2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2007 Brian Paul All Rights Reserved.
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
27 * \file tnl/t_vb_program.c
28 * \brief Pipeline stage for executing vertex programs.
29 * \author Brian Paul, Keith Whitwell
33 #include "main/glheader.h"
34 #include "main/colormac.h"
35 #include "main/context.h"
36 #include "main/macros.h"
37 #include "main/imports.h"
38 #include "shader/prog_instruction.h"
39 #include "shader/prog_statevars.h"
40 #include "shader/prog_execute.h"
41 #include "swrast/s_context.h"
42 #include "swrast/s_texfilter.h"
45 #include "tnl/t_context.h"
46 #include "tnl/t_pipeline.h"
51 * Private storage for the vertex program pipeline stage.
53 struct vp_stage_data
{
54 /** The results of running the vertex program go into these arrays. */
55 GLvector4f results
[VERT_RESULT_MAX
];
57 GLvector4f ndcCoords
; /**< normalized device coords */
58 GLubyte
*clipmask
; /**< clip flags */
59 GLubyte ormask
, andmask
; /**< for clipping */
63 #define VP_STAGE_DATA(stage) ((struct vp_stage_data *)(stage->privatePtr))
67 userclip( GLcontext
*ctx
,
71 GLubyte
*clipandmask
)
75 for (p
= 0; p
< ctx
->Const
.MaxClipPlanes
; p
++) {
76 if (ctx
->Transform
.ClipPlanesEnabled
& (1 << p
)) {
78 const GLfloat a
= ctx
->Transform
._ClipUserPlane
[p
][0];
79 const GLfloat b
= ctx
->Transform
._ClipUserPlane
[p
][1];
80 const GLfloat c
= ctx
->Transform
._ClipUserPlane
[p
][2];
81 const GLfloat d
= ctx
->Transform
._ClipUserPlane
[p
][3];
82 GLfloat
*coord
= (GLfloat
*)clip
->data
;
83 GLuint stride
= clip
->stride
;
84 GLuint count
= clip
->count
;
86 for (nr
= 0, i
= 0 ; i
< count
; i
++) {
87 GLfloat dp
= (coord
[0] * a
+
94 clipmask
[i
] |= CLIP_USER_BIT
;
97 STRIDE_F(coord
, stride
);
101 *clipormask
|= CLIP_USER_BIT
;
103 *clipandmask
|= CLIP_USER_BIT
;
113 do_ndc_cliptest(GLcontext
*ctx
, struct vp_stage_data
*store
)
115 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
116 struct vertex_buffer
*VB
= &tnl
->vb
;
117 /* Cliptest and perspective divide. Clip functions must clear
121 store
->andmask
= CLIP_FRUSTUM_BITS
;
123 if (tnl
->NeedNdcCoords
) {
125 _mesa_clip_tab
[VB
->ClipPtr
->size
]( VB
->ClipPtr
,
133 _mesa_clip_np_tab
[VB
->ClipPtr
->size
]( VB
->ClipPtr
,
140 if (store
->andmask
) {
141 /* All vertices are outside the frustum */
145 /* Test userclip planes. This contributes to VB->ClipMask.
147 /** XXX NEW_SLANG _Enabled ??? */
148 if (ctx
->Transform
.ClipPlanesEnabled
&& (!ctx
->VertexProgram
._Enabled
||
149 ctx
->VertexProgram
.Current
->IsPositionInvariant
)) {
156 if (store
->andmask
) {
161 VB
->ClipAndMask
= store
->andmask
;
162 VB
->ClipOrMask
= store
->ormask
;
163 VB
->ClipMask
= store
->clipmask
;
170 * XXX the texture sampling code in this module is a bit of a hack.
171 * The texture sampling code is in swrast, though it doesn't have any
172 * real dependencies on the rest of swrast. It should probably be
173 * moved into main/ someday.
176 vp_fetch_texel(GLcontext
*ctx
, const GLfloat texcoord
[4], GLfloat lambda
,
177 GLuint unit
, GLfloat color
[4])
180 SWcontext
*swrast
= SWRAST_CONTEXT(ctx
);
182 /* XXX use a float-valued TextureSample routine here!!! */
183 swrast
->TextureSample
[unit
](ctx
, ctx
->Texture
.Unit
[unit
]._Current
,
184 1, (const GLfloat (*)[4]) texcoord
,
186 color
[0] = CHAN_TO_FLOAT(rgba
[0]);
187 color
[1] = CHAN_TO_FLOAT(rgba
[1]);
188 color
[2] = CHAN_TO_FLOAT(rgba
[2]);
189 color
[3] = CHAN_TO_FLOAT(rgba
[3]);
194 * Called via ctx->Driver.ProgramStringNotify() after a new vertex program
195 * string has been parsed.
198 _tnl_program_string(GLcontext
*ctx
, GLenum target
, struct gl_program
*program
)
201 * If we had derived anything from the program that was private to this
202 * stage we'd recompute/validate it here.
208 * Initialize virtual machine state prior to executing vertex program.
211 init_machine(GLcontext
*ctx
, struct gl_program_machine
*machine
)
213 /* Input registers get initialized from the current vertex attribs */
214 MEMCPY(machine
->VertAttribs
, ctx
->Current
.Attrib
,
215 MAX_VERTEX_PROGRAM_ATTRIBS
* 4 * sizeof(GLfloat
));
217 if (ctx
->VertexProgram
._Current
->IsNVProgram
) {
219 /* Output/result regs are initialized to [0,0,0,1] */
220 for (i
= 0; i
< MAX_NV_VERTEX_PROGRAM_OUTPUTS
; i
++) {
221 ASSIGN_4V(machine
->Outputs
[i
], 0.0F
, 0.0F
, 0.0F
, 1.0F
);
223 /* Temp regs are initialized to [0,0,0,0] */
224 for (i
= 0; i
< MAX_NV_VERTEX_PROGRAM_TEMPS
; i
++) {
225 ASSIGN_4V(machine
->Temporaries
[i
], 0.0F
, 0.0F
, 0.0F
, 0.0F
);
227 for (i
= 0; i
< MAX_VERTEX_PROGRAM_ADDRESS_REGS
; i
++) {
228 ASSIGN_4V(machine
->AddressReg
[i
], 0, 0, 0, 0);
232 machine
->NumDeriv
= 0;
234 /* init condition codes */
235 machine
->CondCodes
[0] = COND_EQ
;
236 machine
->CondCodes
[1] = COND_EQ
;
237 machine
->CondCodes
[2] = COND_EQ
;
238 machine
->CondCodes
[3] = COND_EQ
;
240 /* init call stack */
241 machine
->StackDepth
= 0;
243 machine
->FetchTexelLod
= vp_fetch_texel
;
244 machine
->FetchTexelDeriv
= NULL
; /* not used by vertex programs */
249 * Map the texture images which the vertex program will access (if any).
252 map_textures(GLcontext
*ctx
, const struct gl_vertex_program
*vp
)
256 if (!ctx
->Driver
.MapTexture
)
259 for (u
= 0; u
< ctx
->Const
.MaxVertexTextureImageUnits
; u
++) {
260 if (vp
->Base
.TexturesUsed
[u
]) {
261 /* Note: _Current *should* correspond to the target indicated
262 * in TexturesUsed[u].
264 ctx
->Driver
.MapTexture(ctx
, ctx
->Texture
.Unit
[u
]._Current
);
271 * Unmap the texture images which were used by the vertex program (if any).
274 unmap_textures(GLcontext
*ctx
, const struct gl_vertex_program
*vp
)
278 if (!ctx
->Driver
.MapTexture
)
281 for (u
= 0; u
< ctx
->Const
.MaxVertexTextureImageUnits
; u
++) {
282 if (vp
->Base
.TexturesUsed
[u
]) {
283 /* Note: _Current *should* correspond to the target indicated
284 * in TexturesUsed[u].
286 ctx
->Driver
.UnmapTexture(ctx
, ctx
->Texture
.Unit
[u
]._Current
);
293 * This function executes vertex programs
296 run_vp( GLcontext
*ctx
, struct tnl_pipeline_stage
*stage
)
298 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
299 struct vp_stage_data
*store
= VP_STAGE_DATA(stage
);
300 struct vertex_buffer
*VB
= &tnl
->vb
;
301 struct gl_vertex_program
*program
= ctx
->VertexProgram
._Current
;
302 struct gl_program_machine machine
;
303 GLuint outputs
[VERT_RESULT_MAX
], numOutputs
;
309 if (program
->IsNVProgram
) {
310 _mesa_load_tracked_matrices(ctx
);
313 /* ARB program or vertex shader */
314 _mesa_load_state_parameters(ctx
, program
->Base
.Parameters
);
317 /* make list of outputs to save some time below */
319 for (i
= 0; i
< VERT_RESULT_MAX
; i
++) {
320 if (program
->Base
.OutputsWritten
& (1 << i
)) {
321 outputs
[numOutputs
++] = i
;
325 map_textures(ctx
, program
);
327 for (i
= 0; i
< VB
->Count
; i
++) {
330 init_machine(ctx
, &machine
);
333 printf("Input %d: %f, %f, %f, %f\n", i
,
334 VB
->AttribPtr
[0]->data
[i
][0],
335 VB
->AttribPtr
[0]->data
[i
][1],
336 VB
->AttribPtr
[0]->data
[i
][2],
337 VB
->AttribPtr
[0]->data
[i
][3]);
338 printf(" color: %f, %f, %f, %f\n",
339 VB
->AttribPtr
[3]->data
[i
][0],
340 VB
->AttribPtr
[3]->data
[i
][1],
341 VB
->AttribPtr
[3]->data
[i
][2],
342 VB
->AttribPtr
[3]->data
[i
][3]);
343 printf(" normal: %f, %f, %f, %f\n",
344 VB
->AttribPtr
[2]->data
[i
][0],
345 VB
->AttribPtr
[2]->data
[i
][1],
346 VB
->AttribPtr
[2]->data
[i
][2],
347 VB
->AttribPtr
[2]->data
[i
][3]);
350 /* the vertex array case */
351 for (attr
= 0; attr
< VERT_ATTRIB_MAX
; attr
++) {
352 if (program
->Base
.InputsRead
& (1 << attr
)) {
353 const GLubyte
*ptr
= (const GLubyte
*) VB
->AttribPtr
[attr
]->data
;
354 const GLuint size
= VB
->AttribPtr
[attr
]->size
;
355 const GLuint stride
= VB
->AttribPtr
[attr
]->stride
;
356 const GLfloat
*data
= (GLfloat
*) (ptr
+ stride
* i
);
357 COPY_CLEAN_4V(machine
.VertAttribs
[attr
], size
, data
);
361 /* execute the program */
362 _mesa_execute_program(ctx
, &program
->Base
, &machine
);
364 /* copy the output registers into the VB->attribs arrays */
365 for (j
= 0; j
< numOutputs
; j
++) {
366 const GLuint attr
= outputs
[j
];
367 COPY_4V(store
->results
[attr
].data
[i
], machine
.Outputs
[attr
]);
370 printf("HPOS: %f %f %f %f\n",
371 machine
.Outputs
[0][0],
372 machine
.Outputs
[0][1],
373 machine
.Outputs
[0][2],
374 machine
.Outputs
[0][3]);
378 unmap_textures(ctx
, program
);
380 /* Fixup fog and point size results if needed */
381 if (program
->IsNVProgram
) {
382 if (ctx
->Fog
.Enabled
&&
383 (program
->Base
.OutputsWritten
& (1 << VERT_RESULT_FOGC
)) == 0) {
384 for (i
= 0; i
< VB
->Count
; i
++) {
385 store
->results
[VERT_RESULT_FOGC
].data
[i
][0] = 1.0;
389 if (ctx
->VertexProgram
.PointSizeEnabled
&&
390 (program
->Base
.OutputsWritten
& (1 << VERT_RESULT_PSIZ
)) == 0) {
391 for (i
= 0; i
< VB
->Count
; i
++) {
392 store
->results
[VERT_RESULT_PSIZ
].data
[i
][0] = ctx
->Point
.Size
;
397 if (program
->IsPositionInvariant
) {
398 /* We need the exact same transform as in the fixed function path here
399 * to guarantee invariance, depending on compiler optimization flags
400 * results could be different otherwise.
402 VB
->ClipPtr
= TransformRaw( &store
->results
[0],
403 &ctx
->_ModelProjectMatrix
,
406 /* Drivers expect this to be clean to element 4...
408 switch (VB
->ClipPtr
->size
) {
412 _mesa_vector4f_clean_elem( VB
->ClipPtr
, VB
->Count
, 2 );
415 _mesa_vector4f_clean_elem( VB
->ClipPtr
, VB
->Count
, 3 );
422 /* Setup the VB pointers so that the next pipeline stages get
423 * their data from the right place (the program output arrays).
425 VB
->ClipPtr
= &store
->results
[VERT_RESULT_HPOS
];
426 VB
->ClipPtr
->size
= 4;
427 VB
->ClipPtr
->count
= VB
->Count
;
430 VB
->ColorPtr
[0] = &store
->results
[VERT_RESULT_COL0
];
431 VB
->ColorPtr
[1] = &store
->results
[VERT_RESULT_BFC0
];
432 VB
->SecondaryColorPtr
[0] = &store
->results
[VERT_RESULT_COL1
];
433 VB
->SecondaryColorPtr
[1] = &store
->results
[VERT_RESULT_BFC1
];
434 VB
->FogCoordPtr
= &store
->results
[VERT_RESULT_FOGC
];
436 VB
->AttribPtr
[VERT_ATTRIB_COLOR0
] = &store
->results
[VERT_RESULT_COL0
];
437 VB
->AttribPtr
[VERT_ATTRIB_COLOR1
] = &store
->results
[VERT_RESULT_COL1
];
438 VB
->AttribPtr
[VERT_ATTRIB_FOG
] = &store
->results
[VERT_RESULT_FOGC
];
439 VB
->AttribPtr
[_TNL_ATTRIB_POINTSIZE
] = &store
->results
[VERT_RESULT_PSIZ
];
441 for (i
= 0; i
< ctx
->Const
.MaxTextureCoordUnits
; i
++) {
443 VB
->AttribPtr
[_TNL_ATTRIB_TEX0
+ i
]
444 = &store
->results
[VERT_RESULT_TEX0
+ i
];
447 for (i
= 0; i
< ctx
->Const
.MaxVarying
; i
++) {
448 if (program
->Base
.OutputsWritten
& (1 << (VERT_RESULT_VAR0
+ i
))) {
449 /* Note: varying results get put into the generic attributes */
450 VB
->AttribPtr
[VERT_ATTRIB_GENERIC0
+i
]
451 = &store
->results
[VERT_RESULT_VAR0
+ i
];
456 /* Perform NDC and cliptest operations:
458 return do_ndc_cliptest(ctx
, store
);
463 * Called the first time stage->run is called. In effect, don't
464 * allocate data until the first time the stage is run.
467 init_vp(GLcontext
*ctx
, struct tnl_pipeline_stage
*stage
)
469 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
470 struct vertex_buffer
*VB
= &(tnl
->vb
);
471 struct vp_stage_data
*store
;
472 const GLuint size
= VB
->Size
;
475 stage
->privatePtr
= MALLOC(sizeof(*store
));
476 store
= VP_STAGE_DATA(stage
);
480 /* Allocate arrays of vertex output values */
481 for (i
= 0; i
< VERT_RESULT_MAX
; i
++) {
482 _mesa_vector4f_alloc( &store
->results
[i
], 0, size
, 32 );
483 store
->results
[i
].size
= 4;
486 /* a few other misc allocations */
487 _mesa_vector4f_alloc( &store
->ndcCoords
, 0, size
, 32 );
488 store
->clipmask
= (GLubyte
*) ALIGN_MALLOC(sizeof(GLubyte
)*size
, 32 );
495 * Destructor for this pipeline stage.
498 dtr(struct tnl_pipeline_stage
*stage
)
500 struct vp_stage_data
*store
= VP_STAGE_DATA(stage
);
505 /* free the vertex program result arrays */
506 for (i
= 0; i
< VERT_RESULT_MAX
; i
++)
507 _mesa_vector4f_free( &store
->results
[i
] );
509 /* free misc arrays */
510 _mesa_vector4f_free( &store
->ndcCoords
);
511 ALIGN_FREE( store
->clipmask
);
514 stage
->privatePtr
= NULL
;
520 validate_vp_stage(GLcontext
*ctx
, struct tnl_pipeline_stage
*stage
)
522 if (ctx
->VertexProgram
._Current
) {
523 _swrast_update_texture_samplers(ctx
);
530 * Public description of this pipeline stage.
532 const struct tnl_pipeline_stage _tnl_vertex_program_stage
=
535 NULL
, /* private_data */
536 init_vp
, /* create */
538 validate_vp_stage
, /* validate */
539 run_vp
/* run -- initially set to ctr */