glapi: Correctly generate static disatches for X86.
[mesa.git] / src / mesa / vbo / vbo_exec_draw.c
1 /*
2 * Mesa 3-D graphics library
3 * Version: 7.2
4 *
5 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
6 *
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:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
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.
23 *
24 * Authors:
25 * Keith Whitwell <keith@tungstengraphics.com>
26 */
27
28 #include "main/glheader.h"
29 #include "main/bufferobj.h"
30 #include "main/context.h"
31 #include "main/enums.h"
32 #include "main/state.h"
33
34 #include "vbo_context.h"
35
36
37 #if FEATURE_beginend
38
39
40 static void
41 vbo_exec_debug_verts( struct vbo_exec_context *exec )
42 {
43 GLuint count = exec->vtx.vert_count;
44 GLuint i;
45
46 printf("%s: %u vertices %d primitives, %d vertsize\n",
47 __FUNCTION__,
48 count,
49 exec->vtx.prim_count,
50 exec->vtx.vertex_size);
51
52 for (i = 0 ; i < exec->vtx.prim_count ; i++) {
53 struct _mesa_prim *prim = &exec->vtx.prim[i];
54 printf(" prim %d: %s%s %d..%d %s %s\n",
55 i,
56 _mesa_lookup_prim_by_nr(prim->mode),
57 prim->weak ? " (weak)" : "",
58 prim->start,
59 prim->start + prim->count,
60 prim->begin ? "BEGIN" : "(wrap)",
61 prim->end ? "END" : "(wrap)");
62 }
63 }
64
65
66 /*
67 * NOTE: Need to have calculated primitives by this point -- do it on the fly.
68 * NOTE: Old 'parity' issue is gone.
69 */
70 static GLuint
71 vbo_copy_vertices( struct vbo_exec_context *exec )
72 {
73 GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
74 GLuint ovf, i;
75 GLuint sz = exec->vtx.vertex_size;
76 GLfloat *dst = exec->vtx.copied.buffer;
77 const GLfloat *src = (exec->vtx.buffer_map +
78 exec->vtx.prim[exec->vtx.prim_count-1].start *
79 exec->vtx.vertex_size);
80
81
82 switch (exec->ctx->Driver.CurrentExecPrimitive) {
83 case GL_POINTS:
84 return 0;
85 case GL_LINES:
86 ovf = nr&1;
87 for (i = 0 ; i < ovf ; i++)
88 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
89 return i;
90 case GL_TRIANGLES:
91 ovf = nr%3;
92 for (i = 0 ; i < ovf ; i++)
93 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
94 return i;
95 case GL_QUADS:
96 ovf = nr&3;
97 for (i = 0 ; i < ovf ; i++)
98 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
99 return i;
100 case GL_LINE_STRIP:
101 if (nr == 0) {
102 return 0;
103 }
104 else {
105 memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
106 return 1;
107 }
108 case GL_LINE_LOOP:
109 case GL_TRIANGLE_FAN:
110 case GL_POLYGON:
111 if (nr == 0) {
112 return 0;
113 }
114 else if (nr == 1) {
115 memcpy( dst, src+0, sz * sizeof(GLfloat) );
116 return 1;
117 }
118 else {
119 memcpy( dst, src+0, sz * sizeof(GLfloat) );
120 memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
121 return 2;
122 }
123 case GL_TRIANGLE_STRIP:
124 /* no parity issue, but need to make sure the tri is not drawn twice */
125 if (nr & 1) {
126 exec->vtx.prim[exec->vtx.prim_count-1].count--;
127 }
128 /* fallthrough */
129 case GL_QUAD_STRIP:
130 switch (nr) {
131 case 0:
132 ovf = 0;
133 break;
134 case 1:
135 ovf = 1;
136 break;
137 default:
138 ovf = 2 + (nr & 1);
139 break;
140 }
141 for (i = 0 ; i < ovf ; i++)
142 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
143 return i;
144 case PRIM_OUTSIDE_BEGIN_END:
145 return 0;
146 default:
147 assert(0);
148 return 0;
149 }
150 }
151
152
153
154 /* TODO: populate these as the vertex is defined:
155 */
156 static void
157 vbo_exec_bind_arrays( GLcontext *ctx )
158 {
159 struct vbo_context *vbo = vbo_context(ctx);
160 struct vbo_exec_context *exec = &vbo->exec;
161 struct gl_client_array *arrays = exec->vtx.arrays;
162 const GLuint count = exec->vtx.vert_count;
163 const GLubyte *data = (GLubyte *) exec->vtx.buffer_map;
164 const GLuint *map;
165 GLuint attr;
166 GLbitfield varying_inputs = 0x0;
167
168 /* Install the default (ie Current) attributes first, then overlay
169 * all active ones.
170 */
171 switch (get_program_mode(exec->ctx)) {
172 case VP_NONE:
173 for (attr = 0; attr < 16; attr++) {
174 exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
175 }
176 for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
177 ASSERT(attr + 16 < Elements(exec->vtx.inputs));
178 exec->vtx.inputs[attr + 16] = &vbo->mat_currval[attr];
179 }
180 map = vbo->map_vp_none;
181 break;
182 case VP_NV:
183 case VP_ARB:
184 /* The aliasing of attributes for NV vertex programs has already
185 * occurred. NV vertex programs cannot access material values,
186 * nor attributes greater than VERT_ATTRIB_TEX7.
187 */
188 for (attr = 0; attr < 16; attr++) {
189 exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
190 ASSERT(attr + 16 < Elements(exec->vtx.inputs));
191 exec->vtx.inputs[attr + 16] = &vbo->generic_currval[attr];
192 }
193 map = vbo->map_vp_arb;
194
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.
198 */
199 if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
200 (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
201 exec->vtx.inputs[16] = exec->vtx.inputs[0];
202 exec->vtx.attrsz[16] = exec->vtx.attrsz[0];
203 exec->vtx.attrptr[16] = exec->vtx.attrptr[0];
204 exec->vtx.attrsz[0] = 0;
205 }
206 break;
207 default:
208 assert(0);
209 }
210
211 /* Make all active attributes (including edgeflag) available as
212 * arrays of floats.
213 */
214 for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
215 const GLuint src = map[attr];
216
217 if (exec->vtx.attrsz[src]) {
218 /* override the default array set above */
219 ASSERT(attr < Elements(exec->vtx.inputs));
220 ASSERT(attr < Elements(exec->vtx.arrays)); /* arrays[] */
221 exec->vtx.inputs[attr] = &arrays[attr];
222
223 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
224 /* a real buffer obj: Ptr is an offset, not a pointer*/
225 GLsizeiptr offset;
226 assert(exec->vtx.bufferobj->Pointer); /* buf should be mapped */
227 offset = (GLbyte *) data -
228 (GLbyte *) exec->vtx.bufferobj->Pointer +
229 exec->vtx.bufferobj->Offset;
230 assert(offset >= 0);
231 arrays[attr].Ptr = (void *) offset;
232 }
233 else {
234 /* Ptr into ordinary app memory */
235 arrays[attr].Ptr = (void *) data;
236 }
237 arrays[attr].Size = exec->vtx.attrsz[src];
238 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
239 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
240 arrays[attr].Type = GL_FLOAT;
241 arrays[attr].Format = GL_RGBA;
242 arrays[attr].Enabled = 1;
243 _mesa_reference_buffer_object(ctx,
244 &arrays[attr].BufferObj,
245 exec->vtx.bufferobj);
246 arrays[attr]._MaxElement = count; /* ??? */
247
248 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
249 varying_inputs |= 1 << attr;
250 }
251 }
252
253 _mesa_set_varying_vp_inputs( ctx, varying_inputs );
254 }
255
256
257 static void
258 vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
259 {
260 GLenum target = GL_ARRAY_BUFFER_ARB;
261
262 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
263 GLcontext *ctx = exec->ctx;
264
265 if (ctx->Driver.FlushMappedBufferRange) {
266 GLintptr offset = exec->vtx.buffer_used - exec->vtx.bufferobj->Offset;
267 GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) * sizeof(float);
268
269 if (length)
270 ctx->Driver.FlushMappedBufferRange(ctx, target,
271 offset, length,
272 exec->vtx.bufferobj);
273 }
274
275 exec->vtx.buffer_used += (exec->vtx.buffer_ptr -
276 exec->vtx.buffer_map) * sizeof(float);
277
278 assert(exec->vtx.buffer_used <= VBO_VERT_BUFFER_SIZE);
279 assert(exec->vtx.buffer_ptr != NULL);
280
281 ctx->Driver.UnmapBuffer(ctx, target, exec->vtx.bufferobj);
282 exec->vtx.buffer_map = NULL;
283 exec->vtx.buffer_ptr = NULL;
284 exec->vtx.max_vert = 0;
285 }
286 }
287
288
289 void
290 vbo_exec_vtx_map( struct vbo_exec_context *exec )
291 {
292 GLcontext *ctx = exec->ctx;
293 const GLenum target = GL_ARRAY_BUFFER_ARB;
294 const GLenum access = GL_READ_WRITE_ARB; /* for MapBuffer */
295 const GLenum accessRange = GL_MAP_WRITE_BIT | /* for MapBufferRange */
296 GL_MAP_INVALIDATE_RANGE_BIT |
297 GL_MAP_UNSYNCHRONIZED_BIT |
298 GL_MAP_FLUSH_EXPLICIT_BIT |
299 MESA_MAP_NOWAIT_BIT;
300 const GLenum usage = GL_STREAM_DRAW_ARB;
301
302 if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
303 return;
304
305 if (exec->vtx.buffer_map != NULL) {
306 assert(0);
307 exec->vtx.buffer_map = NULL;
308 exec->vtx.buffer_ptr = NULL;
309 }
310
311 if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024 &&
312 ctx->Driver.MapBufferRange) {
313 exec->vtx.buffer_map =
314 (GLfloat *)ctx->Driver.MapBufferRange(ctx,
315 target,
316 exec->vtx.buffer_used,
317 (VBO_VERT_BUFFER_SIZE -
318 exec->vtx.buffer_used),
319 accessRange,
320 exec->vtx.bufferobj);
321 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
322 }
323
324 if (!exec->vtx.buffer_map) {
325 exec->vtx.buffer_used = 0;
326
327 ctx->Driver.BufferData(ctx, target,
328 VBO_VERT_BUFFER_SIZE,
329 NULL, usage, exec->vtx.bufferobj);
330
331
332 if (ctx->Driver.MapBufferRange)
333 exec->vtx.buffer_map =
334 (GLfloat *)ctx->Driver.MapBufferRange(ctx, target,
335 0, VBO_VERT_BUFFER_SIZE,
336 accessRange,
337 exec->vtx.bufferobj);
338 if (!exec->vtx.buffer_map)
339 exec->vtx.buffer_map =
340 (GLfloat *)ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj);
341 assert(exec->vtx.buffer_map);
342 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
343 }
344
345 if (0)
346 printf("map %d..\n", exec->vtx.buffer_used);
347 }
348
349
350
351 /**
352 * Execute the buffer and save copied verts.
353 */
354 void
355 vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap )
356 {
357 if (0)
358 vbo_exec_debug_verts( exec );
359
360 if (exec->vtx.prim_count &&
361 exec->vtx.vert_count) {
362
363 exec->vtx.copied.nr = vbo_copy_vertices( exec );
364
365 if (exec->vtx.copied.nr != exec->vtx.vert_count) {
366 GLcontext *ctx = exec->ctx;
367
368 /* Before the update_state() as this may raise _NEW_ARRAY
369 * from _mesa_set_varying_vp_inputs().
370 */
371 vbo_exec_bind_arrays( ctx );
372
373 if (ctx->NewState)
374 _mesa_update_state( ctx );
375
376 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
377 vbo_exec_vtx_unmap( exec );
378 }
379
380 if (0)
381 printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
382 exec->vtx.vert_count);
383
384 vbo_context(ctx)->draw_prims( ctx,
385 exec->vtx.inputs,
386 exec->vtx.prim,
387 exec->vtx.prim_count,
388 NULL,
389 GL_TRUE,
390 0,
391 exec->vtx.vert_count - 1);
392
393 /* If using a real VBO, get new storage -- unless asked not to.
394 */
395 if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !unmap) {
396 vbo_exec_vtx_map( exec );
397 }
398 }
399 }
400
401 /* May have to unmap explicitly if we didn't draw:
402 */
403 if (unmap &&
404 _mesa_is_bufferobj(exec->vtx.bufferobj) &&
405 exec->vtx.buffer_map) {
406 vbo_exec_vtx_unmap( exec );
407 }
408
409
410 if (unmap || exec->vtx.vertex_size == 0)
411 exec->vtx.max_vert = 0;
412 else
413 exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) /
414 (exec->vtx.vertex_size * sizeof(GLfloat)));
415
416 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
417 exec->vtx.prim_count = 0;
418 exec->vtx.vert_count = 0;
419 }
420
421
422 #endif /* FEATURE_beginend */