i965: Add new SIMD8 VS prog data flag
[mesa.git] / src / mesa / drivers / dri / i965 / brw_context.h
index a8f5d0ffc7b4b9886291547da3e8f6469c30519c..7a6ec2fda944a65e40048b261c8a8ddc366b04a3 100644 (file)
@@ -149,8 +149,21 @@ struct brw_vec4_prog_key;
 struct brw_wm_prog_key;
 struct brw_wm_prog_data;
 
+enum brw_cache_id {
+   BRW_CACHE_FS_PROG,
+   BRW_CACHE_BLORP_BLIT_PROG,
+   BRW_CACHE_SF_PROG,
+   BRW_CACHE_VS_PROG,
+   BRW_CACHE_FF_GS_PROG,
+   BRW_CACHE_GS_PROG,
+   BRW_CACHE_CLIP_PROG,
+
+   BRW_MAX_CACHE
+};
+
 enum brw_state_id {
-   BRW_STATE_URB_FENCE,
+   /* brw_cache_ids must come first - see brw_state_cache.c */
+   BRW_STATE_URB_FENCE = BRW_MAX_CACHE,
    BRW_STATE_FRAGMENT_PROGRAM,
    BRW_STATE_GEOMETRY_PROGRAM,
    BRW_STATE_VERTEX_PROGRAM,
@@ -182,46 +195,92 @@ enum brw_state_id {
    BRW_STATE_INTERPOLATION_MAP,
    BRW_STATE_PUSH_CONSTANT_ALLOCATION,
    BRW_STATE_NUM_SAMPLES,
+   BRW_STATE_TEXTURE_BUFFER,
+   BRW_STATE_GEN4_UNIT_STATE,
+   BRW_STATE_CC_VP,
+   BRW_STATE_SF_VP,
+   BRW_STATE_CLIP_VP,
+   BRW_STATE_SAMPLER_STATE_TABLE,
+   BRW_STATE_VS_ATTRIB_WORKAROUNDS,
    BRW_NUM_STATE_BITS
 };
 
-#define BRW_NEW_URB_FENCE               (1 << BRW_STATE_URB_FENCE)
-#define BRW_NEW_FRAGMENT_PROGRAM        (1 << BRW_STATE_FRAGMENT_PROGRAM)
-#define BRW_NEW_GEOMETRY_PROGRAM        (1 << BRW_STATE_GEOMETRY_PROGRAM)
-#define BRW_NEW_VERTEX_PROGRAM          (1 << BRW_STATE_VERTEX_PROGRAM)
-#define BRW_NEW_CURBE_OFFSETS           (1 << BRW_STATE_CURBE_OFFSETS)
-#define BRW_NEW_REDUCED_PRIMITIVE       (1 << BRW_STATE_REDUCED_PRIMITIVE)
-#define BRW_NEW_PRIMITIVE               (1 << BRW_STATE_PRIMITIVE)
-#define BRW_NEW_CONTEXT                 (1 << BRW_STATE_CONTEXT)
-#define BRW_NEW_PSP                     (1 << BRW_STATE_PSP)
-#define BRW_NEW_SURFACES               (1 << BRW_STATE_SURFACES)
-#define BRW_NEW_VS_BINDING_TABLE       (1 << BRW_STATE_VS_BINDING_TABLE)
-#define BRW_NEW_GS_BINDING_TABLE       (1 << BRW_STATE_GS_BINDING_TABLE)
-#define BRW_NEW_PS_BINDING_TABLE       (1 << BRW_STATE_PS_BINDING_TABLE)
-#define BRW_NEW_INDICES                        (1 << BRW_STATE_INDICES)
-#define BRW_NEW_VERTICES               (1 << BRW_STATE_VERTICES)
+/**
+ * BRW_NEW_*_PROG_DATA and BRW_NEW_*_PROGRAM are similar, but distinct.
+ *
+ * BRW_NEW_*_PROGRAM relates to the gl_shader_program/gl_program structures.
+ * When the currently bound shader program differs from the previous draw
+ * call, these will be flagged.  They cover brw->{stage}_program and
+ * ctx->{Stage}Program->_Current.
+ *
+ * BRW_NEW_*_PROG_DATA is flagged when the effective shaders change, from a
+ * driver perspective.  Even if the same shader is bound at the API level,
+ * we may need to switch between multiple versions of that shader to handle
+ * changes in non-orthagonal state.
+ *
+ * Additionally, multiple shader programs may have identical vertex shaders
+ * (for example), or compile down to the same code in the backend.  We combine
+ * those into a single program cache entry.
+ *
+ * BRW_NEW_*_PROG_DATA occurs when switching program cache entries, which
+ * covers the brw_*_prog_data structures, and brw->*.prog_offset.
+ */
+#define BRW_NEW_FS_PROG_DATA            (1ull << BRW_CACHE_FS_PROG)
+/* XXX: The BRW_NEW_BLORP_BLIT_PROG_DATA dirty bit is unused (as BLORP doesn't
+ * use the normal state upload paths), but the cache is still used.  To avoid
+ * polluting the brw_state_cache code with special cases, we retain the dirty
+ * bit for now.  It should eventually be removed.
+ */
+#define BRW_NEW_BLORP_BLIT_PROG_DATA    (1ull << BRW_CACHE_BLORP_BLIT_PROG)
+#define BRW_NEW_SF_PROG_DATA            (1ull << BRW_CACHE_SF_PROG)
+#define BRW_NEW_VS_PROG_DATA            (1ull << BRW_CACHE_VS_PROG)
+#define BRW_NEW_FF_GS_PROG_DATA         (1ull << BRW_CACHE_FF_GS_PROG)
+#define BRW_NEW_GS_PROG_DATA            (1ull << BRW_CACHE_GS_PROG)
+#define BRW_NEW_CLIP_PROG_DATA          (1ull << BRW_CACHE_CLIP_PROG)
+#define BRW_NEW_URB_FENCE               (1ull << BRW_STATE_URB_FENCE)
+#define BRW_NEW_FRAGMENT_PROGRAM        (1ull << BRW_STATE_FRAGMENT_PROGRAM)
+#define BRW_NEW_GEOMETRY_PROGRAM        (1ull << BRW_STATE_GEOMETRY_PROGRAM)
+#define BRW_NEW_VERTEX_PROGRAM          (1ull << BRW_STATE_VERTEX_PROGRAM)
+#define BRW_NEW_CURBE_OFFSETS           (1ull << BRW_STATE_CURBE_OFFSETS)
+#define BRW_NEW_REDUCED_PRIMITIVE       (1ull << BRW_STATE_REDUCED_PRIMITIVE)
+#define BRW_NEW_PRIMITIVE               (1ull << BRW_STATE_PRIMITIVE)
+#define BRW_NEW_CONTEXT                 (1ull << BRW_STATE_CONTEXT)
+#define BRW_NEW_PSP                     (1ull << BRW_STATE_PSP)
+#define BRW_NEW_SURFACES                (1ull << BRW_STATE_SURFACES)
+#define BRW_NEW_VS_BINDING_TABLE        (1ull << BRW_STATE_VS_BINDING_TABLE)
+#define BRW_NEW_GS_BINDING_TABLE        (1ull << BRW_STATE_GS_BINDING_TABLE)
+#define BRW_NEW_PS_BINDING_TABLE        (1ull << BRW_STATE_PS_BINDING_TABLE)
+#define BRW_NEW_INDICES                 (1ull << BRW_STATE_INDICES)
+#define BRW_NEW_VERTICES                (1ull << BRW_STATE_VERTICES)
 /**
  * Used for any batch entry with a relocated pointer that will be used
  * by any 3D rendering.
  */
-#define BRW_NEW_BATCH                  (1 << BRW_STATE_BATCH)
+#define BRW_NEW_BATCH                   (1ull << BRW_STATE_BATCH)
 /** \see brw.state.depth_region */
-#define BRW_NEW_INDEX_BUFFER           (1 << BRW_STATE_INDEX_BUFFER)
-#define BRW_NEW_VS_CONSTBUF            (1 << BRW_STATE_VS_CONSTBUF)
-#define BRW_NEW_GS_CONSTBUF            (1 << BRW_STATE_GS_CONSTBUF)
-#define BRW_NEW_PROGRAM_CACHE          (1 << BRW_STATE_PROGRAM_CACHE)
-#define BRW_NEW_STATE_BASE_ADDRESS     (1 << BRW_STATE_STATE_BASE_ADDRESS)
-#define BRW_NEW_VUE_MAP_VS             (1 << BRW_STATE_VUE_MAP_VS)
-#define BRW_NEW_VUE_MAP_GEOM_OUT       (1 << BRW_STATE_VUE_MAP_GEOM_OUT)
-#define BRW_NEW_TRANSFORM_FEEDBACK     (1 << BRW_STATE_TRANSFORM_FEEDBACK)
-#define BRW_NEW_RASTERIZER_DISCARD     (1 << BRW_STATE_RASTERIZER_DISCARD)
-#define BRW_NEW_STATS_WM               (1 << BRW_STATE_STATS_WM)
-#define BRW_NEW_UNIFORM_BUFFER          (1 << BRW_STATE_UNIFORM_BUFFER)
-#define BRW_NEW_ATOMIC_BUFFER           (1 << BRW_STATE_ATOMIC_BUFFER)
-#define BRW_NEW_META_IN_PROGRESS        (1 << BRW_STATE_META_IN_PROGRESS)
-#define BRW_NEW_INTERPOLATION_MAP       (1 << BRW_STATE_INTERPOLATION_MAP)
-#define BRW_NEW_PUSH_CONSTANT_ALLOCATION (1 << BRW_STATE_PUSH_CONSTANT_ALLOCATION)
-#define BRW_NEW_NUM_SAMPLES             (1 << BRW_STATE_NUM_SAMPLES)
+#define BRW_NEW_INDEX_BUFFER            (1ull << BRW_STATE_INDEX_BUFFER)
+#define BRW_NEW_VS_CONSTBUF             (1ull << BRW_STATE_VS_CONSTBUF)
+#define BRW_NEW_GS_CONSTBUF             (1ull << BRW_STATE_GS_CONSTBUF)
+#define BRW_NEW_PROGRAM_CACHE           (1ull << BRW_STATE_PROGRAM_CACHE)
+#define BRW_NEW_STATE_BASE_ADDRESS      (1ull << BRW_STATE_STATE_BASE_ADDRESS)
+#define BRW_NEW_VUE_MAP_VS              (1ull << BRW_STATE_VUE_MAP_VS)
+#define BRW_NEW_VUE_MAP_GEOM_OUT        (1ull << BRW_STATE_VUE_MAP_GEOM_OUT)
+#define BRW_NEW_TRANSFORM_FEEDBACK      (1ull << BRW_STATE_TRANSFORM_FEEDBACK)
+#define BRW_NEW_RASTERIZER_DISCARD      (1ull << BRW_STATE_RASTERIZER_DISCARD)
+#define BRW_NEW_STATS_WM                (1ull << BRW_STATE_STATS_WM)
+#define BRW_NEW_UNIFORM_BUFFER          (1ull << BRW_STATE_UNIFORM_BUFFER)
+#define BRW_NEW_ATOMIC_BUFFER           (1ull << BRW_STATE_ATOMIC_BUFFER)
+#define BRW_NEW_META_IN_PROGRESS        (1ull << BRW_STATE_META_IN_PROGRESS)
+#define BRW_NEW_INTERPOLATION_MAP       (1ull << BRW_STATE_INTERPOLATION_MAP)
+#define BRW_NEW_PUSH_CONSTANT_ALLOCATION (1ull << BRW_STATE_PUSH_CONSTANT_ALLOCATION)
+#define BRW_NEW_NUM_SAMPLES             (1ull << BRW_STATE_NUM_SAMPLES)
+#define BRW_NEW_TEXTURE_BUFFER          (1ull << BRW_STATE_TEXTURE_BUFFER)
+#define BRW_NEW_GEN4_UNIT_STATE         (1ull << BRW_STATE_GEN4_UNIT_STATE)
+#define BRW_NEW_CC_VP                   (1ull << BRW_STATE_CC_VP)
+#define BRW_NEW_SF_VP                   (1ull << BRW_STATE_SF_VP)
+#define BRW_NEW_CLIP_VP                 (1ull << BRW_STATE_CLIP_VP)
+#define BRW_NEW_SAMPLER_STATE_TABLE     (1ull << BRW_STATE_SAMPLER_STATE_TABLE)
+#define BRW_NEW_VS_ATTRIB_WORKAROUNDS   (1ull << BRW_STATE_VS_ATTRIB_WORKAROUNDS)
 
 struct brw_state_flags {
    /** State update flags signalled by mesa internals */
@@ -229,16 +288,7 @@ struct brw_state_flags {
    /**
     * State update flags signalled as the result of brw_tracked_state updates
     */
-   GLuint brw;
-   /**
-    * State update flags that used to be signalled by brw_state_cache.c
-    * searches.
-    *
-    * Now almost all of that state is just streamed out on demand, but the
-    * flags for those state blobs updating have stayed in the same bitfield.
-    * brw_state_cache.c still flags CACHE_NEW_*_PROG.
-    */
-   GLuint cache;
+   uint64_t brw;
 };
 
 /** Subclass of Mesa vertex program */
@@ -298,12 +348,17 @@ struct brw_stage_prog_data {
    GLuint nr_params;       /**< number of float params/constants */
    GLuint nr_pull_params;
 
+   unsigned curb_read_length;
+   unsigned total_scratch;
+
    /**
     * Register where the thread expects to find input data from the URB
     * (typically uniforms, followed by vertex or fragment attributes).
     */
    unsigned dispatch_grf_start_reg;
 
+   bool use_alt_mode; /**< Use ALT floating point mode?  Otherwise, IEEE. */
+
    /* Pointers to tracked values (only valid once
     * _mesa_load_state_parameters has been called at runtime).
     *
@@ -325,13 +380,11 @@ struct brw_stage_prog_data {
 struct brw_wm_prog_data {
    struct brw_stage_prog_data base;
 
-   GLuint curb_read_length;
    GLuint num_varying_inputs;
 
    GLuint dispatch_grf_start_reg_16;
    GLuint reg_blocks;
    GLuint reg_blocks_16;
-   GLuint total_scratch;
 
    struct {
       /** @{
@@ -341,10 +394,13 @@ struct brw_wm_prog_data {
       /** @} */
    } binding_table;
 
+   uint8_t computed_depth_mode;
+
    bool no_8;
    bool dual_src_blend;
    bool uses_pos_offset;
    bool uses_omask;
+   bool uses_kill;
    uint32_t prog_offset_16;
 
    /**
@@ -531,16 +587,16 @@ struct brw_vec4_prog_data {
    struct brw_stage_prog_data base;
    struct brw_vue_map vue_map;
 
-   GLuint curb_read_length;
    GLuint urb_read_length;
    GLuint total_grf;
-   GLuint total_scratch;
 
    /* Used for calculating urb partitions.  In the VS, this is the size of the
     * URB entry used for both input and output to the thread.  In the GS, this
     * is the size of the URB entry used for output.
     */
    GLuint urb_entry_size;
+
+   bool simd8;
 };
 
 
@@ -553,46 +609,7 @@ struct brw_vs_prog_data {
    GLbitfield64 inputs_read;
 
    bool uses_vertexid;
-};
-
-
-/* Note: brw_gs_prog_data_compare() must be updated when adding fields to
- * this struct!
- */
-struct brw_gs_prog_data
-{
-   struct brw_vec4_prog_data base;
-
-   /**
-    * Size of an output vertex, measured in HWORDS (32 bytes).
-    */
-   unsigned output_vertex_size_hwords;
-
-   unsigned output_topology;
-
-   /**
-    * Size of the control data (cut bits or StreamID bits), in hwords (32
-    * bytes).  0 if there is no control data.
-    */
-   unsigned control_data_header_size_hwords;
-
-   /**
-    * Format of the control data (either GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_SID
-    * if the control data is StreamID bits, or
-    * GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_CUT if the control data is cut bits).
-    * Ignored if control_data_header_size is 0.
-    */
-   unsigned control_data_format;
-
-   bool include_primitive_id;
-
-   int invocations;
-
-   /**
-    * True if the thread should be dispatched in DUAL_INSTANCE mode, false if
-    * it should be dispatched in DUAL_OBJECT mode.
-    */
-   bool dual_instanced_dispatch;
+   bool uses_instanceid;
 };
 
 /** Number of texture sampler units */
@@ -639,7 +656,77 @@ struct brw_gs_prog_data
                             2 /* shader time, pull constants */)
 
 #define SURF_INDEX_GEN6_SOL_BINDING(t) (t)
-#define BRW_MAX_GEN6_GS_SURFACES       SURF_INDEX_GEN6_SOL_BINDING(BRW_MAX_SOL_BINDINGS)
+
+/* Note: brw_gs_prog_data_compare() must be updated when adding fields to
+ * this struct!
+ */
+struct brw_gs_prog_data
+{
+   struct brw_vec4_prog_data base;
+
+   /**
+    * Size of an output vertex, measured in HWORDS (32 bytes).
+    */
+   unsigned output_vertex_size_hwords;
+
+   unsigned output_topology;
+
+   /**
+    * Size of the control data (cut bits or StreamID bits), in hwords (32
+    * bytes).  0 if there is no control data.
+    */
+   unsigned control_data_header_size_hwords;
+
+   /**
+    * Format of the control data (either GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_SID
+    * if the control data is StreamID bits, or
+    * GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_CUT if the control data is cut bits).
+    * Ignored if control_data_header_size is 0.
+    */
+   unsigned control_data_format;
+
+   bool include_primitive_id;
+
+   int invocations;
+
+   /**
+    * Dispatch mode, can be any of:
+    * GEN7_GS_DISPATCH_MODE_DUAL_OBJECT
+    * GEN7_GS_DISPATCH_MODE_DUAL_INSTANCE
+    * GEN7_GS_DISPATCH_MODE_SINGLE
+    */
+   int dispatch_mode;
+
+   /**
+    * Gen6 transform feedback enabled flag.
+    */
+   bool gen6_xfb_enabled;
+
+   /**
+    * Gen6: Provoking vertex convention for odd-numbered triangles
+    * in tristrips.
+    */
+   GLuint pv_first:1;
+
+   /**
+    * Gen6: Number of varyings that are output to transform feedback.
+    */
+   GLuint num_transform_feedback_bindings:7; /* 0-BRW_MAX_SOL_BINDINGS */
+
+   /**
+    * Gen6: Map from the index of a transform feedback binding table entry to the
+    * gl_varying_slot that should be streamed out through that binding table
+    * entry.
+    */
+   unsigned char transform_feedback_bindings[BRW_MAX_SOL_BINDINGS];
+
+   /**
+    * Gen6: Map from the index of a transform feedback binding table entry to the
+    * swizzles that should be used when streaming out data through that
+    * binding table entry.
+    */
+   unsigned char transform_feedback_swizzles[BRW_MAX_SOL_BINDINGS];
+};
 
 /**
  * Stride in bytes between shader_time entries.
@@ -649,34 +736,10 @@ struct brw_gs_prog_data
  */
 #define SHADER_TIME_STRIDE 64
 
-enum brw_cache_id {
-   BRW_CC_VP,
-   BRW_CC_UNIT,
-   BRW_WM_PROG,
-   BRW_BLORP_BLIT_PROG,
-   BRW_BLORP_CONST_COLOR_PROG,
-   BRW_SAMPLER,
-   BRW_WM_UNIT,
-   BRW_SF_PROG,
-   BRW_SF_VP,
-   BRW_SF_UNIT, /* scissor state on gen6 */
-   BRW_VS_UNIT,
-   BRW_VS_PROG,
-   BRW_FF_GS_UNIT,
-   BRW_FF_GS_PROG,
-   BRW_GS_PROG,
-   BRW_CLIP_VP,
-   BRW_CLIP_UNIT,
-   BRW_CLIP_PROG,
-
-   BRW_MAX_CACHE
-};
-
 struct brw_cache_item {
    /**
     * Effectively part of the key, cache_id identifies what kind of state
-    * buffer is involved, and also which brw->state.dirty.cache flag should
-    * be set when this cache item is chosen.
+    * buffer is involved, and also which dirty flag should set.
     */
    enum brw_cache_id cache_id;
    /** 32-bit hash of the key data */
@@ -743,27 +806,6 @@ enum shader_time_shader_type {
    ST_FS16_RESET,
 };
 
-/* Flags for brw->state.cache.
- */
-#define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)
-#define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)
-#define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)
-#define CACHE_NEW_BLORP_BLIT_PROG        (1<<BRW_BLORP_BLIT_PROG)
-#define CACHE_NEW_BLORP_CONST_COLOR_PROG (1<<BRW_BLORP_CONST_COLOR_PROG)
-#define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)
-#define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)
-#define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)
-#define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)
-#define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)
-#define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)
-#define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)
-#define CACHE_NEW_FF_GS_UNIT             (1<<BRW_FF_GS_UNIT)
-#define CACHE_NEW_FF_GS_PROG             (1<<BRW_FF_GS_PROG)
-#define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)
-#define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)
-#define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)
-#define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)
-
 struct brw_vertex_buffer {
    /** Buffer object containing the uploaded vertex data */
    drm_intel_bo *bo;
@@ -998,7 +1040,6 @@ struct brw_context
    bool always_flush_cache;
    bool disable_throttling;
    bool precompile;
-   bool disable_derivative_optimization;
 
    driOptionCache optionCache;
    /** @} */
@@ -1061,6 +1102,24 @@ struct brw_context
    /* Whether the last depth/stencil packets were both NULL. */
    bool no_depth_or_stencil;
 
+   /* The last PMA stall bits programmed. */
+   uint32_t pma_stall_bits;
+
+   struct {
+      /** Does the current draw use the index buffer? */
+      bool indexed;
+
+      int start_vertex_location;
+      int base_vertex_location;
+
+      /**
+       * Buffer and offset used for GL_ARB_shader_draw_parameters
+       * (for now, only gl_BaseVertex).
+       */
+      drm_intel_bo *draw_params_bo;
+      uint32_t draw_params_offset;
+   } draw;
+
    struct {
       struct brw_vertex_element inputs[VERT_ATTRIB_MAX];
       struct brw_vertex_buffer buffers[VERT_ATTRIB_MAX];
@@ -1078,6 +1137,14 @@ struct brw_context
        * the same VB packed over and over again.
        */
       unsigned int start_vertex_bias;
+
+      /**
+       * Certain vertex attribute formats aren't natively handled by the
+       * hardware and require special VS code to fix up their values.
+       *
+       * These bitfields indicate which workarounds are needed.
+       */
+      uint8_t attrib_wa_flags[VERT_ATTRIB_MAX];
    } vb;
 
    struct {
@@ -1123,6 +1190,7 @@ struct brw_context
     */
    struct {
       GLuint vsize;            /* vertex size plus header in urb registers */
+      GLuint gsize;            /* GS output size in urb registers */
       GLuint csize;            /* constant buffer size in urb registers */
       GLuint sfsize;           /* setup data size in urb registers */
 
@@ -1145,10 +1213,10 @@ struct brw_context
       GLuint cs_start;
       GLuint size; /* Hardware URB size, in KB. */
 
-      /* gen6: True if the most recently sent _3DSTATE_URB message allocated
+      /* True if the most recently sent _3DSTATE_URB message allocated
        * URB space for the GS.
        */
-      bool gen6_gs_previously_active;
+      bool gs_present;
    } urb;
 
 
@@ -1213,7 +1281,12 @@ struct brw_context
       uint32_t state_offset;
 
       uint32_t bind_bo_offset;
-      uint32_t surf_offset[BRW_MAX_GEN6_GS_SURFACES];
+      /**
+       * Surface offsets for the binding table. We only need surfaces to
+       * implement transform feedback so BRW_MAX_SOL_BINDINGS is all that we
+       * need in this case.
+       */
+      uint32_t surf_offset[BRW_MAX_SOL_BINDINGS];
    } ff_gs;
 
    struct {
@@ -1359,6 +1432,8 @@ struct brw_context
       double report_time;
    } shader_time;
 
+   struct brw_fast_clear_state *fast_clear_state;
+
    __DRIcontext *driContext;
    struct intel_screen *intelScreen;
 };
@@ -1420,6 +1495,19 @@ void brw_meta_fbo_stencil_blit(struct brw_context *brw,
 void brw_meta_stencil_updownsample(struct brw_context *brw,
                                    struct intel_mipmap_tree *src,
                                    struct intel_mipmap_tree *dst);
+
+bool brw_meta_fast_clear(struct brw_context *brw,
+                         struct gl_framebuffer *fb,
+                         GLbitfield mask,
+                         bool partial_clear);
+
+void
+brw_meta_resolve_color(struct brw_context *brw,
+                       struct intel_mipmap_tree *mt);
+void
+brw_meta_fast_clear_free(struct brw_context *brw);
+
+
 /*======================================================================
  * brw_misc_state.c
  */
@@ -1532,7 +1620,8 @@ brw_update_sol_surface(struct brw_context *brw,
 void brw_upload_ubo_surfaces(struct brw_context *brw,
                             struct gl_shader *shader,
                              struct brw_stage_state *stage_state,
-                             struct brw_stage_prog_data *prog_data);
+                             struct brw_stage_prog_data *prog_data,
+                             bool dword_pitch);
 void brw_upload_abo_surfaces(struct brw_context *brw,
                              struct gl_shader_program *prog,
                              struct brw_stage_state *stage_state,
@@ -1629,6 +1718,8 @@ gen6_get_sample_position(struct gl_context *ctx,
                          struct gl_framebuffer *fb,
                          GLuint index,
                          GLfloat *result);
+void
+gen6_set_sample_maps(struct gl_context *ctx);
 
 /* gen8_multisample_state.c */
 void gen8_emit_3dstate_multisample(struct brw_context *brw, unsigned num_samp);
@@ -1746,6 +1837,16 @@ brw_emit_depth_stencil_hiz(struct brw_context *brw,
                            uint32_t width, uint32_t height,
                            uint32_t tile_x, uint32_t tile_y);
 
+void
+gen6_emit_depth_stencil_hiz(struct brw_context *brw,
+                            struct intel_mipmap_tree *depth_mt,
+                            uint32_t depth_offset, uint32_t depthbuffer_format,
+                            uint32_t depth_surface_type,
+                            struct intel_mipmap_tree *stencil_mt,
+                            bool hiz, bool separate_stencil,
+                            uint32_t width, uint32_t height,
+                            uint32_t tile_x, uint32_t tile_y);
+
 void
 gen7_emit_depth_stencil_hiz(struct brw_context *brw,
                             struct intel_mipmap_tree *depth_mt,