGenEnabled
);
1292 if (needtgenable
& (R_BIT
)) {
1293 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1295 if (needtgenable
& (Q_BIT
)) {
1296 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1299 tgi
|= R200_TEXGEN_INPUT_OBJ
<< inputshift
;
1300 set_texgen_matrix( rmesa
, unit
,
1301 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->ObjectPlaneS
: I
,
1302 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->ObjectPlaneT
: I
+ 4,
1303 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->ObjectPlaneR
: I
+ 8,
1304 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->ObjectPlaneQ
: I
+ 12);
1308 case GL_EYE_LINEAR
: {
1309 GLuint needtgenable
= r200_need_dis_texgen( texUnit
->TexGenEnabled
,
1310 texUnit
->EyePlaneS
, texUnit
->EyePlaneT
,
1311 texUnit
->EyePlaneR
, texUnit
->EyePlaneQ
);
1312 if (needtgenable
& (S_BIT
| T_BIT
)) {
1313 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1314 fprintf(stderr
, "fallback mixed texgen / eye plane, 0x%x\n",
1315 texUnit
->TexGenEnabled
);
1318 if (needtgenable
& (R_BIT
)) {
1319 tgcm
&= ~(R200_TEXGEN_COMP_R
<< (unit
* 4));
1321 if (needtgenable
& (Q_BIT
)) {
1322 tgcm
&= ~(R200_TEXGEN_COMP_Q
<< (unit
* 4));
1324 tgi
|= R200_TEXGEN_INPUT_EYE
<< inputshift
;
1325 set_texgen_matrix( rmesa
, unit
,
1326 (texUnit
->TexGenEnabled
& S_BIT
) ? texUnit
->EyePlaneS
: I
,
1327 (texUnit
->TexGenEnabled
& T_BIT
) ? texUnit
->EyePlaneT
: I
+ 4,
1328 (texUnit
->TexGenEnabled
& R_BIT
) ? texUnit
->EyePlaneR
: I
+ 8,
1329 (texUnit
->TexGenEnabled
& Q_BIT
) ? texUnit
->EyePlaneQ
: I
+ 12);
1333 case GL_REFLECTION_MAP_NV
:
1334 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1335 tgi
|= R200_TEXGEN_INPUT_EYE_REFLECT
<< inputshift
;
1336 /* pretty weird, must only negate when lighting is enabled? */
1337 if (ctx
->Light
.Enabled
)
1338 set_texgen_matrix( rmesa
, unit
,
1339 (texUnit
->TexGenEnabled
& S_BIT
) ? reflect
: I
,
1340 (texUnit
->TexGenEnabled
& T_BIT
) ? reflect
+ 4 : I
+ 4,
1341 (texUnit
->TexGenEnabled
& R_BIT
) ? reflect
+ 8 : I
+ 8,
1345 case GL_NORMAL_MAP_NV
:
1346 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1347 tgi
|= R200_TEXGEN_INPUT_EYE_NORMAL
<<inputshift
;
1351 rmesa
->TexGenNeedNormals
[unit
] = GL_TRUE
;
1352 tgi
|= R200_TEXGEN_INPUT_SPHERE
<<inputshift
;
1356 /* All texgen units were disabled, so just pass coords through. */
1357 tgi
|= unit
<< inputshift
;
1361 /* Unsupported mode, fallback:
1363 if (R200_DEBUG
& DEBUG_FALLBACKS
)
1364 fprintf(stderr
, "fallback unsupported texgen, %d\n",
1369 rmesa
->TexGenEnabled
|= R200_TEXGEN_TEXMAT_0_ENABLE
<< unit
;
1370 rmesa
->TexGenCompSel
|= R200_OUTPUT_TEX_0
<< unit
;
1372 if (tgi
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] ||
1373 tgcm
!= rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
])
1375 R200_STATECHANGE(rmesa
, tcg
);
1376 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_1
] = tgi
;
1377 rmesa
->hw
.tcg
.cmd
[TCG_TEX_PROC_CTL_2
] = tgcm
;
1383 void set_re_cntl_d3d( GLcontext
*ctx
, int unit
, GLboolean use_d3d
)
1385 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1389 re_cntl
= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] & ~(R200_VTX_STQ0_D3D
<< (2 * unit
));
1391 re_cntl
|= R200_VTX_STQ0_D3D
<< (2 * unit
);
1393 if ( re_cntl
!= rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] ) {
1394 R200_STATECHANGE( rmesa
, set
);
1395 rmesa
->hw
.set
.cmd
[SET_RE_CNTL
] = re_cntl
;
1400 * Compute the cached hardware register values for the given texture object.
1402 * \param rmesa Context pointer
1403 * \param t the r300 texture object
1405 static void setup_hardware_state(r200ContextPtr rmesa
, radeonTexObj
*t
)
1407 const struct gl_texture_image
*firstImage
=
1408 t
->base
.Image
[0][t
->mt
->firstLevel
];
1409 GLint log2Width
, log2Height
, log2Depth
, texelBytes
;
1415 log2Width
= firstImage
->WidthLog2
;
1416 log2Height
= firstImage
->HeightLog2
;
1417 log2Depth
= firstImage
->DepthLog2
;
1418 texelBytes
= firstImage
->TexFormat
->TexelBytes
;
1421 if (!t
->image_override
) {
1422 if (VALID_FORMAT(firstImage
->TexFormat
->MesaFormat
)) {
1423 const struct tx_table
*table
= _mesa_little_endian() ? tx_table_le
:
1426 t
->pp_txformat
&= ~(R200_TXFORMAT_FORMAT_MASK
|
1427 R200_TXFORMAT_ALPHA_IN_MAP
);
1428 t
->pp_txfilter
&= ~R200_YUV_TO_RGB
;
1430 t
->pp_txformat
|= table
[ firstImage
->TexFormat
->MesaFormat
].format
;
1431 t
->pp_txfilter
|= table
[ firstImage
->TexFormat
->MesaFormat
].filter
;
1433 _mesa_problem(NULL
, "unexpected texture format in %s",
1439 t
->pp_txfilter
&= ~R200_MAX_MIP_LEVEL_MASK
;
1440 t
->pp_txfilter
|= (t
->mt
->lastLevel
- t
->mt
->firstLevel
) << R200_MAX_MIP_LEVEL_SHIFT
;
1442 t
->pp_txformat
&= ~(R200_TXFORMAT_WIDTH_MASK
|
1443 R200_TXFORMAT_HEIGHT_MASK
|
1444 R200_TXFORMAT_CUBIC_MAP_ENABLE
|
1445 R200_TXFORMAT_F5_WIDTH_MASK
|
1446 R200_TXFORMAT_F5_HEIGHT_MASK
);
1447 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_WIDTH_SHIFT
) |
1448 (log2Height
<< R200_TXFORMAT_HEIGHT_SHIFT
));
1452 t
->pp_txformat_x
&= ~(R200_DEPTH_LOG2_MASK
| R200_TEXCOORD_MASK
);
1453 if (t
->base
.Target
== GL_TEXTURE_3D
) {
1454 t
->pp_txformat_x
|= (log2Depth
<< R200_DEPTH_LOG2_SHIFT
);
1455 t
->pp_txformat_x
|= R200_TEXCOORD_VOLUME
;
1458 else if (t
->base
.Target
== GL_TEXTURE_CUBE_MAP
) {
1459 ASSERT(log2Width
== log2Height
);
1460 t
->pp_txformat
|= ((log2Width
<< R200_TXFORMAT_F5_WIDTH_SHIFT
) |
1461 (log2Height
<< R200_TXFORMAT_F5_HEIGHT_SHIFT
) |
1462 /* don't think we need this bit, if it exists at all - fglrx does not set it */
1463 (R200_TXFORMAT_CUBIC_MAP_ENABLE
));
1464 t
->pp_txformat_x
|= R200_TEXCOORD_CUBIC_ENV
;
1465 t
->pp_cubic_faces
= ((log2Width
<< R200_FACE_WIDTH_1_SHIFT
) |
1466 (log2Height
<< R200_FACE_HEIGHT_1_SHIFT
) |
1467 (log2Width
<< R200_FACE_WIDTH_2_SHIFT
) |
1468 (log2Height
<< R200_FACE_HEIGHT_2_SHIFT
) |
1469 (log2Width
<< R200_FACE_WIDTH_3_SHIFT
) |
1470 (log2Height
<< R200_FACE_HEIGHT_3_SHIFT
) |
1471 (log2Width
<< R200_FACE_WIDTH_4_SHIFT
) |
1472 (log2Height
<< R200_FACE_HEIGHT_4_SHIFT
));
1475 /* If we don't in fact send enough texture coordinates, q will be 1,
1476 * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
1478 t
->pp_txformat_x
|= R200_TEXCOORD_PROJ
;
1481 t
->pp_txsize
= (((firstImage
->Width
- 1) << R200_PP_TX_WIDTHMASK_SHIFT
)
1482 | ((firstImage
->Height
- 1) << R200_PP_TX_HEIGHTMASK_SHIFT
));
1484 if ( !t
->image_override
) {
1485 if (firstImage
->IsCompressed
)
1486 t
->pp_txpitch
= (firstImage
->Width
+ 63) & ~(63);
1488 t
->pp_txpitch
= ((firstImage
->Width
* texelBytes
) + 63) & ~(63);
1489 t
->pp_txpitch
-= 32;
1492 if (t
->base
.Target
== GL_TEXTURE_RECTANGLE_NV
) {
1493 t
->pp_txformat
|= R200_TXFORMAT_NON_POWER2
;
1498 static GLboolean
r200_validate_texture(GLcontext
*ctx
, struct gl_texture_object
*texObj
, int unit
)
1500 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1501 radeonTexObj
*t
= radeon_tex_obj(texObj
);
1503 if (!radeon_validate_texture_miptree(ctx
, texObj
))
1506 r200_validate_texgen(ctx
, unit
);
1507 /* Configure the hardware registers (more precisely, the cached version
1508 * of the hardware registers). */
1509 setup_hardware_state(rmesa
, t
);
1511 if (texObj
->Target
== GL_TEXTURE_RECTANGLE_NV
||
1512 texObj
->Target
== GL_TEXTURE_2D
||
1513 texObj
->Target
== GL_TEXTURE_1D
)
1514 set_re_cntl_d3d( ctx
, unit
, GL_FALSE
);
1516 set_re_cntl_d3d( ctx
, unit
, GL_TRUE
);
1517 R200_STATECHANGE( rmesa
, ctx
);
1518 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_0_ENABLE
<< unit
;
1520 R200_STATECHANGE( rmesa
, vtx
);
1521 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] &= ~(7 << (unit
* 3));
1522 rmesa
->hw
.vtx
.cmd
[VTX_TCL_OUTPUT_VTXFMT_1
] |= 4 << (unit
* 3);
1524 rmesa
->recheck_texgen
[unit
] = GL_TRUE
;
1525 import_tex_obj_state( rmesa
, unit
, t
);
1527 if (rmesa
->recheck_texgen
[unit
]) {
1528 GLboolean fallback
= !r200_validate_texgen( ctx
, unit
);
1529 TCL_FALLBACK( ctx
, (R200_TCL_FALLBACK_TEXGEN_0
<<unit
), fallback
);
1530 rmesa
->recheck_texgen
[unit
] = 0;
1531 rmesa
->radeon
.NewGLState
|= _NEW_TEXTURE_MATRIX
;
1534 t
->validated
= GL_TRUE
;
1536 FALLBACK( rmesa
, RADEON_FALLBACK_BORDER_MODE
, t
->border_fallback
);
1538 return !t
->border_fallback
;
1541 static GLboolean
r200UpdateTextureUnit(GLcontext
*ctx
, int unit
)
1543 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1544 GLuint unitneeded
= rmesa
->state
.texture
.unit
[unit
].unitneeded
;
1547 /* disable the unit */
1548 disable_tex_obj_state(rmesa
, unit
);
1552 if (!r200_validate_texture(ctx
, ctx
->Texture
.Unit
[unit
]._Current
, unit
)) {
1554 "failed to validate texture for unit %d.\n",
1556 rmesa
->state
.texture
.unit
[unit
].texobj
= NULL
;
1560 rmesa
->state
.texture
.unit
[unit
].texobj
= radeon_tex_obj(ctx
->Texture
.Unit
[unit
]._Current
);
1565 void r200UpdateTextureState( GLcontext
*ctx
)
1567 r200ContextPtr rmesa
= R200_CONTEXT(ctx
);
1571 /* NOTE: must not manipulate rmesa->state.texture.unit[].unitneeded or
1572 rmesa->state.envneeded before a R200_STATECHANGE (or R200_NEWPRIM) since
1573 we use these to determine if we want to emit the corresponding state
1575 R200_NEWPRIM( rmesa
);
1577 if (ctx
->ATIFragmentShader
._Enabled
) {
1579 for (i
= 0; i
< R200_MAX_TEXTURE_UNITS
; i
++) {
1580 rmesa
->state
.texture
.unit
[i
].unitneeded
= ctx
->Texture
.Unit
[i
]._ReallyEnabled
;
1585 ok
= r200UpdateAllTexEnv( ctx
);
1588 ok
= (r200UpdateTextureUnit( ctx
, 0 ) &&
1589 r200UpdateTextureUnit( ctx
, 1 ) &&
1590 r200UpdateTextureUnit( ctx
, 2 ) &&
1591 r200UpdateTextureUnit( ctx
, 3 ) &&
1592 r200UpdateTextureUnit( ctx
, 4 ) &&
1593 r200UpdateTextureUnit( ctx
, 5 ));
1596 if (ok
&& ctx
->ATIFragmentShader
._Enabled
) {
1597 r200UpdateFragmentShader(ctx
);
1600 FALLBACK( rmesa
, R200_FALLBACK_TEXTURE
, !ok
);
1602 if (rmesa
->radeon
.TclFallback
)
1603 r200ChooseVertexState( ctx
);
1606 if (rmesa
->radeon
.radeonScreen
->chip_family
== CHIP_FAMILY_R200
) {
1609 * T0 hang workaround -------------
1610 * not needed for r200 derivatives
1612 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_ENABLE_MASK
) == R200_TEX_0_ENABLE
&&
1613 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1615 R200_STATECHANGE(rmesa
, ctx
);
1616 R200_STATECHANGE(rmesa
, tex
[1]);
1617 rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] |= R200_TEX_1_ENABLE
;
1618 if (!(rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_1_ENABLE
))
1619 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1620 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] |= R200_TXFORMAT_LOOKUP_DISABLE
;
1622 else if (!ctx
->ATIFragmentShader
._Enabled
) {
1623 if ((rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
) &&
1624 (rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] & R200_TXFORMAT_LOOKUP_DISABLE
)) {
1625 R200_STATECHANGE(rmesa
, tex
[1]);
1626 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~R200_TXFORMAT_LOOKUP_DISABLE
;
1629 /* do the same workaround for the first pass of a fragment shader.
1630 * completely unknown if necessary / sufficient.
1632 if ((rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] & R200_PPX_TEX_ENABLE_MASK
) == R200_PPX_TEX_0_ENABLE
&&
1633 (rmesa
->hw
.tex
[0].cmd
[TEX_PP_TXFILTER
] & R200_MIN_FILTER_MASK
) > R200_MIN_FILTER_LINEAR
) {
1635 R200_STATECHANGE(rmesa
, cst
);
1636 R200_STATECHANGE(rmesa
, tex
[1]);
1637 rmesa
->hw
.cst
.cmd
[CST_PP_CNTL_X
] |= R200_PPX_TEX_1_ENABLE
;
1638 if (!(rmesa
->hw
.ctx
.cmd
[CTX_PP_CNTL
] & R200_TEX_1_ENABLE
))
1639 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXFORMAT
] &= ~TEXOBJ_TXFORMAT_MASK
;
1640 rmesa
->hw
.tex
[1].cmd
[TEX_PP_TXMULTI_CTL
] |= R200_PASS1_TXFORMAT_LOOKUP_DISABLE
;
1643 /* maybe needs to be done pairwise due to 2 parallel (physical) tex units ?
1644 looks like that's not the case, if 8500/9100 owners don't complain remove this...
1645 for ( i = 0; i < ctx->Const.MaxTextureUnits; i += 2) {
1646 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & ((R200_TEX_0_ENABLE |
1647 R200_TEX_1_ENABLE ) << i)) == (R200_TEX_0_ENABLE << i)) &&
1648 ((rmesa->hw.tex[i].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK) >
1649 R200_MIN_FILTER_LINEAR)) {
1650 R200_STATECHANGE(rmesa, ctx);
1651 R200_STATECHANGE(rmesa, tex[i+1]);
1652 rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= (R200_TEX_1_ENABLE << i);
1653 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
1654 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] |= 0x08000000;
1657 if ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE << i)) &&
1658 (rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] & 0x08000000)) {
1659 R200_STATECHANGE(rmesa, tex[i+1]);
1660 rmesa->hw.tex[i+1].cmd[TEX_PP_TXFORMAT] &= ~0x08000000;
1666 * Texture cache LRU hang workaround -------------
1667 * not needed for r200 derivatives
1668 * hopefully this covers first pass of a shader as well
1671 /* While the cases below attempt to only enable the workaround in the
1672 * specific cases necessary, they were insufficient. See bugzilla #1519,
1673 * #729, #814. Tests with quake3 showed no impact on performance.
1678 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE )) &&
1679 ((((rmesa->hw.tex[0].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1681 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_2_ENABLE) &&
1682 ((((rmesa->hw.tex[2].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1684 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_4_ENABLE) &&
1685 ((((rmesa->hw.tex[4].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1691 if (((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_1_ENABLE )) &&
1692 ((((rmesa->hw.tex[1].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1694 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_3_ENABLE) &&
1695 ((((rmesa->hw.tex[3].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1697 ((rmesa->hw.ctx.cmd[CTX_PP_CNTL] & R200_TEX_5_ENABLE) &&
1698 ((((rmesa->hw.tex[5].cmd[TEX_PP_TXFILTER] & R200_MIN_FILTER_MASK)) &
1704 if (dbg
!= rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
]) {
1705 R200_STATECHANGE( rmesa
, tam
);
1706 rmesa
->hw
.tam
.cmd
[TAM_DEBUG3
] = dbg
;
1707 if (0) printf("TEXCACHE LRU HANG WORKAROUND %x\n", dbg
);