src/gallium/drivers/svga/svga_state_need_swtnl.c

   1 /**********************************************************
   2  * Copyright 2008-2009 VMware, Inc.  All rights reserved.
   3  *
   4  * Permission is hereby granted, free of charge, to any person
   5  * obtaining a copy of this software and associated documentation
   6  * files (the "Software"), to deal in the Software without
   7  * restriction, including without limitation the rights to use, copy,
   8  * modify, merge, publish, distribute, sublicense, and/or sell copies
   9  * of the Software, and to permit persons to whom the Software is
  10  * furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice shall be
  13  * included in all copies or substantial portions of the Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  22  * SOFTWARE.
  23  *
  24  **********************************************************/
  25
  26 #include "util/u_inlines.h"
  27 #include "pipe/p_state.h"
  28
  29
  30 #include "svga_context.h"
  31 #include "svga_state.h"
  32 #include "svga_debug.h"
  33 #include "svga_hw_reg.h"
  34
  35 /***********************************************************************
  36  */
  37
  38 static INLINE SVGA3dDeclType
  39 svga_translate_vertex_format(enum pipe_format format)
  40 {
  41    switch (format) {
  42    case PIPE_FORMAT_R32_FLOAT:            return SVGA3D_DECLTYPE_FLOAT1;
  43    case PIPE_FORMAT_R32G32_FLOAT:         return SVGA3D_DECLTYPE_FLOAT2;
  44    case PIPE_FORMAT_R32G32B32_FLOAT:      return SVGA3D_DECLTYPE_FLOAT3;
  45    case PIPE_FORMAT_R32G32B32A32_FLOAT:   return SVGA3D_DECLTYPE_FLOAT4;
  46    case PIPE_FORMAT_B8G8R8A8_UNORM:       return SVGA3D_DECLTYPE_D3DCOLOR;
  47    case PIPE_FORMAT_R8G8B8A8_USCALED:     return SVGA3D_DECLTYPE_UBYTE4;
  48    case PIPE_FORMAT_R16G16_SSCALED:       return SVGA3D_DECLTYPE_SHORT2;
  49    case PIPE_FORMAT_R16G16B16A16_SSCALED: return SVGA3D_DECLTYPE_SHORT4;
  50    case PIPE_FORMAT_R8G8B8A8_UNORM:       return SVGA3D_DECLTYPE_UBYTE4N;
  51    case PIPE_FORMAT_R16G16_SNORM:         return SVGA3D_DECLTYPE_SHORT2N;
  52    case PIPE_FORMAT_R16G16B16A16_SNORM:   return SVGA3D_DECLTYPE_SHORT4N;
  53    case PIPE_FORMAT_R16G16_UNORM:         return SVGA3D_DECLTYPE_USHORT2N;
  54    case PIPE_FORMAT_R16G16B16A16_UNORM:   return SVGA3D_DECLTYPE_USHORT4N;
  55    case PIPE_FORMAT_R10G10B10X2_USCALED:  return SVGA3D_DECLTYPE_UDEC3;
  56    case PIPE_FORMAT_R10G10B10X2_SNORM:    return SVGA3D_DECLTYPE_DEC3N;
  57    case PIPE_FORMAT_R16G16_FLOAT:         return SVGA3D_DECLTYPE_FLOAT16_2;
  58    case PIPE_FORMAT_R16G16B16A16_FLOAT:   return SVGA3D_DECLTYPE_FLOAT16_4;
  59
  60    default:
  61       /* There are many formats without hardware support.  This case
  62        * will be hit regularly, meaning we'll need swvfetch.
  63        */
  64       return SVGA3D_DECLTYPE_MAX;
  65    }
  66 }
  67
  68
  69 static int update_need_swvfetch( struct svga_context *svga,
  70                                  unsigned dirty )
  71 {
  72    unsigned i;
  73    boolean need_swvfetch = FALSE;
  74
  75    if (!svga->curr.velems) {
  76       /* No vertex elements bound. */
  77       return 0;
  78    }
  79
  80    for (i = 0; i < svga->curr.velems->count; i++) {
  81       svga->state.sw.ve_format[i] = svga_translate_vertex_format(svga->curr.velems->velem[i].src_format);
  82       if (svga->state.sw.ve_format[i] == SVGA3D_DECLTYPE_MAX) {
  83          need_swvfetch = TRUE;
  84          break;
  85       }
  86    }
  87
  88    if (need_swvfetch != svga->state.sw.need_swvfetch) {
  89       svga->state.sw.need_swvfetch = need_swvfetch;
  90       svga->dirty |= SVGA_NEW_NEED_SWVFETCH;
  91    }
  92
  93    return 0;
  94 }
  95
  96 struct svga_tracked_state svga_update_need_swvfetch =
  97 {
  98    "update need_swvfetch",
  99    ( SVGA_NEW_VELEMENT ),
 100    update_need_swvfetch
 101 };
 102
 103
 104 /***********************************************************************
 105  */
 106
 107 static int update_need_pipeline( struct svga_context *svga,
 108                                  unsigned dirty )
 109 {
 110
 111    boolean need_pipeline = FALSE;
 112    struct svga_vertex_shader *vs = svga->curr.vs;
 113
 114    /* SVGA_NEW_RAST, SVGA_NEW_REDUCED_PRIMITIVE
 115     */
 116    if (svga->curr.rast->need_pipeline & (1 << svga->curr.reduced_prim)) {
 117       SVGA_DBG(DEBUG_SWTNL, "%s: rast need_pipeline (0x%x) & prim (0x%x)\n",
 118                  __FUNCTION__,
 119                  svga->curr.rast->need_pipeline,
 120                  (1 << svga->curr.reduced_prim) );
 121       SVGA_DBG(DEBUG_SWTNL, "%s: rast need_pipeline tris (%s), lines (%s), points (%s)\n",
 122                  __FUNCTION__,
 123                  svga->curr.rast->need_pipeline_tris_str,
 124                  svga->curr.rast->need_pipeline_lines_str,
 125                  svga->curr.rast->need_pipeline_points_str);
 126       need_pipeline = TRUE;
 127    }
 128
 129    /* EDGEFLAGS
 130     */
 131     if (vs->base.info.writes_edgeflag) {
 132       SVGA_DBG(DEBUG_SWTNL, "%s: edgeflags\n", __FUNCTION__);
 133       need_pipeline = TRUE;
 134    }
 135
 136    /* SVGA_NEW_CLIP
 137     */
 138    if (svga->curr.clip.nr) {
 139       SVGA_DBG(DEBUG_SWTNL, "%s: userclip\n", __FUNCTION__);
 140       need_pipeline = TRUE;
 141    }
 142
 143    if (need_pipeline != svga->state.sw.need_pipeline) {
 144       svga->state.sw.need_pipeline = need_pipeline;
 145       svga->dirty |= SVGA_NEW_NEED_PIPELINE;
 146    }
 147
 148    return 0;
 149 }
 150
 151
 152 struct svga_tracked_state svga_update_need_pipeline =
 153 {
 154    "need pipeline",
 155    (SVGA_NEW_RAST |
 156     SVGA_NEW_CLIP |
 157     SVGA_NEW_VS |
 158     SVGA_NEW_REDUCED_PRIMITIVE),
 159    update_need_pipeline
 160 };
 161
 162
 163 /***********************************************************************
 164  */
 165
 166 static int update_need_swtnl( struct svga_context *svga,
 167                               unsigned dirty )
 168 {
 169    boolean need_swtnl;
 170
 171    if (svga->debug.no_swtnl) {
 172       svga->state.sw.need_swvfetch = 0;
 173       svga->state.sw.need_pipeline = 0;
 174    }
 175
 176    need_swtnl = (svga->state.sw.need_swvfetch ||
 177                  svga->state.sw.need_pipeline);
 178
 179    if (svga->debug.force_swtnl) {
 180       need_swtnl = 1;
 181    }
 182
 183    /*
 184     * Some state changes the draw module does makes us belive we
 185     * we don't need swtnl. This causes the vdecl code to pickup
 186     * the wrong buffers and vertex formats. Try trivial/line-wide.
 187     */
 188    if (svga->state.sw.in_swtnl_draw)
 189       need_swtnl = 1;
 190
 191    if (need_swtnl != svga->state.sw.need_swtnl) {
 192       SVGA_DBG(DEBUG_SWTNL|DEBUG_PERF,
 193                "%s: need_swvfetch %s, need_pipeline %s\n",
 194                __FUNCTION__,
 195                svga->state.sw.need_swvfetch ? "true" : "false",
 196                svga->state.sw.need_pipeline ? "true" : "false");
 197
 198       svga->state.sw.need_swtnl = need_swtnl;
 199       svga->dirty |= SVGA_NEW_NEED_SWTNL;
 200       svga->swtnl.new_vdecl = TRUE;
 201    }
 202
 203    return 0;
 204 }
 205
 206
 207 struct svga_tracked_state svga_update_need_swtnl =
 208 {
 209    "need swtnl",
 210    (SVGA_NEW_NEED_PIPELINE |
 211     SVGA_NEW_NEED_SWVFETCH),
 212    update_need_swtnl
 213 };