src/gallium/drivers/i965/brw_gs_emit.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 "brw_batchbuffer.h"
  34
  35 #include "brw_defines.h"
  36 #include "brw_context.h"
  37 #include "brw_eu.h"
  38 #include "brw_gs.h"
  39
  40 static void brw_gs_alloc_regs( struct brw_gs_compile *c,
  41                                GLuint nr_verts )
  42 {
  43    GLuint i = 0,j;
  44
  45    /* Register usage is static, precompute here:
  46     */
  47    c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;
  48
  49    /* Payload vertices plus space for more generated vertices:
  50     */
  51    for (j = 0; j < nr_verts; j++) {
  52       c->reg.vertex[j] = brw_vec4_grf(i, 0);
  53       i += c->nr_regs;
  54    }
  55
  56    c->prog_data.urb_read_length = c->nr_regs;
  57    c->prog_data.total_grf = i;
  58 }
  59
  60
  61 static void brw_gs_emit_vue(struct brw_gs_compile *c,
  62                             struct brw_reg vert,
  63                             GLboolean last,
  64                             GLuint header)
  65 {
  66    struct brw_compile *p = &c->func;
  67    GLboolean allocate = !last;
  68
  69    /* Overwrite PrimType and PrimStart in the message header, for
  70     * each vertex in turn:
  71     */
  72    brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header));
  73
  74    /* Copy the vertex from vertn into m1..mN+1:
  75     */
  76    brw_copy8(p, brw_message_reg(1), vert, c->nr_regs);
  77
  78    /* Send each vertex as a seperate write to the urb.  This is
  79     * different to the concept in brw_sf_emit.c, where subsequent
  80     * writes are used to build up a single urb entry.  Each of these
  81     * writes instantiates a seperate urb entry, and a new one must be
  82     * allocated each time.
  83     */
  84    brw_urb_WRITE(p,
  85                  allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
  86                  0,
  87                  c->reg.R0,
  88                  allocate,
  89                  1,             /* used */
  90                  c->nr_regs + 1, /* msg length */
  91                  allocate ? 1 : 0, /* response length */
  92                  allocate ? 0 : 1, /* eot */
  93                  1,             /* writes_complete */
  94                  0,             /* urb offset */
  95                  BRW_URB_SWIZZLE_NONE);
  96 }
  97
  98 static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim)
  99 {
 100         struct brw_compile *p = &c->func;
 101         brw_MOV(p, get_element_ud(c->reg.R0, 1), brw_imm_ud(num_prim));
 102         brw_ff_sync(p,
 103                                 c->reg.R0,
 104                                 0,
 105                                 c->reg.R0,
 106                                 1,
 107                                 1,              /* used */
 108                                 1,      /* msg length */
 109                                 1,              /* response length */
 110                                 0,              /* eot */
 111                                 1,              /* write compelete */
 112                                 0,              /* urb offset */
 113                                 BRW_URB_SWIZZLE_NONE);
 114 }
 115
 116
 117 void brw_gs_quads( struct brw_gs_compile *c )
 118 {
 119    brw_gs_alloc_regs(c, 4);
 120
 121    /* Use polygons for correct edgeflag behaviour. Note that vertex 3
 122     * is the PV for quads, but vertex 0 for polygons:
 123     */
 124    if (c->need_ff_sync)
 125            brw_gs_ff_sync(c, 1);
 126    brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
 127    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2));
 128    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2));
 129    brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
 130 }
 131
 132 void brw_gs_quad_strip( struct brw_gs_compile *c )
 133 {
 134    brw_gs_alloc_regs(c, 4);
 135
 136    if (c->need_ff_sync)
 137            brw_gs_ff_sync(c, 1);
 138    brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
 139    brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2));
 140    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2));
 141    brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
 142 }
 143
 144 void brw_gs_tris( struct brw_gs_compile *c )
 145 {
 146    brw_gs_alloc_regs(c, 3);
 147
 148    if (c->need_ff_sync)
 149            brw_gs_ff_sync(c, 1);
 150    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START));
 151    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2));
 152    brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END));
 153 }
 154
 155 void brw_gs_lines( struct brw_gs_compile *c )
 156 {
 157    brw_gs_alloc_regs(c, 2);
 158
 159    if (c->need_ff_sync)
 160            brw_gs_ff_sync(c, 1);
 161    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START));
 162    brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END));
 163 }
 164
 165 void brw_gs_points( struct brw_gs_compile *c )
 166 {
 167    brw_gs_alloc_regs(c, 1);
 168
 169    if (c->need_ff_sync)
 170            brw_gs_ff_sync(c, 1);
 171    brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END));
 172 }
 173
 174
 175
 176
 177
 178
 179
 180