1/*
2 * Copyright (C) 2009-2010 Francisco Jerez.
3 * All Rights Reserved.
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining
6 * a copy of this software and associated documentation files (the
7 * "Software"), to deal in the Software without restriction, including
8 * without limitation the rights to use, copy, modify, merge, publish,
9 * distribute, sublicense, and/or sell copies of the Software, and to
10 * permit persons to whom the Software is furnished to do so, subject to
11 * the following conditions:
12 *
13 * The above copyright notice and this permission notice (including the
14 * next paragraph) shall be included in all copies or substantial
15 * portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
20 * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
21 * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
22 * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
23 * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
24 *
25 */
26
27#include "nouveau_driver.h"
28#include "nouveau_bufferobj.h"
29#include "nouveau_util.h"
30
31#include "main/bufferobj.h"
32#include "main/glformats.h"
33#include "main/varray.h"
34#include "main/image.h"
35
36/* Arbitrary pushbuf length we can assume we can get with a single
37 * call to WAIT_RING. */
38#define PUSHBUF_DWORDS 65536
39
40/* Functions to turn GL arrays or index buffers into nouveau_array
41 * structures. */
42
43static int
44get_array_stride(struct gl_context *ctx, const struct tnl_vertex_array *a)
45{
46	struct nouveau_render_state *render = to_render_state(ctx);
47	const struct gl_vertex_buffer_binding *binding = a->BufferBinding;
48
49	if (render->mode == VBO && !_mesa_is_bufferobj(binding->BufferObj)) {
50		const struct gl_array_attributes *attrib = a->VertexAttrib;
51		/* Pack client buffers. */
52		return align(attrib->Format._ElementSize, 4);
53	} else {
54		return binding->Stride;
55	}
56}
57
58static void
59vbo_init_arrays(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
60		const struct tnl_vertex_array *arrays)
61{
62	struct nouveau_render_state *render = to_render_state(ctx);
63	GLboolean imm = (render->mode == IMM);
64	int i, attr;
65
66	if (ib) {
67		GLenum ib_type;
68
69		if (ib->index_size == 4)
70			ib_type = GL_UNSIGNED_INT;
71		else if (ib->index_size == 2)
72			ib_type = GL_UNSIGNED_SHORT;
73		else
74			ib_type = GL_UNSIGNED_BYTE;
75
76		nouveau_init_array(&render->ib, 0, 0, ib->count, ib_type,
77				   ib->obj, ib->ptr, GL_TRUE, ctx);
78	}
79
80	FOR_EACH_BOUND_ATTR(render, i, attr) {
81		const struct tnl_vertex_array *array = &arrays[attr];
82		const struct gl_vertex_buffer_binding *binding =
83			array->BufferBinding;
84		const struct gl_array_attributes *attrib = array->VertexAttrib;
85		const GLubyte *p = _mesa_vertex_attrib_address(attrib, binding);
86
87		nouveau_init_array(&render->attrs[attr], attr,
88				   get_array_stride(ctx, array),
89				   attrib->Format.Size, attrib->Format.Type,
90				   imm ? binding->BufferObj : NULL,
91				   p, imm, ctx);
92	}
93}
94
95static void
96vbo_deinit_arrays(struct gl_context *ctx, const struct _mesa_index_buffer *ib,
97		  const struct tnl_vertex_array *arrays)
98{
99	struct nouveau_render_state *render = to_render_state(ctx);
100	int i, attr;
101
102	if (ib)
103		nouveau_cleanup_array(&render->ib);
104
105	FOR_EACH_BOUND_ATTR(render, i, attr) {
106		struct nouveau_array *a = &render->attrs[attr];
107
108		if (render->mode == IMM)
109			nouveau_bo_ref(NULL, &a->bo);
110
111		nouveau_deinit_array(a);
112		render->map[i] = -1;
113	}
114
115	render->attr_count = 0;
116}
117
118/* Make some rendering decisions from the GL context. */
119
120static void
121vbo_choose_render_mode(struct gl_context *ctx, const struct tnl_vertex_array *arrays)
122{
123	struct nouveau_render_state *render = to_render_state(ctx);
124	int i;
125
126	render->mode = VBO;
127
128	if (ctx->Light.Enabled) {
129		for (i = 0; i < VERT_ATTRIB_MAT_MAX; i++) {
130			if (arrays[VERT_ATTRIB_MAT(i)].BufferBinding->Stride) {
131				render->mode = IMM;
132				break;
133			}
134		}
135	}
136}
137
138static void
139vbo_emit_attr(struct gl_context *ctx, const struct tnl_vertex_array *arrays,
140	      int attr)
141{
142	struct nouveau_pushbuf *push = context_push(ctx);
143	struct nouveau_render_state *render = to_render_state(ctx);
144	const struct tnl_vertex_array *array = &arrays[attr];
145	const struct gl_vertex_buffer_binding *binding = array->BufferBinding;
146	const struct gl_array_attributes *attrib = array->VertexAttrib;
147	const GLubyte *p = _mesa_vertex_attrib_address(attrib, binding);
148	struct nouveau_array *a = &render->attrs[attr];
149	RENDER_LOCALS(ctx);
150
151	if (!binding->Stride) {
152		if (attr >= VERT_ATTRIB_MAT(0))
153			/* nouveau_update_state takes care of materials. */
154			return;
155
156		/* Constant attribute. */
157		nouveau_init_array(a, attr, binding->Stride, attrib->Format.Size,
158				   attrib->Format.Type, binding->BufferObj, p,
159				   GL_TRUE, ctx);
160		EMIT_IMM(ctx, a, 0);
161		nouveau_deinit_array(a);
162
163	} else {
164		/* Varying attribute. */
165		struct nouveau_attr_info *info = &TAG(vertex_attrs)[attr];
166
167		if (render->mode == VBO) {
168			render->map[info->vbo_index] = attr;
169			render->vertex_size += attrib->Format._ElementSize;
170			render->attr_count = MAX2(render->attr_count,
171						  info->vbo_index + 1);
172		} else {
173			render->map[render->attr_count++] = attr;
174			render->vertex_size += 4 * info->imm_fields;
175		}
176	}
177}
178
179#define MAT(a) VERT_ATTRIB_MAT(MAT_ATTRIB_##a)
180
181static void
182vbo_choose_attrs(struct gl_context *ctx, const struct tnl_vertex_array *arrays)
183{
184	struct nouveau_render_state *render = to_render_state(ctx);
185	int i;
186
187	/* Reset the vertex size. */
188	render->vertex_size = 0;
189	render->attr_count = 0;
190
191	vbo_emit_attr(ctx, arrays, VERT_ATTRIB_COLOR0);
192	if (ctx->Fog.ColorSumEnabled && !ctx->Light.Enabled)
193		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_COLOR1);
194
195	for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
196		if (ctx->Texture._EnabledCoordUnits & (1 << i))
197			vbo_emit_attr(ctx, arrays, VERT_ATTRIB_TEX0 + i);
198	}
199
200	if (ctx->Fog.Enabled && ctx->Fog.FogCoordinateSource == GL_FOG_COORD)
201		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_FOG);
202
203	if (ctx->Light.Enabled ||
204	    (ctx->Texture._GenFlags & TEXGEN_NEED_NORMALS))
205		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_NORMAL);
206
207	if (ctx->Light.Enabled && render->mode == IMM) {
208		vbo_emit_attr(ctx, arrays, MAT(FRONT_AMBIENT));
209		vbo_emit_attr(ctx, arrays, MAT(FRONT_DIFFUSE));
210		vbo_emit_attr(ctx, arrays, MAT(FRONT_SPECULAR));
211		vbo_emit_attr(ctx, arrays, MAT(FRONT_SHININESS));
212
213		if (ctx->Light.Model.TwoSide) {
214			vbo_emit_attr(ctx, arrays, MAT(BACK_AMBIENT));
215			vbo_emit_attr(ctx, arrays, MAT(BACK_DIFFUSE));
216			vbo_emit_attr(ctx, arrays, MAT(BACK_SPECULAR));
217			vbo_emit_attr(ctx, arrays, MAT(BACK_SHININESS));
218		}
219	}
220
221	vbo_emit_attr(ctx, arrays, VERT_ATTRIB_POS);
222}
223
224static int
225get_max_client_stride(struct gl_context *ctx, const struct tnl_vertex_array *arrays)
226{
227	struct nouveau_render_state *render = to_render_state(ctx);
228	int i, attr, s = 0;
229
230	FOR_EACH_BOUND_ATTR(render, i, attr) {
231		const struct tnl_vertex_array *a = &arrays[attr];
232
233		if (!_mesa_is_bufferobj(a->BufferBinding->BufferObj))
234			s = MAX2(s, get_array_stride(ctx, a));
235	}
236
237	return s;
238}
239
240static void
241TAG(vbo_render_prims)(struct gl_context *ctx,
242		      const struct tnl_vertex_array *arrays,
243		      const struct _mesa_prim *prims, GLuint nr_prims,
244		      const struct _mesa_index_buffer *ib,
245		      GLboolean index_bounds_valid,
246		      GLuint min_index, GLuint max_index,
247		      struct gl_transform_feedback_object *tfb_vertcount,
248                      unsigned stream,
249		      struct gl_buffer_object *indirect);
250
251static GLboolean
252vbo_maybe_split(struct gl_context *ctx, const struct tnl_vertex_array *arrays,
253	    const struct _mesa_prim *prims, GLuint nr_prims,
254	    const struct _mesa_index_buffer *ib,
255	    GLuint min_index, GLuint max_index)
256{
257	struct nouveau_context *nctx = to_nouveau_context(ctx);
258	struct nouveau_render_state *render = to_render_state(ctx);
259	struct nouveau_bufctx *bufctx = nctx->hw.bufctx;
260	unsigned pushbuf_avail = PUSHBUF_DWORDS - 2 * (bufctx->relocs +
261						       render->attr_count),
262		vert_avail = get_max_vertices(ctx, NULL, pushbuf_avail),
263		idx_avail = get_max_vertices(ctx, ib, pushbuf_avail);
264	int stride;
265
266	/* Try to keep client buffers smaller than the scratch BOs. */
267	if (render->mode == VBO &&
268	    (stride = get_max_client_stride(ctx, arrays)))
269		    vert_avail = MIN2(vert_avail,
270				      NOUVEAU_SCRATCH_SIZE / stride);
271
272	if (max_index - min_index > vert_avail ||
273	    (ib && ib->count > idx_avail)) {
274		struct split_limits limits = {
275			.max_verts = vert_avail,
276			.max_indices = idx_avail,
277			.max_vb_size = ~0,
278		};
279
280		_tnl_split_prims(ctx, arrays, prims, nr_prims, ib, min_index,
281				 max_index, TAG(vbo_render_prims), &limits);
282		return GL_TRUE;
283	}
284
285	return GL_FALSE;
286}
287
288/* VBO rendering path. */
289
290static GLboolean
291check_update_array(struct nouveau_array *a, unsigned offset,
292		   struct nouveau_bo *bo, int *pdelta)
293{
294	int delta = *pdelta;
295	GLboolean dirty;
296
297	if (a->bo == bo) {
298		if (delta < 0)
299			delta = ((int)offset - (int)a->offset) / a->stride;
300
301		dirty = (delta < 0 ||
302			 offset != (a->offset + delta * a->stride));
303	} else {
304		dirty = GL_TRUE;
305	}
306
307	*pdelta = (dirty ? 0 : delta);
308	return dirty;
309}
310
311static void
312vbo_bind_vertices(struct gl_context *ctx, const struct tnl_vertex_array *arrays,
313		  int base, unsigned min_index, unsigned max_index, int *pdelta)
314{
315	struct nouveau_render_state *render = to_render_state(ctx);
316	struct nouveau_pushbuf *push = context_push(ctx);
317	struct nouveau_bo *bo[NUM_VERTEX_ATTRS];
318	unsigned offset[NUM_VERTEX_ATTRS];
319	GLboolean dirty = GL_FALSE;
320	int i, j, attr;
321	RENDER_LOCALS(ctx);
322
323	*pdelta = -1;
324
325	FOR_EACH_BOUND_ATTR(render, i, attr) {
326		const struct tnl_vertex_array *array = &arrays[attr];
327		const struct gl_vertex_buffer_binding *binding =
328			array->BufferBinding;
329		const struct gl_array_attributes *attrib = array->VertexAttrib;
330		const GLubyte *p = _mesa_vertex_attrib_address(attrib, binding);
331		struct gl_buffer_object *obj = binding->BufferObj;
332		struct nouveau_array *a = &render->attrs[attr];
333		unsigned delta = (base + min_index) * binding->Stride;
334
335		bo[i] = NULL;
336
337		if (nouveau_bufferobj_hw(obj)) {
338			/* Array in a buffer obj. */
339			nouveau_bo_ref(to_nouveau_bufferobj(obj)->bo, &bo[i]);
340			offset[i] = delta + (intptr_t)p;
341
342		} else {
343			int n = max_index - min_index + 1;
344			char *sp = (char *)ADD_POINTERS(
345				nouveau_bufferobj_sys(obj), p) + delta;
346			char *dp  = nouveau_get_scratch(ctx, n * a->stride,
347							&bo[i], &offset[i]);
348
349			/* Array in client memory, move it to a
350			 * scratch buffer obj. */
351			for (j = 0; j < n; j++)
352				memcpy(dp + j * a->stride,
353				       sp + j * binding->Stride,
354				       a->stride);
355		}
356
357		dirty |= check_update_array(a, offset[i], bo[i], pdelta);
358	}
359
360	*pdelta -= min_index;
361
362	if (dirty) {
363		/* Buffers changed, update the attribute binding. */
364		FOR_EACH_BOUND_ATTR(render, i, attr) {
365			struct nouveau_array *a = &render->attrs[attr];
366
367			nouveau_bo_ref(NULL, &a->bo);
368			a->offset = offset[i];
369			a->bo = bo[i];
370		}
371
372		TAG(render_release_vertices)(ctx);
373		TAG(render_bind_vertices)(ctx);
374	} else {
375		/* Just cleanup. */
376		FOR_EACH_BOUND_ATTR(render, i, attr)
377			nouveau_bo_ref(NULL, &bo[i]);
378	}
379
380	BATCH_VALIDATE();
381}
382
383static void
384vbo_draw_vbo(struct gl_context *ctx, const struct tnl_vertex_array *arrays,
385	     const struct _mesa_prim *prims, GLuint nr_prims,
386	     const struct _mesa_index_buffer *ib, GLuint min_index,
387	     GLuint max_index)
388{
389	struct nouveau_context *nctx = to_nouveau_context(ctx);
390	struct nouveau_pushbuf *push = context_push(ctx);
391	dispatch_t dispatch = get_array_dispatch(&to_render_state(ctx)->ib);
392	int i, delta = 0, basevertex = 0;
393	RENDER_LOCALS(ctx);
394
395	TAG(render_set_format)(ctx);
396
397	for (i = 0; i < nr_prims; i++) {
398		unsigned start = prims[i].start,
399			count = prims[i].count;
400
401		if (i == 0 || basevertex != prims[i].basevertex) {
402			basevertex = prims[i].basevertex;
403			vbo_bind_vertices(ctx, arrays, basevertex, min_index,
404					  max_index, &delta);
405
406			nouveau_pushbuf_bufctx(push, nctx->hw.bufctx);
407			if (nouveau_pushbuf_validate(push)) {
408				nouveau_pushbuf_bufctx(push, NULL);
409				return;
410			}
411		}
412
413		if (count > get_max_vertices(ctx, ib, PUSH_AVAIL(push)))
414			PUSH_SPACE(push, PUSHBUF_DWORDS);
415
416		BATCH_BEGIN(nvgl_primitive(prims[i].mode));
417		dispatch(ctx, start, delta, count);
418		BATCH_END();
419	}
420
421	nouveau_pushbuf_bufctx(push, NULL);
422	TAG(render_release_vertices)(ctx);
423}
424
425/* Immediate rendering path. */
426
427static unsigned
428extract_id(struct nouveau_array *a, int i, int j)
429{
430	return j;
431}
432
433static void
434vbo_draw_imm(struct gl_context *ctx, const struct tnl_vertex_array *arrays,
435	     const struct _mesa_prim *prims, GLuint nr_prims,
436	     const struct _mesa_index_buffer *ib, GLuint min_index,
437	     GLuint max_index)
438{
439	struct nouveau_render_state *render = to_render_state(ctx);
440	struct nouveau_context *nctx = to_nouveau_context(ctx);
441	struct nouveau_pushbuf *push = context_push(ctx);
442	extract_u_t extract = ib ? render->ib.extract_u : extract_id;
443	int i, j, k, attr;
444	RENDER_LOCALS(ctx);
445
446	nouveau_pushbuf_bufctx(push, nctx->hw.bufctx);
447	if (nouveau_pushbuf_validate(push)) {
448		nouveau_pushbuf_bufctx(push, NULL);
449		return;
450	}
451
452	for (i = 0; i < nr_prims; i++) {
453		unsigned start = prims[i].start,
454			end = start + prims[i].count;
455
456		if (prims[i].count > get_max_vertices(ctx, ib,
457						      PUSH_AVAIL(push)))
458			PUSH_SPACE(push, PUSHBUF_DWORDS);
459
460		BATCH_BEGIN(nvgl_primitive(prims[i].mode));
461
462		for (; start < end; start++) {
463			j = prims[i].basevertex +
464				extract(&render->ib, 0, start);
465
466			FOR_EACH_BOUND_ATTR(render, k, attr)
467				EMIT_IMM(ctx, &render->attrs[attr], j);
468		}
469
470		BATCH_END();
471	}
472
473	nouveau_pushbuf_bufctx(push, NULL);
474}
475
476/* draw_prims entry point when we're doing hw-tnl. */
477
478static void
479TAG(vbo_render_prims)(struct gl_context *ctx,
480		      const struct tnl_vertex_array *arrays,
481		      const struct _mesa_prim *prims, GLuint nr_prims,
482		      const struct _mesa_index_buffer *ib,
483		      GLboolean index_bounds_valid,
484		      GLuint min_index, GLuint max_index,
485		      struct gl_transform_feedback_object *tfb_vertcount,
486                      unsigned stream,
487		      struct gl_buffer_object *indirect)
488{
489	struct nouveau_render_state *render = to_render_state(ctx);
490
491	if (!index_bounds_valid)
492		vbo_get_minmax_indices(ctx, prims, ib, &min_index, &max_index,
493				       nr_prims);
494
495	vbo_choose_render_mode(ctx, arrays);
496	vbo_choose_attrs(ctx, arrays);
497
498	if (vbo_maybe_split(ctx, arrays, prims, nr_prims, ib, min_index,
499			    max_index))
500		return;
501
502	vbo_init_arrays(ctx, ib, arrays);
503
504	if (render->mode == VBO)
505		vbo_draw_vbo(ctx, arrays, prims, nr_prims, ib, min_index,
506			     max_index);
507	else
508		vbo_draw_imm(ctx, arrays, prims, nr_prims, ib, min_index,
509			     max_index);
510
511	vbo_deinit_arrays(ctx, ib, arrays);
512}
513
514/* VBO rendering entry points. */
515
516static void
517TAG(vbo_check_render_prims)(struct gl_context *ctx,
518			    const struct tnl_vertex_array *arrays,
519			    const struct _mesa_prim *prims, GLuint nr_prims,
520			    const struct _mesa_index_buffer *ib,
521			    GLboolean index_bounds_valid,
522			    GLuint min_index, GLuint max_index,
523			    struct gl_transform_feedback_object *tfb_vertcount,
524                            unsigned stream,
525			    struct gl_buffer_object *indirect)
526{
527	struct nouveau_context *nctx = to_nouveau_context(ctx);
528
529	nouveau_validate_framebuffer(ctx);
530
531	if (nctx->fallback == HWTNL)
532		TAG(vbo_render_prims)(ctx, arrays, prims, nr_prims, ib,
533				      index_bounds_valid, min_index, max_index,
534				      tfb_vertcount, stream, indirect);
535
536	if (nctx->fallback == SWTNL)
537		_tnl_draw_prims(ctx, arrays, prims, nr_prims, ib,
538				index_bounds_valid, min_index, max_index,
539				tfb_vertcount, stream, indirect);
540}
541
542static void
543TAG(vbo_draw)(struct gl_context *ctx,
544	      const struct _mesa_prim *prims, GLuint nr_prims,
545	      const struct _mesa_index_buffer *ib,
546	      GLboolean index_bounds_valid,
547	      GLuint min_index, GLuint max_index,
548	      struct gl_transform_feedback_object *tfb_vertcount,
549	      unsigned stream,
550	      struct gl_buffer_object *indirect)
551{
552	/* Borrow and update the inputs list from the tnl context */
553	const struct tnl_vertex_array* arrays = _tnl_bind_inputs(ctx);
554
555	TAG(vbo_check_render_prims)(ctx, arrays,
556				    prims, nr_prims, ib,
557				    index_bounds_valid, min_index, max_index,
558				    tfb_vertcount, stream, indirect);
559}
560
561void
562TAG(vbo_init)(struct gl_context *ctx)
563{
564	struct nouveau_render_state *render = to_render_state(ctx);
565	int i;
566
567	for (i = 0; i < VERT_ATTRIB_MAX; i++)
568		render->map[i] = -1;
569
570	/* Overwrite our draw function */
571	ctx->Driver.Draw = TAG(vbo_draw);
572	vbo_use_buffer_objects(ctx);
573}
574
575void
576TAG(vbo_destroy)(struct gl_context *ctx)
577{
578}
579