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/macros.h"
  35 #include "main/simple_list.h"
  36
  37 #include "vbo/vbo_context.h"
  38
  39 #include "brw_context.h"
  40 #include "brw_defines.h"
  41 #include "brw_draw.h"
  42 #include "brw_state.h"
  43
  44 #include "intel_fbo.h"
  45 #include "intel_mipmap_tree.h"
  46 #include "intel_regions.h"
  47 #include "intel_span.h"
  48 #include "intel_tex.h"
  49 #include "intel_tex_obj.h"
  50
  51 #include "tnl/t_pipeline.h"
  52 #include "glsl/ralloc.h"
  53
  54 /***************************************
  55  * Mesa's Driver Functions
  56  ***************************************/
  57
  58 static void brwInitDriverFunctions(struct intel_screen *screen,
  59                                    struct dd_function_table *functions)
  60 {
  61    intelInitDriverFunctions( functions );
  62
  63    brwInitFragProgFuncs( functions );
  64    brw_init_queryobj_functions(functions);
  65
  66    functions->BeginTransformFeedback = brw_begin_transform_feedback;
  67
  68    if (screen->gen >= 7)
  69       functions->EndTransformFeedback = gen7_end_transform_feedback;
  70    else
  71       functions->EndTransformFeedback = brw_end_transform_feedback;
  72 }
  73
  74 bool
  75 brwCreateContext(int api,
  76                  const struct gl_config *mesaVis,
  77                  __DRIcontext *driContextPriv,
  78                  unsigned major_version,
  79                  unsigned minor_version,
  80                  uint32_t flags,
  81                  unsigned *error,
  82                  void *sharedContextPrivate)
  83 {
  84    __DRIscreen *sPriv = driContextPriv->driScreenPriv;
  85    struct intel_screen *screen = sPriv->driverPrivate;
  86    struct dd_function_table functions;
  87    unsigned i;
  88
  89    /* Filter against the requested API and version.
  90     */
  91    switch (api) {
  92    case API_OPENGL: {
  93 #ifdef TEXTURE_FLOAT_ENABLED
  94       const unsigned max_version =
  95          (screen->gen == 6 ||
  96           (screen->gen == 7 && screen->kernel_has_gen7_sol_reset))
  97          ? 30 : 21;
  98 #else
  99       const unsigned max_version = 21;
 100 #endif
 101       const unsigned req_version = major_version * 10 + minor_version;
 102
 103       if (req_version > max_version) {
 104          *error = __DRI_CTX_ERROR_BAD_VERSION;
 105          return false;
 106       }
 107       break;
 108    }
 109    case API_OPENGLES:
 110    case API_OPENGLES2:
 111       break;
 112    default:
 113       *error = __DRI_CTX_ERROR_BAD_API;
 114       return false;
 115    }
 116
 117    struct brw_context *brw = rzalloc(NULL, struct brw_context);
 118    if (!brw) {
 119       printf("%s: failed to alloc context\n", __FUNCTION__);
 120       *error = __DRI_CTX_ERROR_NO_MEMORY;
 121       return false;
 122    }
 123
 124    brwInitDriverFunctions(screen, &functions);
 125
 126    struct intel_context *intel = &brw->intel;
 127    struct gl_context *ctx = &intel->ctx;
 128
 129    if (!intelInitContext( intel, api, mesaVis, driContextPriv,
 130                           sharedContextPrivate, &functions )) {
 131       printf("%s: failed to init intel context\n", __FUNCTION__);
 132       FREE(brw);
 133       *error = __DRI_CTX_ERROR_NO_MEMORY;
 134       return false;
 135    }
 136
 137    brwInitVtbl( brw );
 138
 139    brw_init_surface_formats(brw);
 140
 141    /* Initialize swrast, tnl driver tables: */
 142    intelInitSpanFuncs(ctx);
 143
 144    TNLcontext *tnl = TNL_CONTEXT(ctx);
 145    if (tnl)
 146       tnl->Driver.RunPipeline = _tnl_run_pipeline;
 147
 148    ctx->Const.MaxDualSourceDrawBuffers = 1;
 149    ctx->Const.MaxDrawBuffers = BRW_MAX_DRAW_BUFFERS;
 150    ctx->Const.MaxTextureImageUnits = BRW_MAX_TEX_UNIT;
 151    ctx->Const.MaxTextureCoordUnits = 8; /* Mesa limit */
 152    ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits,
 153                                      ctx->Const.MaxTextureImageUnits);
 154    ctx->Const.MaxVertexTextureImageUnits = BRW_MAX_TEX_UNIT;
 155    ctx->Const.MaxCombinedTextureImageUnits =
 156       ctx->Const.MaxVertexTextureImageUnits +
 157       ctx->Const.MaxTextureImageUnits;
 158
 159    ctx->Const.MaxTextureLevels = 14; /* 8192 */
 160    if (ctx->Const.MaxTextureLevels > MAX_TEXTURE_LEVELS)
 161            ctx->Const.MaxTextureLevels = MAX_TEXTURE_LEVELS;
 162    ctx->Const.Max3DTextureLevels = 9;
 163    ctx->Const.MaxCubeTextureLevels = 12;
 164
 165    if (intel->gen >= 7)
 166       ctx->Const.MaxArrayTextureLayers = 2048;
 167    else
 168       ctx->Const.MaxArrayTextureLayers = 512;
 169
 170    ctx->Const.MaxTextureRectSize = (1<<12);
 171
 172    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 173
 174    /* Hardware only supports a limited number of transform feedback buffers.
 175     * So we need to override the Mesa default (which is based only on software
 176     * limits).
 177     */
 178    ctx->Const.MaxTransformFeedbackBuffers = BRW_MAX_SOL_BUFFERS;
 179
 180    /* On Gen6, in the worst case, we use up one binding table entry per
 181     * transform feedback component (see comments above the definition of
 182     * BRW_MAX_SOL_BINDINGS, in brw_context.h), so we need to advertise a value
 183     * for MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS equal to
 184     * BRW_MAX_SOL_BINDINGS.
 185     *
 186     * In "separate components" mode, we need to divide this value by
 187     * BRW_MAX_SOL_BUFFERS, so that the total number of binding table entries
 188     * used up by all buffers will not exceed BRW_MAX_SOL_BINDINGS.
 189     */
 190    ctx->Const.MaxTransformFeedbackInterleavedComponents = BRW_MAX_SOL_BINDINGS;
 191    ctx->Const.MaxTransformFeedbackSeparateComponents =
 192       BRW_MAX_SOL_BINDINGS / BRW_MAX_SOL_BUFFERS;
 193
 194    if (intel->gen == 6)
 195       ctx->Const.MaxSamples = 4;
 196    else if (intel->gen >= 7)
 197       ctx->Const.MaxSamples = 8;
 198
 199    /* if conformance mode is set, swrast can handle any size AA point */
 200    ctx->Const.MaxPointSizeAA = 255.0;
 201
 202    /* We want the GLSL compiler to emit code that uses condition codes */
 203    for (i = 0; i <= MESA_SHADER_FRAGMENT; i++) {
 204       ctx->ShaderCompilerOptions[i].MaxIfDepth = intel->gen < 6 ? 16 : UINT_MAX;
 205       ctx->ShaderCompilerOptions[i].EmitCondCodes = true;
 206       ctx->ShaderCompilerOptions[i].EmitNVTempInitialization = true;
 207       ctx->ShaderCompilerOptions[i].EmitNoNoise = true;
 208       ctx->ShaderCompilerOptions[i].EmitNoMainReturn = true;
 209       ctx->ShaderCompilerOptions[i].EmitNoIndirectInput = true;
 210       ctx->ShaderCompilerOptions[i].EmitNoIndirectOutput = true;
 211
 212       ctx->ShaderCompilerOptions[i].EmitNoIndirectUniform =
 213          (i == MESA_SHADER_FRAGMENT);
 214       ctx->ShaderCompilerOptions[i].EmitNoIndirectTemp =
 215          (i == MESA_SHADER_FRAGMENT);
 216       ctx->ShaderCompilerOptions[i].LowerClipDistance = true;
 217    }
 218
 219    ctx->Const.VertexProgram.MaxNativeInstructions = (16 * 1024);
 220    ctx->Const.VertexProgram.MaxAluInstructions = 0;
 221    ctx->Const.VertexProgram.MaxTexInstructions = 0;
 222    ctx->Const.VertexProgram.MaxTexIndirections = 0;
 223    ctx->Const.VertexProgram.MaxNativeAluInstructions = 0;
 224    ctx->Const.VertexProgram.MaxNativeTexInstructions = 0;
 225    ctx->Const.VertexProgram.MaxNativeTexIndirections = 0;
 226    ctx->Const.VertexProgram.MaxNativeAttribs = 16;
 227    ctx->Const.VertexProgram.MaxNativeTemps = 256;
 228    ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
 229    ctx->Const.VertexProgram.MaxNativeParameters = 1024;
 230    ctx->Const.VertexProgram.MaxEnvParams =
 231       MIN2(ctx->Const.VertexProgram.MaxNativeParameters,
 232            ctx->Const.VertexProgram.MaxEnvParams);
 233
 234    ctx->Const.FragmentProgram.MaxNativeInstructions = (16 * 1024);
 235    ctx->Const.FragmentProgram.MaxNativeAluInstructions = (16 * 1024);
 236    ctx->Const.FragmentProgram.MaxNativeTexInstructions = (16 * 1024);
 237    ctx->Const.FragmentProgram.MaxNativeTexIndirections = (16 * 1024);
 238    ctx->Const.FragmentProgram.MaxNativeAttribs = 12;
 239    ctx->Const.FragmentProgram.MaxNativeTemps = 256;
 240    ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
 241    ctx->Const.FragmentProgram.MaxNativeParameters = 1024;
 242    ctx->Const.FragmentProgram.MaxEnvParams =
 243       MIN2(ctx->Const.FragmentProgram.MaxNativeParameters,
 244            ctx->Const.FragmentProgram.MaxEnvParams);
 245
 246    /* Fragment shaders use real, 32-bit twos-complement integers for all
 247     * integer types.
 248     */
 249    ctx->Const.FragmentProgram.LowInt.RangeMin = 31;
 250    ctx->Const.FragmentProgram.LowInt.RangeMax = 30;
 251    ctx->Const.FragmentProgram.LowInt.Precision = 0;
 252    ctx->Const.FragmentProgram.HighInt = ctx->Const.FragmentProgram.MediumInt
 253       = ctx->Const.FragmentProgram.LowInt;
 254
 255    /* Gen6 converts quads to polygon in beginning of 3D pipeline,
 256       but we're not sure how it's actually done for vertex order,
 257       that affect provoking vertex decision. Always use last vertex
 258       convention for quad primitive which works as expected for now. */
 259    if (intel->gen >= 6)
 260        ctx->Const.QuadsFollowProvokingVertexConvention = false;
 261
 262    ctx->Const.QueryCounterBits.Timestamp = 36;
 263
 264    if (intel->is_g4x || intel->gen >= 5) {
 265       brw->CMD_VF_STATISTICS = GM45_3DSTATE_VF_STATISTICS;
 266       brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_GM45;
 267       brw->has_surface_tile_offset = true;
 268       if (intel->gen < 6)
 269           brw->has_compr4 = true;
 270       brw->has_aa_line_parameters = true;
 271       brw->has_pln = true;
 272   } else {
 273       brw->CMD_VF_STATISTICS = GEN4_3DSTATE_VF_STATISTICS;
 274       brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_965;
 275    }
 276
 277    /* WM maximum threads is number of EUs times number of threads per EU. */
 278    if (intel->gen >= 7) {
 279       if (intel->gt == 1) {
 280          brw->max_wm_threads = 48;
 281          brw->max_vs_threads = 36;
 282          brw->max_gs_threads = 36;
 283          brw->urb.size = 128;
 284          brw->urb.max_vs_entries = 512;
 285          brw->urb.max_gs_entries = 192;
 286       } else if (intel->gt == 2) {
 287          brw->max_wm_threads = 172;
 288          brw->max_vs_threads = 128;
 289          brw->max_gs_threads = 128;
 290          brw->urb.size = 256;
 291          brw->urb.max_vs_entries = 704;
 292          brw->urb.max_gs_entries = 320;
 293       } else {
 294          assert(!"Unknown gen7 device.");
 295       }
 296    } else if (intel->gen == 6) {
 297       if (intel->gt == 2) {
 298          brw->max_wm_threads = 80;
 299          brw->max_vs_threads = 60;
 300          brw->max_gs_threads = 60;
 301          brw->urb.size = 64;            /* volume 5c.5 section 5.1 */
 302          brw->urb.max_vs_entries = 256; /* volume 2a (see 3DSTATE_URB) */
 303          brw->urb.max_gs_entries = 256;
 304       } else {
 305          brw->max_wm_threads = 40;
 306          brw->max_vs_threads = 24;
 307          brw->max_gs_threads = 21; /* conservative; 24 if rendering disabled */
 308          brw->urb.size = 32;            /* volume 5c.5 section 5.1 */
 309          brw->urb.max_vs_entries = 256; /* volume 2a (see 3DSTATE_URB) */
 310          brw->urb.max_gs_entries = 256;
 311       }
 312       brw->urb.gen6_gs_previously_active = false;
 313    } else if (intel->gen == 5) {
 314       brw->urb.size = 1024;
 315       brw->max_vs_threads = 72;
 316       brw->max_gs_threads = 32;
 317       brw->max_wm_threads = 12 * 6;
 318    } else if (intel->is_g4x) {
 319       brw->urb.size = 384;
 320       brw->max_vs_threads = 32;
 321       brw->max_gs_threads = 2;
 322       brw->max_wm_threads = 10 * 5;
 323    } else if (intel->gen < 6) {
 324       brw->urb.size = 256;
 325       brw->max_vs_threads = 16;
 326       brw->max_gs_threads = 2;
 327       brw->max_wm_threads = 8 * 4;
 328       brw->has_negative_rhw_bug = true;
 329    }
 330
 331    if (intel->gen <= 7) {
 332       brw->needs_unlit_centroid_workaround = true;
 333    }
 334
 335    brw->prim_restart.in_progress = false;
 336    brw->prim_restart.enable_cut_index = false;
 337    intel->hw_ctx = drm_intel_gem_context_create(intel->bufmgr);
 338
 339    brw_init_state( brw );
 340
 341    brw->curbe.last_buf = calloc(1, 4096);
 342    brw->curbe.next_buf = calloc(1, 4096);
 343
 344    brw->state.dirty.mesa = ~0;
 345    brw->state.dirty.brw = ~0;
 346
 347    brw->emit_state_always = 0;
 348
 349    intel->batch.need_workaround_flush = true;
 350
 351    ctx->VertexProgram._MaintainTnlProgram = true;
 352    ctx->FragmentProgram._MaintainTexEnvProgram = true;
 353
 354    brw_draw_init( brw );
 355
 356    brw->precompile = driQueryOptionb(&intel->optionCache, "shader_precompile");
 357
 358    ctx->Const.NativeIntegers = true;
 359    ctx->Const.UniformBooleanTrue = 1;
 360
 361    ctx->Const.ForceGLSLExtensionsWarn = driQueryOptionb(&intel->optionCache, "force_glsl_extensions_warn");
 362
 363    ctx->Const.ContextFlags = 0;
 364    if ((flags & __DRI_CTX_FLAG_FORWARD_COMPATIBLE) != 0)
 365       ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_FORWARD_COMPATIBLE_BIT;
 366
 367    if ((flags & __DRI_CTX_FLAG_DEBUG) != 0)
 368       ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_DEBUG_BIT;
 369
 370    return true;
 371 }
 372