nvc0: handle nr being 0 in nvc0_set_global_bindings
[mesa.git] / src / gallium / drivers / nouveau / nvc0 / nvc0_state.c
index b6666ca58e8d3c3cb35e228c7d24bad5f3b58d81..076fe1c6decc15ad307bac61f92bdf9d85067b32 100644 (file)
  */
 
 #include "pipe/p_defines.h"
+#include "util/u_framebuffer.h"
 #include "util/u_helpers.h"
 #include "util/u_inlines.h"
 #include "util/u_transfer.h"
 
 #include "tgsi/tgsi_parse.h"
+#include "compiler/nir/nir.h"
+#include "compiler/nir/nir_serialize.h"
 
 #include "nvc0/nvc0_stateobj.h"
 #include "nvc0/nvc0_context.h"
+#include "nvc0/nvc0_query_hw.h"
 
 #include "nvc0/nvc0_3d.xml.h"
-#include "nv50/nv50_texture.xml.h"
 
 #include "nouveau_gldefs.h"
 
-static INLINE uint32_t
+static inline uint32_t
 nvc0_colormask(unsigned mask)
 {
     uint32_t ret = 0;
@@ -55,7 +58,7 @@ nvc0_colormask(unsigned mask)
 #define NVC0_BLEND_FACTOR_CASE(a, b) \
    case PIPE_BLENDFACTOR_##a: return NV50_BLEND_FACTOR_##b
 
-static INLINE uint32_t
+static inline uint32_t
 nvc0_blend_fac(unsigned factor)
 {
    switch (factor) {
@@ -92,8 +95,8 @@ nvc0_blend_state_create(struct pipe_context *pipe,
    int r; /* reference */
    uint32_t ms;
    uint8_t blend_en = 0;
-   boolean indep_masks = FALSE;
-   boolean indep_funcs = FALSE;
+   bool indep_masks = false;
+   bool indep_funcs = false;
 
    so->pipe = *cso;
 
@@ -111,7 +114,7 @@ nvc0_blend_state_create(struct pipe_context *pipe,
              cso->rt[i].alpha_func != cso->rt[r].alpha_func ||
              cso->rt[i].alpha_src_factor != cso->rt[r].alpha_src_factor ||
              cso->rt[i].alpha_dst_factor != cso->rt[r].alpha_dst_factor) {
-            indep_funcs = TRUE;
+            indep_funcs = true;
             break;
          }
       }
@@ -120,7 +123,7 @@ nvc0_blend_state_create(struct pipe_context *pipe,
 
       for (i = 1; i < 8; ++i) {
          if (cso->rt[i].colormask != cso->rt[0].colormask) {
-            indep_masks = TRUE;
+            indep_masks = true;
             break;
          }
       }
@@ -185,7 +188,7 @@ nvc0_blend_state_create(struct pipe_context *pipe,
    SB_BEGIN_3D(so, MULTISAMPLE_CTRL, 1);
    SB_DATA    (so, ms);
 
-   assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
+   assert(so->size <= ARRAY_SIZE(so->state));
    return so;
 }
 
@@ -195,7 +198,7 @@ nvc0_blend_state_bind(struct pipe_context *pipe, void *hwcso)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->blend = hwcso;
-    nvc0->dirty |= NVC0_NEW_BLEND;
+    nvc0->dirty_3d |= NVC0_NEW_3D_BLEND;
 }
 
 static void
@@ -210,6 +213,7 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
                              const struct pipe_rasterizer_state *cso)
 {
     struct nvc0_rasterizer_stateobj *so;
+    uint16_t class_3d = nouveau_screen(pipe->screen)->class_3d;
     uint32_t reg;
 
     so = CALLOC_STRUCT(nvc0_rasterizer_stateobj);
@@ -221,9 +225,6 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
      * always emit 16 commands, one for each scissor rectangle, here.
      */
 
-    SB_BEGIN_3D(so, SHADE_MODEL, 1);
-    SB_DATA    (so, cso->flatshade ? NVC0_3D_SHADE_MODEL_FLAT :
-                                     NVC0_3D_SHADE_MODEL_SMOOTH);
     SB_IMMED_3D(so, PROVOKING_VERTEX_LAST, !cso->flatshade_first);
     SB_IMMED_3D(so, VERTEX_TWO_SIDE_ENABLE, cso->light_twoside);
 
@@ -234,7 +235,7 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
     SB_IMMED_3D(so, MULTISAMPLE_ENABLE, cso->multisample);
 
     SB_IMMED_3D(so, LINE_SMOOTH_ENABLE, cso->line_smooth);
-    if (cso->line_smooth)
+    if (cso->line_smooth || cso->multisample)
        SB_BEGIN_3D(so, LINE_WIDTH_SMOOTH, 1);
     else
        SB_BEGIN_3D(so, LINE_WIDTH_ALIASED, 1);
@@ -263,6 +264,12 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
     SB_IMMED_3D(so, POINT_SPRITE_ENABLE, cso->point_quad_rasterization);
     SB_IMMED_3D(so, POINT_SMOOTH_ENABLE, cso->point_smooth);
 
+    if (class_3d >= GM200_3D_CLASS) {
+       SB_IMMED_3D(so, FILL_RECTANGLE,
+                   cso->fill_front == PIPE_POLYGON_MODE_FILL_RECTANGLE ?
+                   NVC0_3D_FILL_RECTANGLE_ENABLE : 0);
+    }
+
     SB_BEGIN_3D(so, MACRO_POLYGON_MODE_FRONT, 1);
     SB_DATA    (so, nvgl_polygon_mode(cso->fill_front));
     SB_BEGIN_3D(so, MACRO_POLYGON_MODE_BACK, 1);
@@ -295,13 +302,15 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
     if (cso->offset_point || cso->offset_line || cso->offset_tri) {
         SB_BEGIN_3D(so, POLYGON_OFFSET_FACTOR, 1);
         SB_DATA    (so, fui(cso->offset_scale));
-        SB_BEGIN_3D(so, POLYGON_OFFSET_UNITS, 1);
-        SB_DATA    (so, fui(cso->offset_units * 2.0f));
+        if (!cso->offset_units_unscaled) {
+           SB_BEGIN_3D(so, POLYGON_OFFSET_UNITS, 1);
+           SB_DATA    (so, fui(cso->offset_units * 2.0f));
+        }
         SB_BEGIN_3D(so, POLYGON_OFFSET_CLAMP, 1);
         SB_DATA    (so, fui(cso->offset_clamp));
     }
 
-    if (cso->depth_clip)
+    if (cso->depth_clip_near)
        reg = NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1;
     else
        reg =
@@ -317,7 +326,21 @@ nvc0_rasterizer_state_create(struct pipe_context *pipe,
 
     SB_IMMED_3D(so, PIXEL_CENTER_INTEGER, !cso->half_pixel_center);
 
-    assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
+    if (class_3d >= GM200_3D_CLASS) {
+        if (cso->conservative_raster_mode != PIPE_CONSERVATIVE_RASTER_OFF) {
+            bool post_snap = cso->conservative_raster_mode ==
+                PIPE_CONSERVATIVE_RASTER_POST_SNAP;
+            uint32_t state = cso->subpixel_precision_x;
+            state |= cso->subpixel_precision_y << 4;
+            state |= (uint32_t)(cso->conservative_raster_dilate * 4) << 8;
+            state |= (post_snap || class_3d < GP100_3D_CLASS) ? 1 << 10 : 0;
+            SB_IMMED_3D(so, MACRO_CONSERVATIVE_RASTER_STATE, state);
+        } else {
+            SB_IMMED_3D(so, CONSERVATIVE_RASTER, 0);
+        }
+    }
+
+    assert(so->size <= ARRAY_SIZE(so->state));
     return (void *)so;
 }
 
@@ -327,7 +350,7 @@ nvc0_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
    struct nvc0_context *nvc0 = nvc0_context(pipe);
 
    nvc0->rast = hwcso;
-   nvc0->dirty |= NVC0_NEW_RASTERIZER;
+   nvc0->dirty_3d |= NVC0_NEW_3D_RASTERIZER;
 }
 
 static void
@@ -351,6 +374,13 @@ nvc0_zsa_state_create(struct pipe_context *pipe,
       SB_DATA    (so, nvgl_comparison_op(cso->depth.func));
    }
 
+   SB_IMMED_3D(so, DEPTH_BOUNDS_EN, cso->depth.bounds_test);
+   if (cso->depth.bounds_test) {
+      SB_BEGIN_3D(so, DEPTH_BOUNDS(0), 2);
+      SB_DATA    (so, fui(cso->depth.bounds_min));
+      SB_DATA    (so, fui(cso->depth.bounds_max));
+   }
+
    if (cso->stencil[0].enabled) {
       SB_BEGIN_3D(so, STENCIL_ENABLE, 5);
       SB_DATA    (so, 1);
@@ -388,7 +418,7 @@ nvc0_zsa_state_create(struct pipe_context *pipe,
       SB_DATA    (so, nvgl_comparison_op(cso->alpha.func));
    }
 
-   assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
+   assert(so->size <= ARRAY_SIZE(so->state));
    return (void *)so;
 }
 
@@ -398,7 +428,7 @@ nvc0_zsa_state_bind(struct pipe_context *pipe, void *hwcso)
    struct nvc0_context *nvc0 = nvc0_context(pipe);
 
    nvc0->zsa = hwcso;
-   nvc0->dirty |= NVC0_NEW_ZSA;
+   nvc0->dirty_3d |= NVC0_NEW_3D_ZSA;
 }
 
 static void
@@ -413,30 +443,12 @@ nvc0_zsa_state_delete(struct pipe_context *pipe, void *hwcso)
 #define NV50_TSC_WRAP_CASE(n) \
     case PIPE_TEX_WRAP_##n: return NV50_TSC_WRAP_##n
 
-static INLINE unsigned
-nv50_tsc_wrap_mode(unsigned wrap)
-{
-   switch (wrap) {
-   NV50_TSC_WRAP_CASE(REPEAT);
-   NV50_TSC_WRAP_CASE(MIRROR_REPEAT);
-   NV50_TSC_WRAP_CASE(CLAMP_TO_EDGE);
-   NV50_TSC_WRAP_CASE(CLAMP_TO_BORDER);
-   NV50_TSC_WRAP_CASE(CLAMP);
-   NV50_TSC_WRAP_CASE(MIRROR_CLAMP_TO_EDGE);
-   NV50_TSC_WRAP_CASE(MIRROR_CLAMP_TO_BORDER);
-   NV50_TSC_WRAP_CASE(MIRROR_CLAMP);
-   default:
-       NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
-       return NV50_TSC_WRAP_REPEAT;
-   }
-}
-
 static void
 nvc0_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
 {
    unsigned s, i;
 
-   for (s = 0; s < 5; ++s)
+   for (s = 0; s < 6; ++s)
       for (i = 0; i < nvc0_context(pipe)->num_samplers[s]; ++i)
          if (nvc0_context(pipe)->samplers[s][i] == hwcso)
             nvc0_context(pipe)->samplers[s][i] = NULL;
@@ -446,100 +458,47 @@ nvc0_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
    FREE(hwcso);
 }
 
-static INLINE void
-nvc0_stage_sampler_states_bind(struct nvc0_context *nvc0, int s,
-                               unsigned nr, void **hwcso)
+static inline void
+nvc0_stage_sampler_states_bind(struct nvc0_context *nvc0,
+                               unsigned s,
+                               unsigned nr, void **hwcsos)
 {
+   unsigned highest_found = 0;
    unsigned i;
 
    for (i = 0; i < nr; ++i) {
+      struct nv50_tsc_entry *hwcso = hwcsos ? nv50_tsc_entry(hwcsos[i]) : NULL;
       struct nv50_tsc_entry *old = nvc0->samplers[s][i];
 
-      if (hwcso[i] == old)
+      if (hwcso)
+         highest_found = i;
+
+      if (hwcso == old)
          continue;
       nvc0->samplers_dirty[s] |= 1 << i;
 
-      nvc0->samplers[s][i] = nv50_tsc_entry(hwcso[i]);
+      nvc0->samplers[s][i] = hwcso;
       if (old)
          nvc0_screen_tsc_unlock(nvc0->screen, old);
    }
-   for (; i < nvc0->num_samplers[s]; ++i) {
-      if (nvc0->samplers[s][i]) {
-         nvc0_screen_tsc_unlock(nvc0->screen, nvc0->samplers[s][i]);
-         nvc0->samplers[s][i] = NULL;
-      }
-   }
-
-   nvc0->num_samplers[s] = nr;
-
-   nvc0->dirty |= NVC0_NEW_SAMPLERS;
+   if (nr >= nvc0->num_samplers[s])
+      nvc0->num_samplers[s] = highest_found + 1;
 }
 
 static void
-nvc0_stage_sampler_states_bind_range(struct nvc0_context *nvc0,
-                                     const unsigned s,
-                                     unsigned start, unsigned nr, void **cso)
+nvc0_bind_sampler_states(struct pipe_context *pipe,
+                         enum pipe_shader_type shader,
+                         unsigned start, unsigned nr, void **samplers)
 {
-   const unsigned end = start + nr;
-   int last_valid = -1;
-   unsigned i;
-
-   if (cso) {
-      for (i = start; i < end; ++i) {
-         const unsigned p = i - start;
-         if (cso[p])
-            last_valid = i;
-         if (cso[p] == nvc0->samplers[s][i])
-            continue;
-         nvc0->samplers_dirty[s] |= 1 << i;
-
-         if (nvc0->samplers[s][i])
-            nvc0_screen_tsc_unlock(nvc0->screen, nvc0->samplers[s][i]);
-         nvc0->samplers[s][i] = cso[p];
-      }
-   } else {
-      for (i = start; i < end; ++i) {
-         if (nvc0->samplers[s][i]) {
-            nvc0_screen_tsc_unlock(nvc0->screen, nvc0->samplers[s][i]);
-            nvc0->samplers[s][i] = NULL;
-            nvc0->samplers_dirty[s] |= 1 << i;
-         }
-      }
-   }
+   const unsigned s = nvc0_shader_stage(shader);
 
-   if (nvc0->num_samplers[s] <= end) {
-      if (last_valid < 0) {
-         for (i = start; i && !nvc0->samplers[s][i - 1]; --i);
-         nvc0->num_samplers[s] = i;
-      } else {
-         nvc0->num_samplers[s] = last_valid + 1;
-      }
-   }
-}
+   assert(start == 0);
+   nvc0_stage_sampler_states_bind(nvc0_context(pipe), s, nr, samplers);
 
-static void
-nvc0_bind_sampler_states(struct pipe_context *pipe, unsigned shader,
-                         unsigned start, unsigned nr, void **s)
-{
-   switch (shader) {
-   case PIPE_SHADER_VERTEX:
-      assert(start == 0);
-      nvc0_stage_sampler_states_bind(nvc0_context(pipe), 0, nr, s);
-      break;
-   case PIPE_SHADER_GEOMETRY:
-      assert(start == 0);
-      nvc0_stage_sampler_states_bind(nvc0_context(pipe), 3, nr, s);
-      break;
-   case PIPE_SHADER_FRAGMENT:
-      assert(start == 0);
-      nvc0_stage_sampler_states_bind(nvc0_context(pipe), 4, nr, s);
-      break;
-   case PIPE_SHADER_COMPUTE:
-      nvc0_stage_sampler_states_bind_range(nvc0_context(pipe), 5,
-                                           start, nr, s);
+   if (s == 5)
       nvc0_context(pipe)->dirty_cp |= NVC0_NEW_CP_SAMPLERS;
-      break;
-   }
+   else
+      nvc0_context(pipe)->dirty_3d |= NVC0_NEW_3D_SAMPLERS;
 }
 
 
@@ -555,7 +514,7 @@ nvc0_sampler_view_destroy(struct pipe_context *pipe,
    FREE(nv50_tic_entry(view));
 }
 
-static INLINE void
+static inline void
 nvc0_stage_set_sampler_views(struct nvc0_context *nvc0, int s,
                              unsigned nr,
                              struct pipe_sampler_view **views)
@@ -563,111 +522,66 @@ nvc0_stage_set_sampler_views(struct nvc0_context *nvc0, int s,
    unsigned i;
 
    for (i = 0; i < nr; ++i) {
+      struct pipe_sampler_view *view = views ? views[i] : NULL;
       struct nv50_tic_entry *old = nv50_tic_entry(nvc0->textures[s][i]);
 
-      if (views[i] == nvc0->textures[s][i])
+      if (view == nvc0->textures[s][i])
          continue;
       nvc0->textures_dirty[s] |= 1 << i;
 
+      if (view && view->texture) {
+         struct pipe_resource *res = view->texture;
+         if (res->target == PIPE_BUFFER &&
+             (res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT))
+            nvc0->textures_coherent[s] |= 1 << i;
+         else
+            nvc0->textures_coherent[s] &= ~(1 << i);
+      } else {
+         nvc0->textures_coherent[s] &= ~(1 << i);
+      }
+
       if (old) {
-         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TEX(s, i));
+         if (s == 5)
+            nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_TEX(i));
+         else
+            nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_TEX(s, i));
          nvc0_screen_tic_unlock(nvc0->screen, old);
       }
 
-      pipe_sampler_view_reference(&nvc0->textures[s][i], views[i]);
+      pipe_sampler_view_reference(&nvc0->textures[s][i], view);
    }
 
    for (i = nr; i < nvc0->num_textures[s]; ++i) {
       struct nv50_tic_entry *old = nv50_tic_entry(nvc0->textures[s][i]);
       if (old) {
-         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TEX(s, i));
+         if (s == 5)
+            nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_TEX(i));
+         else
+            nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_TEX(s, i));
          nvc0_screen_tic_unlock(nvc0->screen, old);
          pipe_sampler_view_reference(&nvc0->textures[s][i], NULL);
       }
    }
 
    nvc0->num_textures[s] = nr;
-
-   nvc0->dirty |= NVC0_NEW_TEXTURES;
 }
 
 static void
-nvc0_stage_set_sampler_views_range(struct nvc0_context *nvc0, const unsigned s,
-                                   unsigned start, unsigned nr,
-                                   struct pipe_sampler_view **views)
-{
-   struct nouveau_bufctx *bctx = (s == 5) ? nvc0->bufctx_cp : nvc0->bufctx_3d;
-   const unsigned end = start + nr;
-   const unsigned bin = (s == 5) ? NVC0_BIND_CP_TEX(0) : NVC0_BIND_TEX(s, 0);
-   int last_valid = -1;
-   unsigned i;
-
-   if (views) {
-      for (i = start; i < end; ++i) {
-         const unsigned p = i - start;
-         if (views[p])
-            last_valid = i;
-         if (views[p] == nvc0->textures[s][i])
-            continue;
-         nvc0->textures_dirty[s] |= 1 << i;
-
-         if (nvc0->textures[s][i]) {
-            struct nv50_tic_entry *old = nv50_tic_entry(nvc0->textures[s][i]);
-            nouveau_bufctx_reset(bctx, bin + i);
-            nvc0_screen_tic_unlock(nvc0->screen, old);
-         }
-         pipe_sampler_view_reference(&nvc0->textures[s][i], views[p]);
-      }
-   } else {
-      for (i = start; i < end; ++i) {
-         struct nv50_tic_entry *old = nv50_tic_entry(nvc0->textures[s][i]);
-         if (!old)
-            continue;
-         nvc0->textures_dirty[s] |= 1 << i;
-
-         nvc0_screen_tic_unlock(nvc0->screen, old);
-         pipe_sampler_view_reference(&nvc0->textures[s][i], NULL);
-         nouveau_bufctx_reset(bctx, bin + i);
-      }
-   }
-
-   if (nvc0->num_textures[s] <= end) {
-      if (last_valid < 0) {
-         for (i = start; i && !nvc0->textures[s][i - 1]; --i);
-         nvc0->num_textures[s] = i;
-      } else {
-         nvc0->num_textures[s] = last_valid + 1;
-      }
-   }
-}
-
-static void
-nvc0_set_sampler_views(struct pipe_context *pipe, unsigned shader,
+nvc0_set_sampler_views(struct pipe_context *pipe, enum pipe_shader_type shader,
                        unsigned start, unsigned nr,
                        struct pipe_sampler_view **views)
 {
+   const unsigned s = nvc0_shader_stage(shader);
+
    assert(start == 0);
-   switch (shader) {
-   case PIPE_SHADER_VERTEX:
-      nvc0_stage_set_sampler_views(nvc0_context(pipe), 0, nr, views);
-      break;
-   case PIPE_SHADER_GEOMETRY:
-      nvc0_stage_set_sampler_views(nvc0_context(pipe), 3, nr, views);
-      break;
-   case PIPE_SHADER_FRAGMENT:
-      nvc0_stage_set_sampler_views(nvc0_context(pipe), 4, nr, views);
-      break;
-   case PIPE_SHADER_COMPUTE:
-      nvc0_stage_set_sampler_views_range(nvc0_context(pipe), 5,
-                                         start, nr, views);
+   nvc0_stage_set_sampler_views(nvc0_context(pipe), s, nr, views);
+
+   if (s == 5)
       nvc0_context(pipe)->dirty_cp |= NVC0_NEW_CP_TEXTURES;
-      break;
-   default:
-      ;
-   }
+   else
+      nvc0_context(pipe)->dirty_3d |= NVC0_NEW_3D_TEXTURES;
 }
 
-
 /* ============================= SHADERS =======================================
  */
 
@@ -682,13 +596,28 @@ nvc0_sp_state_create(struct pipe_context *pipe,
       return NULL;
 
    prog->type = type;
+   prog->pipe.type = cso->type;
 
-   if (cso->tokens)
+   switch(cso->type) {
+   case PIPE_SHADER_IR_TGSI:
       prog->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+      break;
+   case PIPE_SHADER_IR_NIR:
+      prog->pipe.ir.nir = cso->ir.nir;
+      break;
+   default:
+      assert(!"unsupported IR!");
+      free(prog);
+      return NULL;
+   }
 
    if (cso->stream_output.num_outputs)
       prog->pipe.stream_output = cso->stream_output;
 
+   prog->translated = nvc0_program_translate(
+      prog, nvc0_context(pipe)->screen->base.device->chipset,
+      &nouveau_context(pipe)->debug);
+
    return (void *)prog;
 }
 
@@ -699,7 +628,10 @@ nvc0_sp_state_delete(struct pipe_context *pipe, void *hwcso)
 
    nvc0_program_destroy(nvc0_context(pipe), prog);
 
-   FREE((void *)prog->pipe.tokens);
+   if (prog->pipe.type == PIPE_SHADER_IR_TGSI)
+      FREE((void *)prog->pipe.tokens);
+   else if (prog->pipe.type == PIPE_SHADER_IR_NIR)
+      ralloc_free(prog->pipe.ir.nir);
    FREE(prog);
 }
 
@@ -716,7 +648,7 @@ nvc0_vp_state_bind(struct pipe_context *pipe, void *hwcso)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->vertprog = hwcso;
-    nvc0->dirty |= NVC0_NEW_VERTPROG;
+    nvc0->dirty_3d |= NVC0_NEW_3D_VERTPROG;
 }
 
 static void *
@@ -732,7 +664,7 @@ nvc0_fp_state_bind(struct pipe_context *pipe, void *hwcso)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->fragprog = hwcso;
-    nvc0->dirty |= NVC0_NEW_FRAGPROG;
+    nvc0->dirty_3d |= NVC0_NEW_3D_FRAGPROG;
 }
 
 static void *
@@ -748,7 +680,39 @@ nvc0_gp_state_bind(struct pipe_context *pipe, void *hwcso)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->gmtyprog = hwcso;
-    nvc0->dirty |= NVC0_NEW_GMTYPROG;
+    nvc0->dirty_3d |= NVC0_NEW_3D_GMTYPROG;
+}
+
+static void *
+nvc0_tcp_state_create(struct pipe_context *pipe,
+                     const struct pipe_shader_state *cso)
+{
+   return nvc0_sp_state_create(pipe, cso, PIPE_SHADER_TESS_CTRL);
+}
+
+static void
+nvc0_tcp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+    struct nvc0_context *nvc0 = nvc0_context(pipe);
+
+    nvc0->tctlprog = hwcso;
+    nvc0->dirty_3d |= NVC0_NEW_3D_TCTLPROG;
+}
+
+static void *
+nvc0_tep_state_create(struct pipe_context *pipe,
+                     const struct pipe_shader_state *cso)
+{
+   return nvc0_sp_state_create(pipe, cso, PIPE_SHADER_TESS_EVAL);
+}
+
+static void
+nvc0_tep_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+    struct nvc0_context *nvc0 = nvc0_context(pipe);
+
+    nvc0->tevlprog = hwcso;
+    nvc0->dirty_3d |= NVC0_NEW_3D_TEVLPROG;
 }
 
 static void *
@@ -761,12 +725,37 @@ nvc0_cp_state_create(struct pipe_context *pipe,
    if (!prog)
       return NULL;
    prog->type = PIPE_SHADER_COMPUTE;
+   prog->pipe.type = cso->ir_type;
 
    prog->cp.smem_size = cso->req_local_mem;
    prog->cp.lmem_size = cso->req_private_mem;
    prog->parm_size = cso->req_input_mem;
 
-   prog->pipe.tokens = tgsi_dup_tokens((const struct tgsi_token *)cso->prog);
+   switch(cso->ir_type) {
+   case PIPE_SHADER_IR_TGSI:
+      prog->pipe.tokens = tgsi_dup_tokens((const struct tgsi_token *)cso->prog);
+      break;
+   case PIPE_SHADER_IR_NIR:
+      prog->pipe.ir.nir = (nir_shader *)cso->prog;
+      break;
+   case PIPE_SHADER_IR_NIR_SERIALIZED: {
+      struct blob_reader reader;
+      const struct pipe_binary_program_header *hdr = cso->prog;
+
+      blob_reader_init(&reader, hdr->blob, hdr->num_bytes);
+      prog->pipe.ir.nir = nir_deserialize(NULL, pipe->screen->get_compiler_options(pipe->screen, PIPE_SHADER_IR_NIR, PIPE_SHADER_COMPUTE), &reader);
+      prog->pipe.type = PIPE_SHADER_IR_NIR;
+      break;
+   }
+   default:
+      assert(!"unsupported IR!");
+      free(prog);
+      return NULL;
+   }
+
+   prog->translated = nvc0_program_translate(
+      prog, nvc0_context(pipe)->screen->base.device->chipset,
+      &nouveau_context(pipe)->debug);
 
    return (void *)prog;
 }
@@ -781,8 +770,9 @@ nvc0_cp_state_bind(struct pipe_context *pipe, void *hwcso)
 }
 
 static void
-nvc0_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
-                         struct pipe_constant_buffer *cb)
+nvc0_set_constant_buffer(struct pipe_context *pipe,
+                         enum pipe_shader_type shader, uint index,
+                         const struct pipe_constant_buffer *cb)
 {
    struct nvc0_context *nvc0 = nvc0_context(pipe);
    struct pipe_resource *res = cb ? cb->buffer : NULL;
@@ -790,7 +780,9 @@ nvc0_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
    const unsigned i = index;
 
    if (unlikely(shader == PIPE_SHADER_COMPUTE)) {
-      assert(!cb || !cb->user_buffer);
+      if (nvc0->constbuf[s][i].user)
+         nvc0->constbuf[s][i].u.buf = NULL;
+      else
       if (nvc0->constbuf[s][i].u.buf)
          nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_CB(i));
 
@@ -800,27 +792,35 @@ nvc0_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
          nvc0->constbuf[s][i].u.buf = NULL;
       else
       if (nvc0->constbuf[s][i].u.buf)
-         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_CB(s, i));
+         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_CB(s, i));
 
-      nvc0->dirty |= NVC0_NEW_CONSTBUF;
+      nvc0->dirty_3d |= NVC0_NEW_3D_CONSTBUF;
    }
    nvc0->constbuf_dirty[s] |= 1 << i;
 
+   if (nvc0->constbuf[s][i].u.buf)
+      nv04_resource(nvc0->constbuf[s][i].u.buf)->cb_bindings[s] &= ~(1 << i);
    pipe_resource_reference(&nvc0->constbuf[s][i].u.buf, res);
 
-   nvc0->constbuf[s][i].user = (cb && cb->user_buffer) ? TRUE : FALSE;
+   nvc0->constbuf[s][i].user = (cb && cb->user_buffer) ? true : false;
    if (nvc0->constbuf[s][i].user) {
       nvc0->constbuf[s][i].u.data = cb->user_buffer;
-      nvc0->constbuf[s][i].size = cb->buffer_size;
+      nvc0->constbuf[s][i].size = MIN2(cb->buffer_size, 0x10000);
       nvc0->constbuf_valid[s] |= 1 << i;
+      nvc0->constbuf_coherent[s] &= ~(1 << i);
    } else
    if (cb) {
       nvc0->constbuf[s][i].offset = cb->buffer_offset;
-      nvc0->constbuf[s][i].size = align(cb->buffer_size, 0x100);
+      nvc0->constbuf[s][i].size = MIN2(align(cb->buffer_size, 0x100), 0x10000);
       nvc0->constbuf_valid[s] |= 1 << i;
+      if (res && res->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+         nvc0->constbuf_coherent[s] |= 1 << i;
+      else
+         nvc0->constbuf_coherent[s] &= ~(1 << i);
    }
    else {
       nvc0->constbuf_valid[s] &= ~(1 << i);
+      nvc0->constbuf_coherent[s] &= ~(1 << i);
    }
 }
 
@@ -834,7 +834,7 @@ nvc0_set_blend_color(struct pipe_context *pipe,
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->blend_colour = *bcol;
-    nvc0->dirty |= NVC0_NEW_BLEND_COLOUR;
+    nvc0->dirty_3d |= NVC0_NEW_3D_BLEND_COLOUR;
 }
 
 static void
@@ -844,7 +844,7 @@ nvc0_set_stencil_ref(struct pipe_context *pipe,
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->stencil_ref = *sr;
-    nvc0->dirty |= NVC0_NEW_STENCIL_REF;
+    nvc0->dirty_3d |= NVC0_NEW_3D_STENCIL_REF;
 }
 
 static void
@@ -855,7 +855,7 @@ nvc0_set_clip_state(struct pipe_context *pipe,
 
     memcpy(nvc0->clip.ucp, clip->ucp, sizeof(clip->ucp));
 
-    nvc0->dirty |= NVC0_NEW_CLIP;
+    nvc0->dirty_3d |= NVC0_NEW_3D_CLIP;
 }
 
 static void
@@ -864,7 +864,7 @@ nvc0_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->sample_mask = sample_mask;
-    nvc0->dirty |= NVC0_NEW_SAMPLE_MASK;
+    nvc0->dirty_3d |= NVC0_NEW_3D_SAMPLE_MASK;
 }
 
 static void
@@ -874,7 +874,7 @@ nvc0_set_min_samples(struct pipe_context *pipe, unsigned min_samples)
 
    if (nvc0->min_samples != min_samples) {
       nvc0->min_samples = min_samples;
-      nvc0->dirty |= NVC0_NEW_MIN_SAMPLES;
+      nvc0->dirty_3d |= NVC0_NEW_3D_MIN_SAMPLES;
    }
 }
 
@@ -883,23 +883,28 @@ nvc0_set_framebuffer_state(struct pipe_context *pipe,
                            const struct pipe_framebuffer_state *fb)
 {
     struct nvc0_context *nvc0 = nvc0_context(pipe);
-    unsigned i;
 
-    nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_FB);
+    nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_FB);
 
-    for (i = 0; i < fb->nr_cbufs; ++i)
-       pipe_surface_reference(&nvc0->framebuffer.cbufs[i], fb->cbufs[i]);
-    for (; i < nvc0->framebuffer.nr_cbufs; ++i)
-       pipe_surface_reference(&nvc0->framebuffer.cbufs[i], NULL);
+    util_copy_framebuffer_state(&nvc0->framebuffer, fb);
 
-    nvc0->framebuffer.nr_cbufs = fb->nr_cbufs;
+    nvc0->dirty_3d |= NVC0_NEW_3D_FRAMEBUFFER | NVC0_NEW_3D_SAMPLE_LOCATIONS |
+       NVC0_NEW_3D_TEXTURES;
+    nvc0->dirty_cp |= NVC0_NEW_CP_TEXTURES;
+}
 
-    nvc0->framebuffer.width = fb->width;
-    nvc0->framebuffer.height = fb->height;
+static void
+nvc0_set_sample_locations(struct pipe_context *pipe,
+                          size_t size, const uint8_t *locations)
+{
+    struct nvc0_context *nvc0 = nvc0_context(pipe);
 
-    pipe_surface_reference(&nvc0->framebuffer.zsbuf, fb->zsbuf);
+    nvc0->sample_locations_enabled = size && locations;
+    if (size > sizeof(nvc0->sample_locations))
+       size = sizeof(nvc0->sample_locations);
+    memcpy(nvc0->sample_locations, locations, size);
 
-    nvc0->dirty |= NVC0_NEW_FRAMEBUFFER;
+    nvc0->dirty_3d |= NVC0_NEW_3D_SAMPLE_LOCATIONS;
 }
 
 static void
@@ -909,7 +914,7 @@ nvc0_set_polygon_stipple(struct pipe_context *pipe,
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->stipple = *stipple;
-    nvc0->dirty |= NVC0_NEW_STIPPLE;
+    nvc0->dirty_3d |= NVC0_NEW_3D_STIPPLE;
 }
 
 static void
@@ -927,7 +932,7 @@ nvc0_set_scissor_states(struct pipe_context *pipe,
          continue;
       nvc0->scissors[start_slot + i] = scissor[i];
       nvc0->scissors_dirty |= 1 << (start_slot + i);
-      nvc0->dirty |= NVC0_NEW_SCISSOR;
+      nvc0->dirty_3d |= NVC0_NEW_3D_SCISSOR;
    }
 }
 
@@ -946,11 +951,39 @@ nvc0_set_viewport_states(struct pipe_context *pipe,
          continue;
       nvc0->viewports[start_slot + i] = vpt[i];
       nvc0->viewports_dirty |= 1 << (start_slot + i);
-      nvc0->dirty |= NVC0_NEW_VIEWPORT;
+      nvc0->dirty_3d |= NVC0_NEW_3D_VIEWPORT;
    }
 
 }
 
+static void
+nvc0_set_window_rectangles(struct pipe_context *pipe,
+                           bool include,
+                           unsigned num_rectangles,
+                           const struct pipe_scissor_state *rectangles)
+{
+   struct nvc0_context *nvc0 = nvc0_context(pipe);
+
+   nvc0->window_rect.inclusive = include;
+   nvc0->window_rect.rects = MIN2(num_rectangles, NVC0_MAX_WINDOW_RECTANGLES);
+   memcpy(nvc0->window_rect.rect, rectangles,
+          sizeof(struct pipe_scissor_state) * nvc0->window_rect.rects);
+
+   nvc0->dirty_3d |= NVC0_NEW_3D_WINDOW_RECTS;
+}
+
+static void
+nvc0_set_tess_state(struct pipe_context *pipe,
+                    const float default_tess_outer[4],
+                    const float default_tess_inner[2])
+{
+   struct nvc0_context *nvc0 = nvc0_context(pipe);
+
+   memcpy(nvc0->default_tess_outer, default_tess_outer, 4 * sizeof(float));
+   memcpy(nvc0->default_tess_inner, default_tess_inner, 2 * sizeof(float));
+   nvc0->dirty_3d |= NVC0_NEW_3D_TESSFACTOR;
+}
+
 static void
 nvc0_set_vertex_buffers(struct pipe_context *pipe,
                         unsigned start_slot, unsigned count,
@@ -959,56 +992,39 @@ nvc0_set_vertex_buffers(struct pipe_context *pipe,
     struct nvc0_context *nvc0 = nvc0_context(pipe);
     unsigned i;
 
+    nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_VTX);
+    nvc0->dirty_3d |= NVC0_NEW_3D_ARRAYS;
+
     util_set_vertex_buffers_count(nvc0->vtxbuf, &nvc0->num_vtxbufs, vb,
                                   start_slot, count);
 
     if (!vb) {
        nvc0->vbo_user &= ~(((1ull << count) - 1) << start_slot);
        nvc0->constant_vbos &= ~(((1ull << count) - 1) << start_slot);
+       nvc0->vtxbufs_coherent &= ~(((1ull << count) - 1) << start_slot);
        return;
     }
 
     for (i = 0; i < count; ++i) {
        unsigned dst_index = start_slot + i;
 
-       if (vb[i].user_buffer) {
+       if (vb[i].is_user_buffer) {
           nvc0->vbo_user |= 1 << dst_index;
           if (!vb[i].stride && nvc0->screen->eng3d->oclass < GM107_3D_CLASS)
              nvc0->constant_vbos |= 1 << dst_index;
           else
              nvc0->constant_vbos &= ~(1 << dst_index);
+          nvc0->vtxbufs_coherent &= ~(1 << dst_index);
        } else {
           nvc0->vbo_user &= ~(1 << dst_index);
           nvc0->constant_vbos &= ~(1 << dst_index);
-       }
-    }
-
-    nvc0->dirty |= NVC0_NEW_ARRAYS;
-    nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_VTX);
-}
-
-static void
-nvc0_set_index_buffer(struct pipe_context *pipe,
-                      const struct pipe_index_buffer *ib)
-{
-    struct nvc0_context *nvc0 = nvc0_context(pipe);
-
-    if (nvc0->idxbuf.buffer)
-       nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_IDX);
 
-    if (ib) {
-       pipe_resource_reference(&nvc0->idxbuf.buffer, ib->buffer);
-       nvc0->idxbuf.index_size = ib->index_size;
-       if (ib->buffer) {
-          nvc0->idxbuf.offset = ib->offset;
-          nvc0->dirty |= NVC0_NEW_IDXBUF;
-       } else {
-          nvc0->idxbuf.user_buffer = ib->user_buffer;
-          nvc0->dirty &= ~NVC0_NEW_IDXBUF;
+          if (vb[i].buffer.resource &&
+              vb[i].buffer.resource->flags & PIPE_RESOURCE_FLAG_MAP_COHERENT)
+             nvc0->vtxbufs_coherent |= (1 << dst_index);
+          else
+             nvc0->vtxbufs_coherent &= ~(1 << dst_index);
        }
-    } else {
-       nvc0->dirty &= ~NVC0_NEW_IDXBUF;
-       pipe_resource_reference(&nvc0->idxbuf.buffer, NULL);
     }
 }
 
@@ -1018,7 +1034,7 @@ nvc0_vertex_state_bind(struct pipe_context *pipe, void *hwcso)
     struct nvc0_context *nvc0 = nvc0_context(pipe);
 
     nvc0->vertex = hwcso;
-    nvc0->dirty |= NVC0_NEW_VERTEX;
+    nvc0->dirty_3d |= NVC0_NEW_3D_VERTEX;
 }
 
 static struct pipe_stream_output_target *
@@ -1031,12 +1047,12 @@ nvc0_so_target_create(struct pipe_context *pipe,
    if (!targ)
       return NULL;
 
-   targ->pq = pipe->create_query(pipe, NVC0_QUERY_TFB_BUFFER_OFFSET, 0);
+   targ->pq = pipe->create_query(pipe, NVC0_HW_QUERY_TFB_BUFFER_OFFSET, 0);
    if (!targ->pq) {
       FREE(targ);
       return NULL;
    }
-   targ->clean = TRUE;
+   targ->clean = true;
 
    targ->pipe.buffer_size = size;
    targ->pipe.buffer_offset = offset;
@@ -1046,11 +1062,30 @@ nvc0_so_target_create(struct pipe_context *pipe,
    pipe_reference_init(&targ->pipe.reference, 1);
 
    assert(buf->base.target == PIPE_BUFFER);
-   util_range_add(&buf->valid_buffer_range, offset, offset + size);
+   util_range_add(&buf->base, &buf->valid_buffer_range, offset, offset + size);
 
    return &targ->pipe;
 }
 
+static void
+nvc0_so_target_save_offset(struct pipe_context *pipe,
+                           struct pipe_stream_output_target *ptarg,
+                           unsigned index, bool *serialize)
+{
+   struct nvc0_so_target *targ = nvc0_so_target(ptarg);
+
+   if (*serialize) {
+      *serialize = false;
+      PUSH_SPACE(nvc0_context(pipe)->base.pushbuf, 1);
+      IMMED_NVC0(nvc0_context(pipe)->base.pushbuf, NVC0_3D(SERIALIZE), 0);
+
+      NOUVEAU_DRV_STAT(nouveau_screen(pipe->screen), gpu_serialize_count, 1);
+   }
+
+   nvc0_query(targ->pq)->index = index;
+   pipe->end_query(pipe, targ->pq);
+}
+
 static void
 nvc0_so_target_destroy(struct pipe_context *pipe,
                        struct pipe_stream_output_target *ptarg)
@@ -1069,13 +1104,13 @@ nvc0_set_transform_feedback_targets(struct pipe_context *pipe,
 {
    struct nvc0_context *nvc0 = nvc0_context(pipe);
    unsigned i;
-   boolean serialize = TRUE;
+   bool serialize = true;
 
    assert(num_targets <= 4);
 
    for (i = 0; i < num_targets; ++i) {
-      const boolean changed = nvc0->tfbbuf[i] != targets[i];
-      const boolean append = (offsets[i] == ((unsigned)-1));
+      const bool changed = nvc0->tfbbuf[i] != targets[i];
+      const bool append = (offsets[i] == ((unsigned)-1));
       if (!changed && append)
          continue;
       nvc0->tfbbuf_dirty |= 1 << i;
@@ -1084,19 +1119,23 @@ nvc0_set_transform_feedback_targets(struct pipe_context *pipe,
          nvc0_so_target_save_offset(pipe, nvc0->tfbbuf[i], i, &serialize);
 
       if (targets[i] && !append)
-         nvc0_so_target(targets[i])->clean = TRUE;
+         nvc0_so_target(targets[i])->clean = true;
 
       pipe_so_target_reference(&nvc0->tfbbuf[i], targets[i]);
    }
    for (; i < nvc0->num_tfbbufs; ++i) {
-      nvc0->tfbbuf_dirty |= 1 << i;
-      nvc0_so_target_save_offset(pipe, nvc0->tfbbuf[i], i, &serialize);
-      pipe_so_target_reference(&nvc0->tfbbuf[i], NULL);
+      if (nvc0->tfbbuf[i]) {
+         nvc0->tfbbuf_dirty |= 1 << i;
+         nvc0_so_target_save_offset(pipe, nvc0->tfbbuf[i], i, &serialize);
+         pipe_so_target_reference(&nvc0->tfbbuf[i], NULL);
+      }
    }
    nvc0->num_tfbbufs = num_targets;
 
-   if (nvc0->tfbbuf_dirty)
-      nvc0->dirty |= NVC0_NEW_TFB_TARGETS;
+   if (nvc0->tfbbuf_dirty) {
+      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_TFB);
+      nvc0->dirty_3d |= NVC0_NEW_3D_TFB_TARGETS;
+   }
 }
 
 static void
@@ -1125,7 +1164,7 @@ nvc0_bind_surfaces_range(struct nvc0_context *nvc0, const unsigned t,
    nvc0->surfaces_dirty[t] |= mask;
 
    if (t == 0)
-      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_SUF);
+      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_SUF);
    else
       nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_SUF);
 }
@@ -1140,17 +1179,178 @@ nvc0_set_compute_resources(struct pipe_context *pipe,
    nvc0_context(pipe)->dirty_cp |= NVC0_NEW_CP_SURFACES;
 }
 
+static bool
+nvc0_bind_images_range(struct nvc0_context *nvc0, const unsigned s,
+                       unsigned start, unsigned nr,
+                       const struct pipe_image_view *pimages)
+{
+   const unsigned end = start + nr;
+   unsigned mask = 0;
+   unsigned i;
+
+   assert(s < 6);
+
+   if (pimages) {
+      for (i = start; i < end; ++i) {
+         struct pipe_image_view *img = &nvc0->images[s][i];
+         const unsigned p = i - start;
+
+         if (img->resource == pimages[p].resource &&
+             img->format == pimages[p].format &&
+             img->access == pimages[p].access) {
+            if (img->resource == NULL)
+               continue;
+            if (img->resource->target == PIPE_BUFFER &&
+                img->u.buf.offset == pimages[p].u.buf.offset &&
+                img->u.buf.size == pimages[p].u.buf.size)
+               continue;
+            if (img->resource->target != PIPE_BUFFER &&
+                img->u.tex.first_layer == pimages[p].u.tex.first_layer &&
+                img->u.tex.last_layer == pimages[p].u.tex.last_layer &&
+                img->u.tex.level == pimages[p].u.tex.level)
+               continue;
+         }
+
+         mask |= (1 << i);
+         if (pimages[p].resource)
+            nvc0->images_valid[s] |= (1 << i);
+         else
+            nvc0->images_valid[s] &= ~(1 << i);
+
+         img->format = pimages[p].format;
+         img->access = pimages[p].access;
+         if (pimages[p].resource && pimages[p].resource->target == PIPE_BUFFER)
+            img->u.buf = pimages[p].u.buf;
+         else
+            img->u.tex = pimages[p].u.tex;
+
+         pipe_resource_reference(
+               &img->resource, pimages[p].resource);
+
+         if (nvc0->screen->base.class_3d >= GM107_3D_CLASS) {
+            if (nvc0->images_tic[s][i]) {
+               struct nv50_tic_entry *old =
+                  nv50_tic_entry(nvc0->images_tic[s][i]);
+               nvc0_screen_tic_unlock(nvc0->screen, old);
+               pipe_sampler_view_reference(&nvc0->images_tic[s][i], NULL);
+            }
+
+            nvc0->images_tic[s][i] =
+               gm107_create_texture_view_from_image(&nvc0->base.pipe,
+                                                    &pimages[p]);
+         }
+      }
+      if (!mask)
+         return false;
+   } else {
+      mask = ((1 << nr) - 1) << start;
+      if (!(nvc0->images_valid[s] & mask))
+         return false;
+      for (i = start; i < end; ++i) {
+         pipe_resource_reference(&nvc0->images[s][i].resource, NULL);
+         if (nvc0->screen->base.class_3d >= GM107_3D_CLASS) {
+            struct nv50_tic_entry *old = nv50_tic_entry(nvc0->images_tic[s][i]);
+            if (old) {
+               nvc0_screen_tic_unlock(nvc0->screen, old);
+               pipe_sampler_view_reference(&nvc0->images_tic[s][i], NULL);
+            }
+         }
+      }
+      nvc0->images_valid[s] &= ~mask;
+   }
+   nvc0->images_dirty[s] |= mask;
+
+   if (s == 5)
+      nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_SUF);
+   else
+      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_SUF);
+
+   return true;
+}
+
+static void
+nvc0_set_shader_images(struct pipe_context *pipe,
+                       enum pipe_shader_type shader,
+                       unsigned start, unsigned nr,
+                       const struct pipe_image_view *images)
+{
+   const unsigned s = nvc0_shader_stage(shader);
+   if (!nvc0_bind_images_range(nvc0_context(pipe), s, start, nr, images))
+      return;
+
+   if (s == 5)
+      nvc0_context(pipe)->dirty_cp |= NVC0_NEW_CP_SURFACES;
+   else
+      nvc0_context(pipe)->dirty_3d |= NVC0_NEW_3D_SURFACES;
+}
+
+static bool
+nvc0_bind_buffers_range(struct nvc0_context *nvc0, const unsigned t,
+                        unsigned start, unsigned nr,
+                        const struct pipe_shader_buffer *pbuffers)
+{
+   const unsigned end = start + nr;
+   unsigned mask = 0;
+   unsigned i;
+
+   assert(t < 6);
+
+   if (pbuffers) {
+      for (i = start; i < end; ++i) {
+         struct pipe_shader_buffer *buf = &nvc0->buffers[t][i];
+         const unsigned p = i - start;
+         if (buf->buffer == pbuffers[p].buffer &&
+             buf->buffer_offset == pbuffers[p].buffer_offset &&
+             buf->buffer_size == pbuffers[p].buffer_size)
+            continue;
+
+         mask |= (1 << i);
+         if (pbuffers[p].buffer)
+            nvc0->buffers_valid[t] |= (1 << i);
+         else
+            nvc0->buffers_valid[t] &= ~(1 << i);
+         buf->buffer_offset = pbuffers[p].buffer_offset;
+         buf->buffer_size = pbuffers[p].buffer_size;
+         pipe_resource_reference(&buf->buffer, pbuffers[p].buffer);
+      }
+      if (!mask)
+         return false;
+   } else {
+      mask = ((1 << nr) - 1) << start;
+      if (!(nvc0->buffers_valid[t] & mask))
+         return false;
+      for (i = start; i < end; ++i)
+         pipe_resource_reference(&nvc0->buffers[t][i].buffer, NULL);
+      nvc0->buffers_valid[t] &= ~mask;
+   }
+   nvc0->buffers_dirty[t] |= mask;
+
+   if (t == 5)
+      nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_BUF);
+   else
+      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_BUF);
+
+   return true;
+}
+
 static void
-nvc0_set_shader_resources(struct pipe_context *pipe,
-                          unsigned start, unsigned nr,
-                          struct pipe_surface **resources)
+nvc0_set_shader_buffers(struct pipe_context *pipe,
+                        enum pipe_shader_type shader,
+                        unsigned start, unsigned nr,
+                        const struct pipe_shader_buffer *buffers,
+                        unsigned writable_bitmask)
 {
-   nvc0_bind_surfaces_range(nvc0_context(pipe), 0, start, nr, resources);
+   const unsigned s = nvc0_shader_stage(shader);
+   if (!nvc0_bind_buffers_range(nvc0_context(pipe), s, start, nr, buffers))
+      return;
 
-   nvc0_context(pipe)->dirty |= NVC0_NEW_SURFACES;
+   if (s == 5)
+      nvc0_context(pipe)->dirty_cp |= NVC0_NEW_CP_BUFFERS;
+   else
+      nvc0_context(pipe)->dirty_3d |= NVC0_NEW_3D_BUFFERS;
 }
 
-static INLINE void
+static inline void
 nvc0_set_global_handle(uint32_t *phandle, struct pipe_resource *res)
 {
    struct nv04_resource *buf = nv04_resource(res);
@@ -1179,12 +1379,18 @@ nvc0_set_global_bindings(struct pipe_context *pipe,
    unsigned i;
    const unsigned end = start + nr;
 
+   if (!nr)
+      return;
+
    if (nvc0->global_residents.size <= (end * sizeof(struct pipe_resource *))) {
       const unsigned old_size = nvc0->global_residents.size;
-      const unsigned req_size = end * sizeof(struct pipe_resource *);
-      util_dynarray_resize(&nvc0->global_residents, req_size);
-      memset((uint8_t *)nvc0->global_residents.data + old_size, 0,
-             req_size - old_size);
+      if (util_dynarray_resize(&nvc0->global_residents, struct pipe_resource *, end)) {
+         memset((uint8_t *)nvc0->global_residents.data + old_size, 0,
+                nvc0->global_residents.size - old_size);
+      } else {
+         NOUVEAU_ERR("Could not resize global residents array\n");
+         return;
+      }
    }
 
    if (resources) {
@@ -1203,7 +1409,7 @@ nvc0_set_global_bindings(struct pipe_context *pipe,
 
    nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_GLOBAL);
 
-   nvc0->dirty_cp = NVC0_NEW_CP_GLOBALS;
+   nvc0->dirty_cp |= NVC0_NEW_CP_GLOBALS;
 }
 
 void
@@ -1234,12 +1440,18 @@ nvc0_init_state_functions(struct nvc0_context *nvc0)
    pipe->create_vs_state = nvc0_vp_state_create;
    pipe->create_fs_state = nvc0_fp_state_create;
    pipe->create_gs_state = nvc0_gp_state_create;
+   pipe->create_tcs_state = nvc0_tcp_state_create;
+   pipe->create_tes_state = nvc0_tep_state_create;
    pipe->bind_vs_state = nvc0_vp_state_bind;
    pipe->bind_fs_state = nvc0_fp_state_bind;
    pipe->bind_gs_state = nvc0_gp_state_bind;
+   pipe->bind_tcs_state = nvc0_tcp_state_bind;
+   pipe->bind_tes_state = nvc0_tep_state_bind;
    pipe->delete_vs_state = nvc0_sp_state_delete;
    pipe->delete_fs_state = nvc0_sp_state_delete;
    pipe->delete_gs_state = nvc0_sp_state_delete;
+   pipe->delete_tcs_state = nvc0_sp_state_delete;
+   pipe->delete_tes_state = nvc0_sp_state_delete;
 
    pipe->create_compute_state = nvc0_cp_state_create;
    pipe->bind_compute_state = nvc0_cp_state_bind;
@@ -1252,16 +1464,18 @@ nvc0_init_state_functions(struct nvc0_context *nvc0)
    pipe->set_min_samples = nvc0_set_min_samples;
    pipe->set_constant_buffer = nvc0_set_constant_buffer;
    pipe->set_framebuffer_state = nvc0_set_framebuffer_state;
+   pipe->set_sample_locations = nvc0_set_sample_locations;
    pipe->set_polygon_stipple = nvc0_set_polygon_stipple;
    pipe->set_scissor_states = nvc0_set_scissor_states;
    pipe->set_viewport_states = nvc0_set_viewport_states;
+   pipe->set_window_rectangles = nvc0_set_window_rectangles;
+   pipe->set_tess_state = nvc0_set_tess_state;
 
    pipe->create_vertex_elements_state = nvc0_vertex_state_create;
    pipe->delete_vertex_elements_state = nvc0_vertex_state_delete;
    pipe->bind_vertex_elements_state = nvc0_vertex_state_bind;
 
    pipe->set_vertex_buffers = nvc0_set_vertex_buffers;
-   pipe->set_index_buffer = nvc0_set_index_buffer;
 
    pipe->create_stream_output_target = nvc0_so_target_create;
    pipe->stream_output_target_destroy = nvc0_so_target_destroy;
@@ -1269,8 +1483,15 @@ nvc0_init_state_functions(struct nvc0_context *nvc0)
 
    pipe->set_global_binding = nvc0_set_global_bindings;
    pipe->set_compute_resources = nvc0_set_compute_resources;
-   pipe->set_shader_resources = nvc0_set_shader_resources;
+   pipe->set_shader_images = nvc0_set_shader_images;
+   pipe->set_shader_buffers = nvc0_set_shader_buffers;
 
    nvc0->sample_mask = ~0;
    nvc0->min_samples = 1;
+   nvc0->default_tess_outer[0] =
+   nvc0->default_tess_outer[1] =
+   nvc0->default_tess_outer[2] =
+   nvc0->default_tess_outer[3] = 1.0;
+   nvc0->default_tess_inner[0] =
+   nvc0->default_tess_inner[1] = 1.0;
 }