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 "main/glheader.h"
  34 #include "main/context.h"
  35 #include "main/macros.h"
  36 #include "main/imports.h"
  37 #include "main/mtypes.h"
  38 #include "main/enums.h"
  39
  40 #include "math/m_xform.h"
  41
  42 #include "tnl/t_context.h"
  43 #include "tnl/t_vertex.h"
  44 #include "tnl/t_pipeline.h"
  45
  46 #include "intel_screen.h"
  47 #include "intel_context.h"
  48 #include "intel_tris.h"
  49 #include "intel_batchbuffer.h"
  50 #include "intel_reg.h"
  51
  52 /*
  53  * Render unclipped vertex buffers by emitting vertices directly to
  54  * dma buffers.  Use strip/fan hardware primitives where possible.
  55  * Try to simulate missing primitives with indexed vertices.
  56  */
  57 #define HAVE_POINTS      0      /* Has it, but can't use because subpixel has to
  58                                  * be adjusted for points on the INTEL/I845G
  59                                  */
  60 #define HAVE_LINES       1
  61 #define HAVE_LINE_STRIPS 1
  62 #define HAVE_TRIANGLES   1
  63 #define HAVE_TRI_STRIPS  1
  64 #define HAVE_TRI_STRIP_1 0      /* has it, template can't use it yet */
  65 #define HAVE_TRI_FANS    1
  66 #define HAVE_POLYGONS    1
  67 #define HAVE_QUADS       0
  68 #define HAVE_QUAD_STRIPS 0
  69
  70 #define HAVE_ELTS        0
  71
  72 static uint32_t hw_prim[GL_POLYGON + 1] = {
  73    0,
  74    PRIM3D_LINELIST,
  75    PRIM3D_LINESTRIP,
  76    PRIM3D_LINESTRIP,
  77    PRIM3D_TRILIST,
  78    PRIM3D_TRISTRIP,
  79    PRIM3D_TRIFAN,
  80    0,
  81    0,
  82    PRIM3D_POLY
  83 };
  84
  85 static const GLenum reduced_prim[GL_POLYGON + 1] = {
  86    GL_POINTS,
  87    GL_LINES,
  88    GL_LINES,
  89    GL_LINES,
  90    GL_TRIANGLES,
  91    GL_TRIANGLES,
  92    GL_TRIANGLES,
  93    GL_TRIANGLES,
  94    GL_TRIANGLES,
  95    GL_TRIANGLES
  96 };
  97
  98 static const int scale_prim[GL_POLYGON + 1] = {
  99    0,                           /* fallback case */
 100    1,
 101    2,
 102    2,
 103    1,
 104    3,
 105    3,
 106    0,                           /* fallback case */
 107    0,                           /* fallback case */
 108    3
 109 };
 110
 111
 112 static void
 113 intelDmaPrimitive(struct intel_context *intel, GLenum prim)
 114 {
 115    if (0)
 116       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
 117    INTEL_FIREVERTICES(intel);
 118    intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]);
 119    intel_set_prim(intel, hw_prim[prim]);
 120 }
 121
 122 static INLINE GLuint intel_get_vb_max(struct intel_context *intel)
 123 {
 124    GLuint ret;
 125
 126    if (intel->intelScreen->no_vbo)
 127       ret = intel->batch->size - 1500;
 128    else
 129       ret = INTEL_VB_SIZE;
 130    ret /= (intel->vertex_size * 4);
 131    return ret;
 132 }
 133
 134 static INLINE GLuint intel_get_current_max(struct intel_context *intel)
 135 {
 136
 137    if (intel->intelScreen->no_vbo)
 138       return intel_get_vb_max(intel);
 139    else
 140       return (INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4);
 141 }
 142
 143 #define LOCAL_VARS struct intel_context *intel = intel_context(ctx)
 144 #define INIT( prim )                            \
 145 do {                                            \
 146    intelDmaPrimitive( intel, prim );            \
 147 } while (0)
 148
 149 #define FLUSH() INTEL_FIREVERTICES(intel)
 150
 151 #define GET_SUBSEQUENT_VB_MAX_VERTS() intel_get_vb_max(intel)
 152 #define GET_CURRENT_VB_MAX_VERTS() intel_get_current_max(intel)
 153
 154 #define ALLOC_VERTS(nr) intel_get_prim_space(intel, nr)
 155
 156 #define EMIT_VERTS( ctx, j, nr, buf ) \
 157   _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf )
 158
 159 #define TAG(x) intel_##x
 160 #include "tnl_dd/t_dd_dmatmp.h"
 161
 162
 163 /**********************************************************************/
 164 /*                          Render pipeline stage                     */
 165 /**********************************************************************/
 166
 167 /* Heuristic to choose between the two render paths:
 168  */
 169 static GLboolean
 170 choose_render(struct intel_context *intel, struct vertex_buffer *VB)
 171 {
 172    int vertsz = intel->vertex_size;
 173    int cost_render = 0;
 174    int cost_fallback = 0;
 175    int nr_prims = 0;
 176    int nr_rprims = 0;
 177    int nr_rverts = 0;
 178    int rprim = intel->reduced_primitive;
 179    int i = 0;
 180
 181    for (i = 0; i < VB->PrimitiveCount; i++) {
 182       GLuint prim = VB->Primitive[i].mode;
 183       GLuint length = VB->Primitive[i].count;
 184
 185       if (!length)
 186          continue;
 187
 188       nr_prims++;
 189       nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK];
 190
 191       if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) {
 192          nr_rprims++;
 193          rprim = reduced_prim[prim & PRIM_MODE_MASK];
 194       }
 195    }
 196
 197    /* One point for each generated primitive:
 198     */
 199    cost_render = nr_prims;
 200    cost_fallback = nr_rprims;
 201
 202    /* One point for every 1024 dwords (4k) of dma:
 203     */
 204    cost_render += (vertsz * i) / 1024;
 205    cost_fallback += (vertsz * nr_rverts) / 1024;
 206
 207    if (0)
 208       fprintf(stderr, "cost render: %d fallback: %d\n",
 209               cost_render, cost_fallback);
 210
 211    if (cost_render > cost_fallback)
 212       return GL_FALSE;
 213
 214    return GL_TRUE;
 215 }
 216
 217
 218 static GLboolean
 219 intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage)
 220 {
 221    struct intel_context *intel = intel_context(ctx);
 222    TNLcontext *tnl = TNL_CONTEXT(ctx);
 223    struct vertex_buffer *VB = &tnl->vb;
 224    GLuint i;
 225
 226    intel->vtbl.render_prevalidate( intel );
 227
 228    /* Don't handle clipping or indexed vertices.
 229     */
 230    if (intel->RenderIndex != 0 ||
 231        !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) {
 232       return GL_TRUE;
 233    }
 234
 235    tnl->clipspace.new_inputs |= VERT_BIT_POS;
 236
 237    tnl->Driver.Render.Start(ctx);
 238
 239    for (i = 0; i < VB->PrimitiveCount; i++) {
 240       GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 241       GLuint start = VB->Primitive[i].start;
 242       GLuint length = VB->Primitive[i].count;
 243
 244       if (!length)
 245          continue;
 246
 247       intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start,
 248                                                      start + length, prim);
 249    }
 250
 251    tnl->Driver.Render.Finish(ctx);
 252
 253    INTEL_FIREVERTICES(intel);
 254
 255    return GL_FALSE;             /* finished the pipe */
 256 }
 257
 258 static const struct tnl_pipeline_stage _intel_render_stage = {
 259    "intel render",
 260    NULL,
 261    NULL,
 262    NULL,
 263    NULL,
 264    intel_run_render             /* run */
 265 };
 266
 267 const struct tnl_pipeline_stage *intel_pipeline[] = {
 268    &_tnl_vertex_transform_stage,
 269    &_tnl_normal_transform_stage,
 270    &_tnl_lighting_stage,
 271    &_tnl_fog_coordinate_stage,
 272    &_tnl_texgen_stage,
 273    &_tnl_texture_transform_stage,
 274    &_tnl_point_attenuation_stage,
 275    &_tnl_vertex_program_stage,
 276 #if 1
 277    &_intel_render_stage,        /* ADD: unclipped rastersetup-to-dma */
 278 #endif
 279    &_tnl_render_stage,
 280    0,
 281 };