0ef30819e225233ce06c96e51664b998b58c7128
[mesa.git] / src / mesa / vbo / vbo_exec_draw.c
1 /*
2 * Mesa 3-D graphics library
3 *
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
5 *
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:
12 *
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
15 *
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.
23 *
24 * Authors:
25 * Keith Whitwell <keithw@vmware.com>
26 */
27
28 #include <stdio.h>
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"
36
37 #include "vbo_context.h"
38 #include "vbo_noop.h"
39
40
41 static void
42 vbo_exec_debug_verts( struct vbo_exec_context *exec )
43 {
44 GLuint count = exec->vtx.vert_count;
45 GLuint i;
46
47 printf("%s: %u vertices %d primitives, %d vertsize\n",
48 __func__,
49 count,
50 exec->vtx.prim_count,
51 exec->vtx.vertex_size);
52
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",
56 i,
57 _mesa_lookup_prim_by_nr(prim->mode),
58 prim->weak ? " (weak)" : "",
59 prim->start,
60 prim->start + prim->count,
61 prim->begin ? "BEGIN" : "(wrap)",
62 prim->end ? "END" : "(wrap)");
63 }
64 }
65
66
67 /**
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.
74 */
75 static GLuint
76 vbo_copy_vertices( struct vbo_exec_context *exec )
77 {
78 struct _mesa_prim *last_prim = &exec->vtx.prim[exec->vtx.prim_count - 1];
79 const GLuint nr = last_prim->count;
80 GLuint ovf, i;
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;
84
85 switch (exec->ctx->Driver.CurrentExecPrimitive) {
86 case GL_POINTS:
87 return 0;
88 case GL_LINES:
89 ovf = nr&1;
90 for (i = 0 ; i < ovf ; i++)
91 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
92 return i;
93 case GL_TRIANGLES:
94 ovf = nr%3;
95 for (i = 0 ; i < ovf ; i++)
96 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
97 return i;
98 case GL_QUADS:
99 ovf = nr&3;
100 for (i = 0 ; i < ovf ; i++)
101 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
102 return i;
103 case GL_LINE_STRIP:
104 if (nr == 0) {
105 return 0;
106 }
107 else {
108 memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
109 return 1;
110 }
111 case GL_LINE_LOOP:
112 if (last_prim->begin == 0) {
113 /* We're dealing with the second or later section of a split/wrapped
114 * GL_LINE_LOOP. Since we're converting line loops to line strips,
115 * we've already increment the last_prim->start counter by one to
116 * skip the 0th vertex in the loop. We need to undo that (effectively
117 * subtract one from last_prim->start) so that we copy the 0th vertex
118 * to the next vertex buffer.
119 */
120 assert(last_prim->start > 0);
121 src -= sz;
122 }
123 /* fall-through */
124 case GL_TRIANGLE_FAN:
125 case GL_POLYGON:
126 if (nr == 0) {
127 return 0;
128 }
129 else if (nr == 1) {
130 memcpy( dst, src+0, sz * sizeof(GLfloat) );
131 return 1;
132 }
133 else {
134 memcpy( dst, src+0, sz * sizeof(GLfloat) );
135 memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
136 return 2;
137 }
138 case GL_TRIANGLE_STRIP:
139 /* no parity issue, but need to make sure the tri is not drawn twice */
140 if (nr & 1) {
141 last_prim->count--;
142 }
143 /* fallthrough */
144 case GL_QUAD_STRIP:
145 switch (nr) {
146 case 0:
147 ovf = 0;
148 break;
149 case 1:
150 ovf = 1;
151 break;
152 default:
153 ovf = 2 + (nr & 1);
154 break;
155 }
156 for (i = 0 ; i < ovf ; i++)
157 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
158 return i;
159 case PRIM_OUTSIDE_BEGIN_END:
160 return 0;
161 default:
162 assert(0);
163 return 0;
164 }
165 }
166
167
168
169 /* TODO: populate these as the vertex is defined:
170 */
171 static void
172 vbo_exec_bind_arrays( struct gl_context *ctx )
173 {
174 struct vbo_context *vbo = vbo_context(ctx);
175 struct vbo_exec_context *exec = &vbo->exec;
176 struct gl_client_array *arrays = exec->vtx.arrays;
177 const GLuint *map;
178 GLuint attr;
179 GLbitfield64 varying_inputs = 0x0;
180
181 /* Install the default (ie Current) attributes first, then overlay
182 * all active ones.
183 */
184 switch (get_program_mode(exec->ctx)) {
185 case VP_NONE:
186 for (attr = 0; attr < VERT_ATTRIB_FF_MAX; attr++) {
187 exec->vtx.inputs[attr] = &vbo->currval[VBO_ATTRIB_POS+attr];
188 }
189 for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
190 assert(VERT_ATTRIB_GENERIC(attr) < ARRAY_SIZE(exec->vtx.inputs));
191 exec->vtx.inputs[VERT_ATTRIB_GENERIC(attr)] =
192 &vbo->currval[VBO_ATTRIB_MAT_FRONT_AMBIENT+attr];
193 }
194 map = vbo->map_vp_none;
195 break;
196 case VP_ARB:
197 for (attr = 0; attr < VERT_ATTRIB_FF_MAX; attr++) {
198 exec->vtx.inputs[attr] = &vbo->currval[VBO_ATTRIB_POS+attr];
199 }
200 for (attr = 0; attr < VERT_ATTRIB_GENERIC_MAX; attr++) {
201 assert(VERT_ATTRIB_GENERIC(attr) < ARRAY_SIZE(exec->vtx.inputs));
202 exec->vtx.inputs[VERT_ATTRIB_GENERIC(attr)] =
203 &vbo->currval[VBO_ATTRIB_GENERIC0+attr];
204 }
205 map = vbo->map_vp_arb;
206
207 /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
208 * In that case we effectively need to route the data from
209 * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
210 */
211 if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
212 (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
213 exec->vtx.inputs[VERT_ATTRIB_GENERIC0] = exec->vtx.inputs[0];
214 exec->vtx.attrsz[VERT_ATTRIB_GENERIC0] = exec->vtx.attrsz[0];
215 exec->vtx.attrptr[VERT_ATTRIB_GENERIC0] = exec->vtx.attrptr[0];
216 vbo_reset_attr(exec, VERT_ATTRIB_POS);
217 exec->vtx.enabled &= (~BITFIELD64_BIT(VBO_ATTRIB_POS));
218 exec->vtx.enabled |= BITFIELD64_BIT(VBO_ATTRIB_GENERIC0);
219 }
220 break;
221 default:
222 assert(0);
223 }
224
225 for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
226 const GLuint src = map[attr];
227
228 if (exec->vtx.attrsz[src]) {
229 GLsizeiptr offset = (GLbyte *)exec->vtx.attrptr[src] -
230 (GLbyte *)exec->vtx.vertex;
231
232 /* override the default array set above */
233 assert(attr < ARRAY_SIZE(exec->vtx.inputs));
234 assert(attr < ARRAY_SIZE(exec->vtx.arrays)); /* arrays[] */
235 exec->vtx.inputs[attr] = &arrays[attr];
236
237 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
238 /* a real buffer obj: Ptr is an offset, not a pointer */
239 assert(exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Pointer);
240 assert(offset >= 0);
241 arrays[attr].Ptr = (GLubyte *)
242 exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Offset + offset;
243 }
244 else {
245 /* Ptr into ordinary app memory */
246 arrays[attr].Ptr = (GLubyte *)exec->vtx.buffer_map + offset;
247 }
248 arrays[attr].Size = exec->vtx.attrsz[src];
249 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
250 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
251 arrays[attr].Type = exec->vtx.attrtype[src];
252 arrays[attr].Integer =
253 vbo_attrtype_to_integer_flag(exec->vtx.attrtype[src]);
254 arrays[attr].Format = GL_RGBA;
255 arrays[attr].Enabled = 1;
256 arrays[attr]._ElementSize = arrays[attr].Size * sizeof(GLfloat);
257 _mesa_reference_buffer_object(ctx,
258 &arrays[attr].BufferObj,
259 exec->vtx.bufferobj);
260
261 varying_inputs |= VERT_BIT(attr);
262 }
263 }
264
265 _mesa_set_varying_vp_inputs( ctx, varying_inputs );
266 ctx->NewDriverState |= ctx->DriverFlags.NewArray;
267 }
268
269
270 /**
271 * Unmap the VBO. This is called before drawing.
272 */
273 static void
274 vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
275 {
276 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
277 struct gl_context *ctx = exec->ctx;
278
279 if (ctx->Driver.FlushMappedBufferRange) {
280 GLintptr offset = exec->vtx.buffer_used -
281 exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Offset;
282 GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) *
283 sizeof(float);
284
285 if (length)
286 ctx->Driver.FlushMappedBufferRange(ctx, offset, length,
287 exec->vtx.bufferobj,
288 MAP_INTERNAL);
289 }
290
291 exec->vtx.buffer_used += (exec->vtx.buffer_ptr -
292 exec->vtx.buffer_map) * sizeof(float);
293
294 assert(exec->vtx.buffer_used <= VBO_VERT_BUFFER_SIZE);
295 assert(exec->vtx.buffer_ptr != NULL);
296
297 ctx->Driver.UnmapBuffer(ctx, exec->vtx.bufferobj, MAP_INTERNAL);
298 exec->vtx.buffer_map = NULL;
299 exec->vtx.buffer_ptr = NULL;
300 exec->vtx.max_vert = 0;
301 }
302 }
303
304
305 /**
306 * Map the vertex buffer to begin storing glVertex, glColor, etc data.
307 */
308 void
309 vbo_exec_vtx_map( struct vbo_exec_context *exec )
310 {
311 struct gl_context *ctx = exec->ctx;
312 const GLenum accessRange = GL_MAP_WRITE_BIT | /* for MapBufferRange */
313 GL_MAP_INVALIDATE_RANGE_BIT |
314 GL_MAP_UNSYNCHRONIZED_BIT |
315 GL_MAP_FLUSH_EXPLICIT_BIT |
316 MESA_MAP_NOWAIT_BIT;
317 const GLenum usage = GL_STREAM_DRAW_ARB;
318
319 if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
320 return;
321
322 assert(!exec->vtx.buffer_map);
323 assert(!exec->vtx.buffer_ptr);
324
325 if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024) {
326 /* The VBO exists and there's room for more */
327 if (exec->vtx.bufferobj->Size > 0) {
328 exec->vtx.buffer_map =
329 (fi_type *)ctx->Driver.MapBufferRange(ctx,
330 exec->vtx.buffer_used,
331 (VBO_VERT_BUFFER_SIZE -
332 exec->vtx.buffer_used),
333 accessRange,
334 exec->vtx.bufferobj,
335 MAP_INTERNAL);
336 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
337 }
338 else {
339 exec->vtx.buffer_ptr = exec->vtx.buffer_map = NULL;
340 }
341 }
342
343 if (!exec->vtx.buffer_map) {
344 /* Need to allocate a new VBO */
345 exec->vtx.buffer_used = 0;
346
347 if (ctx->Driver.BufferData(ctx, GL_ARRAY_BUFFER_ARB,
348 VBO_VERT_BUFFER_SIZE,
349 NULL, usage,
350 GL_MAP_WRITE_BIT |
351 GL_DYNAMIC_STORAGE_BIT |
352 GL_CLIENT_STORAGE_BIT,
353 exec->vtx.bufferobj)) {
354 /* buffer allocation worked, now map the buffer */
355 exec->vtx.buffer_map =
356 (fi_type *)ctx->Driver.MapBufferRange(ctx,
357 0, VBO_VERT_BUFFER_SIZE,
358 accessRange,
359 exec->vtx.bufferobj,
360 MAP_INTERNAL);
361 }
362 else {
363 _mesa_error(ctx, GL_OUT_OF_MEMORY, "VBO allocation");
364 exec->vtx.buffer_map = NULL;
365 }
366 }
367
368 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
369
370 if (!exec->vtx.buffer_map) {
371 /* out of memory */
372 _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt_noop );
373 }
374 else {
375 if (_mesa_using_noop_vtxfmt(ctx->Exec)) {
376 /* The no-op functions are installed so switch back to regular
377 * functions. We do this test just to avoid frequent and needless
378 * calls to _mesa_install_exec_vtxfmt().
379 */
380 _mesa_install_exec_vtxfmt(ctx, &exec->vtxfmt);
381 }
382 }
383
384 if (0)
385 printf("map %d..\n", exec->vtx.buffer_used);
386 }
387
388
389
390 /**
391 * Execute the buffer and save copied verts.
392 * \param keep_unmapped if true, leave the VBO unmapped when we're done.
393 */
394 void
395 vbo_exec_vtx_flush(struct vbo_exec_context *exec, GLboolean keepUnmapped)
396 {
397 if (0)
398 vbo_exec_debug_verts( exec );
399
400 if (exec->vtx.prim_count &&
401 exec->vtx.vert_count) {
402
403 exec->vtx.copied.nr = vbo_copy_vertices( exec );
404
405 if (exec->vtx.copied.nr != exec->vtx.vert_count) {
406 struct gl_context *ctx = exec->ctx;
407
408 /* Before the update_state() as this may raise _NEW_VARYING_VP_INPUTS
409 * from _mesa_set_varying_vp_inputs().
410 */
411 vbo_exec_bind_arrays( ctx );
412
413 if (ctx->NewState)
414 _mesa_update_state( ctx );
415
416 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
417 vbo_exec_vtx_unmap( exec );
418 }
419
420 if (0)
421 printf("%s %d %d\n", __func__, exec->vtx.prim_count,
422 exec->vtx.vert_count);
423
424 vbo_context(ctx)->draw_prims( ctx,
425 exec->vtx.prim,
426 exec->vtx.prim_count,
427 NULL,
428 GL_TRUE,
429 0,
430 exec->vtx.vert_count - 1,
431 NULL, 0, NULL);
432
433 /* If using a real VBO, get new storage -- unless asked not to.
434 */
435 if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !keepUnmapped) {
436 vbo_exec_vtx_map( exec );
437 }
438 }
439 }
440
441 /* May have to unmap explicitly if we didn't draw:
442 */
443 if (keepUnmapped &&
444 _mesa_is_bufferobj(exec->vtx.bufferobj) &&
445 exec->vtx.buffer_map) {
446 vbo_exec_vtx_unmap( exec );
447 }
448
449 if (keepUnmapped || exec->vtx.vertex_size == 0)
450 exec->vtx.max_vert = 0;
451 else
452 exec->vtx.max_vert = vbo_compute_max_verts(exec);
453
454 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
455 exec->vtx.prim_count = 0;
456 exec->vtx.vert_count = 0;
457 }