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 "util/u_math.h"
  33 #include "lp_debug.h"
  34 #include "lp_rast_priv.h"
  35 #include "lp_tile_soa.h"
  36
  37
  38 /**
  39  * Add a 4x4 block of pixels to the block list.
  40  * All pixels are known to be inside the triangle's bounds.
  41  */
  42 static void
  43 block_full_4( struct lp_rasterizer_task *rast_task, int x, int y )
  44 {
  45    const unsigned i = rast_task->nr_blocks;
  46    assert(x % 4 == 0);
  47    assert(y % 4 == 0);
  48    rast_task->blocks[i].x = x;
  49    rast_task->blocks[i].y = y;
  50    rast_task->blocks[i].mask = ~0;
  51    rast_task->nr_blocks++;
  52 }
  53
  54
  55 /**
  56  * Add a 16x16 block of pixels to the block list.
  57  * All pixels are known to be inside the triangle's bounds.
  58  */
  59 static void
  60 block_full_16( struct lp_rasterizer_task *rast_task, 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(rast_task, x + ix, y + iy);
  68 }
  69
  70
  71 /**
  72  * Evaluate each pixel in a 4x4 block to determine if it lies within
  73  * the triangle's bounds.
  74  * Generate a mask of in/out flags and add the block to the blocks list.
  75  */
  76 static void
  77 do_block_4( struct lp_rasterizer_task *rast_task,
  78             const struct lp_rast_triangle *tri,
  79             int x, int y,
  80             int c1,
  81             int c2,
  82             int c3 )
  83 {
  84    int i;
  85    unsigned mask = 0;
  86
  87    assert(x % 4 == 0);
  88    assert(y % 4 == 0);
  89
  90    for (i = 0; i < 16; i++) {
  91       int any_negative = ((c1 + tri->step[0][i]) |
  92                           (c2 + tri->step[1][i]) |
  93                           (c3 + tri->step[2][i])) >> 31;
  94       mask |= (~any_negative) & (1 << i);
  95    }
  96
  97    /* As we do trivial reject already, masks should rarely be all zero:
  98     */
  99    if (mask) {
 100       const unsigned i = rast_task->nr_blocks;
 101       rast_task->blocks[i].x = x;
 102       rast_task->blocks[i].y = y;
 103       rast_task->blocks[i].mask = mask;
 104       rast_task->nr_blocks++;
 105    }
 106 }
 107
 108
 109 /**
 110  * Evaluate a 16x16 block of pixels to determine which 4x4 subblocks are in/out
 111  * of the triangle's bounds.
 112  */
 113 static void
 114 do_block_16( struct lp_rasterizer_task *rast_task,
 115              const struct lp_rast_triangle *tri,
 116              int x, int y,
 117              int c1,
 118              int c2,
 119              int c3 )
 120 {
 121    int ix, iy, i = 0;
 122
 123    int ei1 = tri->ei1 * 4;
 124    int ei2 = tri->ei2 * 4;
 125    int ei3 = tri->ei3 * 4;
 126
 127    int eo1 = tri->eo1 * 4;
 128    int eo2 = tri->eo2 * 4;
 129    int eo3 = tri->eo3 * 4;
 130
 131    assert(x % 16 == 0);
 132    assert(y % 16 == 0);
 133
 134    for (iy = 0; iy < 16; iy+=4) {
 135       for (ix = 0; ix < 16; ix+=4, i++) {
 136          int cx1 = c1 + (tri->step[0][i] * 4);
 137          int cx2 = c2 + (tri->step[1][i] * 4);
 138          int cx3 = c3 + (tri->step[2][i] * 4);
 139
 140          if (cx1 + eo1 < 0 ||
 141              cx2 + eo2 < 0 ||
 142              cx3 + eo3 < 0) {
 143             /* the block is completely outside the triangle - nop */
 144          }
 145          else if (cx1 + ei1 > 0 &&
 146                   cx2 + ei2 > 0 &&
 147                   cx3 + ei3 > 0) {
 148             /* the block is completely inside the triangle */
 149             block_full_4(rast_task, x+ix, y+iy);
 150          }
 151          else {
 152             /* the block is partially in/out of the triangle */
 153             do_block_4(rast_task, tri, x+ix, y+iy, cx1, cx2, cx3);
 154          }
 155       }
 156    }
 157 }
 158
 159
 160 /**
 161  * Scan the tile in chunks and figure out which pixels to rasterize
 162  * for this triangle.
 163  */
 164 void
 165 lp_rast_triangle( struct lp_rasterizer *rast,
 166                   unsigned thread_index,
 167                   const union lp_rast_cmd_arg arg )
 168 {
 169    struct lp_rasterizer_task *rast_task = &rast->tasks[thread_index];
 170    const struct lp_rast_triangle *tri = arg.triangle;
 171
 172    int x = rast_task->x;
 173    int y = rast_task->y;
 174    int ix, iy;
 175    unsigned i = 0;
 176
 177    int c1 = tri->c1 + tri->dx12 * y - tri->dy12 * x;
 178    int c2 = tri->c2 + tri->dx23 * y - tri->dy23 * x;
 179    int c3 = tri->c3 + tri->dx31 * y - tri->dy31 * x;
 180
 181    int ei1 = tri->ei1 * 16;
 182    int ei2 = tri->ei2 * 16;
 183    int ei3 = tri->ei3 * 16;
 184
 185    int eo1 = tri->eo1 * 16;
 186    int eo2 = tri->eo2 * 16;
 187    int eo3 = tri->eo3 * 16;
 188
 189    assert(Elements(rast_task->blocks) == (TILE_SIZE * TILE_SIZE) / (4*4));
 190
 191    LP_DBG(DEBUG_RAST, "lp_rast_triangle\n");
 192
 193    rast_task->nr_blocks = 0;
 194
 195    /* Walk over the tile to build a list of 4x4 pixel blocks which will
 196     * be filled/shaded.  We do this at two granularities: 16x16 blocks
 197     * and then 4x4 blocks.
 198     */
 199    for (iy = 0; iy < TILE_SIZE; iy += 16) {
 200       for (ix = 0; ix < TILE_SIZE; ix += 16, i++) {
 201          int cx1 = c1 + (tri->step[0][i] * 16);
 202          int cx2 = c2 + (tri->step[1][i] * 16);
 203          int cx3 = c3 + (tri->step[2][i] * 16);
 204
 205          if (cx1 + eo1 < 0 ||
 206              cx2 + eo2 < 0 ||
 207              cx3 + eo3 < 0) {
 208             /* the block is completely outside the triangle - nop */
 209          }
 210          else if (cx1 + ei1 > 0 &&
 211                   cx2 + ei2 > 0 &&
 212                   cx3 + ei3 > 0) {
 213             /* the block is completely inside the triangle */
 214             block_full_16(rast_task, x+ix, y+iy);
 215          }
 216          else {
 217             /* the block is partially in/out of the triangle */
 218             do_block_16(rast_task, tri, x+ix, y+iy, cx1, cx2, cx3);
 219          }
 220       }
 221    }
 222
 223    assert(rast_task->nr_blocks <= Elements(rast_task->blocks));
 224
 225    /* Shade the 4x4 pixel blocks */
 226    for (i = 0; i < rast_task->nr_blocks; i++)
 227       lp_rast_shade_quads(rast,
 228                           thread_index,
 229                           &tri->inputs,
 230                           rast_task->blocks[i].x,
 231                           rast_task->blocks[i].y,
 232                           rast_task->blocks[i].mask);
 233 }