Merge remote-tracking branch 'mesa-public/master' into vulkan
[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 src -= sz;
121 }
122 /* fall-through */
123 case GL_TRIANGLE_FAN:
124 case GL_POLYGON:
125 if (nr == 0) {
126 return 0;
127 }
128 else if (nr == 1) {
129 memcpy( dst, src+0, sz * sizeof(GLfloat) );
130 return 1;
131 }
132 else {
133 memcpy( dst, src+0, sz * sizeof(GLfloat) );
134 memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
135 return 2;
136 }
137 case GL_TRIANGLE_STRIP:
138 /* no parity issue, but need to make sure the tri is not drawn twice */
139 if (nr & 1) {
140 last_prim->count--;
141 }
142 /* fallthrough */
143 case GL_QUAD_STRIP:
144 switch (nr) {
145 case 0:
146 ovf = 0;
147 break;
148 case 1:
149 ovf = 1;
150 break;
151 default:
152 ovf = 2 + (nr & 1);
153 break;
154 }
155 for (i = 0 ; i < ovf ; i++)
156 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
157 return i;
158 case PRIM_OUTSIDE_BEGIN_END:
159 return 0;
160 default:
161 assert(0);
162 return 0;
163 }
164 }
165
166
167
168 /* TODO: populate these as the vertex is defined:
169 */
170 static void
171 vbo_exec_bind_arrays( struct gl_context *ctx )
172 {
173 struct vbo_context *vbo = vbo_context(ctx);
174 struct vbo_exec_context *exec = &vbo->exec;
175 struct gl_client_array *arrays = exec->vtx.arrays;
176 const GLuint *map;
177 GLuint attr;
178 GLbitfield64 varying_inputs = 0x0;
179
180 /* Install the default (ie Current) attributes first, then overlay
181 * all active ones.
182 */
183 switch (get_program_mode(exec->ctx)) {
184 case VP_NONE:
185 for (attr = 0; attr < VERT_ATTRIB_FF_MAX; attr++) {
186 exec->vtx.inputs[attr] = &vbo->currval[VBO_ATTRIB_POS+attr];
187 }
188 for (attr = 0; attr < MAT_ATTRIB_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_MAT_FRONT_AMBIENT+attr];
192 }
193 map = vbo->map_vp_none;
194 break;
195 case VP_ARB:
196 for (attr = 0; attr < VERT_ATTRIB_FF_MAX; attr++) {
197 exec->vtx.inputs[attr] = &vbo->currval[VBO_ATTRIB_POS+attr];
198 }
199 for (attr = 0; attr < VERT_ATTRIB_GENERIC_MAX; attr++) {
200 assert(VERT_ATTRIB_GENERIC(attr) < ARRAY_SIZE(exec->vtx.inputs));
201 exec->vtx.inputs[VERT_ATTRIB_GENERIC(attr)] =
202 &vbo->currval[VBO_ATTRIB_GENERIC0+attr];
203 }
204 map = vbo->map_vp_arb;
205
206 /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
207 * In that case we effectively need to route the data from
208 * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
209 */
210 if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
211 (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
212 exec->vtx.inputs[VERT_ATTRIB_GENERIC0] = exec->vtx.inputs[0];
213 exec->vtx.attrsz[VERT_ATTRIB_GENERIC0] = exec->vtx.attrsz[0];
214 exec->vtx.attrptr[VERT_ATTRIB_GENERIC0] = exec->vtx.attrptr[0];
215 exec->vtx.attrsz[0] = 0;
216 }
217 break;
218 default:
219 assert(0);
220 }
221
222 for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
223 const GLuint src = map[attr];
224
225 if (exec->vtx.attrsz[src]) {
226 GLsizeiptr offset = (GLbyte *)exec->vtx.attrptr[src] -
227 (GLbyte *)exec->vtx.vertex;
228
229 /* override the default array set above */
230 assert(attr < ARRAY_SIZE(exec->vtx.inputs));
231 assert(attr < ARRAY_SIZE(exec->vtx.arrays)); /* arrays[] */
232 exec->vtx.inputs[attr] = &arrays[attr];
233
234 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
235 /* a real buffer obj: Ptr is an offset, not a pointer */
236 assert(exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Pointer);
237 assert(offset >= 0);
238 arrays[attr].Ptr = (GLubyte *)
239 exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Offset + offset;
240 }
241 else {
242 /* Ptr into ordinary app memory */
243 arrays[attr].Ptr = (GLubyte *)exec->vtx.buffer_map + offset;
244 }
245 arrays[attr].Size = exec->vtx.attrsz[src];
246 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
247 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
248 arrays[attr].Type = exec->vtx.attrtype[src];
249 arrays[attr].Integer =
250 vbo_attrtype_to_integer_flag(exec->vtx.attrtype[src]);
251 arrays[attr].Format = GL_RGBA;
252 arrays[attr].Enabled = 1;
253 arrays[attr]._ElementSize = arrays[attr].Size * sizeof(GLfloat);
254 _mesa_reference_buffer_object(ctx,
255 &arrays[attr].BufferObj,
256 exec->vtx.bufferobj);
257
258 varying_inputs |= VERT_BIT(attr);
259 }
260 }
261
262 _mesa_set_varying_vp_inputs( ctx, varying_inputs );
263 ctx->NewDriverState |= ctx->DriverFlags.NewArray;
264 }
265
266
267 /**
268 * Unmap the VBO. This is called before drawing.
269 */
270 static void
271 vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
272 {
273 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
274 struct gl_context *ctx = exec->ctx;
275
276 if (ctx->Driver.FlushMappedBufferRange) {
277 GLintptr offset = exec->vtx.buffer_used -
278 exec->vtx.bufferobj->Mappings[MAP_INTERNAL].Offset;
279 GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) *
280 sizeof(float);
281
282 if (length)
283 ctx->Driver.FlushMappedBufferRange(ctx, offset, length,
284 exec->vtx.bufferobj,
285 MAP_INTERNAL);
286 }
287
288 exec->vtx.buffer_used += (exec->vtx.buffer_ptr -
289 exec->vtx.buffer_map) * sizeof(float);
290
291 assert(exec->vtx.buffer_used <= VBO_VERT_BUFFER_SIZE);
292 assert(exec->vtx.buffer_ptr != NULL);
293
294 ctx->Driver.UnmapBuffer(ctx, exec->vtx.bufferobj, MAP_INTERNAL);
295 exec->vtx.buffer_map = NULL;
296 exec->vtx.buffer_ptr = NULL;
297 exec->vtx.max_vert = 0;
298 }
299 }
300
301
302 /**
303 * Map the vertex buffer to begin storing glVertex, glColor, etc data.
304 */
305 void
306 vbo_exec_vtx_map( struct vbo_exec_context *exec )
307 {
308 struct gl_context *ctx = exec->ctx;
309 const GLenum accessRange = GL_MAP_WRITE_BIT | /* for MapBufferRange */
310 GL_MAP_INVALIDATE_RANGE_BIT |
311 GL_MAP_UNSYNCHRONIZED_BIT |
312 GL_MAP_FLUSH_EXPLICIT_BIT |
313 MESA_MAP_NOWAIT_BIT;
314 const GLenum usage = GL_STREAM_DRAW_ARB;
315
316 if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
317 return;
318
319 assert(!exec->vtx.buffer_map);
320 assert(!exec->vtx.buffer_ptr);
321
322 if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024) {
323 /* The VBO exists and there's room for more */
324 if (exec->vtx.bufferobj->Size > 0) {
325 exec->vtx.buffer_map =
326 (fi_type *)ctx->Driver.MapBufferRange(ctx,
327 exec->vtx.buffer_used,
328 (VBO_VERT_BUFFER_SIZE -
329 exec->vtx.buffer_used),
330 accessRange,
331 exec->vtx.bufferobj,
332 MAP_INTERNAL);
333 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
334 }
335 else {
336 exec->vtx.buffer_ptr = exec->vtx.buffer_map = NULL;
337 }
338 }
339
340 if (!exec->vtx.buffer_map) {
341 /* Need to allocate a new VBO */
342 exec->vtx.buffer_used = 0;
343
344 if (ctx->Driver.BufferData(ctx, GL_ARRAY_BUFFER_ARB,
345 VBO_VERT_BUFFER_SIZE,
346 NULL, usage,
347 GL_MAP_WRITE_BIT |
348 GL_DYNAMIC_STORAGE_BIT |
349 GL_CLIENT_STORAGE_BIT,
350 exec->vtx.bufferobj)) {
351 /* buffer allocation worked, now map the buffer */
352 exec->vtx.buffer_map =
353 (fi_type *)ctx->Driver.MapBufferRange(ctx,
354 0, VBO_VERT_BUFFER_SIZE,
355 accessRange,
356 exec->vtx.bufferobj,
357 MAP_INTERNAL);
358 }
359 else {
360 _mesa_error(ctx, GL_OUT_OF_MEMORY, "VBO allocation");
361 exec->vtx.buffer_map = NULL;
362 }
363 }
364
365 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
366
367 if (!exec->vtx.buffer_map) {
368 /* out of memory */
369 _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt_noop );
370 }
371 else {
372 if (_mesa_using_noop_vtxfmt(ctx->Exec)) {
373 /* The no-op functions are installed so switch back to regular
374 * functions. We do this test just to avoid frequent and needless
375 * calls to _mesa_install_exec_vtxfmt().
376 */
377 _mesa_install_exec_vtxfmt(ctx, &exec->vtxfmt);
378 }
379 }
380
381 if (0)
382 printf("map %d..\n", exec->vtx.buffer_used);
383 }
384
385
386
387 /**
388 * Execute the buffer and save copied verts.
389 * \param keep_unmapped if true, leave the VBO unmapped when we're done.
390 */
391 void
392 vbo_exec_vtx_flush(struct vbo_exec_context *exec, GLboolean keepUnmapped)
393 {
394 if (0)
395 vbo_exec_debug_verts( exec );
396
397 if (exec->vtx.prim_count &&
398 exec->vtx.vert_count) {
399
400 exec->vtx.copied.nr = vbo_copy_vertices( exec );
401
402 if (exec->vtx.copied.nr != exec->vtx.vert_count) {
403 struct gl_context *ctx = exec->ctx;
404
405 /* Before the update_state() as this may raise _NEW_VARYING_VP_INPUTS
406 * from _mesa_set_varying_vp_inputs().
407 */
408 vbo_exec_bind_arrays( ctx );
409
410 if (ctx->NewState)
411 _mesa_update_state( ctx );
412
413 if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
414 vbo_exec_vtx_unmap( exec );
415 }
416
417 if (0)
418 printf("%s %d %d\n", __func__, exec->vtx.prim_count,
419 exec->vtx.vert_count);
420
421 vbo_context(ctx)->draw_prims( ctx,
422 exec->vtx.prim,
423 exec->vtx.prim_count,
424 NULL,
425 GL_TRUE,
426 0,
427 exec->vtx.vert_count - 1,
428 NULL, 0, NULL);
429
430 /* If using a real VBO, get new storage -- unless asked not to.
431 */
432 if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !keepUnmapped) {
433 vbo_exec_vtx_map( exec );
434 }
435 }
436 }
437
438 /* May have to unmap explicitly if we didn't draw:
439 */
440 if (keepUnmapped &&
441 _mesa_is_bufferobj(exec->vtx.bufferobj) &&
442 exec->vtx.buffer_map) {
443 vbo_exec_vtx_unmap( exec );
444 }
445
446 if (keepUnmapped || exec->vtx.vertex_size == 0)
447 exec->vtx.max_vert = 0;
448 else
449 exec->vtx.max_vert = vbo_compute_max_verts(exec);
450
451 exec->vtx.buffer_ptr = exec->vtx.buffer_map;
452 exec->vtx.prim_count = 0;
453 exec->vtx.vert_count = 0;
454 }