src/mesa/drivers/dri/i965/brw_context.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 #include "main/imports.h"
  34 #include "main/api_noop.h"
  35 #include "main/macros.h"
  36 #include "main/simple_list.h"
  37 #include "brw_context.h"
  38 #include "brw_defines.h"
  39 #include "brw_draw.h"
  40 #include "brw_state.h"
  41 #include "intel_span.h"
  42 #include "tnl/t_pipeline.h"
  43 #include "glsl/ralloc.h"
  44
  45 /***************************************
  46  * Mesa's Driver Functions
  47  ***************************************/
  48
  49 static void brwInitDriverFunctions( struct dd_function_table *functions )
  50 {
  51    intelInitDriverFunctions( functions );
  52
  53    brwInitFragProgFuncs( functions );
  54    brw_init_queryobj_functions(functions);
  55 }
  56
  57 GLboolean brwCreateContext( int api,
  58                             const struct gl_config *mesaVis,
  59                             __DRIcontext *driContextPriv,
  60                             void *sharedContextPrivate)
  61 {
  62    struct dd_function_table functions;
  63    struct brw_context *brw = rzalloc(NULL, struct brw_context);
  64    struct intel_context *intel = &brw->intel;
  65    struct gl_context *ctx = &intel->ctx;
  66    unsigned i;
  67
  68    if (!brw) {
  69       printf("%s: failed to alloc context\n", __FUNCTION__);
  70       return GL_FALSE;
  71    }
  72
  73    brwInitVtbl( brw );
  74    brwInitDriverFunctions( &functions );
  75
  76    if (!intelInitContext( intel, api, mesaVis, driContextPriv,
  77                           sharedContextPrivate, &functions )) {
  78       printf("%s: failed to init intel context\n", __FUNCTION__);
  79       FREE(brw);
  80       return GL_FALSE;
  81    }
  82
  83    /* Initialize swrast, tnl driver tables: */
  84    intelInitSpanFuncs(ctx);
  85
  86    TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
  87
  88    ctx->Const.MaxDrawBuffers = BRW_MAX_DRAW_BUFFERS;
  89    ctx->Const.MaxTextureImageUnits = BRW_MAX_TEX_UNIT;
  90    ctx->Const.MaxTextureCoordUnits = 8; /* Mesa limit */
  91    ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits,
  92                                      ctx->Const.MaxTextureImageUnits);
  93    ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */
  94    ctx->Const.MaxCombinedTextureImageUnits =
  95       ctx->Const.MaxVertexTextureImageUnits +
  96       ctx->Const.MaxTextureImageUnits;
  97
  98    ctx->Const.MaxTextureLevels = 14; /* 8192 */
  99    if (ctx->Const.MaxTextureLevels > MAX_TEXTURE_LEVELS)
 100            ctx->Const.MaxTextureLevels = MAX_TEXTURE_LEVELS;
 101    ctx->Const.Max3DTextureLevels = 9;
 102    ctx->Const.MaxCubeTextureLevels = 12;
 103    /* minimum maximum.  Users are likely to run into memory problems
 104     * even at this size, since 64 * 2048 * 2048 * 4 = 1GB and we can't
 105     * address that much.
 106     */
 107    ctx->Const.MaxArrayTextureLayers = 64;
 108    ctx->Const.MaxTextureRectSize = (1<<12);
 109
 110    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 111
 112    /* if conformance mode is set, swrast can handle any size AA point */
 113    ctx->Const.MaxPointSizeAA = 255.0;
 114
 115    /* We want the GLSL compiler to emit code that uses condition codes */
 116    for (i = 0; i <= MESA_SHADER_FRAGMENT; i++) {
 117       ctx->ShaderCompilerOptions[i].EmitCondCodes = GL_TRUE;
 118       ctx->ShaderCompilerOptions[i].EmitNVTempInitialization = GL_TRUE;
 119       ctx->ShaderCompilerOptions[i].EmitNoNoise = GL_TRUE;
 120       ctx->ShaderCompilerOptions[i].EmitNoMainReturn = GL_TRUE;
 121       ctx->ShaderCompilerOptions[i].EmitNoIndirectInput = GL_TRUE;
 122       ctx->ShaderCompilerOptions[i].EmitNoIndirectOutput = GL_TRUE;
 123
 124       ctx->ShaderCompilerOptions[i].EmitNoIndirectUniform =
 125          (i == MESA_SHADER_FRAGMENT);
 126       ctx->ShaderCompilerOptions[i].EmitNoIndirectTemp =
 127          (i == MESA_SHADER_FRAGMENT);
 128       ctx->ShaderCompilerOptions[i].LowerClipDistance = GL_TRUE;
 129    }
 130
 131    ctx->Const.VertexProgram.MaxNativeInstructions = (16 * 1024);
 132    ctx->Const.VertexProgram.MaxAluInstructions = 0;
 133    ctx->Const.VertexProgram.MaxTexInstructions = 0;
 134    ctx->Const.VertexProgram.MaxTexIndirections = 0;
 135    ctx->Const.VertexProgram.MaxNativeAluInstructions = 0;
 136    ctx->Const.VertexProgram.MaxNativeTexInstructions = 0;
 137    ctx->Const.VertexProgram.MaxNativeTexIndirections = 0;
 138    ctx->Const.VertexProgram.MaxNativeAttribs = 16;
 139    ctx->Const.VertexProgram.MaxNativeTemps = 256;
 140    ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
 141    ctx->Const.VertexProgram.MaxNativeParameters = 1024;
 142    ctx->Const.VertexProgram.MaxEnvParams =
 143       MIN2(ctx->Const.VertexProgram.MaxNativeParameters,
 144            ctx->Const.VertexProgram.MaxEnvParams);
 145
 146    ctx->Const.FragmentProgram.MaxNativeInstructions = (16 * 1024);
 147    ctx->Const.FragmentProgram.MaxNativeAluInstructions = (16 * 1024);
 148    ctx->Const.FragmentProgram.MaxNativeTexInstructions = (16 * 1024);
 149    ctx->Const.FragmentProgram.MaxNativeTexIndirections = (16 * 1024);
 150    ctx->Const.FragmentProgram.MaxNativeAttribs = 12;
 151    ctx->Const.FragmentProgram.MaxNativeTemps = 256;
 152    ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
 153    ctx->Const.FragmentProgram.MaxNativeParameters = 1024;
 154    ctx->Const.FragmentProgram.MaxEnvParams =
 155       MIN2(ctx->Const.FragmentProgram.MaxNativeParameters,
 156            ctx->Const.FragmentProgram.MaxEnvParams);
 157
 158    /* Fragment shaders use real, 32-bit twos-complement integers for all
 159     * integer types.
 160     */
 161    ctx->Const.FragmentProgram.LowInt.RangeMin = 31;
 162    ctx->Const.FragmentProgram.LowInt.RangeMax = 30;
 163    ctx->Const.FragmentProgram.LowInt.Precision = 0;
 164    ctx->Const.FragmentProgram.HighInt = ctx->Const.FragmentProgram.MediumInt
 165       = ctx->Const.FragmentProgram.LowInt;
 166
 167    /* Gen6 converts quads to polygon in beginning of 3D pipeline,
 168       but we're not sure how it's actually done for vertex order,
 169       that affect provoking vertex decision. Always use last vertex
 170       convention for quad primitive which works as expected for now. */
 171    if (intel->gen >= 6)
 172        ctx->Const.QuadsFollowProvokingVertexConvention = GL_FALSE;
 173
 174    if (intel->is_g4x || intel->gen >= 5) {
 175       brw->CMD_VF_STATISTICS = GM45_3DSTATE_VF_STATISTICS;
 176       brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_GM45;
 177       brw->has_surface_tile_offset = GL_TRUE;
 178       if (intel->gen < 6)
 179           brw->has_compr4 = GL_TRUE;
 180       brw->has_aa_line_parameters = GL_TRUE;
 181       brw->has_pln = GL_TRUE;
 182   } else {
 183       brw->CMD_VF_STATISTICS = GEN4_3DSTATE_VF_STATISTICS;
 184       brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_965;
 185    }
 186
 187    /* WM maximum threads is number of EUs times number of threads per EU. */
 188    if (intel->gen >= 7) {
 189       if (intel->gt == 1) {
 190          brw->wm_max_threads = 86;
 191          brw->vs_max_threads = 36;
 192          brw->urb.size = 128;
 193          brw->urb.max_vs_entries = 512;
 194          brw->urb.max_gs_entries = 192;
 195       } else if (intel->gt == 2) {
 196          brw->wm_max_threads = 86;
 197          brw->vs_max_threads = 128;
 198          brw->urb.size = 256;
 199          brw->urb.max_vs_entries = 704;
 200          brw->urb.max_gs_entries = 320;
 201       } else {
 202          assert(!"Unknown gen7 device.");
 203       }
 204    } else if (intel->gen == 6) {
 205       if (intel->gt == 2) {
 206          /* This could possibly be 80, but is supposed to require
 207           * disabling of WIZ hashing (bit 6 of GT_MODE, 0x20d0) and a
 208           * GPU reset to change.
 209           */
 210          brw->wm_max_threads = 40;
 211          brw->vs_max_threads = 60;
 212          brw->urb.size = 64;            /* volume 5c.5 section 5.1 */
 213          brw->urb.max_vs_entries = 256; /* volume 2a (see 3DSTATE_URB) */
 214       } else {
 215          brw->wm_max_threads = 40;
 216          brw->vs_max_threads = 24;
 217          brw->urb.size = 32;            /* volume 5c.5 section 5.1 */
 218          brw->urb.max_vs_entries = 128; /* volume 2a (see 3DSTATE_URB) */
 219       }
 220    } else if (intel->gen == 5) {
 221       brw->urb.size = 1024;
 222       brw->vs_max_threads = 72;
 223       brw->wm_max_threads = 12 * 6;
 224    } else if (intel->is_g4x) {
 225       brw->urb.size = 384;
 226       brw->vs_max_threads = 32;
 227       brw->wm_max_threads = 10 * 5;
 228    } else if (intel->gen < 6) {
 229       brw->urb.size = 256;
 230       brw->vs_max_threads = 16;
 231       brw->wm_max_threads = 8 * 4;
 232       brw->has_negative_rhw_bug = GL_TRUE;
 233    }
 234
 235    if (INTEL_DEBUG & DEBUG_SINGLE_THREAD) {
 236       brw->vs_max_threads = 1;
 237       brw->wm_max_threads = 1;
 238    }
 239
 240    brw_init_state( brw );
 241
 242    brw->curbe.last_buf = calloc(1, 4096);
 243    brw->curbe.next_buf = calloc(1, 4096);
 244
 245    brw->state.dirty.mesa = ~0;
 246    brw->state.dirty.brw = ~0;
 247
 248    brw->emit_state_always = 0;
 249
 250    intel->batch.need_workaround_flush = true;
 251
 252    ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
 253    ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 254
 255    brw_draw_init( brw );
 256
 257    brw->new_vs_backend = (getenv("INTEL_OLD_VS") == NULL);
 258
 259    /* If we're using the new shader backend, we require integer uniforms
 260     * stored as actual integers.
 261     */
 262    if (brw->new_vs_backend) {
 263       ctx->Const.NativeIntegers = true;
 264       ctx->Const.UniformBooleanTrue = 1;
 265    }
 266
 267    return GL_TRUE;
 268 }
 269