src/mesa/drivers/dri/i915/intel_render.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
   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 TUNGSTEN GRAPHICS 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  * Render unclipped vertex buffers by emitting vertices directly to
  30  * dma buffers.  Use strip/fan hardware acceleration where possible.
  31  *
  32  */
  33 #include "glheader.h"
  34 #include "context.h"
  35 #include "macros.h"
  36 #include "imports.h"
  37 #include "mtypes.h"
  38 #include "enums.h"
  39
  40 #include "tnl/t_context.h"
  41 #include "tnl/t_vertex.h"
  42
  43 #include "intel_screen.h"
  44 #include "intel_context.h"
  45 #include "intel_tris.h"
  46 #include "intel_batchbuffer.h"
  47 #include "intel_reg.h"
  48
  49 /*
  50  * Render unclipped vertex buffers by emitting vertices directly to
  51  * dma buffers.  Use strip/fan hardware primitives where possible.
  52  * Try to simulate missing primitives with indexed vertices.
  53  */
  54 #define HAVE_POINTS      0  /* Has it, but can't use because subpixel has to
  55                              * be adjusted for points on the INTEL/I845G
  56                              */
  57 #define HAVE_LINES       1
  58 #define HAVE_LINE_STRIPS 1
  59 #define HAVE_TRIANGLES   1
  60 #define HAVE_TRI_STRIPS  1
  61 #define HAVE_TRI_STRIP_1 0  /* has it, template can't use it yet */
  62 #define HAVE_TRI_FANS    1
  63 #define HAVE_POLYGONS    1
  64 #define HAVE_QUADS       0
  65 #define HAVE_QUAD_STRIPS 0
  66
  67 #define HAVE_ELTS        0
  68
  69 static GLuint hw_prim[GL_POLYGON+1] = {
  70    0,
  71    PRIM3D_LINELIST,
  72    PRIM3D_LINESTRIP,
  73    PRIM3D_LINESTRIP,
  74    PRIM3D_TRILIST,
  75    PRIM3D_TRISTRIP,
  76    PRIM3D_TRIFAN,
  77    0,
  78    0,
  79    PRIM3D_POLY
  80 };
  81
  82 static const GLenum reduced_prim[GL_POLYGON+1] = {
  83    GL_POINTS,
  84    GL_LINES,
  85    GL_LINES,
  86    GL_LINES,
  87    GL_TRIANGLES,
  88    GL_TRIANGLES,
  89    GL_TRIANGLES,
  90    GL_TRIANGLES,
  91    GL_TRIANGLES,
  92    GL_TRIANGLES
  93 };
  94
  95 static const int scale_prim[GL_POLYGON+1] = {
  96    0,                           /* fallback case */
  97    1,
  98    2,
  99    2,
 100    1,
 101    3,
 102    3,
 103    0,                           /* fallback case */
 104    0,                           /* fallback case */
 105    3
 106 };
 107
 108
 109 static void intelDmaPrimitive( intelContextPtr intel, GLenum prim )
 110 {
 111    if (0) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
 112    INTEL_FIREVERTICES(intel);
 113    intel->vtbl.reduced_primitive_state( intel, reduced_prim[prim] );
 114    intelStartInlinePrimitive( intel, hw_prim[prim] );
 115 }
 116
 117
 118 #define LOCAL_VARS intelContextPtr intel = INTEL_CONTEXT(ctx)
 119 #define INIT( prim )                            \
 120 do {                                            \
 121    intelDmaPrimitive( intel, prim );            \
 122 } while (0)
 123 #define FLUSH()  INTEL_FIREVERTICES( intel )
 124
 125 #define GET_SUBSEQUENT_VB_MAX_VERTS() \
 126   (((intel->alloc.size / 2) - 1500) / (intel->vertex_size*4))
 127 #define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS()
 128
 129 #define ALLOC_VERTS( nr ) \
 130    intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size )
 131
 132 #define EMIT_VERTS( ctx, j, nr, buf ) \
 133   _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf )
 134
 135 #define TAG(x) intel_##x
 136 #include "tnl_dd/t_dd_dmatmp.h"
 137
 138
 139 /**********************************************************************/
 140 /*                          Render pipeline stage                     */
 141 /**********************************************************************/
 142
 143 /* Heuristic to choose between the two render paths:
 144  */
 145 static GLboolean choose_render( intelContextPtr intel,
 146                                 struct vertex_buffer *VB )
 147 {
 148    int vertsz = intel->vertex_size;
 149    int cost_render = 0;
 150    int cost_fallback = 0;
 151    int nr_prims = 0;
 152    int nr_rprims = 0;
 153    int nr_rverts = 0;
 154    int rprim = intel->reduced_primitive;
 155    int i = 0;
 156
 157    for (i = 0 ; i < VB->PrimitiveCount ; i++) {
 158       GLuint prim = VB->Primitive[i].mode;
 159       GLuint length = VB->Primitive[i].count;
 160
 161       if (!length)
 162          continue;
 163
 164       nr_prims++;
 165       nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK];
 166
 167       if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) {
 168          nr_rprims++;
 169          rprim = reduced_prim[prim & PRIM_MODE_MASK];
 170       }
 171    }
 172
 173    /* One point for each generated primitive:
 174     */
 175    cost_render = nr_prims;
 176    cost_fallback = nr_rprims;
 177
 178    /* One point for every 1024 dwords (4k) of dma:
 179     */
 180    cost_render += (vertsz * i) / 1024;
 181    cost_fallback += (vertsz * nr_rverts) / 1024;
 182
 183    if (0)
 184       fprintf(stderr, "cost render: %d fallback: %d\n",
 185               cost_render, cost_fallback);
 186
 187    if (cost_render > cost_fallback)
 188       return GL_FALSE;
 189
 190    return GL_TRUE;
 191 }
 192
 193
 194 static GLboolean intel_run_render( GLcontext *ctx,
 195                                  struct tnl_pipeline_stage *stage )
 196 {
 197    intelContextPtr intel = INTEL_CONTEXT(ctx);
 198    TNLcontext *tnl = TNL_CONTEXT(ctx);
 199    struct vertex_buffer *VB = &tnl->vb;
 200    GLuint i;
 201
 202    intel->vtbl.render_prevalidate( intel );
 203
 204    /* Don't handle clipping or indexed vertices.
 205     */
 206    if (intel->RenderIndex != 0 ||
 207        !intel_validate_render( ctx, VB ) ||
 208        !choose_render( intel, VB )) {
 209       return GL_TRUE;
 210    }
 211
 212    tnl->clipspace.new_inputs |= VERT_BIT_POS;
 213
 214    tnl->Driver.Render.Start( ctx );
 215
 216    for (i = 0 ; i < VB->PrimitiveCount ; i++)
 217    {
 218       GLuint prim = VB->Primitive[i].mode;
 219       GLuint start = VB->Primitive[i].start;
 220       GLuint length = VB->Primitive[i].count;
 221
 222       if (!length)
 223          continue;
 224
 225       intel_render_tab_verts[prim & PRIM_MODE_MASK]( ctx, start, start + length,
 226                                                      prim );
 227    }
 228
 229    tnl->Driver.Render.Finish( ctx );
 230
 231    return GL_FALSE;     /* finished the pipe */
 232 }
 233
 234 const struct tnl_pipeline_stage _intel_render_stage =
 235 {
 236    "intel render",
 237    NULL,
 238    NULL,
 239    NULL,
 240    NULL,
 241    intel_run_render     /* run */
 242 };