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 "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    memset(key, 0, sizeof(*key));
  53
  54    /* CACHE_NEW_VS_PROG */
  55    key->total_grf = brw->vs.prog_data->total_grf;
  56    key->urb_entry_read_length = brw->vs.prog_data->urb_read_length;
  57    key->curb_entry_read_length = brw->vs.prog_data->curb_read_length;
  58
  59    /* BRW_NEW_URB_FENCE */
  60    key->nr_urb_entries = brw->urb.nr_vs_entries;
  61    key->urb_size = brw->urb.vsize;
  62
  63    /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */
  64    if (brw->attribs.Transform->ClipPlanesEnabled) {
  65       /* Note that we read in the userclip planes as well, hence
  66        * clip_start:
  67        */
  68       key->curbe_offset = brw->curbe.clip_start;
  69    }
  70    else {
  71       key->curbe_offset = brw->curbe.vs_start;
  72    }
  73 }
  74
  75 static dri_bo *
  76 vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
  77 {
  78    struct brw_vs_unit_state vs;
  79    dri_bo *bo;
  80
  81    memset(&vs, 0, sizeof(vs));
  82
  83    vs.thread0.kernel_start_pointer = brw->vs.prog_bo->offset >> 6; /* reloc */
  84    vs.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
  85    vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
  86    vs.thread3.urb_entry_read_length = key->urb_entry_read_length;
  87    vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
  88    vs.thread3.dispatch_grf_start_reg = 1;
  89    vs.thread3.urb_entry_read_offset = 0;
  90    vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
  91
  92    vs.thread4.nr_urb_entries = key->nr_urb_entries;
  93    vs.thread4.urb_entry_allocation_size = key->urb_size - 1;
  94    vs.thread4.max_threads = MIN2(MAX2(0, (key->nr_urb_entries - 6) / 2 - 1),
  95                                  15);
  96
  97    if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
  98       vs.thread4.max_threads = 0;
  99
 100    /* No samplers for ARB_vp programs:
 101     */
 102    vs.vs5.sampler_count = 0;
 103
 104    if (INTEL_DEBUG & DEBUG_STATS)
 105       vs.thread4.stats_enable = 1;
 106
 107    /* Vertex program always enabled:
 108     */
 109    vs.vs6.vs_enable = 1;
 110
 111    bo = brw_upload_cache(&brw->cache, BRW_VS_UNIT,
 112                          key, sizeof(*key),
 113                          &brw->vs.prog_bo, 1,
 114                          &vs, sizeof(vs),
 115                          NULL, NULL);
 116
 117    /* Emit VS program relocation */
 118    intel_bo_emit_reloc(bo,
 119                        I915_GEM_DOMAIN_INSTRUCTION, 0,
 120                        vs.thread0.grf_reg_count << 1,
 121                        offsetof(struct brw_vs_unit_state, thread0),
 122                        brw->vs.prog_bo);
 123
 124    return bo;
 125 }
 126
 127 static void prepare_vs_unit(struct brw_context *brw)
 128 {
 129    struct brw_vs_unit_key key;
 130
 131    vs_unit_populate_key(brw, &key);
 132
 133    dri_bo_unreference(brw->vs.state_bo);
 134    brw->vs.state_bo = brw_search_cache(&brw->cache, BRW_VS_UNIT,
 135                                        &key, sizeof(key),
 136                                        &brw->vs.prog_bo, 1,
 137                                        NULL);
 138    if (brw->vs.state_bo == NULL) {
 139       brw->vs.state_bo = vs_unit_create_from_key(brw, &key);
 140    }
 141 }
 142
 143 const struct brw_tracked_state brw_vs_unit = {
 144    .dirty = {
 145       .mesa  = _NEW_TRANSFORM,
 146       .brw   = (BRW_NEW_CURBE_OFFSETS |
 147                 BRW_NEW_URB_FENCE),
 148       .cache = CACHE_NEW_VS_PROG
 149    },
 150    .prepare = prepare_vs_unit,
 151 };