lp_rast_tri_tmp.h revision 9f6e8e1d6b8696a3ee96cba01b2466ba7a1a8ef6 
1/**************************************************************************
2 *
3 * Copyright 2007-2010 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/*
29 * Rasterization for binned triangles within a tile
30 */
31
32
33
34/**
35 * Prototype for a 8 plane rasterizer function.  Will codegenerate
36 * several of these.
37 *
38 * XXX: Varients for more/fewer planes.
39 * XXX: Need ways of dropping planes as we descend.
40 * XXX: SIMD
41 */
42static void
43TAG(do_block_4)(struct lp_rasterizer_task *task,
44                const struct lp_rast_triangle *tri,
45                const struct lp_rast_plane *plane,
46                int x, int y,
47                const int *c)
48{
49   unsigned mask = 0xffff;
50   int j;
51
52   for (j = 0; j < NR_PLANES; j++) {
53      mask &= ~build_mask(c[j] - 1,
54			  -plane[j].dcdx,
55			  plane[j].dcdy);
56   }
57
58   /* Now pass to the shader:
59    */
60   if (mask)
61      lp_rast_shade_quads_mask(task, &tri->inputs, x, y, mask);
62}
63
64/**
65 * Evaluate a 16x16 block of pixels to determine which 4x4 subblocks are in/out
66 * of the triangle's bounds.
67 */
68static void
69TAG(do_block_16)(struct lp_rasterizer_task *task,
70                 const struct lp_rast_triangle *tri,
71                 const struct lp_rast_plane *plane,
72                 int x, int y,
73                 const int *c)
74{
75   unsigned outmask, inmask, partmask, partial_mask;
76   unsigned j;
77
78   outmask = 0;                 /* outside one or more trivial reject planes */
79   partmask = 0;                /* outside one or more trivial accept planes */
80
81   for (j = 0; j < NR_PLANES; j++) {
82      const int dcdx = -plane[j].dcdx * 4;
83      const int dcdy = plane[j].dcdy * 4;
84      const int cox = plane[j].eo * 4;
85      const int cio = plane[j].ei * 4 - 1;
86
87      build_masks(c[j] + cox,
88		  cio - cox,
89		  dcdx, dcdy,
90		  &outmask,   /* sign bits from c[i][0..15] + cox */
91		  &partmask); /* sign bits from c[i][0..15] + cio */
92   }
93
94   if (outmask == 0xffff)
95      return;
96
97   /* Mask of sub-blocks which are inside all trivial accept planes:
98    */
99   inmask = ~partmask & 0xffff;
100
101   /* Mask of sub-blocks which are inside all trivial reject planes,
102    * but outside at least one trivial accept plane:
103    */
104   partial_mask = partmask & ~outmask;
105
106   assert((partial_mask & inmask) == 0);
107
108   LP_COUNT_ADD(nr_empty_4, util_bitcount(0xffff & ~(partial_mask | inmask)));
109
110   /* Iterate over partials:
111    */
112   while (partial_mask) {
113      int i = ffs(partial_mask) - 1;
114      int ix = (i & 3) * 4;
115      int iy = (i >> 2) * 4;
116      int px = x + ix;
117      int py = y + iy;
118      int cx[NR_PLANES];
119
120      partial_mask &= ~(1 << i);
121
122      LP_COUNT(nr_partially_covered_4);
123
124      for (j = 0; j < NR_PLANES; j++)
125         cx[j] = (c[j]
126		  - plane[j].dcdx * ix
127		  + plane[j].dcdy * iy);
128
129      TAG(do_block_4)(task, tri, plane, px, py, cx);
130   }
131
132   /* Iterate over fulls:
133    */
134   while (inmask) {
135      int i = ffs(inmask) - 1;
136      int ix = (i & 3) * 4;
137      int iy = (i >> 2) * 4;
138      int px = x + ix;
139      int py = y + iy;
140
141      inmask &= ~(1 << i);
142
143      LP_COUNT(nr_fully_covered_4);
144      block_full_4(task, tri, px, py);
145   }
146}
147
148
149/**
150 * Scan the tile in chunks and figure out which pixels to rasterize
151 * for this triangle.
152 */
153void
154TAG(lp_rast_triangle)(struct lp_rasterizer_task *task,
155                      const union lp_rast_cmd_arg arg)
156{
157   const struct lp_rast_triangle *tri = arg.triangle.tri;
158   unsigned plane_mask = arg.triangle.plane_mask;
159   const int x = task->x, y = task->y;
160   struct lp_rast_plane plane[NR_PLANES];
161   int c[NR_PLANES];
162   unsigned outmask, inmask, partmask, partial_mask;
163   unsigned j = 0;
164
165   if (tri->inputs.disable) {
166      /* This triangle was partially binned and has been disabled */
167      return;
168   }
169
170   outmask = 0;                 /* outside one or more trivial reject planes */
171   partmask = 0;                /* outside one or more trivial accept planes */
172
173   if (tri->inputs.disable) {
174      /* This triangle was partially binned and has been disabled */
175      return;
176   }
177
178   while (plane_mask) {
179      int i = ffs(plane_mask) - 1;
180      plane[j] = tri->plane[i];
181      plane_mask &= ~(1 << i);
182      c[j] = plane[j].c + plane[j].dcdy * y - plane[j].dcdx * x;
183
184      {
185	 const int dcdx = -plane[j].dcdx * 16;
186	 const int dcdy = plane[j].dcdy * 16;
187	 const int cox = plane[j].eo * 16;
188	 const int cio = plane[j].ei * 16 - 1;
189
190	 build_masks(c[j] + cox,
191		     cio - cox,
192		     dcdx, dcdy,
193		     &outmask,   /* sign bits from c[i][0..15] + cox */
194		     &partmask); /* sign bits from c[i][0..15] + cio */
195      }
196
197      j++;
198   }
199
200   if (outmask == 0xffff)
201      return;
202
203   /* Mask of sub-blocks which are inside all trivial accept planes:
204    */
205   inmask = ~partmask & 0xffff;
206
207   /* Mask of sub-blocks which are inside all trivial reject planes,
208    * but outside at least one trivial accept plane:
209    */
210   partial_mask = partmask & ~outmask;
211
212   assert((partial_mask & inmask) == 0);
213
214   LP_COUNT_ADD(nr_empty_16, util_bitcount(0xffff & ~(partial_mask | inmask)));
215
216   /* Iterate over partials:
217    */
218   while (partial_mask) {
219      int i = ffs(partial_mask) - 1;
220      int ix = (i & 3) * 16;
221      int iy = (i >> 2) * 16;
222      int px = x + ix;
223      int py = y + iy;
224      int cx[NR_PLANES];
225
226      for (j = 0; j < NR_PLANES; j++)
227         cx[j] = (c[j]
228		  - plane[j].dcdx * ix
229		  + plane[j].dcdy * iy);
230
231      partial_mask &= ~(1 << i);
232
233      LP_COUNT(nr_partially_covered_16);
234      TAG(do_block_16)(task, tri, plane, px, py, cx);
235   }
236
237   /* Iterate over fulls:
238    */
239   while (inmask) {
240      int i = ffs(inmask) - 1;
241      int ix = (i & 3) * 16;
242      int iy = (i >> 2) * 16;
243      int px = x + ix;
244      int py = y + iy;
245
246      inmask &= ~(1 << i);
247
248      LP_COUNT(nr_fully_covered_16);
249      block_full_16(task, tri, px, py);
250   }
251}
252
253#undef TAG
254#undef NR_PLANES
255
256