src/mesa/drivers/dri/i965/brw_gs.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 #include "main/glheader.h"
  33 #include "main/macros.h"
  34 #include "main/enums.h"
  35
  36 #include "intel_batchbuffer.h"
  37
  38 #include "brw_defines.h"
  39 #include "brw_context.h"
  40 #include "brw_eu.h"
  41 #include "brw_util.h"
  42 #include "brw_state.h"
  43 #include "brw_gs.h"
  44
  45 #include "glsl/ralloc.h"
  46
  47 static void compile_gs_prog( struct brw_context *brw,
  48                              struct brw_gs_prog_key *key )
  49 {
  50    struct intel_context *intel = &brw->intel;
  51    struct brw_gs_compile c;
  52    const GLuint *program;
  53    void *mem_ctx;
  54    GLuint program_size;
  55
  56    /* Gen6: VF has already converted into polygon, and LINELOOP is
  57     * converted to LINESTRIP at the beginning of the 3D pipeline.
  58     */
  59    if (intel->gen >= 6)
  60       return;
  61
  62    memset(&c, 0, sizeof(c));
  63
  64    c.key = *key;
  65    /* The geometry shader needs to access the entire VUE. */
  66    struct brw_vue_map vue_map;
  67    brw_compute_vue_map(&vue_map, intel, c.key.nr_userclip, c.key.attrs);
  68    c.nr_regs = (vue_map.num_slots + 1)/2;
  69
  70    mem_ctx = NULL;
  71
  72    /* Begin the compilation:
  73     */
  74    brw_init_compile(brw, &c.func, mem_ctx);
  75
  76    c.func.single_program_flow = 1;
  77
  78    /* For some reason the thread is spawned with only 4 channels
  79     * unmasked.
  80     */
  81    brw_set_mask_control(&c.func, BRW_MASK_DISABLE);
  82
  83
  84    /* Note that primitives which don't require a GS program have
  85     * already been weeded out by this stage:
  86     */
  87
  88    switch (key->primitive) {
  89    case GL_QUADS:
  90       brw_gs_quads( &c, key );
  91       break;
  92    case GL_QUAD_STRIP:
  93       brw_gs_quad_strip( &c, key );
  94       break;
  95    case GL_LINE_LOOP:
  96       brw_gs_lines( &c );
  97       break;
  98    default:
  99       ralloc_free(mem_ctx);
 100       return;
 101    }
 102
 103    /* get the program
 104     */
 105    program = brw_get_program(&c.func, &program_size);
 106
 107    if (unlikely(INTEL_DEBUG & DEBUG_GS)) {
 108       int i;
 109
 110       printf("gs:\n");
 111       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
 112          brw_disasm(stdout, &((struct brw_instruction *)program)[i],
 113                     intel->gen);
 114       printf("\n");
 115     }
 116
 117    brw_upload_cache(&brw->cache, BRW_GS_PROG,
 118                     &c.key, sizeof(c.key),
 119                     program, program_size,
 120                     &c.prog_data, sizeof(c.prog_data),
 121                     &brw->gs.prog_offset, &brw->gs.prog_data);
 122    ralloc_free(mem_ctx);
 123 }
 124
 125 static const GLenum gs_prim[GL_POLYGON+1] = {
 126    GL_POINTS,
 127    GL_LINES,
 128    GL_LINE_LOOP,
 129    GL_LINES,
 130    GL_TRIANGLES,
 131    GL_TRIANGLES,
 132    GL_TRIANGLES,
 133    GL_QUADS,
 134    GL_QUAD_STRIP,
 135    GL_TRIANGLES
 136 };
 137
 138 static void populate_key( struct brw_context *brw,
 139                           struct brw_gs_prog_key *key )
 140 {
 141    struct gl_context *ctx = &brw->intel.ctx;
 142    struct intel_context *intel = &brw->intel;
 143
 144    memset(key, 0, sizeof(*key));
 145
 146    /* CACHE_NEW_VS_PROG */
 147    key->attrs = brw->vs.prog_data->outputs_written;
 148
 149    /* BRW_NEW_PRIMITIVE */
 150    key->primitive = gs_prim[brw->primitive];
 151
 152    /* _NEW_LIGHT */
 153    key->pv_first = (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION);
 154    if (key->primitive == GL_QUADS && ctx->Light.ShadeModel != GL_FLAT) {
 155       /* Provide consistent primitive order with brw_set_prim's
 156        * optimization of single quads to trifans.
 157        */
 158       key->pv_first = GL_TRUE;
 159    }
 160
 161    /* _NEW_TRANSFORM */
 162    key->nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
 163
 164    key->need_gs_prog = (intel->gen >= 6)
 165       ? 0
 166       : (brw->primitive == GL_QUADS ||
 167          brw->primitive == GL_QUAD_STRIP ||
 168          brw->primitive == GL_LINE_LOOP);
 169 }
 170
 171 /* Calculate interpolants for triangle and line rasterization.
 172  */
 173 static void prepare_gs_prog(struct brw_context *brw)
 174 {
 175    struct brw_gs_prog_key key;
 176    /* Populate the key:
 177     */
 178    populate_key(brw, &key);
 179
 180    if (brw->gs.prog_active != key.need_gs_prog) {
 181       brw->state.dirty.cache |= CACHE_NEW_GS_PROG;
 182       brw->gs.prog_active = key.need_gs_prog;
 183    }
 184
 185    if (brw->gs.prog_active) {
 186       if (!brw_search_cache(&brw->cache, BRW_GS_PROG,
 187                             &key, sizeof(key),
 188                             &brw->gs.prog_offset, &brw->gs.prog_data)) {
 189          compile_gs_prog( brw, &key );
 190       }
 191    }
 192 }
 193
 194
 195 const struct brw_tracked_state brw_gs_prog = {
 196    .dirty = {
 197       .mesa  = (_NEW_LIGHT |
 198                 _NEW_TRANSFORM),
 199       .brw   = BRW_NEW_PRIMITIVE,
 200       .cache = CACHE_NEW_VS_PROG
 201    },
 202    .prepare = prepare_gs_prog
 203 };