1/**************************************************************************
2 *
3 * Copyright 2010 VMware, Inc.
4 * 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
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28#include "draw/draw_private.h"
29#include "draw/draw_vs.h"
30#include "draw/draw_gs.h"
31#include "draw/draw_tess.h"
32#include "draw/draw_context.h"
33#include "draw/draw_vbuf.h"
34#include "draw/draw_vertex.h"
35#include "draw/draw_pt.h"
36
37#include "pipe/p_state.h"
38
39#include "util/u_math.h"
40#include "util/u_prim.h"
41#include "util/u_memory.h"
42
43struct pt_so_emit {
44   struct draw_context *draw;
45
46   unsigned input_vertex_stride;
47   const float (*inputs)[4];
48   const float *pre_clip_pos;
49   boolean has_so;
50   boolean use_pre_clip_pos;
51   int pos_idx;
52   unsigned emitted_primitives;
53   unsigned generated_primitives;
54   unsigned stream;
55};
56
57static const struct pipe_stream_output_info *
58draw_so_info(const struct draw_context *draw)
59{
60   const struct pipe_stream_output_info *state = NULL;
61
62   if (draw->gs.geometry_shader) {
63      state = &draw->gs.geometry_shader->state.stream_output;
64   } else if (draw->tes.tess_eval_shader) {
65      state = &draw->tes.tess_eval_shader->state.stream_output;
66   } else {
67      state = &draw->vs.vertex_shader->state.stream_output;
68   }
69
70   return state;
71}
72
73static inline boolean
74draw_has_so(const struct draw_context *draw)
75{
76   const struct pipe_stream_output_info *state = draw_so_info(draw);
77
78   if (state && state->num_outputs > 0)
79      return TRUE;
80
81   return FALSE;
82}
83
84void draw_pt_so_emit_prepare(struct pt_so_emit *emit, boolean use_pre_clip_pos)
85{
86   struct draw_context *draw = emit->draw;
87
88   emit->use_pre_clip_pos = use_pre_clip_pos;
89   emit->has_so = draw_has_so(draw);
90   if (use_pre_clip_pos)
91      emit->pos_idx = draw_current_shader_position_output(draw);
92
93   /* if we have a state with outputs make sure we have
94    * buffers to output to */
95   if (emit->has_so) {
96      boolean has_valid_buffer = FALSE;
97      unsigned i;
98      for (i = 0; i < draw->so.num_targets; ++i) {
99         if (draw->so.targets[i]) {
100            has_valid_buffer = TRUE;
101            break;
102         }
103      }
104      emit->has_so = has_valid_buffer;
105   }
106
107   if (!emit->has_so)
108      return;
109
110   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
111    */
112   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
113}
114
115static void so_emit_prim(struct pt_so_emit *so,
116                         unsigned *indices,
117                         unsigned num_vertices)
118{
119   unsigned slot, i;
120   unsigned input_vertex_stride = so->input_vertex_stride;
121   struct draw_context *draw = so->draw;
122   const float (*input_ptr)[4];
123   const float *pcp_ptr = NULL;
124   const struct pipe_stream_output_info *state = draw_so_info(draw);
125   float *buffer;
126   int buffer_total_bytes[PIPE_MAX_SO_BUFFERS];
127   boolean buffer_written[PIPE_MAX_SO_BUFFERS] = {0};
128
129   input_ptr = so->inputs;
130   if (so->use_pre_clip_pos)
131      pcp_ptr = so->pre_clip_pos;
132
133   ++so->generated_primitives;
134
135   for (i = 0; i < draw->so.num_targets; i++) {
136      struct draw_so_target *target = draw->so.targets[i];
137      if (target) {
138         buffer_total_bytes[i] = target->internal_offset;
139      } else {
140         buffer_total_bytes[i] = 0;
141      }
142   }
143
144   /* check have we space to emit prim first - if not don't do anything */
145   for (i = 0; i < num_vertices; ++i) {
146      unsigned ob;
147      for (slot = 0; slot < state->num_outputs; ++slot) {
148         unsigned num_comps = state->output[slot].num_components;
149         int ob = state->output[slot].output_buffer;
150         unsigned dst_offset = state->output[slot].dst_offset * sizeof(float);
151         unsigned write_size = num_comps * sizeof(float);
152
153         if (state->output[slot].stream != so->stream)
154            continue;
155         /* If a buffer is missing then that's equivalent to
156          * an overflow */
157         if (!draw->so.targets[ob]) {
158            return;
159         }
160         if ((buffer_total_bytes[ob] + write_size + dst_offset) >
161             draw->so.targets[ob]->target.buffer_size) {
162            return;
163         }
164      }
165      for (ob = 0; ob < draw->so.num_targets; ++ob) {
166         buffer_total_bytes[ob] += state->stride[ob] * sizeof(float);
167      }
168   }
169
170   for (i = 0; i < num_vertices; ++i) {
171      const float (*input)[4];
172      const float *pre_clip_pos = NULL;
173      unsigned  ob;
174
175      input = (const float (*)[4])(
176         (const char *)input_ptr + (indices[i] * input_vertex_stride));
177
178      if (pcp_ptr)
179         pre_clip_pos = (const float *)(
180         (const char *)pcp_ptr + (indices[i] * input_vertex_stride));
181
182      for (slot = 0; slot < state->num_outputs; ++slot) {
183         unsigned idx = state->output[slot].register_index;
184         unsigned start_comp = state->output[slot].start_component;
185         unsigned num_comps = state->output[slot].num_components;
186         unsigned stream = state->output[slot].stream;
187
188         if (stream != so->stream)
189            continue;
190         ob = state->output[slot].output_buffer;
191         buffer_written[ob] = TRUE;
192
193         buffer = (float *)((char *)draw->so.targets[ob]->mapping +
194                            draw->so.targets[ob]->target.buffer_offset +
195                            draw->so.targets[ob]->internal_offset) +
196            state->output[slot].dst_offset;
197
198         if (idx == so->pos_idx && pcp_ptr && so->stream == 0)
199            memcpy(buffer, &pre_clip_pos[start_comp],
200                   num_comps * sizeof(float));
201         else
202            memcpy(buffer, &input[idx][start_comp],
203                   num_comps * sizeof(float));
204#if 0
205         {
206            int j;
207            debug_printf("VERT[%d], stream = %d, offset = %d, slot[%d] sc = %d, num_c = %d, idx = %d = [",
208                         i, stream,
209                         draw->so.targets[ob]->internal_offset,
210                         slot, start_comp, num_comps, idx);
211            for (j = 0; j < num_comps; ++j) {
212               unsigned *ubuffer = (unsigned*)buffer;
213               debug_printf("%d (0x%x), ", ubuffer[j], ubuffer[j]);
214            }
215            debug_printf("]\n");
216         }
217#endif
218      }
219      for (ob = 0; ob < draw->so.num_targets; ++ob) {
220         struct draw_so_target *target = draw->so.targets[ob];
221         if (target && buffer_written[ob]) {
222            target->internal_offset += state->stride[ob] * sizeof(float);
223         }
224      }
225   }
226   ++so->emitted_primitives;
227}
228
229static void so_point(struct pt_so_emit *so, int idx)
230{
231   unsigned indices[1];
232
233   indices[0] = idx;
234
235   so_emit_prim(so, indices, 1);
236}
237
238static void so_line(struct pt_so_emit *so, int i0, int i1)
239{
240   unsigned indices[2];
241
242   indices[0] = i0;
243   indices[1] = i1;
244
245   so_emit_prim(so, indices, 2);
246}
247
248static void so_tri(struct pt_so_emit *so, int i0, int i1, int i2)
249{
250   unsigned indices[3];
251
252   indices[0] = i0;
253   indices[1] = i1;
254   indices[2] = i2;
255
256   so_emit_prim(so, indices, 3);
257}
258
259
260#define FUNC         so_run_linear
261#define GET_ELT(idx) (start + (idx))
262#include "draw_so_emit_tmp.h"
263
264
265#define FUNC         so_run_elts
266#define LOCAL_VARS   const ushort *elts = input_prims->elts;
267#define GET_ELT(idx) (elts[start + (idx)])
268#include "draw_so_emit_tmp.h"
269
270
271void draw_pt_so_emit( struct pt_so_emit *emit,
272                      int num_vertex_streams,
273                      const struct draw_vertex_info *input_verts,
274                      const struct draw_prim_info *input_prims )
275{
276   struct draw_context *draw = emit->draw;
277   struct vbuf_render *render = draw->render;
278   unsigned start, i, stream;
279
280   if (!emit->has_so) {
281      if (draw->collect_primgen) {
282         unsigned i;
283         unsigned total = 0;
284         for (i = 0; i < input_prims->primitive_count; i++) {
285            total +=
286               u_decomposed_prims_for_vertices(input_prims->prim,
287                                               input_prims->primitive_lengths[i]);
288         }
289         render->set_stream_output_info(render,
290                                        0, 0, total);
291      }
292      return;
293   }
294
295   if (!draw->so.num_targets)
296      return;
297
298   /* XXX: need to flush to get prim_vbuf.c to release its allocation??*/
299   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
300
301   for (stream = 0; stream < num_vertex_streams; stream++) {
302      emit->emitted_primitives = 0;
303      emit->generated_primitives = 0;
304      if (emit->use_pre_clip_pos)
305         emit->pre_clip_pos = input_verts[stream].verts->clip_pos;
306
307      emit->input_vertex_stride = input_verts[stream].stride;
308      emit->inputs = (const float (*)[4])input_verts[stream].verts->data;
309      emit->stream = stream;
310      for (start = i = 0; i < input_prims[stream].primitive_count;
311           start += input_prims[stream].primitive_lengths[i], i++)
312      {
313         unsigned count = input_prims[stream].primitive_lengths[i];
314
315         if (input_prims->linear) {
316            so_run_linear(emit, &input_prims[stream], &input_verts[stream],
317                          start, count);
318         } else {
319            so_run_elts(emit, &input_prims[stream], &input_verts[stream],
320                        start, count);
321         }
322      }
323      render->set_stream_output_info(render,
324                                     stream,
325                                     emit->emitted_primitives,
326                                     emit->generated_primitives);
327   }
328}
329
330
331struct pt_so_emit *draw_pt_so_emit_create( struct draw_context *draw )
332{
333   struct pt_so_emit *emit = CALLOC_STRUCT(pt_so_emit);
334   if (!emit)
335      return NULL;
336
337   emit->draw = draw;
338
339   return emit;
340}
341
342void draw_pt_so_emit_destroy( struct pt_so_emit *emit )
343{
344   FREE(emit);
345}
346