src/mesa/drivers/dri/i965/brw_vs_state.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
  34 #include "brw_context.h"
  35 #include "brw_state.h"
  36 #include "brw_defines.h"
  37 #include "main/macros.h"
  38
  39 struct brw_vs_unit_key {
  40    unsigned int total_grf;
  41    unsigned int urb_entry_read_length;
  42    unsigned int curb_entry_read_length;
  43
  44    unsigned int curbe_offset;
  45
  46    unsigned int nr_urb_entries, urb_size;
  47 };
  48
  49 static void
  50 vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key)
  51 {
  52    GLcontext *ctx = &brw->intel.ctx;
  53
  54    memset(key, 0, sizeof(*key));
  55
  56    /* CACHE_NEW_VS_PROG */
  57    key->total_grf = brw->vs.prog_data->total_grf;
  58    key->urb_entry_read_length = brw->vs.prog_data->urb_read_length;
  59    key->curb_entry_read_length = brw->vs.prog_data->curb_read_length;
  60
  61    /* BRW_NEW_URB_FENCE */
  62    key->nr_urb_entries = brw->urb.nr_vs_entries;
  63    key->urb_size = brw->urb.vsize;
  64
  65    /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */
  66    if (ctx->Transform.ClipPlanesEnabled) {
  67       /* Note that we read in the userclip planes as well, hence
  68        * clip_start:
  69        */
  70       key->curbe_offset = brw->curbe.clip_start;
  71    }
  72    else {
  73       key->curbe_offset = brw->curbe.vs_start;
  74    }
  75 }
  76
  77 static dri_bo *
  78 vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
  79 {
  80    struct brw_vs_unit_state vs;
  81    dri_bo *bo;
  82    int chipset_max_threads;
  83
  84    memset(&vs, 0, sizeof(vs));
  85
  86    vs.thread0.kernel_start_pointer = brw->vs.prog_bo->offset >> 6; /* reloc */
  87    vs.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
  88    vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
  89    /* Choosing multiple program flow means that we may get 2-vertex threads,
  90     * which will have the channel mask for dwords 4-7 enabled in the thread,
  91     * and those dwords will be written to the second URB handle when we
  92     * brw_urb_WRITE() results.
  93     */
  94    vs.thread1.single_program_flow = 0;
  95    vs.thread3.urb_entry_read_length = key->urb_entry_read_length;
  96    vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
  97    vs.thread3.dispatch_grf_start_reg = 1;
  98    vs.thread3.urb_entry_read_offset = 0;
  99    vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
 100
 101    vs.thread4.nr_urb_entries = key->nr_urb_entries;
 102    vs.thread4.urb_entry_allocation_size = key->urb_size - 1;
 103
 104    if (BRW_IS_G4X(brw))
 105       chipset_max_threads = 32;
 106    else
 107       chipset_max_threads = 16;
 108    vs.thread4.max_threads = CLAMP(key->nr_urb_entries / 2,
 109                                   1, chipset_max_threads) - 1;
 110
 111    if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
 112       vs.thread4.max_threads = 0;
 113
 114    /* No samplers for ARB_vp programs:
 115     */
 116    vs.vs5.sampler_count = 0;
 117
 118    if (INTEL_DEBUG & DEBUG_STATS)
 119       vs.thread4.stats_enable = 1;
 120
 121    /* Vertex program always enabled:
 122     */
 123    vs.vs6.vs_enable = 1;
 124
 125    bo = brw_upload_cache(&brw->cache, BRW_VS_UNIT,
 126                          key, sizeof(*key),
 127                          &brw->vs.prog_bo, 1,
 128                          &vs, sizeof(vs),
 129                          NULL, NULL);
 130
 131    /* Emit VS program relocation */
 132    dri_bo_emit_reloc(bo,
 133                      I915_GEM_DOMAIN_INSTRUCTION, 0,
 134                      vs.thread0.grf_reg_count << 1,
 135                      offsetof(struct brw_vs_unit_state, thread0),
 136                      brw->vs.prog_bo);
 137
 138    return bo;
 139 }
 140
 141 static void prepare_vs_unit(struct brw_context *brw)
 142 {
 143    struct brw_vs_unit_key key;
 144
 145    vs_unit_populate_key(brw, &key);
 146
 147    dri_bo_unreference(brw->vs.state_bo);
 148    brw->vs.state_bo = brw_search_cache(&brw->cache, BRW_VS_UNIT,
 149                                        &key, sizeof(key),
 150                                        &brw->vs.prog_bo, 1,
 151                                        NULL);
 152    if (brw->vs.state_bo == NULL) {
 153       brw->vs.state_bo = vs_unit_create_from_key(brw, &key);
 154    }
 155 }
 156
 157 const struct brw_tracked_state brw_vs_unit = {
 158    .dirty = {
 159       .mesa  = _NEW_TRANSFORM,
 160       .brw   = (BRW_NEW_CURBE_OFFSETS |
 161                 BRW_NEW_URB_FENCE),
 162       .cache = CACHE_NEW_VS_PROG
 163    },
 164    .prepare = prepare_vs_unit,
 165 };