src/mesa/drivers/dri/i965/brw_clip_util.c

   1 /*
   2  Copyright (C) Intel Corp.  2006.  All Rights Reserved.
   3  Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
   4  develop this 3D driver.
   5
   6  Permission is hereby granted, free of charge, to any person obtaining
   7  a 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, sublicense, 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
  16  portions of the Software.
  17
  18  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  19  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
  21  IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
  22  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  23  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  24  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25
  26  **********************************************************************/
  27  /*
  28   * Authors:
  29   *   Keith Whitwell <keith@tungstengraphics.com>
  30   */
  31
  32
  33 #include "glheader.h"
  34 #include "macros.h"
  35 #include "enums.h"
  36
  37 #include "shader/program.h"
  38 #include "intel_batchbuffer.h"
  39
  40 #include "brw_defines.h"
  41 #include "brw_context.h"
  42 #include "brw_eu.h"
  43 #include "brw_util.h"
  44 #include "brw_clip.h"
  45
  46
  47
  48
  49
  50 static struct brw_reg get_tmp( struct brw_clip_compile *c )
  51 {
  52    struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0);
  53
  54    if (++c->last_tmp > c->prog_data.total_grf)
  55       c->prog_data.total_grf = c->last_tmp;
  56
  57    return tmp;
  58 }
  59
  60 static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp )
  61 {
  62    if (tmp.nr == c->last_tmp-1)
  63       c->last_tmp--;
  64 }
  65
  66
  67 static struct brw_reg make_plane_ud(GLuint x, GLuint y, GLuint z, GLuint w)
  68 {
  69    return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x);
  70 }
  71
  72
  73 void brw_clip_init_planes( struct brw_clip_compile *c )
  74 {
  75    struct brw_compile *p = &c->func;
  76
  77    if (!c->key.nr_userclip) {
  78       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0,    0, 0xff, 1));
  79       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0,    0,    1, 1));
  80       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff,    0, 1));
  81       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0,    1,    0, 1));
  82       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff,  0,    0, 1));
  83       brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1,    0,    0, 1));
  84    }
  85 }
  86
  87
  88
  89 #define W 3
  90
  91 /* Project 'pos' to screen space (or back again), overwrite with results:
  92  */
  93 static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos )
  94 {
  95    struct brw_compile *p = &c->func;
  96
  97    /* calc rhw
  98     */
  99    brw_math_invert(p, get_element(pos, W), get_element(pos, W));
 100
 101    /* value.xyz *= value.rhw
 102     */
 103    brw_set_access_mode(p, BRW_ALIGN_16);
 104    brw_MUL(p, brw_writemask(pos, WRITEMASK_XYZ), pos, brw_swizzle1(pos, W));
 105    brw_set_access_mode(p, BRW_ALIGN_1);
 106 }
 107
 108
 109 static void brw_clip_project_vertex( struct brw_clip_compile *c,
 110                                      struct brw_indirect vert_addr )
 111 {
 112    struct brw_compile *p = &c->func;
 113    struct brw_reg tmp = get_tmp(c);
 114
 115    /* Fixup position.  Extract from the original vertex and re-project
 116     * to screen space:
 117     */
 118    brw_MOV(p, tmp, deref_4f(vert_addr, c->offset[VERT_RESULT_HPOS]));
 119    brw_clip_project_position(c, tmp);
 120    brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp);
 121
 122    release_tmp(c, tmp);
 123 }
 124
 125
 126
 127
 128 /* Interpolate between two vertices and put the result into a0.0.
 129  * Increment a0.0 accordingly.
 130  */
 131 void brw_clip_interp_vertex( struct brw_clip_compile *c,
 132                              struct brw_indirect dest_ptr,
 133                              struct brw_indirect v0_ptr, /* from */
 134                              struct brw_indirect v1_ptr, /* to */
 135                              struct brw_reg t0,
 136                              GLboolean force_edgeflag)
 137 {
 138    struct brw_compile *p = &c->func;
 139    struct brw_reg tmp = get_tmp(c);
 140    GLuint i;
 141
 142    /* Just copy the vertex header:
 143     */
 144    brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
 145
 146    /* Iterate over each attribute (could be done in pairs?)
 147     */
 148    for (i = 0; i < c->nr_attrs; i++) {
 149       GLuint delta = i*16 + 32;
 150
 151       if (delta == c->offset[VERT_RESULT_EDGE]) {
 152          if (force_edgeflag)
 153             brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1));
 154          else
 155             brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta));
 156       }
 157       else {
 158          /* Interpolate:
 159           *
 160           *        New = attr0 + t*attr1 - t*attr0
 161           */
 162          brw_MUL(p,
 163                  vec4(brw_null_reg()),
 164                  deref_4f(v1_ptr, delta),
 165                  t0);
 166
 167          brw_MAC(p,
 168                  tmp,
 169                  negate(deref_4f(v0_ptr, delta)),
 170                  t0);
 171
 172          brw_ADD(p,
 173                  deref_4f(dest_ptr, delta),
 174                  deref_4f(v0_ptr, delta),
 175                  tmp);
 176       }
 177    }
 178
 179    if (i & 1) {
 180       GLuint delta = i*16 + 32;
 181       brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
 182    }
 183
 184    release_tmp(c, tmp);
 185
 186    /* Recreate the projected (NDC) coordinate in the new vertex
 187     * header:
 188     */
 189    brw_clip_project_vertex(c, dest_ptr );
 190 }
 191
 192
 193
 194
 195 #define MAX_MRF 16
 196
 197 void brw_clip_emit_vue(struct brw_clip_compile *c,
 198                        struct brw_indirect vert,
 199                        GLboolean allocate,
 200                        GLboolean eot,
 201                        GLuint header)
 202 {
 203    struct brw_compile *p = &c->func;
 204    GLuint start = c->last_mrf;
 205
 206    assert(!(allocate && eot));
 207
 208    /* Cycle through mrf regs - probably futile as we have to wait for
 209     * the allocation response anyway.  Also, the order this function
 210     * is invoked doesn't correspond to the order the instructions will
 211     * be executed, so it won't have any effect in many cases.
 212     */
 213 #if 0
 214    if (start + c->nr_regs + 1 >= MAX_MRF)
 215       start = 0;
 216
 217    c->last_mrf = start + c->nr_regs + 1;
 218 #endif
 219
 220    /* Copy the vertex from vertn into m1..mN+1:
 221     */
 222    brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs);
 223
 224    /* Overwrite PrimType and PrimStart in the message header, for
 225     * each vertex in turn:
 226     */
 227    brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header));
 228
 229
 230    /* Send each vertex as a seperate write to the urb.  This
 231     * is different to the concept in brw_sf_emit.c, where
 232     * subsequent writes are used to build up a single urb
 233     * entry.  Each of these writes instantiates a seperate
 234     * urb entry - (I think... what about 'allocate'?)
 235     */
 236    brw_urb_WRITE(p,
 237                  allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
 238                  start,
 239                  c->reg.R0,
 240                  allocate,
 241                  1,             /* used */
 242                  c->nr_regs + 1, /* msg length */
 243                  allocate ? 1 : 0, /* response_length */
 244                  eot,           /* eot */
 245                  1,             /* writes_complete */
 246                  0,             /* urb offset */
 247                  BRW_URB_SWIZZLE_NONE);
 248 }
 249
 250
 251
 252 void brw_clip_kill_thread(struct brw_clip_compile *c)
 253 {
 254    struct brw_compile *p = &c->func;
 255
 256    /* Send an empty message to kill the thread and release any
 257     * allocated urb entry:
 258     */
 259    brw_urb_WRITE(p,
 260                  retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
 261                  0,
 262                  c->reg.R0,
 263                  0,             /* allocate */
 264                  0,             /* used */
 265                  0,             /* msg len */
 266                  0,             /* response len */
 267                  1,             /* eot */
 268                  1,             /* writes complete */
 269                  0,
 270                  BRW_URB_SWIZZLE_NONE);
 271 }
 272
 273
 274
 275
 276 struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c )
 277 {
 278    return brw_address(c->reg.fixed_planes);
 279 }
 280
 281
 282 struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c )
 283 {
 284    if (c->key.nr_userclip) {
 285       return brw_imm_uw(16);
 286    }
 287    else {
 288       return brw_imm_uw(4);
 289    }
 290 }
 291
 292
 293 /* If flatshading, distribute color from provoking vertex prior to
 294  * clipping.
 295  */
 296 void brw_clip_copy_colors( struct brw_clip_compile *c,
 297                            GLuint to, GLuint from )
 298 {
 299    struct brw_compile *p = &c->func;
 300
 301    if (c->offset[VERT_RESULT_COL0])
 302       brw_MOV(p,
 303               byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL0]),
 304               byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL0]));
 305
 306    if (c->offset[VERT_RESULT_COL1])
 307       brw_MOV(p,
 308               byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL1]),
 309               byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL1]));
 310
 311    if (c->offset[VERT_RESULT_BFC0])
 312       brw_MOV(p,
 313               byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC0]),
 314               byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC0]));
 315
 316    if (c->offset[VERT_RESULT_BFC1])
 317       brw_MOV(p,
 318               byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC1]),
 319               byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC1]));
 320 }
 321
 322
 323
 324 void brw_clip_init_clipmask( struct brw_clip_compile *c )
 325 {
 326    struct brw_compile *p = &c->func;
 327    struct brw_reg incoming = get_element_ud(c->reg.R0, 2);
 328
 329    /* Shift so that lowest outcode bit is rightmost:
 330     */
 331    brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26));
 332
 333    if (c->key.nr_userclip) {
 334       struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD);
 335
 336       /* Rearrange userclip outcodes so that they come directly after
 337        * the fixed plane bits.
 338        */
 339       brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14));
 340       brw_SHR(p, tmp, tmp, brw_imm_ud(8));
 341       brw_OR(p, c->reg.planemask, c->reg.planemask, tmp);
 342
 343       release_tmp(c, tmp);
 344    }
 345 }
 346