2 * Mesa 3-D graphics library
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 * OTHER DEALINGS IN THE SOFTWARE.
25 * Keith Whitwell <keithw@vmware.com>
29 #include "main/glheader.h"
30 #include "main/bufferobj.h"
31 #include "main/compiler.h"
32 #include "main/context.h"
33 #include "main/enums.h"
34 #include "main/state.h"
35 #include "main/vtxfmt.h"
37 #include "vbo_context.h"
42 vbo_exec_debug_verts( struct vbo_exec_context
*exec
)
44 GLuint count
= exec
->vtx
.vert_count
;
47 printf("%s: %u vertices %d primitives, %d vertsize\n",
51 exec
->vtx
.vertex_size
);
53 for (i
= 0 ; i
< exec
->vtx
.prim_count
; i
++) {
54 struct _mesa_prim
*prim
= &exec
->vtx
.prim
[i
];
55 printf(" prim %d: %s%s %d..%d %s %s\n",
57 _mesa_lookup_prim_by_nr(prim
->mode
),
58 prim
->weak
? " (weak)" : "",
60 prim
->start
+ prim
->count
,
61 prim
->begin
? "BEGIN" : "(wrap)",
62 prim
->end
? "END" : "(wrap)");
68 * Copy zero, one or two vertices from the current vertex buffer into
69 * the temporary "copy" buffer.
70 * This is used when a single primitive overflows a vertex buffer and
71 * we need to continue the primitive in a new vertex buffer.
72 * The temporary "copy" buffer holds the vertices which need to get
73 * copied from the old buffer to the new one.
76 vbo_copy_vertices( struct vbo_exec_context
*exec
)
78 struct _mesa_prim
*last_prim
= &exec
->vtx
.prim
[exec
->vtx
.prim_count
- 1];
79 const GLuint nr
= last_prim
->count
;
81 const GLuint sz
= exec
->vtx
.vertex_size
;
82 fi_type
*dst
= exec
->vtx
.copied
.buffer
;
83 const fi_type
*src
= exec
->vtx
.buffer_map
+ last_prim
->start
* sz
;
85 switch (exec
->ctx
->Driver
.CurrentExecPrimitive
) {
90 for (i
= 0 ; i
< ovf
; i
++)
91 memcpy( dst
+i
*sz
, src
+(nr
-ovf
+i
)*sz
, sz
* sizeof(GLfloat
) );
95 for (i
= 0 ; i
< ovf
; i
++)
96 memcpy( dst
+i
*sz
, src
+(nr
-ovf
+i
)*sz
, sz
* sizeof(GLfloat
) );
100 for (i
= 0 ; i
< ovf
; i
++)
101 memcpy( dst
+i
*sz
, src
+(nr
-ovf
+i
)*sz
, sz
* sizeof(GLfloat
) );
108 memcpy( dst
, src
+(nr
-1)*sz
, sz
* sizeof(GLfloat
) );
112 case GL_TRIANGLE_FAN
:
118 memcpy( dst
, src
+0, sz
* sizeof(GLfloat
) );
122 memcpy( dst
, src
+0, sz
* sizeof(GLfloat
) );
123 memcpy( dst
+sz
, src
+(nr
-1)*sz
, sz
* sizeof(GLfloat
) );
126 case GL_TRIANGLE_STRIP
:
127 /* no parity issue, but need to make sure the tri is not drawn twice */
144 for (i
= 0 ; i
< ovf
; i
++)
145 memcpy( dst
+i
*sz
, src
+(nr
-ovf
+i
)*sz
, sz
* sizeof(GLfloat
) );
147 case PRIM_OUTSIDE_BEGIN_END
:
157 /* TODO: populate these as the vertex is defined:
160 vbo_exec_bind_arrays( struct gl_context
*ctx
)
162 struct vbo_context
*vbo
= vbo_context(ctx
);
163 struct vbo_exec_context
*exec
= &vbo
->exec
;
164 struct gl_client_array
*arrays
= exec
->vtx
.arrays
;
167 GLbitfield64 varying_inputs
= 0x0;
169 /* Install the default (ie Current) attributes first, then overlay
172 switch (get_program_mode(exec
->ctx
)) {
174 for (attr
= 0; attr
< VERT_ATTRIB_FF_MAX
; attr
++) {
175 exec
->vtx
.inputs
[attr
] = &vbo
->currval
[VBO_ATTRIB_POS
+attr
];
177 for (attr
= 0; attr
< MAT_ATTRIB_MAX
; attr
++) {
178 assert(VERT_ATTRIB_GENERIC(attr
) < ARRAY_SIZE(exec
->vtx
.inputs
));
179 exec
->vtx
.inputs
[VERT_ATTRIB_GENERIC(attr
)] =
180 &vbo
->currval
[VBO_ATTRIB_MAT_FRONT_AMBIENT
+attr
];
182 map
= vbo
->map_vp_none
;
185 for (attr
= 0; attr
< VERT_ATTRIB_FF_MAX
; attr
++) {
186 exec
->vtx
.inputs
[attr
] = &vbo
->currval
[VBO_ATTRIB_POS
+attr
];
188 for (attr
= 0; attr
< VERT_ATTRIB_GENERIC_MAX
; attr
++) {
189 assert(VERT_ATTRIB_GENERIC(attr
) < ARRAY_SIZE(exec
->vtx
.inputs
));
190 exec
->vtx
.inputs
[VERT_ATTRIB_GENERIC(attr
)] =
191 &vbo
->currval
[VBO_ATTRIB_GENERIC0
+attr
];
193 map
= vbo
->map_vp_arb
;
195 /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
196 * In that case we effectively need to route the data from
197 * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
199 if ((ctx
->VertexProgram
._Current
->Base
.InputsRead
& VERT_BIT_POS
) == 0 &&
200 (ctx
->VertexProgram
._Current
->Base
.InputsRead
& VERT_BIT_GENERIC0
)) {
201 exec
->vtx
.inputs
[VERT_ATTRIB_GENERIC0
] = exec
->vtx
.inputs
[0];
202 exec
->vtx
.attrsz
[VERT_ATTRIB_GENERIC0
] = exec
->vtx
.attrsz
[0];
203 exec
->vtx
.attrptr
[VERT_ATTRIB_GENERIC0
] = exec
->vtx
.attrptr
[0];
204 exec
->vtx
.attrsz
[0] = 0;
211 for (attr
= 0; attr
< VERT_ATTRIB_MAX
; attr
++) {
212 const GLuint src
= map
[attr
];
214 if (exec
->vtx
.attrsz
[src
]) {
215 GLsizeiptr offset
= (GLbyte
*)exec
->vtx
.attrptr
[src
] -
216 (GLbyte
*)exec
->vtx
.vertex
;
218 /* override the default array set above */
219 assert(attr
< ARRAY_SIZE(exec
->vtx
.inputs
));
220 assert(attr
< ARRAY_SIZE(exec
->vtx
.arrays
)); /* arrays[] */
221 exec
->vtx
.inputs
[attr
] = &arrays
[attr
];
223 if (_mesa_is_bufferobj(exec
->vtx
.bufferobj
)) {
224 /* a real buffer obj: Ptr is an offset, not a pointer */
225 assert(exec
->vtx
.bufferobj
->Mappings
[MAP_INTERNAL
].Pointer
);
227 arrays
[attr
].Ptr
= (GLubyte
*)
228 exec
->vtx
.bufferobj
->Mappings
[MAP_INTERNAL
].Offset
+ offset
;
231 /* Ptr into ordinary app memory */
232 arrays
[attr
].Ptr
= (GLubyte
*)exec
->vtx
.buffer_map
+ offset
;
234 arrays
[attr
].Size
= exec
->vtx
.attrsz
[src
];
235 arrays
[attr
].StrideB
= exec
->vtx
.vertex_size
* sizeof(GLfloat
);
236 arrays
[attr
].Stride
= exec
->vtx
.vertex_size
* sizeof(GLfloat
);
237 arrays
[attr
].Type
= exec
->vtx
.attrtype
[src
];
238 arrays
[attr
].Integer
=
239 vbo_attrtype_to_integer_flag(exec
->vtx
.attrtype
[src
]);
240 arrays
[attr
].Format
= GL_RGBA
;
241 arrays
[attr
].Enabled
= 1;
242 arrays
[attr
]._ElementSize
= arrays
[attr
].Size
* sizeof(GLfloat
);
243 _mesa_reference_buffer_object(ctx
,
244 &arrays
[attr
].BufferObj
,
245 exec
->vtx
.bufferobj
);
247 varying_inputs
|= VERT_BIT(attr
);
251 _mesa_set_varying_vp_inputs( ctx
, varying_inputs
);
252 ctx
->NewDriverState
|= ctx
->DriverFlags
.NewArray
;
257 * Unmap the VBO. This is called before drawing.
260 vbo_exec_vtx_unmap( struct vbo_exec_context
*exec
)
262 if (_mesa_is_bufferobj(exec
->vtx
.bufferobj
)) {
263 struct gl_context
*ctx
= exec
->ctx
;
265 if (ctx
->Driver
.FlushMappedBufferRange
) {
266 GLintptr offset
= exec
->vtx
.buffer_used
-
267 exec
->vtx
.bufferobj
->Mappings
[MAP_INTERNAL
].Offset
;
268 GLsizeiptr length
= (exec
->vtx
.buffer_ptr
- exec
->vtx
.buffer_map
) *
272 ctx
->Driver
.FlushMappedBufferRange(ctx
, offset
, length
,
277 exec
->vtx
.buffer_used
+= (exec
->vtx
.buffer_ptr
-
278 exec
->vtx
.buffer_map
) * sizeof(float);
280 assert(exec
->vtx
.buffer_used
<= VBO_VERT_BUFFER_SIZE
);
281 assert(exec
->vtx
.buffer_ptr
!= NULL
);
283 ctx
->Driver
.UnmapBuffer(ctx
, exec
->vtx
.bufferobj
, MAP_INTERNAL
);
284 exec
->vtx
.buffer_map
= NULL
;
285 exec
->vtx
.buffer_ptr
= NULL
;
286 exec
->vtx
.max_vert
= 0;
292 * Map the vertex buffer to begin storing glVertex, glColor, etc data.
295 vbo_exec_vtx_map( struct vbo_exec_context
*exec
)
297 struct gl_context
*ctx
= exec
->ctx
;
298 const GLenum accessRange
= GL_MAP_WRITE_BIT
| /* for MapBufferRange */
299 GL_MAP_INVALIDATE_RANGE_BIT
|
300 GL_MAP_UNSYNCHRONIZED_BIT
|
301 GL_MAP_FLUSH_EXPLICIT_BIT
|
303 const GLenum usage
= GL_STREAM_DRAW_ARB
;
305 if (!_mesa_is_bufferobj(exec
->vtx
.bufferobj
))
308 assert(!exec
->vtx
.buffer_map
);
309 assert(!exec
->vtx
.buffer_ptr
);
311 if (VBO_VERT_BUFFER_SIZE
> exec
->vtx
.buffer_used
+ 1024) {
312 /* The VBO exists and there's room for more */
313 if (exec
->vtx
.bufferobj
->Size
> 0) {
314 exec
->vtx
.buffer_map
=
315 (fi_type
*)ctx
->Driver
.MapBufferRange(ctx
,
316 exec
->vtx
.buffer_used
,
317 (VBO_VERT_BUFFER_SIZE
-
318 exec
->vtx
.buffer_used
),
322 exec
->vtx
.buffer_ptr
= exec
->vtx
.buffer_map
;
325 exec
->vtx
.buffer_ptr
= exec
->vtx
.buffer_map
= NULL
;
329 if (!exec
->vtx
.buffer_map
) {
330 /* Need to allocate a new VBO */
331 exec
->vtx
.buffer_used
= 0;
333 if (ctx
->Driver
.BufferData(ctx
, GL_ARRAY_BUFFER_ARB
,
334 VBO_VERT_BUFFER_SIZE
,
337 GL_DYNAMIC_STORAGE_BIT
|
338 GL_CLIENT_STORAGE_BIT
,
339 exec
->vtx
.bufferobj
)) {
340 /* buffer allocation worked, now map the buffer */
341 exec
->vtx
.buffer_map
=
342 (fi_type
*)ctx
->Driver
.MapBufferRange(ctx
,
343 0, VBO_VERT_BUFFER_SIZE
,
349 _mesa_error(ctx
, GL_OUT_OF_MEMORY
, "VBO allocation");
350 exec
->vtx
.buffer_map
= NULL
;
354 exec
->vtx
.buffer_ptr
= exec
->vtx
.buffer_map
;
356 if (!exec
->vtx
.buffer_map
) {
358 _mesa_install_exec_vtxfmt( ctx
, &exec
->vtxfmt_noop
);
361 if (_mesa_using_noop_vtxfmt(ctx
->Exec
)) {
362 /* The no-op functions are installed so switch back to regular
363 * functions. We do this test just to avoid frequent and needless
364 * calls to _mesa_install_exec_vtxfmt().
366 _mesa_install_exec_vtxfmt(ctx
, &exec
->vtxfmt
);
371 printf("map %d..\n", exec
->vtx
.buffer_used
);
377 * Execute the buffer and save copied verts.
378 * \param keep_unmapped if true, leave the VBO unmapped when we're done.
381 vbo_exec_vtx_flush(struct vbo_exec_context
*exec
, GLboolean keepUnmapped
)
384 vbo_exec_debug_verts( exec
);
386 if (exec
->vtx
.prim_count
&&
387 exec
->vtx
.vert_count
) {
389 exec
->vtx
.copied
.nr
= vbo_copy_vertices( exec
);
391 if (exec
->vtx
.copied
.nr
!= exec
->vtx
.vert_count
) {
392 struct gl_context
*ctx
= exec
->ctx
;
394 /* Before the update_state() as this may raise _NEW_VARYING_VP_INPUTS
395 * from _mesa_set_varying_vp_inputs().
397 vbo_exec_bind_arrays( ctx
);
400 _mesa_update_state( ctx
);
402 if (_mesa_is_bufferobj(exec
->vtx
.bufferobj
)) {
403 vbo_exec_vtx_unmap( exec
);
407 printf("%s %d %d\n", __func__
, exec
->vtx
.prim_count
,
408 exec
->vtx
.vert_count
);
410 vbo_context(ctx
)->draw_prims( ctx
,
412 exec
->vtx
.prim_count
,
416 exec
->vtx
.vert_count
- 1,
419 /* If using a real VBO, get new storage -- unless asked not to.
421 if (_mesa_is_bufferobj(exec
->vtx
.bufferobj
) && !keepUnmapped
) {
422 vbo_exec_vtx_map( exec
);
427 /* May have to unmap explicitly if we didn't draw:
430 _mesa_is_bufferobj(exec
->vtx
.bufferobj
) &&
431 exec
->vtx
.buffer_map
) {
432 vbo_exec_vtx_unmap( exec
);
435 if (keepUnmapped
|| exec
->vtx
.vertex_size
== 0)
436 exec
->vtx
.max_vert
= 0;
438 exec
->vtx
.max_vert
= ((VBO_VERT_BUFFER_SIZE
- exec
->vtx
.buffer_used
) /
439 (exec
->vtx
.vertex_size
* sizeof(GLfloat
)));
441 exec
->vtx
.buffer_ptr
= exec
->vtx
.buffer_map
;
442 exec
->vtx
.prim_count
= 0;
443 exec
->vtx
.vert_count
= 0;