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