i965/gs: Add a case to brwNewProgram() for geometry shaders.

[mesa.git] / src / mesa / drivers / dri / i965 / brw_state_upload.c

diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index fa0e6028e4b59b18998d6b78726b8da3c1ca8928..64e0bcf8e8f2cb97da516f67c484ec482a7db0db 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -41,6 +41,9 @@ static const struct brw_tracked_state *gen4_atoms[] =
 {
    &brw_vs_prog, /* must do before GS prog, state base address. */
    &brw_gs_prog, /* must do before state base address */
+
+   &brw_interpolation_map,
+
    &brw_clip_prog, /* must do before state base address */
    &brw_sf_prog, /* must do before state base address */
    &brw_wm_prog, /* must do before state base address */
@@ -65,7 +68,8 @@ static const struct brw_tracked_state *gen4_atoms[] =
    &brw_vs_binding_table,
    &brw_wm_binding_table,
 
-   &brw_samplers,
+   &brw_fs_samplers,
+   &brw_vs_samplers,
 
    /* These set up state for brw_psp_urb_cbs */
    &brw_wm_unit,
@@ -111,7 +115,6 @@ static const struct brw_tracked_state *gen6_atoms[] =
    &gen6_sf_vp,
 
    /* Command packets: */
-   &brw_invariant_state,
 
    /* must do before binding table pointers, cc state ptrs */
    &brw_state_base_address,
@@ -123,7 +126,6 @@ static const struct brw_tracked_state *gen6_atoms[] =
    &gen6_blend_state,          /* must do before cc unit */
    &gen6_color_calc_state,     /* must do before cc unit */
    &gen6_depth_stencil_state,  /* must do before cc unit */
-   &gen6_cc_state_pointers,
 
    &gen6_vs_push_constants, /* Before vs_state */
    &gen6_wm_push_constants, /* Before wm_state */
@@ -142,7 +144,8 @@ static const struct brw_tracked_state *gen6_atoms[] =
    &gen6_gs_binding_table,
    &brw_wm_binding_table,
 
-   &brw_samplers,
+   &brw_fs_samplers,
+   &brw_vs_samplers,
    &gen6_sampler_state,
    &gen6_multisample_state,
 
@@ -177,8 +180,6 @@ static const struct brw_tracked_state *gen7_atoms[] =
    &brw_wm_prog,
 
    /* Command packets: */
-   &brw_invariant_state,
-   &gen7_push_constant_alloc,
 
    /* must do before binding table pointers, cc state ptrs */
    &brw_state_base_address,
@@ -191,9 +192,6 @@ static const struct brw_tracked_state *gen7_atoms[] =
    &gen6_blend_state,          /* must do before cc unit */
    &gen6_color_calc_state,     /* must do before cc unit */
    &gen6_depth_stencil_state,  /* must do before cc unit */
-   &gen7_blend_state_pointer,
-   &gen7_cc_state_pointer,
-   &gen7_depth_stencil_state_pointer,
 
    &gen6_vs_push_constants, /* Before vs_state */
    &gen6_wm_push_constants, /* Before wm_surfaces and constant_buffer */
@@ -210,7 +208,8 @@ static const struct brw_tracked_state *gen7_atoms[] =
    &brw_vs_binding_table,
    &brw_wm_binding_table,
 
-   &gen7_samplers,
+   &brw_fs_samplers,
+   &brw_vs_samplers,
    &gen6_multisample_state,
 
    &gen7_disable_stages,
@@ -241,6 +240,22 @@ static const struct brw_tracked_state *gen7_atoms[] =
    &haswell_cut_index,
 };
 
+static void
+brw_upload_initial_gpu_state(struct brw_context *brw)
+{
+   /* On platforms with hardware contexts, we can set our initial GPU state
+    * right away rather than doing it via state atoms.  This saves a small
+    * amount of overhead on every draw call.
+    */
+   if (!brw->hw_ctx)
+      return;
+
+   brw_upload_invariant_state(brw);
+
+   if (brw->gen >= 7) {
+      gen7_allocate_push_constants(brw);
+   }
+}
 
 void brw_init_state( struct brw_context *brw )
 {
@@ -249,10 +264,10 @@ void brw_init_state( struct brw_context *brw )
 
    brw_init_caches(brw);
 
-   if (brw->intel.gen >= 7) {
+   if (brw->gen >= 7) {
       atoms = gen7_atoms;
       num_atoms = ARRAY_SIZE(gen7_atoms);
-   } else if (brw->intel.gen == 6) {
+   } else if (brw->gen == 6) {
       atoms = gen6_atoms;
       num_atoms = ARRAY_SIZE(gen6_atoms);
    } else {
@@ -270,6 +285,8 @@ void brw_init_state( struct brw_context *brw )
       assert((*atoms)->emit);
       atoms++;
    }
+
+   brw_upload_initial_gpu_state(brw);
 }
 
 
@@ -281,8 +298,8 @@ void brw_destroy_state( struct brw_context *brw )
 /***********************************************************************
  */
 
-static GLuint check_state( const struct brw_state_flags *a,
-                          const struct brw_state_flags *b )
+static bool
+check_state(const struct brw_state_flags *a, const struct brw_state_flags *b)
 {
    return ((a->mesa & b->mesa) |
           (a->brw & b->brw) |
@@ -373,13 +390,11 @@ static struct dirty_bit_map brw_bits[] = {
    DEFINE_BIT(BRW_NEW_RASTERIZER_DISCARD),
    DEFINE_BIT(BRW_NEW_UNIFORM_BUFFER),
    DEFINE_BIT(BRW_NEW_META_IN_PROGRESS),
+   DEFINE_BIT(BRW_NEW_INTERPOLATION_MAP),
    {0, 0, 0}
 };
 
 static struct dirty_bit_map cache_bits[] = {
-   DEFINE_BIT(CACHE_NEW_BLEND_STATE),
-   DEFINE_BIT(CACHE_NEW_DEPTH_STENCIL_STATE),
-   DEFINE_BIT(CACHE_NEW_COLOR_CALC_STATE),
    DEFINE_BIT(CACHE_NEW_CC_VP),
    DEFINE_BIT(CACHE_NEW_CC_UNIT),
    DEFINE_BIT(CACHE_NEW_WM_PROG),
@@ -414,7 +429,7 @@ brw_update_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
 }
 
 static void
-brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
+brw_print_dirty_count(struct dirty_bit_map *bit_map)
 {
    int i;
 
@@ -432,14 +447,13 @@ brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
  */
 void brw_upload_state(struct brw_context *brw)
 {
-   struct gl_context *ctx = &brw->intel.ctx;
-   struct intel_context *intel = &brw->intel;
+   struct gl_context *ctx = &brw->ctx;
    struct brw_state_flags *state = &brw->state.dirty;
    int i;
    static int dirty_count = 0;
 
-   state->mesa |= brw->intel.NewGLState;
-   brw->intel.NewGLState = 0;
+   state->mesa |= brw->NewGLState;
+   brw->NewGLState = 0;
 
    state->brw |= ctx->NewDriverState;
    ctx->NewDriverState = 0;
@@ -455,6 +469,11 @@ void brw_upload_state(struct brw_context *brw)
       brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
    }
 
+   if (brw->geometry_program != ctx->GeometryProgram._Current) {
+      brw->geometry_program = ctx->GeometryProgram._Current;
+      brw->state.dirty.brw |= BRW_NEW_GEOMETRY_PROGRAM;
+   }
+
    if (brw->vertex_program != ctx->VertexProgram._Current) {
       brw->vertex_program = ctx->VertexProgram._Current;
       brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
@@ -468,7 +487,7 @@ void brw_upload_state(struct brw_context *brw)
    if ((state->mesa | state->cache | state->brw) == 0)
       return;
 
-   intel_check_front_buffer_rendering(intel);
+   intel_check_front_buffer_rendering(brw);
 
    if (unlikely(INTEL_DEBUG)) {
       /* Debug version which enforces various sanity checks on the
@@ -513,9 +532,9 @@ void brw_upload_state(struct brw_context *brw)
       brw_update_dirty_count(brw_bits, state->brw);
       brw_update_dirty_count(cache_bits, state->cache);
       if (dirty_count++ % 1000 == 0) {
-        brw_print_dirty_count(mesa_bits, state->mesa);
-        brw_print_dirty_count(brw_bits, state->brw);
-        brw_print_dirty_count(cache_bits, state->cache);
+        brw_print_dirty_count(mesa_bits);
+        brw_print_dirty_count(brw_bits);
+        brw_print_dirty_count(cache_bits);
         fprintf(stderr, "\n");
       }
    }