1 /**************************************************************************
3 Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
4 VA Linux Systems Inc., Fremont, California.
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 <keithw@vmware.com>
35 #include "main/glheader.h"
36 #include "main/mtypes.h"
37 #include "main/colormac.h"
38 #include "main/enums.h"
39 #include "main/imports.h"
40 #include "main/macros.h"
41 #include "main/simple_list.h"
43 #include "math/m_xform.h"
45 #include "swrast_setup/swrast_setup.h"
48 #include "tnl/t_context.h"
49 #include "tnl/t_pipeline.h"
51 #include "radeon_context.h"
52 #include "radeon_ioctl.h"
53 #include "radeon_state.h"
54 #include "radeon_swtcl.h"
55 #include "radeon_tcl.h"
56 #include "radeon_debug.h"
59 /* R100: xyzw, c0, c1/fog, stq[0..2] = 4+1+1+3*3 = 15 right? */
60 /* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */
61 #define RADEON_MAX_TNL_VERTEX_SIZE (15 * sizeof(GLfloat)) /* for mesa _tnl stage */
63 /***********************************************************************
65 ***********************************************************************/
67 #define EMIT_ATTR( ATTR, STYLE, F0 ) \
69 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \
70 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \
71 rmesa->radeon.swtcl.vertex_attr_count++; \
75 #define EMIT_PAD( N ) \
77 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0; \
78 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD; \
79 rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N); \
80 rmesa->radeon.swtcl.vertex_attr_count++; \
83 static GLuint radeon_cp_vc_frmts
[3][2] =
85 { RADEON_CP_VC_FRMT_ST0
, RADEON_CP_VC_FRMT_ST0
| RADEON_CP_VC_FRMT_Q0
},
86 { RADEON_CP_VC_FRMT_ST1
, RADEON_CP_VC_FRMT_ST1
| RADEON_CP_VC_FRMT_Q1
},
87 { RADEON_CP_VC_FRMT_ST2
, RADEON_CP_VC_FRMT_ST2
| RADEON_CP_VC_FRMT_Q2
},
90 static void radeonSetVertexFormat( struct gl_context
*ctx
)
92 r100ContextPtr rmesa
= R100_CONTEXT( ctx
);
93 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
94 struct vertex_buffer
*VB
= &tnl
->vb
;
95 GLbitfield64 index_bitset
= tnl
->render_inputs_bitset
;
101 if ( VB
->NdcPtr
!= NULL
) {
102 VB
->AttribPtr
[VERT_ATTRIB_POS
] = VB
->NdcPtr
;
105 VB
->AttribPtr
[VERT_ATTRIB_POS
] = VB
->ClipPtr
;
108 assert( VB
->AttribPtr
[VERT_ATTRIB_POS
] != NULL
);
109 rmesa
->radeon
.swtcl
.vertex_attr_count
= 0;
111 /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
112 * build up a hardware vertex.
114 if ( !rmesa
->swtcl
.needproj
||
115 (index_bitset
& BITFIELD64_RANGE(_TNL_ATTRIB_TEX0
, _TNL_NUM_TEX
))) {
117 EMIT_ATTR( _TNL_ATTRIB_POS
, EMIT_4F
,
118 RADEON_CP_VC_FRMT_XY
| RADEON_CP_VC_FRMT_Z
| RADEON_CP_VC_FRMT_W0
);
122 EMIT_ATTR( _TNL_ATTRIB_POS
, EMIT_3F
,
123 RADEON_CP_VC_FRMT_XY
| RADEON_CP_VC_FRMT_Z
);
127 rmesa
->swtcl
.coloroffset
= offset
;
128 #if MESA_LITTLE_ENDIAN
129 EMIT_ATTR( _TNL_ATTRIB_COLOR0
, EMIT_4UB_4F_RGBA
,
130 RADEON_CP_VC_FRMT_PKCOLOR
);
132 EMIT_ATTR( _TNL_ATTRIB_COLOR0
, EMIT_4UB_4F_ABGR
,
133 RADEON_CP_VC_FRMT_PKCOLOR
);
137 rmesa
->swtcl
.specoffset
= 0;
139 (BITFIELD64_BIT(_TNL_ATTRIB_COLOR1
) | BITFIELD64_BIT(_TNL_ATTRIB_FOG
))) {
141 #if MESA_LITTLE_ENDIAN
142 if (index_bitset
& BITFIELD64_BIT(_TNL_ATTRIB_COLOR1
)) {
143 rmesa
->swtcl
.specoffset
= offset
;
144 EMIT_ATTR( _TNL_ATTRIB_COLOR1
, EMIT_3UB_3F_RGB
,
145 RADEON_CP_VC_FRMT_PKSPEC
);
151 if (index_bitset
& BITFIELD64_BIT(_TNL_ATTRIB_FOG
)) {
152 EMIT_ATTR( _TNL_ATTRIB_FOG
, EMIT_1UB_1F
,
153 RADEON_CP_VC_FRMT_PKSPEC
);
159 if (index_bitset
& BITFIELD64_BIT(_TNL_ATTRIB_FOG
)) {
160 EMIT_ATTR( _TNL_ATTRIB_FOG
, EMIT_1UB_1F
,
161 RADEON_CP_VC_FRMT_PKSPEC
);
167 if (index_bitset
& BITFIELD64_BIT(_TNL_ATTRIB_COLOR1
)) {
168 rmesa
->swtcl
.specoffset
= offset
;
169 EMIT_ATTR( _TNL_ATTRIB_COLOR1
, EMIT_3UB_3F_BGR
,
170 RADEON_CP_VC_FRMT_PKSPEC
);
178 if (index_bitset
& BITFIELD64_RANGE(_TNL_ATTRIB_TEX0
, _TNL_NUM_TEX
)) {
181 for (i
= 0; i
< ctx
->Const
.MaxTextureUnits
; i
++) {
182 if (index_bitset
& BITFIELD64_BIT(_TNL_ATTRIB_TEX(i
))) {
183 GLuint sz
= VB
->AttribPtr
[_TNL_ATTRIB_TEX0
+ i
]->size
;
188 EMIT_ATTR( _TNL_ATTRIB_TEX0
+i
, EMIT_2F
,
189 radeon_cp_vc_frmts
[i
][0] );
192 if (ctx
->Texture
.Unit
[i
]._Current
&&
193 ctx
->Texture
.Unit
[i
]._Current
->Target
== GL_TEXTURE_CUBE_MAP
) {
194 EMIT_ATTR( _TNL_ATTRIB_TEX0
+i
, EMIT_3F
,
195 radeon_cp_vc_frmts
[i
][1] );
197 EMIT_ATTR( _TNL_ATTRIB_TEX0
+i
, EMIT_2F
,
198 radeon_cp_vc_frmts
[i
][0] );
202 if (ctx
->Texture
.Unit
[i
]._Current
&&
203 ctx
->Texture
.Unit
[i
]._Current
->Target
== GL_TEXTURE_CUBE_MAP
) {
204 EMIT_ATTR( _TNL_ATTRIB_TEX0
+i
, EMIT_3F
,
205 radeon_cp_vc_frmts
[i
][1] );
207 EMIT_ATTR( _TNL_ATTRIB_TEX0
+i
, EMIT_3F_XYW
,
208 radeon_cp_vc_frmts
[i
][1] );
218 if (rmesa
->radeon
.tnl_index_bitset
!= index_bitset
||
219 fmt_0
!= rmesa
->swtcl
.vertex_format
) {
220 RADEON_NEWPRIM(rmesa
);
221 rmesa
->swtcl
.vertex_format
= fmt_0
;
222 rmesa
->radeon
.swtcl
.vertex_size
=
223 _tnl_install_attrs( ctx
,
224 rmesa
->radeon
.swtcl
.vertex_attrs
,
225 rmesa
->radeon
.swtcl
.vertex_attr_count
,
227 rmesa
->radeon
.swtcl
.vertex_size
/= 4;
228 rmesa
->radeon
.tnl_index_bitset
= index_bitset
;
229 radeon_print(RADEON_SWRENDER
, RADEON_VERBOSE
,
230 "%s: vertex_size= %d floats\n", __FUNCTION__
, rmesa
->radeon
.swtcl
.vertex_size
);
234 static void radeon_predict_emit_size( r100ContextPtr rmesa
)
237 if (!rmesa
->radeon
.swtcl
.emit_prediction
) {
238 const int state_size
= radeonCountStateEmitSize( &rmesa
->radeon
);
239 const int scissor_size
= 8;
240 const int prims_size
= 8;
241 const int vertex_size
= 7;
243 if (rcommonEnsureCmdBufSpace(&rmesa
->radeon
,
245 (scissor_size
+ prims_size
+ vertex_size
),
247 rmesa
->radeon
.swtcl
.emit_prediction
= radeonCountStateEmitSize( &rmesa
->radeon
);
249 rmesa
->radeon
.swtcl
.emit_prediction
= state_size
;
250 rmesa
->radeon
.swtcl
.emit_prediction
+= scissor_size
+ prims_size
+ vertex_size
251 + rmesa
->radeon
.cmdbuf
.cs
->cdw
;
255 static void radeonRenderStart( struct gl_context
*ctx
)
257 r100ContextPtr rmesa
= R100_CONTEXT( ctx
);
259 radeonSetVertexFormat( ctx
);
261 if (rmesa
->radeon
.dma
.flush
!= 0 &&
262 rmesa
->radeon
.dma
.flush
!= rcommon_flush_last_swtcl_prim
)
263 rmesa
->radeon
.dma
.flush( ctx
);
268 * Set vertex state for SW TCL. The primary purpose of this function is to
269 * determine in advance whether or not the hardware can / should do the
270 * projection divide or Mesa should do it.
272 void radeonChooseVertexState( struct gl_context
*ctx
)
274 r100ContextPtr rmesa
= R100_CONTEXT( ctx
);
275 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
277 GLuint se_coord_fmt
= rmesa
->hw
.set
.cmd
[SET_SE_COORDFMT
];
278 GLboolean unfilled
= (ctx
->Polygon
.FrontMode
!= GL_FILL
||
279 ctx
->Polygon
.BackMode
!= GL_FILL
);
280 GLboolean twosided
= ctx
->Light
.Enabled
&& ctx
->Light
.Model
.TwoSide
;
282 se_coord_fmt
&= ~(RADEON_VTX_XY_PRE_MULT_1_OVER_W0
|
283 RADEON_VTX_Z_PRE_MULT_1_OVER_W0
|
284 RADEON_VTX_W0_IS_NOT_1_OVER_W0
);
286 /* We must ensure that we don't do _tnl_need_projected_coords while in a
287 * rasterization fallback. As this function will be called again when we
288 * leave a rasterization fallback, we can just skip it for now.
290 if (rmesa
->radeon
.Fallback
!= 0)
293 /* HW perspective divide is a win, but tiny vertex formats are a
297 if ((0 == (tnl
->render_inputs_bitset
&
298 (BITFIELD64_RANGE(_TNL_ATTRIB_TEX0
, _TNL_NUM_TEX
)
299 | BITFIELD64_BIT(_TNL_ATTRIB_COLOR1
))))
302 rmesa
->swtcl
.needproj
= GL_TRUE
;
303 se_coord_fmt
|= (RADEON_VTX_XY_PRE_MULT_1_OVER_W0
|
304 RADEON_VTX_Z_PRE_MULT_1_OVER_W0
);
307 rmesa
->swtcl
.needproj
= GL_FALSE
;
308 se_coord_fmt
|= (RADEON_VTX_W0_IS_NOT_1_OVER_W0
);
311 _tnl_need_projected_coords( ctx
, rmesa
->swtcl
.needproj
);
313 if ( se_coord_fmt
!= rmesa
->hw
.set
.cmd
[SET_SE_COORDFMT
] ) {
314 RADEON_STATECHANGE( rmesa
, set
);
315 rmesa
->hw
.set
.cmd
[SET_SE_COORDFMT
] = se_coord_fmt
;
319 void r100_swtcl_flush(struct gl_context
*ctx
, uint32_t current_offset
)
321 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
325 radeonEmitState(&rmesa
->radeon
);
326 radeonEmitVertexAOS( rmesa
,
327 rmesa
->radeon
.swtcl
.vertex_size
,
328 rmesa
->radeon
.swtcl
.bo
,
332 radeonEmitVbufPrim( rmesa
,
333 rmesa
->swtcl
.vertex_format
,
334 rmesa
->radeon
.swtcl
.hw_primitive
,
335 rmesa
->radeon
.swtcl
.numverts
);
336 if ( rmesa
->radeon
.swtcl
.emit_prediction
< rmesa
->radeon
.cmdbuf
.cs
->cdw
)
337 WARN_ONCE("Rendering was %d commands larger than predicted size."
338 " We might overflow command buffer.\n",
339 rmesa
->radeon
.cmdbuf
.cs
->cdw
- rmesa
->radeon
.swtcl
.emit_prediction
);
342 rmesa
->radeon
.swtcl
.emit_prediction
= 0;
347 * Render unclipped vertex buffers by emitting vertices directly to
348 * dma buffers. Use strip/fan hardware primitives where possible.
349 * Try to simulate missing primitives with indexed vertices.
351 #define HAVE_POINTS 1
353 #define HAVE_LINE_STRIPS 1
354 #define HAVE_TRIANGLES 1
355 #define HAVE_TRI_STRIPS 1
356 #define HAVE_TRI_STRIP_1 0
357 #define HAVE_TRI_FANS 1
359 #define HAVE_QUAD_STRIPS 0
360 #define HAVE_POLYGONS 0
361 /* \todo: is it possible to make "ELTS" work with t_vertex code ? */
364 static const GLuint hw_prim
[GL_POLYGON
+1] = {
365 RADEON_CP_VC_CNTL_PRIM_TYPE_POINT
,
366 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE
,
368 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE_STRIP
,
369 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
370 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_STRIP
,
371 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_FAN
,
378 radeonDmaPrimitive( r100ContextPtr rmesa
, GLenum prim
)
380 RADEON_NEWPRIM( rmesa
);
381 rmesa
->radeon
.swtcl
.hw_primitive
= hw_prim
[prim
];
382 // assert(rmesa->radeon.dma.current.ptr == rmesa->radeon.dma.current.start);
385 static void* radeon_alloc_verts( r100ContextPtr rmesa
, GLuint nr
, GLuint size
)
389 radeon_predict_emit_size( rmesa
);
390 rv
= rcommonAllocDmaLowVerts( &rmesa
->radeon
, nr
, size
);
395 #define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx)
396 #define INIT( prim ) radeonDmaPrimitive( rmesa, prim )
397 #define FLUSH() RADEON_NEWPRIM( rmesa )
398 #define GET_CURRENT_VB_MAX_VERTS() 10\
399 // (((int)rmesa->radeon.dma.current.end - (int)rmesa->radeon.dma.current.ptr) / (rmesa->radeon.swtcl.vertex_size*4))
400 #define GET_SUBSEQUENT_VB_MAX_VERTS() \
401 ((RADEON_BUFFER_SIZE) / (rmesa->radeon.swtcl.vertex_size*4))
402 #define ALLOC_VERTS( nr ) radeon_alloc_verts( rmesa, nr, rmesa->radeon.swtcl.vertex_size * 4 )
403 #define EMIT_VERTS( ctx, j, nr, buf ) \
404 _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf)
406 #define TAG(x) radeon_dma_##x
407 #include "tnl_dd/t_dd_dmatmp.h"
410 /**********************************************************************/
411 /* Render pipeline stage */
412 /**********************************************************************/
415 static GLboolean
radeon_run_render( struct gl_context
*ctx
,
416 struct tnl_pipeline_stage
*stage
)
418 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
419 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
420 struct vertex_buffer
*VB
= &tnl
->vb
;
421 tnl_render_func
*tab
= TAG(render_tab_verts
);
424 if (rmesa
->radeon
.swtcl
.RenderIndex
!= 0 ||
425 !radeon_dma_validate_render( ctx
, VB
))
428 radeon_prepare_render(&rmesa
->radeon
);
429 if (rmesa
->radeon
.NewGLState
)
430 radeonValidateState( ctx
);
432 tnl
->Driver
.Render
.Start( ctx
);
434 for (i
= 0 ; i
< VB
->PrimitiveCount
; i
++)
436 GLuint prim
= VB
->Primitive
[i
].mode
;
437 GLuint start
= VB
->Primitive
[i
].start
;
438 GLuint length
= VB
->Primitive
[i
].count
;
443 radeon_print(RADEON_SWRENDER
, RADEON_NORMAL
,
444 "radeon_render.c: prim %s %d..%d\n",
445 _mesa_lookup_enum_by_nr(prim
& PRIM_MODE_MASK
),
446 start
, start
+length
);
449 tab
[prim
& PRIM_MODE_MASK
]( ctx
, start
, start
+ length
, prim
);
452 tnl
->Driver
.Render
.Finish( ctx
);
454 return GL_FALSE
; /* finished the pipe */
459 const struct tnl_pipeline_stage _radeon_render_stage
=
466 radeon_run_render
/* run */
470 /**************************************************************************/
473 static const GLuint reduced_hw_prim
[GL_POLYGON
+1] = {
474 RADEON_CP_VC_CNTL_PRIM_TYPE_POINT
,
475 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE
,
476 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE
,
477 RADEON_CP_VC_CNTL_PRIM_TYPE_LINE
,
478 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
479 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
480 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
481 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
482 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
,
483 RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST
486 static void radeonRasterPrimitive( struct gl_context
*ctx
, GLuint hwprim
);
487 static void radeonRenderPrimitive( struct gl_context
*ctx
, GLenum prim
);
488 static void radeonResetLineStipple( struct gl_context
*ctx
);
491 /***********************************************************************
492 * Emit primitives as inline vertices *
493 ***********************************************************************/
497 #define CTX_ARG r100ContextPtr rmesa
498 #define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
499 #define ALLOC_VERTS( n, size ) radeon_alloc_verts( rmesa, n, (size) * 4 )
502 r100ContextPtr rmesa = R100_CONTEXT(ctx); \
503 const char *radeonverts = (char *)rmesa->radeon.swtcl.verts;
504 #define VERT(x) (radeonVertex *)(radeonverts + ((x) * (vertsize) * sizeof(int)))
505 #define VERTEX radeonVertex
507 #define TAG(x) radeon_##x
508 #include "tnl_dd/t_dd_triemit.h"
511 /***********************************************************************
512 * Macros for t_dd_tritmp.h to draw basic primitives *
513 ***********************************************************************/
515 #define QUAD( a, b, c, d ) radeon_quad( rmesa, a, b, c, d )
516 #define TRI( a, b, c ) radeon_triangle( rmesa, a, b, c )
517 #define LINE( a, b ) radeon_line( rmesa, a, b )
518 #define POINT( a ) radeon_point( rmesa, a )
520 /***********************************************************************
521 * Build render functions from dd templates *
522 ***********************************************************************/
524 #define RADEON_TWOSIDE_BIT 0x01
525 #define RADEON_UNFILLED_BIT 0x02
526 #define RADEON_MAX_TRIFUNC 0x04
530 tnl_points_func points
;
532 tnl_triangle_func triangle
;
534 } rast_tab
[RADEON_MAX_TRIFUNC
];
537 #define DO_FALLBACK 0
539 #define DO_UNFILLED (IND & RADEON_UNFILLED_BIT)
540 #define DO_TWOSIDE (IND & RADEON_TWOSIDE_BIT)
546 #define DO_FULL_QUAD 1
549 #define HAVE_BACK_COLORS 0
550 #define HAVE_HW_FLATSHADE 1
553 #define DEPTH_SCALE 1.0
554 #define UNFILLED_TRI unfilled_tri
555 #define UNFILLED_QUAD unfilled_quad
556 #define VERT_X(_v) _v->v.x
557 #define VERT_Y(_v) _v->v.y
558 #define VERT_Z(_v) _v->v.z
559 #define AREA_IS_CCW( a ) (a < 0)
560 #define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + ((e) * rmesa->radeon.swtcl.vertex_size * sizeof(int)))
562 #define VERT_SET_RGBA( v, c ) \
564 radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]); \
565 UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \
566 UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \
567 UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \
568 UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]); \
571 #define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
573 #define VERT_SET_SPEC( v, c ) \
576 radeon_color_t *spec = (radeon_color_t *)&((v)->ui[specoffset]); \
577 UNCLAMPED_FLOAT_TO_UBYTE(spec->red, (c)[0]); \
578 UNCLAMPED_FLOAT_TO_UBYTE(spec->green, (c)[1]); \
579 UNCLAMPED_FLOAT_TO_UBYTE(spec->blue, (c)[2]); \
582 #define VERT_COPY_SPEC( v0, v1 ) \
585 radeon_color_t *spec0 = (radeon_color_t *)&((v0)->ui[specoffset]); \
586 radeon_color_t *spec1 = (radeon_color_t *)&((v1)->ui[specoffset]); \
587 spec0->red = spec1->red; \
588 spec0->green = spec1->green; \
589 spec0->blue = spec1->blue; \
593 /* These don't need LE32_TO_CPU() as they used to save and restore
594 * colors which are already in the correct format.
596 #define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset]
597 #define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
598 #define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset]
599 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
605 #define LOCAL_VARS(n) \
606 r100ContextPtr rmesa = R100_CONTEXT(ctx); \
607 GLuint color[n] = {0}, spec[n] = {0}; \
608 GLuint coloroffset = rmesa->swtcl.coloroffset; \
609 GLuint specoffset = rmesa->swtcl.specoffset; \
610 (void) color; (void) spec; (void) coloroffset; (void) specoffset;
612 /***********************************************************************
613 * Helpers for rendering unfilled primitives *
614 ***********************************************************************/
616 #define RASTERIZE(x) radeonRasterPrimitive( ctx, reduced_hw_prim[x] )
617 #define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
620 #include "tnl_dd/t_dd_unfilled.h"
624 /***********************************************************************
625 * Generate GL render functions *
626 ***********************************************************************/
631 #include "tnl_dd/t_dd_tritmp.h"
633 #define IND (RADEON_TWOSIDE_BIT)
634 #define TAG(x) x##_twoside
635 #include "tnl_dd/t_dd_tritmp.h"
637 #define IND (RADEON_UNFILLED_BIT)
638 #define TAG(x) x##_unfilled
639 #include "tnl_dd/t_dd_tritmp.h"
641 #define IND (RADEON_TWOSIDE_BIT|RADEON_UNFILLED_BIT)
642 #define TAG(x) x##_twoside_unfilled
643 #include "tnl_dd/t_dd_tritmp.h"
646 static void init_rast_tab( void )
651 init_twoside_unfilled();
654 /**********************************************************************/
655 /* Render unclipped begin/end objects */
656 /**********************************************************************/
658 #define RENDER_POINTS( start, count ) \
659 for ( ; start < count ; start++) \
660 radeon_point( rmesa, VERT(start) )
661 #define RENDER_LINE( v0, v1 ) \
662 radeon_line( rmesa, VERT(v0), VERT(v1) )
663 #define RENDER_TRI( v0, v1, v2 ) \
664 radeon_triangle( rmesa, VERT(v0), VERT(v1), VERT(v2) )
665 #define RENDER_QUAD( v0, v1, v2, v3 ) \
666 radeon_quad( rmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) )
668 #define INIT(x) do { \
669 radeonRenderPrimitive( ctx, x ); \
673 r100ContextPtr rmesa = R100_CONTEXT(ctx); \
674 const GLuint vertsize = rmesa->radeon.swtcl.vertex_size; \
675 const char *radeonverts = (char *)rmesa->radeon.swtcl.verts; \
676 const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \
677 const GLboolean stipple = ctx->Line.StippleFlag; \
678 (void) elt; (void) stipple;
679 #define RESET_STIPPLE if ( stipple ) radeonResetLineStipple( ctx );
680 #define RESET_OCCLUSION
681 #define PRESERVE_VB_DEFS
683 #define TAG(x) radeon_##x##_verts
684 #include "tnl/t_vb_rendertmp.h"
687 #define TAG(x) radeon_##x##_elts
688 #define ELT(x) elt[x]
689 #include "tnl/t_vb_rendertmp.h"
693 /**********************************************************************/
694 /* Choose render functions */
695 /**********************************************************************/
697 void radeonChooseRenderState( struct gl_context
*ctx
)
699 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
700 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
702 GLboolean unfilled
= (ctx
->Polygon
.FrontMode
!= GL_FILL
||
703 ctx
->Polygon
.BackMode
!= GL_FILL
);
704 GLboolean twosided
= ctx
->Light
.Enabled
&& ctx
->Light
.Model
.TwoSide
;
706 if (!rmesa
->radeon
.TclFallback
|| rmesa
->radeon
.Fallback
)
710 index
|= RADEON_TWOSIDE_BIT
;
712 index
|= RADEON_UNFILLED_BIT
;
714 if (index
!= rmesa
->radeon
.swtcl
.RenderIndex
) {
715 tnl
->Driver
.Render
.Points
= rast_tab
[index
].points
;
716 tnl
->Driver
.Render
.Line
= rast_tab
[index
].line
;
717 tnl
->Driver
.Render
.ClippedLine
= rast_tab
[index
].line
;
718 tnl
->Driver
.Render
.Triangle
= rast_tab
[index
].triangle
;
719 tnl
->Driver
.Render
.Quad
= rast_tab
[index
].quad
;
722 tnl
->Driver
.Render
.PrimTabVerts
= radeon_render_tab_verts
;
723 tnl
->Driver
.Render
.PrimTabElts
= radeon_render_tab_elts
;
724 tnl
->Driver
.Render
.ClippedPolygon
= radeon_fast_clipped_poly
;
726 tnl
->Driver
.Render
.PrimTabVerts
= _tnl_render_tab_verts
;
727 tnl
->Driver
.Render
.PrimTabElts
= _tnl_render_tab_elts
;
728 tnl
->Driver
.Render
.ClippedPolygon
= _tnl_RenderClippedPolygon
;
731 rmesa
->radeon
.swtcl
.RenderIndex
= index
;
736 /**********************************************************************/
737 /* High level hooks for t_vb_render.c */
738 /**********************************************************************/
741 static void radeonRasterPrimitive( struct gl_context
*ctx
, GLuint hwprim
)
743 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
745 if (rmesa
->radeon
.swtcl
.hw_primitive
!= hwprim
) {
746 RADEON_NEWPRIM( rmesa
);
747 rmesa
->radeon
.swtcl
.hw_primitive
= hwprim
;
751 static void radeonRenderPrimitive( struct gl_context
*ctx
, GLenum prim
)
753 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
754 GLboolean unfilled
= (ctx
->Polygon
.FrontMode
!= GL_FILL
||
755 ctx
->Polygon
.BackMode
!= GL_FILL
);
757 rmesa
->radeon
.swtcl
.render_primitive
= prim
;
758 if (prim
< GL_TRIANGLES
|| !unfilled
)
759 radeonRasterPrimitive( ctx
, reduced_hw_prim
[prim
] );
762 static void radeonRenderFinish( struct gl_context
*ctx
)
766 static void radeonResetLineStipple( struct gl_context
*ctx
)
768 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
769 RADEON_STATECHANGE( rmesa
, lin
);
773 /**********************************************************************/
774 /* Transition to/from hardware rasterization. */
775 /**********************************************************************/
777 static const char * const fallbackStrings
[] = {
779 "glDrawBuffer(GL_FRONT_AND_BACK)",
780 "glEnable(GL_STENCIL) without hw stencil buffer",
781 "glRenderMode(selection or feedback)",
785 "Mixing GL_CLAMP_TO_BORDER and GL_CLAMP (or GL_MIRROR_CLAMP_ATI)"
789 static const char *getFallbackString(GLuint bit
)
796 return fallbackStrings
[i
];
800 void radeonFallback( struct gl_context
*ctx
, GLuint bit
, GLboolean mode
)
802 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
803 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
804 GLuint oldfallback
= rmesa
->radeon
.Fallback
;
807 rmesa
->radeon
.Fallback
|= bit
;
808 if (oldfallback
== 0) {
809 radeon_firevertices(&rmesa
->radeon
);
810 TCL_FALLBACK( ctx
, RADEON_TCL_FALLBACK_RASTER
, GL_TRUE
);
811 _swsetup_Wakeup( ctx
);
812 rmesa
->radeon
.swtcl
.RenderIndex
= ~0;
813 if (RADEON_DEBUG
& RADEON_FALLBACKS
) {
814 fprintf(stderr
, "Radeon begin rasterization fallback: 0x%x %s\n",
815 bit
, getFallbackString(bit
));
820 rmesa
->radeon
.Fallback
&= ~bit
;
821 if (oldfallback
== bit
) {
822 _swrast_flush( ctx
);
823 tnl
->Driver
.Render
.Start
= radeonRenderStart
;
824 tnl
->Driver
.Render
.PrimitiveNotify
= radeonRenderPrimitive
;
825 tnl
->Driver
.Render
.Finish
= radeonRenderFinish
;
827 tnl
->Driver
.Render
.BuildVertices
= _tnl_build_vertices
;
828 tnl
->Driver
.Render
.CopyPV
= _tnl_copy_pv
;
829 tnl
->Driver
.Render
.Interp
= _tnl_interp
;
831 tnl
->Driver
.Render
.ResetLineStipple
= radeonResetLineStipple
;
832 TCL_FALLBACK( ctx
, RADEON_TCL_FALLBACK_RASTER
, GL_FALSE
);
833 if (rmesa
->radeon
.TclFallback
) {
834 /* These are already done if rmesa->radeon.TclFallback goes to
835 * zero above. But not if it doesn't (RADEON_NO_TCL for
838 _tnl_invalidate_vertex_state( ctx
, ~0 );
839 _tnl_invalidate_vertices( ctx
, ~0 );
840 rmesa
->radeon
.tnl_index_bitset
= 0;
841 radeonChooseVertexState( ctx
);
842 radeonChooseRenderState( ctx
);
844 if (RADEON_DEBUG
& RADEON_FALLBACKS
) {
845 fprintf(stderr
, "Radeon end rasterization fallback: 0x%x %s\n",
846 bit
, getFallbackString(bit
));
853 /**********************************************************************/
854 /* Initialization. */
855 /**********************************************************************/
857 void radeonInitSwtcl( struct gl_context
*ctx
)
859 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
860 r100ContextPtr rmesa
= R100_CONTEXT(ctx
);
861 static int firsttime
= 1;
867 rmesa
->radeon
.swtcl
.emit_prediction
= 0;
869 tnl
->Driver
.Render
.Start
= radeonRenderStart
;
870 tnl
->Driver
.Render
.Finish
= radeonRenderFinish
;
871 tnl
->Driver
.Render
.PrimitiveNotify
= radeonRenderPrimitive
;
872 tnl
->Driver
.Render
.ResetLineStipple
= radeonResetLineStipple
;
873 tnl
->Driver
.Render
.BuildVertices
= _tnl_build_vertices
;
874 tnl
->Driver
.Render
.CopyPV
= _tnl_copy_pv
;
875 tnl
->Driver
.Render
.Interp
= _tnl_interp
;
877 _tnl_init_vertices( ctx
, ctx
->Const
.MaxArrayLockSize
+ 12,
878 RADEON_MAX_TNL_VERTEX_SIZE
);
880 rmesa
->radeon
.swtcl
.verts
= (GLubyte
*)tnl
->clipspace
.vertex_buf
;
881 rmesa
->radeon
.swtcl
.RenderIndex
= ~0;
882 rmesa
->radeon
.swtcl
.render_primitive
= GL_TRIANGLES
;
883 rmesa
->radeon
.swtcl
.hw_primitive
= 0;