src/gallium/drivers/llvmpipe/lp_rast_tri.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2007-2009 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 #include <limits.h>
  33 #include "util/u_math.h"
  34 #include "lp_debug.h"
  35 #include "lp_perf.h"
  36 #include "lp_rast_priv.h"
  37 #include "lp_tile_soa.h"
  38
  39
  40
  41
  42 /**
  43  * Shade all pixels in a 4x4 block.
  44  */
  45 static void
  46 block_full_4(struct lp_rasterizer_task *task,
  47              const struct lp_rast_triangle *tri,
  48              int x, int y)
  49 {
  50    lp_rast_shade_quads_all(task, &tri->inputs, x, y);
  51 }
  52
  53
  54 /**
  55  * Shade all pixels in a 16x16 block.
  56  */
  57 static void
  58 block_full_16(struct lp_rasterizer_task *task,
  59               const struct lp_rast_triangle *tri,
  60               int x, int y)
  61 {
  62    unsigned ix, iy;
  63    assert(x % 16 == 0);
  64    assert(y % 16 == 0);
  65    for (iy = 0; iy < 16; iy += 4)
  66       for (ix = 0; ix < 16; ix += 4)
  67          block_full_4(task, tri, x + ix, y + iy);
  68 }
  69
  70
  71 static INLINE unsigned
  72 build_mask(int c, int dcdx, int dcdy)
  73 {
  74    int mask = 0;
  75
  76    int c0 = c;
  77    int c1 = c0 + dcdx;
  78    int c2 = c1 + dcdx;
  79    int c3 = c2 + dcdx;
  80
  81    mask |= ((c0 + 0 * dcdy) >> 31) & (1 << 0);
  82    mask |= ((c0 + 1 * dcdy) >> 31) & (1 << 2);
  83    mask |= ((c0 + 2 * dcdy) >> 31) & (1 << 8);
  84    mask |= ((c0 + 3 * dcdy) >> 31) & (1 << 10);
  85    mask |= ((c1 + 0 * dcdy) >> 31) & (1 << 1);
  86    mask |= ((c1 + 1 * dcdy) >> 31) & (1 << 3);
  87    mask |= ((c1 + 2 * dcdy) >> 31) & (1 << 9);
  88    mask |= ((c1 + 3 * dcdy) >> 31) & (1 << 11);
  89    mask |= ((c2 + 0 * dcdy) >> 31) & (1 << 4);
  90    mask |= ((c2 + 1 * dcdy) >> 31) & (1 << 6);
  91    mask |= ((c2 + 2 * dcdy) >> 31) & (1 << 12);
  92    mask |= ((c2 + 3 * dcdy) >> 31) & (1 << 14);
  93    mask |= ((c3 + 0 * dcdy) >> 31) & (1 << 5);
  94    mask |= ((c3 + 1 * dcdy) >> 31) & (1 << 7);
  95    mask |= ((c3 + 2 * dcdy) >> 31) & (1 << 13);
  96    mask |= ((c3 + 3 * dcdy) >> 31) & (1 << 15);
  97
  98    return mask;
  99 }
 100
 101 static INLINE unsigned
 102 build_mask_linear(int c, int dcdx, int dcdy)
 103 {
 104    int mask = 0;
 105
 106    int c0 = c;
 107    int c1 = c0 + dcdy;
 108    int c2 = c1 + dcdy;
 109    int c3 = c2 + dcdy;
 110
 111    mask |= ((c0 + 0 * dcdx) >> 31) & (1 << 0);
 112    mask |= ((c0 + 1 * dcdx) >> 31) & (1 << 1);
 113    mask |= ((c0 + 2 * dcdx) >> 31) & (1 << 2);
 114    mask |= ((c0 + 3 * dcdx) >> 31) & (1 << 3);
 115    mask |= ((c1 + 0 * dcdx) >> 31) & (1 << 4);
 116    mask |= ((c1 + 1 * dcdx) >> 31) & (1 << 5);
 117    mask |= ((c1 + 2 * dcdx) >> 31) & (1 << 6);
 118    mask |= ((c1 + 3 * dcdx) >> 31) & (1 << 7);
 119    mask |= ((c2 + 0 * dcdx) >> 31) & (1 << 8);
 120    mask |= ((c2 + 1 * dcdx) >> 31) & (1 << 9);
 121    mask |= ((c2 + 2 * dcdx) >> 31) & (1 << 10);
 122    mask |= ((c2 + 3 * dcdx) >> 31) & (1 << 11);
 123    mask |= ((c3 + 0 * dcdx) >> 31) & (1 << 12);
 124    mask |= ((c3 + 1 * dcdx) >> 31) & (1 << 13);
 125    mask |= ((c3 + 2 * dcdx) >> 31) & (1 << 14);
 126    mask |= ((c3 + 3 * dcdx) >> 31) & (1 << 15);
 127
 128    return mask;
 129 }
 130
 131
 132 #define TAG(x) x##_1
 133 #define NR_PLANES 1
 134 #include "lp_rast_tri_tmp.h"
 135
 136 #define TAG(x) x##_2
 137 #define NR_PLANES 2
 138 #include "lp_rast_tri_tmp.h"
 139
 140 #define TAG(x) x##_3
 141 #define NR_PLANES 3
 142 #include "lp_rast_tri_tmp.h"
 143
 144 #define TAG(x) x##_4
 145 #define NR_PLANES 4
 146 #include "lp_rast_tri_tmp.h"
 147
 148 #define TAG(x) x##_5
 149 #define NR_PLANES 5
 150 #include "lp_rast_tri_tmp.h"
 151
 152 #define TAG(x) x##_6
 153 #define NR_PLANES 6
 154 #include "lp_rast_tri_tmp.h"
 155
 156 #define TAG(x) x##_7
 157 #define NR_PLANES 7
 158 #include "lp_rast_tri_tmp.h"
 159
 160
 161 /* Special case for 3 plane triangle which is contained entirely
 162  * within a 16x16 block.
 163  */
 164 void
 165 lp_rast_triangle_3_16(struct lp_rasterizer_task *task,
 166                       const union lp_rast_cmd_arg arg)
 167 {
 168    const struct lp_rast_triangle *tri = arg.triangle.tri;
 169    const struct lp_rast_plane *plane = tri->plane;
 170    unsigned mask = arg.triangle.plane_mask;
 171    const int x = task->x + (mask & 0xf) * 16;
 172    const int y = task->y + (mask >> 4) * 16;
 173    unsigned outmask, inmask, partmask, partial_mask;
 174    unsigned j;
 175    int c[3];
 176
 177    outmask = 0;                 /* outside one or more trivial reject planes */
 178    partmask = 0;                /* outside one or more trivial accept planes */
 179
 180    for (j = 0; j < 3; j++) {
 181       c[j] = plane[j].c + plane[j].dcdy * y - plane[j].dcdx * x;
 182
 183       {
 184          const int dcdx = -plane[j].dcdx * 4;
 185          const int dcdy = plane[j].dcdy * 4;
 186          const int cox = c[j] + plane[j].eo * 4;
 187          const int cio = c[j] + plane[j].ei * 4 - 1;
 188
 189          outmask |= build_mask_linear(cox, dcdx, dcdy);
 190          partmask |= build_mask_linear(cio, dcdx, dcdy);
 191       }
 192    }
 193
 194    if (outmask == 0xffff)
 195       return;
 196
 197    /* Mask of sub-blocks which are inside all trivial accept planes:
 198     */
 199    inmask = ~partmask & 0xffff;
 200
 201    /* Mask of sub-blocks which are inside all trivial reject planes,
 202     * but outside at least one trivial accept plane:
 203     */
 204    partial_mask = partmask & ~outmask;
 205
 206    assert((partial_mask & inmask) == 0);
 207
 208    /* Iterate over partials:
 209     */
 210    while (partial_mask) {
 211       int i = ffs(partial_mask) - 1;
 212       int ix = (i & 3) * 4;
 213       int iy = (i >> 2) * 4;
 214       int px = x + ix;
 215       int py = y + iy;
 216       int cx[3];
 217
 218       partial_mask &= ~(1 << i);
 219
 220       for (j = 0; j < 3; j++)
 221          cx[j] = (c[j]
 222                   - plane[j].dcdx * ix
 223                   + plane[j].dcdy * iy);
 224
 225       do_block_4_3(task, tri, plane, px, py, cx);
 226    }
 227
 228    /* Iterate over fulls:
 229     */
 230    while (inmask) {
 231       int i = ffs(inmask) - 1;
 232       int ix = (i & 3) * 4;
 233       int iy = (i >> 2) * 4;
 234       int px = x + ix;
 235       int py = y + iy;
 236
 237       inmask &= ~(1 << i);
 238
 239       block_full_4(task, tri, px, py);
 240    }
 241 }