gallium: change pipe_sampler_view::first_element/last_element -> offset/size
[mesa.git] / src / gallium / drivers / ilo / ilo_state.c
index caad308ab3883d9bcf2697d4ed32a1e305f0152f..7693133a1bf4cf43cc76d924f95a8119497de4e3 100644 (file)
  *    Chia-I Wu <olv@lunarg.com>
  */
 
+#include "util/u_dual_blend.h"
+#include "util/u_dynarray.h"
 #include "util/u_framebuffer.h"
 #include "util/u_helpers.h"
+#include "util/u_resource.h"
 #include "util/u_upload_mgr.h"
 
 #include "ilo_context.h"
+#include "ilo_format.h"
 #include "ilo_resource.h"
 #include "ilo_shader.h"
 #include "ilo_state.h"
 
+/**
+ * Translate a pipe primitive type to the matching hardware primitive type.
+ */
+static enum gen_3dprim_type
+ilo_translate_draw_mode(unsigned mode)
+{
+   static const enum gen_3dprim_type prim_mapping[PIPE_PRIM_MAX] = {
+      [PIPE_PRIM_POINTS]                     = GEN6_3DPRIM_POINTLIST,
+      [PIPE_PRIM_LINES]                      = GEN6_3DPRIM_LINELIST,
+      [PIPE_PRIM_LINE_LOOP]                  = GEN6_3DPRIM_LINELOOP,
+      [PIPE_PRIM_LINE_STRIP]                 = GEN6_3DPRIM_LINESTRIP,
+      [PIPE_PRIM_TRIANGLES]                  = GEN6_3DPRIM_TRILIST,
+      [PIPE_PRIM_TRIANGLE_STRIP]             = GEN6_3DPRIM_TRISTRIP,
+      [PIPE_PRIM_TRIANGLE_FAN]               = GEN6_3DPRIM_TRIFAN,
+      [PIPE_PRIM_QUADS]                      = GEN6_3DPRIM_QUADLIST,
+      [PIPE_PRIM_QUAD_STRIP]                 = GEN6_3DPRIM_QUADSTRIP,
+      [PIPE_PRIM_POLYGON]                    = GEN6_3DPRIM_POLYGON,
+      [PIPE_PRIM_LINES_ADJACENCY]            = GEN6_3DPRIM_LINELIST_ADJ,
+      [PIPE_PRIM_LINE_STRIP_ADJACENCY]       = GEN6_3DPRIM_LINESTRIP_ADJ,
+      [PIPE_PRIM_TRIANGLES_ADJACENCY]        = GEN6_3DPRIM_TRILIST_ADJ,
+      [PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY]   = GEN6_3DPRIM_TRISTRIP_ADJ,
+   };
+
+   assert(prim_mapping[mode]);
+
+   return prim_mapping[mode];
+}
+
+static enum gen_index_format
+ilo_translate_index_size(unsigned index_size)
+{
+   switch (index_size) {
+   case 1:                             return GEN6_INDEX_BYTE;
+   case 2:                             return GEN6_INDEX_WORD;
+   case 4:                             return GEN6_INDEX_DWORD;
+   default:
+      assert(!"unknown index size");
+      return GEN6_INDEX_BYTE;
+   }
+}
+
+static enum gen_mip_filter
+ilo_translate_mip_filter(unsigned filter)
+{
+   switch (filter) {
+   case PIPE_TEX_MIPFILTER_NEAREST:    return GEN6_MIPFILTER_NEAREST;
+   case PIPE_TEX_MIPFILTER_LINEAR:     return GEN6_MIPFILTER_LINEAR;
+   case PIPE_TEX_MIPFILTER_NONE:       return GEN6_MIPFILTER_NONE;
+   default:
+      assert(!"unknown mipfilter");
+      return GEN6_MIPFILTER_NONE;
+   }
+}
+
+static int
+ilo_translate_img_filter(unsigned filter)
+{
+   switch (filter) {
+   case PIPE_TEX_FILTER_NEAREST:       return GEN6_MAPFILTER_NEAREST;
+   case PIPE_TEX_FILTER_LINEAR:        return GEN6_MAPFILTER_LINEAR;
+   default:
+      assert(!"unknown sampler filter");
+      return GEN6_MAPFILTER_NEAREST;
+   }
+}
+
+static enum gen_texcoord_mode
+ilo_translate_address_wrap(unsigned wrap)
+{
+   switch (wrap) {
+   case PIPE_TEX_WRAP_CLAMP:           return GEN8_TEXCOORDMODE_HALF_BORDER;
+   case PIPE_TEX_WRAP_REPEAT:          return GEN6_TEXCOORDMODE_WRAP;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:   return GEN6_TEXCOORDMODE_CLAMP;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER: return GEN6_TEXCOORDMODE_CLAMP_BORDER;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:   return GEN6_TEXCOORDMODE_MIRROR;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+   default:
+      assert(!"unknown sampler wrap mode");
+      return GEN6_TEXCOORDMODE_WRAP;
+   }
+}
+
+static enum gen_aniso_ratio
+ilo_translate_max_anisotropy(unsigned max_anisotropy)
+{
+   switch (max_anisotropy) {
+   case 0: case 1: case 2:             return GEN6_ANISORATIO_2;
+   case 3: case 4:                     return GEN6_ANISORATIO_4;
+   case 5: case 6:                     return GEN6_ANISORATIO_6;
+   case 7: case 8:                     return GEN6_ANISORATIO_8;
+   case 9: case 10:                    return GEN6_ANISORATIO_10;
+   case 11: case 12:                   return GEN6_ANISORATIO_12;
+   case 13: case 14:                   return GEN6_ANISORATIO_14;
+   default:                            return GEN6_ANISORATIO_16;
+   }
+}
+
+static enum gen_prefilter_op
+ilo_translate_shadow_func(unsigned func)
+{
+   /*
+    * For PIPE_FUNC_x, the reference value is on the left-hand side of the
+    * comparison, and 1.0 is returned when the comparison is true.
+    *
+    * For GEN6_PREFILTEROP_x, the reference value is on the right-hand side of
+    * the comparison, and 0.0 is returned when the comparison is true.
+    */
+   switch (func) {
+   case PIPE_FUNC_NEVER:               return GEN6_PREFILTEROP_ALWAYS;
+   case PIPE_FUNC_LESS:                return GEN6_PREFILTEROP_LEQUAL;
+   case PIPE_FUNC_EQUAL:               return GEN6_PREFILTEROP_NOTEQUAL;
+   case PIPE_FUNC_LEQUAL:              return GEN6_PREFILTEROP_LESS;
+   case PIPE_FUNC_GREATER:             return GEN6_PREFILTEROP_GEQUAL;
+   case PIPE_FUNC_NOTEQUAL:            return GEN6_PREFILTEROP_EQUAL;
+   case PIPE_FUNC_GEQUAL:              return GEN6_PREFILTEROP_GREATER;
+   case PIPE_FUNC_ALWAYS:              return GEN6_PREFILTEROP_NEVER;
+   default:
+      assert(!"unknown shadow compare function");
+      return GEN6_PREFILTEROP_NEVER;
+   }
+}
+
+static enum gen_front_winding
+ilo_translate_front_ccw(unsigned front_ccw)
+{
+   return (front_ccw) ? GEN6_FRONTWINDING_CCW : GEN6_FRONTWINDING_CW;
+}
+
+static enum gen_cull_mode
+ilo_translate_cull_face(unsigned cull_face)
+{
+   switch (cull_face) {
+   case PIPE_FACE_NONE:                return GEN6_CULLMODE_NONE;
+   case PIPE_FACE_FRONT:               return GEN6_CULLMODE_FRONT;
+   case PIPE_FACE_BACK:                return GEN6_CULLMODE_BACK;
+   case PIPE_FACE_FRONT_AND_BACK:      return GEN6_CULLMODE_BOTH;
+   default:
+      assert(!"unknown face culling");
+      return GEN6_CULLMODE_NONE;
+   }
+}
+
+static enum gen_fill_mode
+ilo_translate_poly_mode(unsigned poly_mode)
+{
+   switch (poly_mode) {
+   case PIPE_POLYGON_MODE_FILL:        return GEN6_FILLMODE_SOLID;
+   case PIPE_POLYGON_MODE_LINE:        return GEN6_FILLMODE_WIREFRAME;
+   case PIPE_POLYGON_MODE_POINT:       return GEN6_FILLMODE_POINT;
+   default:
+      assert(!"unknown polygon mode");
+      return GEN6_FILLMODE_SOLID;
+   }
+}
+
+static enum gen_pixel_location
+ilo_translate_half_pixel_center(bool half_pixel_center)
+{
+   return (half_pixel_center) ? GEN6_PIXLOC_CENTER : GEN6_PIXLOC_UL_CORNER;
+}
+
+static enum gen_compare_function
+ilo_translate_compare_func(unsigned func)
+{
+   switch (func) {
+   case PIPE_FUNC_NEVER:               return GEN6_COMPAREFUNCTION_NEVER;
+   case PIPE_FUNC_LESS:                return GEN6_COMPAREFUNCTION_LESS;
+   case PIPE_FUNC_EQUAL:               return GEN6_COMPAREFUNCTION_EQUAL;
+   case PIPE_FUNC_LEQUAL:              return GEN6_COMPAREFUNCTION_LEQUAL;
+   case PIPE_FUNC_GREATER:             return GEN6_COMPAREFUNCTION_GREATER;
+   case PIPE_FUNC_NOTEQUAL:            return GEN6_COMPAREFUNCTION_NOTEQUAL;
+   case PIPE_FUNC_GEQUAL:              return GEN6_COMPAREFUNCTION_GEQUAL;
+   case PIPE_FUNC_ALWAYS:              return GEN6_COMPAREFUNCTION_ALWAYS;
+   default:
+      assert(!"unknown compare function");
+      return GEN6_COMPAREFUNCTION_NEVER;
+   }
+}
+
+static enum gen_stencil_op
+ilo_translate_stencil_op(unsigned stencil_op)
+{
+   switch (stencil_op) {
+   case PIPE_STENCIL_OP_KEEP:          return GEN6_STENCILOP_KEEP;
+   case PIPE_STENCIL_OP_ZERO:          return GEN6_STENCILOP_ZERO;
+   case PIPE_STENCIL_OP_REPLACE:       return GEN6_STENCILOP_REPLACE;
+   case PIPE_STENCIL_OP_INCR:          return GEN6_STENCILOP_INCRSAT;
+   case PIPE_STENCIL_OP_DECR:          return GEN6_STENCILOP_DECRSAT;
+   case PIPE_STENCIL_OP_INCR_WRAP:     return GEN6_STENCILOP_INCR;
+   case PIPE_STENCIL_OP_DECR_WRAP:     return GEN6_STENCILOP_DECR;
+   case PIPE_STENCIL_OP_INVERT:        return GEN6_STENCILOP_INVERT;
+   default:
+      assert(!"unknown stencil op");
+      return GEN6_STENCILOP_KEEP;
+   }
+}
+
+static enum gen_logic_op
+ilo_translate_logicop(unsigned logicop)
+{
+   switch (logicop) {
+   case PIPE_LOGICOP_CLEAR:            return GEN6_LOGICOP_CLEAR;
+   case PIPE_LOGICOP_NOR:              return GEN6_LOGICOP_NOR;
+   case PIPE_LOGICOP_AND_INVERTED:     return GEN6_LOGICOP_AND_INVERTED;
+   case PIPE_LOGICOP_COPY_INVERTED:    return GEN6_LOGICOP_COPY_INVERTED;
+   case PIPE_LOGICOP_AND_REVERSE:      return GEN6_LOGICOP_AND_REVERSE;
+   case PIPE_LOGICOP_INVERT:           return GEN6_LOGICOP_INVERT;
+   case PIPE_LOGICOP_XOR:              return GEN6_LOGICOP_XOR;
+   case PIPE_LOGICOP_NAND:             return GEN6_LOGICOP_NAND;
+   case PIPE_LOGICOP_AND:              return GEN6_LOGICOP_AND;
+   case PIPE_LOGICOP_EQUIV:            return GEN6_LOGICOP_EQUIV;
+   case PIPE_LOGICOP_NOOP:             return GEN6_LOGICOP_NOOP;
+   case PIPE_LOGICOP_OR_INVERTED:      return GEN6_LOGICOP_OR_INVERTED;
+   case PIPE_LOGICOP_COPY:             return GEN6_LOGICOP_COPY;
+   case PIPE_LOGICOP_OR_REVERSE:       return GEN6_LOGICOP_OR_REVERSE;
+   case PIPE_LOGICOP_OR:               return GEN6_LOGICOP_OR;
+   case PIPE_LOGICOP_SET:              return GEN6_LOGICOP_SET;
+   default:
+      assert(!"unknown logicop function");
+      return GEN6_LOGICOP_CLEAR;
+   }
+}
+
+static int
+ilo_translate_blend_func(unsigned blend)
+{
+   switch (blend) {
+   case PIPE_BLEND_ADD:                return GEN6_BLENDFUNCTION_ADD;
+   case PIPE_BLEND_SUBTRACT:           return GEN6_BLENDFUNCTION_SUBTRACT;
+   case PIPE_BLEND_REVERSE_SUBTRACT:   return GEN6_BLENDFUNCTION_REVERSE_SUBTRACT;
+   case PIPE_BLEND_MIN:                return GEN6_BLENDFUNCTION_MIN;
+   case PIPE_BLEND_MAX:                return GEN6_BLENDFUNCTION_MAX;
+   default:
+      assert(!"unknown blend function");
+      return GEN6_BLENDFUNCTION_ADD;
+   }
+}
+
+static int
+ilo_translate_blend_factor(unsigned factor)
+{
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ONE:                return GEN6_BLENDFACTOR_ONE;
+   case PIPE_BLENDFACTOR_SRC_COLOR:          return GEN6_BLENDFACTOR_SRC_COLOR;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:          return GEN6_BLENDFACTOR_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_DST_ALPHA:          return GEN6_BLENDFACTOR_DST_ALPHA;
+   case PIPE_BLENDFACTOR_DST_COLOR:          return GEN6_BLENDFACTOR_DST_COLOR;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: return GEN6_BLENDFACTOR_SRC_ALPHA_SATURATE;
+   case PIPE_BLENDFACTOR_CONST_COLOR:        return GEN6_BLENDFACTOR_CONST_COLOR;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:        return GEN6_BLENDFACTOR_CONST_ALPHA;
+   case PIPE_BLENDFACTOR_SRC1_COLOR:         return GEN6_BLENDFACTOR_SRC1_COLOR;
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:         return GEN6_BLENDFACTOR_SRC1_ALPHA;
+   case PIPE_BLENDFACTOR_ZERO:               return GEN6_BLENDFACTOR_ZERO;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:      return GEN6_BLENDFACTOR_INV_SRC_COLOR;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:      return GEN6_BLENDFACTOR_INV_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:      return GEN6_BLENDFACTOR_INV_DST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:      return GEN6_BLENDFACTOR_INV_DST_COLOR;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:    return GEN6_BLENDFACTOR_INV_CONST_COLOR;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:    return GEN6_BLENDFACTOR_INV_CONST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:     return GEN6_BLENDFACTOR_INV_SRC1_COLOR;
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:     return GEN6_BLENDFACTOR_INV_SRC1_ALPHA;
+   default:
+      assert(!"unknown blend factor");
+      return GEN6_BLENDFACTOR_ONE;
+   }
+}
+
 static void
-finalize_shader_states(struct ilo_context *ilo)
+finalize_shader_states(struct ilo_state_vector *vec)
 {
    unsigned type;
 
@@ -45,15 +318,15 @@ finalize_shader_states(struct ilo_context *ilo)
 
       switch (type) {
       case PIPE_SHADER_VERTEX:
-         shader = ilo->vs;
+         shader = vec->vs;
          state = ILO_DIRTY_VS;
          break;
       case PIPE_SHADER_GEOMETRY:
-         shader = ilo->gs;
+         shader = vec->gs;
          state = ILO_DIRTY_GS;
          break;
       case PIPE_SHADER_FRAGMENT:
-         shader = ilo->fs;
+         shader = vec->fs;
          state = ILO_DIRTY_FS;
          break;
       default:
@@ -66,20 +339,20 @@ finalize_shader_states(struct ilo_context *ilo)
          continue;
 
       /* compile if the shader or the states it depends on changed */
-      if (ilo->dirty & state) {
-         ilo_shader_select_kernel(shader, ilo, ILO_DIRTY_ALL);
+      if (vec->dirty & state) {
+         ilo_shader_select_kernel(shader, vec, ILO_DIRTY_ALL);
       }
-      else if (ilo_shader_select_kernel(shader, ilo, ilo->dirty)) {
+      else if (ilo_shader_select_kernel(shader, vec, vec->dirty)) {
          /* mark the state dirty if a new kernel is selected */
-         ilo->dirty |= state;
+         vec->dirty |= state;
       }
 
       /* need to setup SBE for FS */
-      if (type == PIPE_SHADER_FRAGMENT && ilo->dirty &
+      if (type == PIPE_SHADER_FRAGMENT && vec->dirty &
             (state | ILO_DIRTY_GS | ILO_DIRTY_VS | ILO_DIRTY_RASTERIZER)) {
-         if (ilo_shader_select_kernel_routing(shader,
-               (ilo->gs) ? ilo->gs : ilo->vs, ilo->rasterizer))
-            ilo->dirty |= state;
+         if (ilo_shader_select_kernel_sbe(shader,
+               (vec->gs) ? vec->gs : vec->vs, vec->rasterizer))
+            vec->dirty |= state;
       }
    }
 }
@@ -96,7 +369,6 @@ finalize_cbuf_state(struct ilo_context *ilo,
       ~ilo_shader_get_kernel_param(sh, ILO_KERNEL_SKIP_CBUF0_UPLOAD);
 
    while (upload_mask) {
-      const enum pipe_format elem_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
       unsigned offset, i;
 
       i = u_bit_scan(&upload_mask);
@@ -104,83 +376,406 @@ finalize_cbuf_state(struct ilo_context *ilo,
       if (cbuf->cso[i].resource)
          continue;
 
-      u_upload_data(ilo->uploader, 0, cbuf->cso[i].user_buffer_size,
+      u_upload_data(ilo->uploader, 0, cbuf->cso[i].info.size, 16,
             cbuf->cso[i].user_buffer, &offset, &cbuf->cso[i].resource);
 
-      ilo_gpe_init_view_surface_for_buffer(ilo->dev,
-            ilo_buffer(cbuf->cso[i].resource),
-            offset, cbuf->cso[i].user_buffer_size,
-            util_format_get_blocksize(elem_format), elem_format,
-            false, false, &cbuf->cso[i].surface);
+      cbuf->cso[i].info.vma = ilo_resource_get_vma(cbuf->cso[i].resource);
+      cbuf->cso[i].info.offset = offset;
+
+      memset(&cbuf->cso[i].surface, 0, sizeof(cbuf->cso[i].surface));
+      ilo_state_surface_init_for_buffer(&cbuf->cso[i].surface,
+            ilo->dev, &cbuf->cso[i].info);
 
-      ilo->dirty |= ILO_DIRTY_CBUF;
+      ilo->state_vector.dirty |= ILO_DIRTY_CBUF;
    }
 }
 
 static void
 finalize_constant_buffers(struct ilo_context *ilo)
 {
-   if (ilo->dirty & (ILO_DIRTY_CBUF | ILO_DIRTY_VS))
-      finalize_cbuf_state(ilo, &ilo->cbuf[PIPE_SHADER_VERTEX], ilo->vs);
+   struct ilo_state_vector *vec = &ilo->state_vector;
 
-   if (ilo->dirty & (ILO_DIRTY_CBUF | ILO_DIRTY_FS))
-      finalize_cbuf_state(ilo, &ilo->cbuf[PIPE_SHADER_FRAGMENT], ilo->fs);
+   if (vec->dirty & (ILO_DIRTY_CBUF | ILO_DIRTY_VS))
+      finalize_cbuf_state(ilo, &vec->cbuf[PIPE_SHADER_VERTEX], vec->vs);
+
+   if (ilo->state_vector.dirty & (ILO_DIRTY_CBUF | ILO_DIRTY_FS))
+      finalize_cbuf_state(ilo, &vec->cbuf[PIPE_SHADER_FRAGMENT], vec->fs);
 }
 
 static void
 finalize_index_buffer(struct ilo_context *ilo)
 {
-   const bool need_upload = (ilo->draw->indexed &&
-         (ilo->ib.user_buffer || ilo->ib.offset % ilo->ib.index_size));
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   const bool need_upload = (vec->draw->indexed &&
+         (vec->ib.state.user_buffer ||
+          vec->ib.state.offset % vec->ib.state.index_size));
    struct pipe_resource *current_hw_res = NULL;
+   struct ilo_state_index_buffer_info info;
+   int64_t vertex_start_bias = 0;
 
-   if (!(ilo->dirty & ILO_DIRTY_IB) && !need_upload)
+   if (!(vec->dirty & ILO_DIRTY_IB) && !need_upload)
       return;
 
-   pipe_resource_reference(&current_hw_res, ilo->ib.hw_resource);
+   /* make sure vec->ib.hw_resource changes when reallocated */
+   pipe_resource_reference(&current_hw_res, vec->ib.hw_resource);
 
    if (need_upload) {
-      const unsigned offset = ilo->ib.index_size * ilo->draw->start;
-      const unsigned size = ilo->ib.index_size * ilo->draw->count;
+      const unsigned offset = vec->ib.state.index_size * vec->draw->start;
+      const unsigned size = vec->ib.state.index_size * vec->draw->count;
       unsigned hw_offset;
 
-      if (ilo->ib.user_buffer) {
-         u_upload_data(ilo->uploader, 0, size,
-               ilo->ib.user_buffer + offset, &hw_offset, &ilo->ib.hw_resource);
-      }
-      else {
-         u_upload_buffer(ilo->uploader, 0, ilo->ib.offset + offset, size,
-               ilo->ib.buffer, &hw_offset, &ilo->ib.hw_resource);
+      if (vec->ib.state.user_buffer) {
+         u_upload_data(ilo->uploader, 0, size, 16,
+               vec->ib.state.user_buffer + offset,
+               &hw_offset, &vec->ib.hw_resource);
+      } else {
+         u_upload_buffer(ilo->uploader, 0,
+               vec->ib.state.offset + offset, size, 16, vec->ib.state.buffer,
+               &hw_offset, &vec->ib.hw_resource);
       }
 
       /* the HW offset should be aligned */
-      assert(hw_offset % ilo->ib.index_size == 0);
-      ilo->ib.draw_start_offset = hw_offset / ilo->ib.index_size;
+      assert(hw_offset % vec->ib.state.index_size == 0);
+      vertex_start_bias = hw_offset / vec->ib.state.index_size;
 
       /*
-       * INDEX[ilo->draw->start] in the original buffer is INDEX[0] in the HW
+       * INDEX[vec->draw->start] in the original buffer is INDEX[0] in the HW
        * resource
        */
-      ilo->ib.draw_start_offset -= ilo->draw->start;
-   }
-   else {
-      pipe_resource_reference(&ilo->ib.hw_resource, ilo->ib.buffer);
+      vertex_start_bias -= vec->draw->start;
+   } else {
+      pipe_resource_reference(&vec->ib.hw_resource, vec->ib.state.buffer);
 
       /* note that index size may be zero when the draw is not indexed */
-      if (ilo->draw->indexed)
-         ilo->ib.draw_start_offset = ilo->ib.offset / ilo->ib.index_size;
-      else
-         ilo->ib.draw_start_offset = 0;
+      if (vec->draw->indexed)
+         vertex_start_bias = vec->ib.state.offset / vec->ib.state.index_size;
    }
 
+   vec->draw_info.vertex_start += vertex_start_bias;
+
    /* treat the IB as clean if the HW states do not change */
-   if (ilo->ib.hw_resource == current_hw_res &&
-       ilo->ib.hw_index_size == ilo->ib.index_size)
-      ilo->dirty &= ~ILO_DIRTY_IB;
+   if (vec->ib.hw_resource == current_hw_res &&
+       vec->ib.hw_index_size == vec->ib.state.index_size)
+      vec->dirty &= ~ILO_DIRTY_IB;
    else
-      ilo->ib.hw_index_size = ilo->ib.index_size;
+      vec->ib.hw_index_size = vec->ib.state.index_size;
 
    pipe_resource_reference(&current_hw_res, NULL);
+
+   memset(&info, 0, sizeof(info));
+   if (vec->ib.hw_resource) {
+      info.vma = ilo_resource_get_vma(vec->ib.hw_resource);
+      info.size = info.vma->vm_size;
+      info.format = ilo_translate_index_size(vec->ib.hw_index_size);
+   }
+
+   ilo_state_index_buffer_set_info(&vec->ib.ib, dev, &info);
+}
+
+static void
+finalize_vertex_elements(struct ilo_context *ilo)
+{
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   struct ilo_ve_state *ve = vec->ve;
+   const bool last_element_edge_flag = (vec->vs &&
+         ilo_shader_get_kernel_param(vec->vs, ILO_KERNEL_VS_INPUT_EDGEFLAG));
+   const bool prepend_vertexid = (vec->vs &&
+         ilo_shader_get_kernel_param(vec->vs, ILO_KERNEL_VS_INPUT_VERTEXID));
+   const bool prepend_instanceid = (vec->vs &&
+         ilo_shader_get_kernel_param(vec->vs,
+            ILO_KERNEL_VS_INPUT_INSTANCEID));
+   const enum gen_index_format index_format = (vec->draw->indexed) ?
+      ilo_translate_index_size(vec->ib.state.index_size) : GEN6_INDEX_DWORD;
+
+   /* check for non-orthogonal states */
+   if (ve->vf_params.cv_topology != vec->draw_info.topology ||
+       ve->vf_params.prepend_vertexid != prepend_vertexid ||
+       ve->vf_params.prepend_instanceid != prepend_instanceid ||
+       ve->vf_params.last_element_edge_flag != last_element_edge_flag ||
+       ve->vf_params.cv_index_format != index_format ||
+       ve->vf_params.cut_index_enable != vec->draw->primitive_restart ||
+       ve->vf_params.cut_index != vec->draw->restart_index) {
+      ve->vf_params.cv_topology = vec->draw_info.topology;
+      ve->vf_params.prepend_vertexid = prepend_vertexid;
+      ve->vf_params.prepend_instanceid = prepend_instanceid;
+      ve->vf_params.last_element_edge_flag = last_element_edge_flag;
+      ve->vf_params.cv_index_format = index_format;
+      ve->vf_params.cut_index_enable = vec->draw->primitive_restart;
+      ve->vf_params.cut_index = vec->draw->restart_index;
+
+      ilo_state_vf_set_params(&ve->vf, dev, &ve->vf_params);
+
+      vec->dirty |= ILO_DIRTY_VE;
+   }
+}
+
+static void
+finalize_vertex_buffers(struct ilo_context *ilo)
+{
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   struct ilo_state_vertex_buffer_info info;
+   unsigned i;
+
+   if (!(vec->dirty & (ILO_DIRTY_VE | ILO_DIRTY_VB)))
+      return;
+
+   memset(&info, 0, sizeof(info));
+
+   for (i = 0; i < vec->ve->vb_count; i++) {
+      const unsigned pipe_idx = vec->ve->vb_mapping[i];
+      const struct pipe_vertex_buffer *cso = &vec->vb.states[pipe_idx];
+
+      if (cso->buffer) {
+         info.vma = ilo_resource_get_vma(cso->buffer);
+         info.offset = cso->buffer_offset;
+         info.size = info.vma->vm_size - cso->buffer_offset;
+
+         info.stride = cso->stride;
+      } else {
+         memset(&info, 0, sizeof(info));
+      }
+
+      ilo_state_vertex_buffer_set_info(&vec->vb.vb[i], dev, &info);
+   }
+}
+
+static void
+finalize_urb(struct ilo_context *ilo)
+{
+   const uint16_t attr_size = sizeof(uint32_t) * 4;
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   struct ilo_state_urb_info info;
+
+   if (!(vec->dirty & (ILO_DIRTY_VE | ILO_DIRTY_VS |
+                       ILO_DIRTY_GS | ILO_DIRTY_FS)))
+      return;
+
+   memset(&info, 0, sizeof(info));
+
+   info.ve_entry_size = attr_size * ilo_state_vf_get_attr_count(&vec->ve->vf);
+
+   if (vec->vs) {
+      info.vs_const_data = (bool)
+         (ilo_shader_get_kernel_param(vec->vs, ILO_KERNEL_PCB_CBUF0_SIZE) +
+          ilo_shader_get_kernel_param(vec->vs, ILO_KERNEL_VS_PCB_UCP_SIZE));
+      info.vs_entry_size = attr_size *
+         ilo_shader_get_kernel_param(vec->vs, ILO_KERNEL_OUTPUT_COUNT);
+   }
+
+   if (vec->gs) {
+      info.gs_const_data = (bool)
+         ilo_shader_get_kernel_param(vec->gs, ILO_KERNEL_PCB_CBUF0_SIZE);
+
+      /*
+       * From the Ivy Bridge PRM, volume 2 part 1, page 189:
+       *
+       *     "All outputs of a GS thread will be stored in the single GS
+       *      thread output URB entry."
+       *
+       * TODO
+       */
+      info.gs_entry_size = attr_size *
+         ilo_shader_get_kernel_param(vec->gs, ILO_KERNEL_OUTPUT_COUNT);
+   }
+
+   if (vec->fs) {
+      info.ps_const_data = (bool)
+         ilo_shader_get_kernel_param(vec->fs, ILO_KERNEL_PCB_CBUF0_SIZE);
+   }
+
+   ilo_state_urb_set_info(&vec->urb, dev, &info);
+}
+
+static void
+finalize_viewport(struct ilo_context *ilo)
+{
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+
+   if (vec->dirty & ILO_DIRTY_VIEWPORT) {
+      ilo_state_viewport_set_params(&vec->viewport.vp,
+            dev, &vec->viewport.params, false);
+   } else if (vec->dirty & ILO_DIRTY_SCISSOR) {
+      ilo_state_viewport_set_params(&vec->viewport.vp,
+            dev, &vec->viewport.params, true);
+      vec->dirty |= ILO_DIRTY_VIEWPORT;
+   }
+}
+
+static bool
+can_enable_gb_test(const struct ilo_rasterizer_state *rasterizer,
+                   const struct ilo_viewport_state *viewport,
+                   const struct ilo_fb_state *fb)
+{
+   unsigned i;
+
+   /*
+    * There are several reasons that guard band test should be disabled
+    *
+    *  - GL wide points (to avoid partially visibie object)
+    *  - GL wide or AA lines (to avoid partially visibie object)
+    *  - missing 2D clipping
+    */
+   if (rasterizer->state.point_size_per_vertex ||
+       rasterizer->state.point_size > 1.0f ||
+       rasterizer->state.line_width > 1.0f ||
+       rasterizer->state.line_smooth)
+      return false;
+
+   for (i = 0; i < viewport->params.count; i++) {
+      const struct ilo_state_viewport_matrix_info *mat =
+         &viewport->matrices[i];
+      float min_x, max_x, min_y, max_y;
+
+      min_x = -1.0f * fabsf(mat->scale[0]) + mat->translate[0];
+      max_x =  1.0f * fabsf(mat->scale[0]) + mat->translate[0];
+      min_y = -1.0f * fabsf(mat->scale[1]) + mat->translate[1];
+      max_y =  1.0f * fabsf(mat->scale[1]) + mat->translate[1];
+
+      if (min_x > 0.0f || max_x < fb->state.width ||
+          min_y > 0.0f || max_y < fb->state.height)
+         return false;
+   }
+
+   return true;
+}
+
+static void
+finalize_rasterizer(struct ilo_context *ilo)
+{
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   struct ilo_rasterizer_state *rasterizer = vec->rasterizer;
+   struct ilo_state_raster_info *info = &vec->rasterizer->info;
+   const bool gb_test_enable =
+      can_enable_gb_test(rasterizer, &vec->viewport, &vec->fb);
+   const bool multisample =
+      (rasterizer->state.multisample && vec->fb.num_samples > 1);
+   const uint8_t barycentric_interps = ilo_shader_get_kernel_param(vec->fs,
+         ILO_KERNEL_FS_BARYCENTRIC_INTERPOLATIONS);
+
+   /* check for non-orthogonal states */
+   if (info->clip.viewport_count != vec->viewport.params.count ||
+       info->clip.gb_test_enable != gb_test_enable ||
+       info->setup.msaa_enable != multisample ||
+       info->setup.line_msaa_enable != multisample ||
+       info->tri.depth_offset_format != vec->fb.depth_offset_format ||
+       info->scan.sample_count != vec->fb.num_samples ||
+       info->scan.sample_mask != vec->sample_mask ||
+       info->scan.barycentric_interps != barycentric_interps ||
+       info->params.any_integer_rt != vec->fb.has_integer_rt ||
+       info->params.hiz_enable != vec->fb.has_hiz) {
+      info->clip.viewport_count = vec->viewport.params.count;
+      info->clip.gb_test_enable = gb_test_enable;
+      info->setup.msaa_enable = multisample;
+      info->setup.line_msaa_enable = multisample;
+      info->tri.depth_offset_format = vec->fb.depth_offset_format;
+      info->scan.sample_count = vec->fb.num_samples;
+      info->scan.sample_mask = vec->sample_mask;
+      info->scan.barycentric_interps = barycentric_interps;
+      info->params.any_integer_rt = vec->fb.has_integer_rt;
+      info->params.hiz_enable = vec->fb.has_hiz;
+
+      ilo_state_raster_set_info(&rasterizer->rs, dev, &rasterizer->info);
+
+      vec->dirty |= ILO_DIRTY_RASTERIZER;
+   }
+}
+
+static bool
+finalize_blend_rt(struct ilo_context *ilo)
+{
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   const struct ilo_fb_state *fb = &vec->fb;
+   struct ilo_blend_state *blend = vec->blend;
+   struct ilo_state_cc_blend_info *info = &vec->blend->info.blend;
+   bool changed = false;
+   unsigned i;
+
+   if (!(vec->dirty & (ILO_DIRTY_FB | ILO_DIRTY_BLEND)))
+      return false;
+
+   /* set up one for dummy RT writes */
+   if (!fb->state.nr_cbufs) {
+      if (info->rt != &blend->dummy_rt) {
+         info->rt = &blend->dummy_rt;
+         info->rt_count = 1;
+         changed = true;
+      }
+
+      return changed;
+   }
+
+   if (info->rt != blend->effective_rt ||
+       info->rt_count != fb->state.nr_cbufs) {
+      info->rt = blend->effective_rt;
+      info->rt_count = fb->state.nr_cbufs;
+      changed = true;
+   }
+
+   for (i = 0; i < fb->state.nr_cbufs; i++) {
+      const struct ilo_fb_blend_caps *caps = &fb->blend_caps[i];
+      struct ilo_state_cc_blend_rt_info *rt = &blend->effective_rt[i];
+      /* ignore logicop when not UNORM */
+      const bool logicop_enable =
+         (blend->rt[i].logicop_enable && caps->is_unorm);
+
+      if (rt->cv_is_unorm != caps->is_unorm ||
+          rt->cv_is_integer != caps->is_integer ||
+          rt->logicop_enable != logicop_enable ||
+          rt->force_dst_alpha_one != caps->force_dst_alpha_one) {
+         rt->cv_is_unorm = caps->is_unorm;
+         rt->cv_is_integer = caps->is_integer;
+         rt->logicop_enable = logicop_enable;
+         rt->force_dst_alpha_one = caps->force_dst_alpha_one;
+
+         changed = true;
+      }
+   }
+
+   return changed;
+}
+
+static void
+finalize_blend(struct ilo_context *ilo)
+{
+   const struct ilo_dev *dev = ilo->dev;
+   struct ilo_state_vector *vec = &ilo->state_vector;
+   struct ilo_blend_state *blend = vec->blend;
+   struct ilo_state_cc_info *info = &blend->info;
+   const bool sample_count_one = (vec->fb.num_samples <= 1);
+   const bool float_source0_alpha =
+      (!vec->fb.state.nr_cbufs || !vec->fb.state.cbufs[0] ||
+       !util_format_is_pure_integer(vec->fb.state.cbufs[0]->format));
+
+   /* check for non-orthogonal states */
+   if (finalize_blend_rt(ilo) ||
+       info->alpha.cv_sample_count_one != sample_count_one ||
+       info->alpha.cv_float_source0_alpha != float_source0_alpha ||
+       info->alpha.test_enable != vec->dsa->alpha_test ||
+       info->alpha.test_func != vec->dsa->alpha_func ||
+       memcmp(&info->stencil, &vec->dsa->stencil, sizeof(info->stencil)) ||
+       memcmp(&info->depth, &vec->dsa->depth, sizeof(info->depth)) ||
+       memcmp(&info->params, &vec->cc_params, sizeof(info->params))) {
+      info->alpha.cv_sample_count_one = sample_count_one;
+      info->alpha.cv_float_source0_alpha = float_source0_alpha;
+      info->alpha.test_enable = vec->dsa->alpha_test;
+      info->alpha.test_func = vec->dsa->alpha_func;
+      info->stencil = vec->dsa->stencil;
+      info->depth = vec->dsa->depth;
+      info->params = vec->cc_params;
+
+      ilo_state_cc_set_info(&blend->cc, dev, info);
+
+      blend->alpha_may_kill = (info->alpha.alpha_to_coverage ||
+                               info->alpha.test_enable);
+
+      vec->dirty |= ILO_DIRTY_BLEND;
+   }
 }
 
 /**
@@ -191,26 +786,142 @@ void
 ilo_finalize_3d_states(struct ilo_context *ilo,
                        const struct pipe_draw_info *draw)
 {
-   ilo->draw = draw;
+   ilo->state_vector.draw = draw;
+
+   ilo->state_vector.draw_info.topology = ilo_translate_draw_mode(draw->mode);
+   ilo->state_vector.draw_info.indexed = draw->indexed;
+   ilo->state_vector.draw_info.vertex_count = draw->count;
+   ilo->state_vector.draw_info.vertex_start = draw->start;
+   ilo->state_vector.draw_info.instance_count = draw->instance_count;
+   ilo->state_vector.draw_info.instance_start = draw->start_instance;
+   ilo->state_vector.draw_info.vertex_base = draw->index_bias;
 
-   finalize_shader_states(ilo);
+   finalize_blend(ilo);
+   finalize_shader_states(&ilo->state_vector);
    finalize_constant_buffers(ilo);
    finalize_index_buffer(ilo);
+   finalize_vertex_elements(ilo);
+   finalize_vertex_buffers(ilo);
+
+   finalize_urb(ilo);
+   finalize_rasterizer(ilo);
+   finalize_viewport(ilo);
 
    u_upload_unmap(ilo->uploader);
 }
 
+static void
+finalize_global_binding(struct ilo_state_vector *vec)
+{
+   struct ilo_shader_state *cs = vec->cs;
+   int base, count, shift;
+   int i;
+
+   count = ilo_shader_get_kernel_param(cs,
+         ILO_KERNEL_CS_SURFACE_GLOBAL_COUNT);
+   if (!count)
+      return;
+
+   base = ilo_shader_get_kernel_param(cs, ILO_KERNEL_CS_SURFACE_GLOBAL_BASE);
+   shift = 32 - util_last_bit(base + count - 1);
+
+   if (count > vec->global_binding.count)
+      count = vec->global_binding.count;
+
+   for (i = 0; i < count; i++) {
+      struct ilo_global_binding_cso *cso =
+         util_dynarray_element(&vec->global_binding.bindings,
+               struct ilo_global_binding_cso, i);
+      const uint32_t offset = *cso->handle & ((1 << shift) - 1);
+
+      *cso->handle = ((base + i) << shift) | offset;
+   }
+}
+
+void
+ilo_finalize_compute_states(struct ilo_context *ilo)
+{
+   finalize_global_binding(&ilo->state_vector);
+}
+
 static void *
 ilo_create_blend_state(struct pipe_context *pipe,
                        const struct pipe_blend_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_cc_info *info;
    struct ilo_blend_state *blend;
+   int i;
 
-   blend = MALLOC_STRUCT(ilo_blend_state);
+   blend = CALLOC_STRUCT(ilo_blend_state);
    assert(blend);
 
-   ilo_gpe_init_blend(ilo->dev, state, blend);
+   info = &blend->info;
+
+   info->alpha.cv_float_source0_alpha = true;
+   info->alpha.cv_sample_count_one = true;
+   info->alpha.alpha_to_one = state->alpha_to_one;
+   info->alpha.alpha_to_coverage = state->alpha_to_coverage;
+   info->alpha.test_enable = false;
+   info->alpha.test_func = GEN6_COMPAREFUNCTION_ALWAYS;
+
+   info->stencil.cv_has_buffer = true;
+   info->depth.cv_has_buffer= true;
+
+   info->blend.rt = blend->effective_rt;
+   info->blend.rt_count = 1;
+   info->blend.dither_enable = state->dither;
+
+   for (i = 0; i < ARRAY_SIZE(blend->rt); i++) {
+      const struct pipe_rt_blend_state *rt = &state->rt[i];
+      struct ilo_state_cc_blend_rt_info *rt_info = &blend->rt[i];
+
+      rt_info->cv_has_buffer = true;
+      rt_info->cv_is_unorm = true;
+      rt_info->cv_is_integer = false;
+
+      /* logic op takes precedence over blending */
+      if (state->logicop_enable) {
+         rt_info->logicop_enable = true;
+         rt_info->logicop_func = ilo_translate_logicop(state->logicop_func);
+      } else if (rt->blend_enable) {
+         rt_info->blend_enable = true;
+
+         rt_info->rgb_src = ilo_translate_blend_factor(rt->rgb_src_factor);
+         rt_info->rgb_dst = ilo_translate_blend_factor(rt->rgb_dst_factor);
+         rt_info->rgb_func = ilo_translate_blend_func(rt->rgb_func);
+
+         rt_info->a_src = ilo_translate_blend_factor(rt->alpha_src_factor);
+         rt_info->a_dst = ilo_translate_blend_factor(rt->alpha_dst_factor);
+         rt_info->a_func = ilo_translate_blend_func(rt->alpha_func);
+      }
+
+      if (!(rt->colormask & PIPE_MASK_A))
+         rt_info->argb_write_disables |= (1 << 3);
+      if (!(rt->colormask & PIPE_MASK_R))
+         rt_info->argb_write_disables |= (1 << 2);
+      if (!(rt->colormask & PIPE_MASK_G))
+         rt_info->argb_write_disables |= (1 << 1);
+      if (!(rt->colormask & PIPE_MASK_B))
+         rt_info->argb_write_disables |= (1 << 0);
+
+      if (!state->independent_blend_enable) {
+         for (i = 1; i < ARRAY_SIZE(blend->rt); i++)
+            blend->rt[i] = *rt_info;
+         break;
+      }
+   }
+
+   memcpy(blend->effective_rt, blend->rt, sizeof(blend->rt));
+
+   blend->dummy_rt.argb_write_disables = 0xf;
+
+   if (!ilo_state_cc_init(&blend->cc, dev, &blend->info)) {
+      FREE(blend);
+      return NULL;
+   }
+
+   blend->dual_blend = util_blend_state_is_dual(state, 0);
 
    return blend;
 }
@@ -218,11 +929,11 @@ ilo_create_blend_state(struct pipe_context *pipe,
 static void
 ilo_bind_blend_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->blend = state;
+   vec->blend = state;
 
-   ilo->dirty |= ILO_DIRTY_BLEND;
+   vec->dirty |= ILO_DIRTY_BLEND;
 }
 
 static void
@@ -235,13 +946,107 @@ static void *
 ilo_create_sampler_state(struct pipe_context *pipe,
                          const struct pipe_sampler_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
    struct ilo_sampler_cso *sampler;
+   struct ilo_state_sampler_info info;
+   struct ilo_state_sampler_border_info border;
 
-   sampler = MALLOC_STRUCT(ilo_sampler_cso);
+   sampler = CALLOC_STRUCT(ilo_sampler_cso);
    assert(sampler);
 
-   ilo_gpe_init_sampler_cso(ilo->dev, state, sampler);
+   memset(&info, 0, sizeof(info));
+
+   info.non_normalized = !state->normalized_coords;
+   if (state->normalized_coords) {
+      info.lod_bias = state->lod_bias;
+      info.min_lod = state->min_lod;
+      info.max_lod = state->max_lod;
+
+      info.mip_filter = ilo_translate_mip_filter(state->min_mip_filter);
+   } else {
+      /* work around a bug in util_blitter */
+      info.mip_filter = GEN6_MIPFILTER_NONE;
+   }
+
+   if (state->max_anisotropy) {
+      info.min_filter = GEN6_MAPFILTER_ANISOTROPIC;
+      info.mag_filter = GEN6_MAPFILTER_ANISOTROPIC;
+   } else {
+      info.min_filter = ilo_translate_img_filter(state->min_img_filter);
+      info.mag_filter = ilo_translate_img_filter(state->mag_img_filter);
+   }
+
+   info.max_anisotropy = ilo_translate_max_anisotropy(state->max_anisotropy);
+
+   /* use LOD 0 when no mipmapping (see sampler_set_gen6_SAMPLER_STATE()) */
+   if (info.mip_filter == GEN6_MIPFILTER_NONE && info.min_lod > 0.0f) {
+      info.min_lod = 0.0f;
+      info.mag_filter = info.min_filter;
+   }
+
+   if (state->seamless_cube_map) {
+      if (state->min_img_filter == PIPE_TEX_FILTER_NEAREST ||
+          state->mag_img_filter == PIPE_TEX_FILTER_NEAREST) {
+         info.tcx_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+         info.tcy_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+         info.tcz_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+      } else {
+         info.tcx_ctrl = GEN6_TEXCOORDMODE_CUBE;
+         info.tcy_ctrl = GEN6_TEXCOORDMODE_CUBE;
+         info.tcz_ctrl = GEN6_TEXCOORDMODE_CUBE;
+      }
+   } else {
+      info.tcx_ctrl = ilo_translate_address_wrap(state->wrap_s);
+      info.tcy_ctrl = ilo_translate_address_wrap(state->wrap_t);
+      info.tcz_ctrl = ilo_translate_address_wrap(state->wrap_r);
+
+      if (ilo_dev_gen(dev) < ILO_GEN(8)) {
+         /*
+          * For nearest filtering, PIPE_TEX_WRAP_CLAMP means
+          * PIPE_TEX_WRAP_CLAMP_TO_EDGE;  for linear filtering,
+          * PIPE_TEX_WRAP_CLAMP means PIPE_TEX_WRAP_CLAMP_TO_BORDER while
+          * additionally clamping the texture coordinates to [0.0, 1.0].
+          *
+          * PIPE_TEX_WRAP_CLAMP is not supported natively until Gen8.  The
+          * clamping has to be taken care of in the shaders.  There are two
+          * filters here, but let the minification one has a say.
+          */
+         const bool clamp_is_to_edge =
+            (state->min_img_filter == PIPE_TEX_FILTER_NEAREST);
+
+         if (clamp_is_to_edge) {
+            if (info.tcx_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER)
+               info.tcx_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+            if (info.tcy_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER)
+               info.tcy_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+            if (info.tcz_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER)
+               info.tcz_ctrl = GEN6_TEXCOORDMODE_CLAMP;
+         } else {
+            if (info.tcx_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER) {
+               info.tcx_ctrl = GEN6_TEXCOORDMODE_CLAMP_BORDER;
+               sampler->saturate_s = true;
+            }
+            if (info.tcy_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER) {
+               info.tcy_ctrl = GEN6_TEXCOORDMODE_CLAMP_BORDER;
+               sampler->saturate_t = true;
+            }
+            if (info.tcz_ctrl == GEN8_TEXCOORDMODE_HALF_BORDER) {
+               info.tcz_ctrl = GEN6_TEXCOORDMODE_CLAMP_BORDER;
+               sampler->saturate_r = true;
+            }
+         }
+      }
+   }
+
+   if (state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE)
+      info.shadow_func = ilo_translate_shadow_func(state->compare_func);
+
+   ilo_state_sampler_init(&sampler->sampler, dev, &info);
+
+   memset(&border, 0, sizeof(border));
+   memcpy(border.rgba.f, state->border_color.f, sizeof(border.rgba.f));
+
+   ilo_state_sampler_border_init(&sampler->border, dev, &border);
 
    return sampler;
 }
@@ -250,12 +1055,12 @@ static void
 ilo_bind_sampler_states(struct pipe_context *pipe, unsigned shader,
                         unsigned start, unsigned count, void **samplers)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_sampler_state *dst = &ilo->sampler[shader];
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_sampler_state *dst = &vec->sampler[shader];
    bool changed = false;
    unsigned i;
 
-   assert(start + count <= Elements(dst->cso));
+   assert(start + count <= ARRAY_SIZE(dst->cso));
 
    if (samplers) {
       for (i = 0; i < count; i++) {
@@ -277,114 +1082,24 @@ ilo_bind_sampler_states(struct pipe_context *pipe, unsigned shader,
          dst->cso[start + i] = NULL;
    }
 
-   if (dst->count <= start + count) {
-      if (samplers)
-         count += start;
-      else
-         count = start;
-
-      while (count > 0 && !dst->cso[count - 1])
-         count--;
-
-      dst->count = count;
-   }
-
    if (changed) {
       switch (shader) {
       case PIPE_SHADER_VERTEX:
-         ilo->dirty |= ILO_DIRTY_SAMPLER_VS;
+         vec->dirty |= ILO_DIRTY_SAMPLER_VS;
          break;
       case PIPE_SHADER_GEOMETRY:
-         ilo->dirty |= ILO_DIRTY_SAMPLER_GS;
+         vec->dirty |= ILO_DIRTY_SAMPLER_GS;
          break;
       case PIPE_SHADER_FRAGMENT:
-         ilo->dirty |= ILO_DIRTY_SAMPLER_FS;
+         vec->dirty |= ILO_DIRTY_SAMPLER_FS;
          break;
       case PIPE_SHADER_COMPUTE:
-         ilo->dirty |= ILO_DIRTY_SAMPLER_CS;
+         vec->dirty |= ILO_DIRTY_SAMPLER_CS;
          break;
       }
    }
 }
 
-static void
-ilo_bind_fragment_sampler_states(struct pipe_context *pipe,
-                                 unsigned num_samplers,
-                                 void **samplers)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_bind_sampler_states(pipe, PIPE_SHADER_FRAGMENT,
-         0, num_samplers, samplers);
-
-   if (ilo->sampler[PIPE_SHADER_FRAGMENT].count > num_samplers) {
-      ilo_bind_sampler_states(pipe, PIPE_SHADER_FRAGMENT, num_samplers,
-            ilo->sampler[PIPE_SHADER_FRAGMENT].count - num_samplers, NULL);
-   }
-}
-
-static void
-ilo_bind_vertex_sampler_states(struct pipe_context *pipe,
-                               unsigned num_samplers,
-                               void **samplers)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_bind_sampler_states(pipe, PIPE_SHADER_VERTEX,
-         0, num_samplers, samplers);
-
-   if (ilo->sampler[PIPE_SHADER_VERTEX].count > num_samplers) {
-      ilo_bind_sampler_states(pipe, PIPE_SHADER_VERTEX, num_samplers,
-            ilo->sampler[PIPE_SHADER_VERTEX].count - num_samplers, NULL);
-   }
-}
-
-static void
-ilo_bind_geometry_sampler_states(struct pipe_context *pipe,
-                                 unsigned num_samplers,
-                                 void **samplers)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_bind_sampler_states(pipe, PIPE_SHADER_GEOMETRY,
-            0, num_samplers, samplers);
-
-   if (ilo->sampler[PIPE_SHADER_GEOMETRY].count > num_samplers) {
-      ilo_bind_sampler_states(pipe, PIPE_SHADER_GEOMETRY, num_samplers,
-            ilo->sampler[PIPE_SHADER_GEOMETRY].count - num_samplers, NULL);
-   }
-}
-
-static void
-ilo_bind_compute_sampler_states(struct pipe_context *pipe,
-                                unsigned start_slot,
-                                unsigned num_samplers,
-                                void **samplers)
-{
-   ilo_bind_sampler_states(pipe, PIPE_SHADER_COMPUTE,
-            start_slot, num_samplers, samplers);
-}
-
-static void
-ilo_bind_sampler_states2(struct pipe_context *pipe, unsigned shader,
-                         unsigned start, unsigned count, void **samplers)
-{
-   switch (shader) {
-   case PIPE_SHADER_VERTEX:
-      ilo_bind_vertex_sampler_states(pipe, count, samplers);
-      break;
-   case PIPE_SHADER_GEOMETRY:
-      ilo_bind_geometry_sampler_states(pipe, count, samplers);
-      break;
-   case PIPE_SHADER_FRAGMENT:
-      ilo_bind_fragment_sampler_states(pipe, count, samplers);
-      break;
-   case PIPE_SHADER_COMPUTE:
-      ilo_bind_compute_sampler_states(pipe, start, count, samplers);
-      break;
-   }
-}
-
 static void
 ilo_delete_sampler_state(struct pipe_context *pipe, void *state)
 {
@@ -395,14 +1110,76 @@ static void *
 ilo_create_rasterizer_state(struct pipe_context *pipe,
                             const struct pipe_rasterizer_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
    struct ilo_rasterizer_state *rast;
+   struct ilo_state_raster_info *info;
 
-   rast = MALLOC_STRUCT(ilo_rasterizer_state);
+   rast = CALLOC_STRUCT(ilo_rasterizer_state);
    assert(rast);
 
    rast->state = *state;
-   ilo_gpe_init_rasterizer(ilo->dev, state, rast);
+
+   info = &rast->info;
+
+   info->clip.clip_enable = true;
+   info->clip.stats_enable = true;
+   info->clip.viewport_count = 1;
+   info->clip.force_rtaindex_zero = true;
+   info->clip.user_clip_enables = state->clip_plane_enable;
+   info->clip.gb_test_enable = true;
+   info->clip.xy_test_enable = true;
+   info->clip.z_far_enable = state->depth_clip;
+   info->clip.z_near_enable = state->depth_clip;
+   info->clip.z_near_zero = state->clip_halfz;
+
+   info->setup.first_vertex_provoking = state->flatshade_first;
+   info->setup.viewport_transform = true;
+   info->setup.scissor_enable = state->scissor;
+   info->setup.msaa_enable = false;
+   info->setup.line_msaa_enable = false;
+   info->point.aa_enable = state->point_smooth;
+   info->point.programmable_width = state->point_size_per_vertex;
+   info->line.aa_enable = state->line_smooth;
+   info->line.stipple_enable = state->line_stipple_enable;
+   info->line.giq_enable = true;
+   info->line.giq_last_pixel = state->line_last_pixel;
+   info->tri.front_winding = ilo_translate_front_ccw(state->front_ccw);
+   info->tri.cull_mode = ilo_translate_cull_face(state->cull_face);
+   info->tri.fill_mode_front = ilo_translate_poly_mode(state->fill_front);
+   info->tri.fill_mode_back = ilo_translate_poly_mode(state->fill_back);
+   info->tri.depth_offset_format = GEN6_ZFORMAT_D24_UNORM_X8_UINT;
+   info->tri.depth_offset_solid = state->offset_tri;
+   info->tri.depth_offset_wireframe = state->offset_line;
+   info->tri.depth_offset_point = state->offset_point;
+   info->tri.poly_stipple_enable = state->poly_stipple_enable;
+
+   info->scan.stats_enable = true;
+   info->scan.sample_count = 1;
+   info->scan.pixloc =
+      ilo_translate_half_pixel_center(state->half_pixel_center);
+   info->scan.sample_mask = ~0u;
+   info->scan.zw_interp = GEN6_ZW_INTERP_PIXEL;
+   info->scan.barycentric_interps = GEN6_INTERP_PERSPECTIVE_PIXEL;
+   info->scan.earlyz_control = GEN7_EDSC_NORMAL;
+   info->scan.earlyz_op = ILO_STATE_RASTER_EARLYZ_NORMAL;
+   info->scan.earlyz_stencil_clear = false;
+
+   info->params.any_integer_rt = false;
+   info->params.hiz_enable = true;
+   info->params.point_width =
+      (state->point_size == 0.0f) ? 1.0f : state->point_size;
+   info->params.line_width =
+      (state->line_width == 0.0f) ? 1.0f : state->line_width;
+
+   info->params.depth_offset_scale = state->offset_scale;
+   /*
+    * Scale the constant term.  The minimum representable value used by the HW
+    * is not large enouch to be the minimum resolvable difference.
+    */
+   info->params.depth_offset_const = state->offset_units * 2.0f;
+   info->params.depth_offset_clamp = state->offset_clamp;
+
+   ilo_state_raster_init(&rast->rs, dev, info);
 
    return rast;
 }
@@ -410,11 +1187,21 @@ ilo_create_rasterizer_state(struct pipe_context *pipe,
 static void
 ilo_bind_rasterizer_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+
+   vec->rasterizer = state;
 
-   ilo->rasterizer = state;
+   if (vec->rasterizer) {
+      struct ilo_state_line_stipple_info info;
 
-   ilo->dirty |= ILO_DIRTY_RASTERIZER;
+      info.pattern = vec->rasterizer->state.line_stipple_pattern;
+      info.repeat_count = vec->rasterizer->state.line_stipple_factor + 1;
+
+      ilo_state_line_stipple_set_info(&vec->line_stipple, dev, &info);
+   }
+
+   vec->dirty |= ILO_DIRTY_RASTERIZER;
 }
 
 static void
@@ -427,13 +1214,48 @@ static void *
 ilo_create_depth_stencil_alpha_state(struct pipe_context *pipe,
                                      const struct pipe_depth_stencil_alpha_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
    struct ilo_dsa_state *dsa;
+   int i;
 
-   dsa = MALLOC_STRUCT(ilo_dsa_state);
+   dsa = CALLOC_STRUCT(ilo_dsa_state);
    assert(dsa);
 
-   ilo_gpe_init_dsa(ilo->dev, state, dsa);
+   dsa->depth.cv_has_buffer = true;
+   dsa->depth.test_enable = state->depth.enabled;
+   dsa->depth.write_enable = state->depth.writemask;
+   dsa->depth.test_func = ilo_translate_compare_func(state->depth.func);
+
+   dsa->stencil.cv_has_buffer = true;
+   for (i = 0; i < ARRAY_SIZE(state->stencil); i++) {
+      const struct pipe_stencil_state *stencil = &state->stencil[i];
+      struct ilo_state_cc_stencil_op_info *op;
+
+      if (!stencil->enabled)
+         break;
+
+      if (i == 0) {
+         dsa->stencil.test_enable = true;
+         dsa->stencil_front.test_mask = stencil->valuemask;
+         dsa->stencil_front.write_mask = stencil->writemask;
+
+         op = &dsa->stencil.front;
+      } else {
+         dsa->stencil.twosided_enable = true;
+         dsa->stencil_back.test_mask = stencil->valuemask;
+         dsa->stencil_back.write_mask = stencil->writemask;
+
+         op = &dsa->stencil.back;
+      }
+
+      op->test_func = ilo_translate_compare_func(stencil->func);
+      op->fail_op = ilo_translate_stencil_op(stencil->fail_op);
+      op->zfail_op = ilo_translate_stencil_op(stencil->zfail_op);
+      op->zpass_op = ilo_translate_stencil_op(stencil->zpass_op);
+   }
+
+   dsa->alpha_test = state->alpha.enabled;
+   dsa->alpha_ref = state->alpha.ref_value;
+   dsa->alpha_func = ilo_translate_compare_func(state->alpha.func);
 
    return dsa;
 }
@@ -441,11 +1263,22 @@ ilo_create_depth_stencil_alpha_state(struct pipe_context *pipe,
 static void
 ilo_bind_depth_stencil_alpha_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo->dsa = state;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+
+   vec->dsa = state;
+   if (vec->dsa) {
+      vec->cc_params.alpha_ref = vec->dsa->alpha_ref;
+      vec->cc_params.stencil_front.test_mask =
+         vec->dsa->stencil_front.test_mask;
+      vec->cc_params.stencil_front.write_mask =
+         vec->dsa->stencil_front.write_mask;
+      vec->cc_params.stencil_back.test_mask =
+         vec->dsa->stencil_back.test_mask;
+      vec->cc_params.stencil_back.write_mask =
+         vec->dsa->stencil_back.write_mask;
+   }
 
-   ilo->dirty |= ILO_DIRTY_DSA;
+   vec->dirty |= ILO_DIRTY_DSA;
 }
 
 static void
@@ -461,7 +1294,7 @@ ilo_create_fs_state(struct pipe_context *pipe,
    struct ilo_context *ilo = ilo_context(pipe);
    struct ilo_shader_state *shader;
 
-   shader = ilo_shader_create_fs(ilo->dev, state, ilo);
+   shader = ilo_shader_create_fs(ilo->dev, state, &ilo->state_vector);
    assert(shader);
 
    ilo_shader_cache_add(ilo->shader_cache, shader);
@@ -472,11 +1305,11 @@ ilo_create_fs_state(struct pipe_context *pipe,
 static void
 ilo_bind_fs_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->fs = state;
+   vec->fs = state;
 
-   ilo->dirty |= ILO_DIRTY_FS;
+   vec->dirty |= ILO_DIRTY_FS;
 }
 
 static void
@@ -496,7 +1329,7 @@ ilo_create_vs_state(struct pipe_context *pipe,
    struct ilo_context *ilo = ilo_context(pipe);
    struct ilo_shader_state *shader;
 
-   shader = ilo_shader_create_vs(ilo->dev, state, ilo);
+   shader = ilo_shader_create_vs(ilo->dev, state, &ilo->state_vector);
    assert(shader);
 
    ilo_shader_cache_add(ilo->shader_cache, shader);
@@ -507,11 +1340,11 @@ ilo_create_vs_state(struct pipe_context *pipe,
 static void
 ilo_bind_vs_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->vs = state;
+   vec->vs = state;
 
-   ilo->dirty |= ILO_DIRTY_VS;
+   vec->dirty |= ILO_DIRTY_VS;
 }
 
 static void
@@ -531,7 +1364,7 @@ ilo_create_gs_state(struct pipe_context *pipe,
    struct ilo_context *ilo = ilo_context(pipe);
    struct ilo_shader_state *shader;
 
-   shader = ilo_shader_create_gs(ilo->dev, state, ilo);
+   shader = ilo_shader_create_gs(ilo->dev, state, &ilo->state_vector);
    assert(shader);
 
    ilo_shader_cache_add(ilo->shader_cache, shader);
@@ -542,15 +1375,15 @@ ilo_create_gs_state(struct pipe_context *pipe,
 static void
 ilo_bind_gs_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
    /* util_blitter may set this unnecessarily */
-   if (ilo->gs == state)
+   if (vec->gs == state)
       return;
 
-   ilo->gs = state;
+   vec->gs = state;
 
-   ilo->dirty |= ILO_DIRTY_GS;
+   vec->dirty |= ILO_DIRTY_GS;
 }
 
 static void
@@ -568,13 +1401,61 @@ ilo_create_vertex_elements_state(struct pipe_context *pipe,
                                  unsigned num_elements,
                                  const struct pipe_vertex_element *elements)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_vf_element_info vf_elements[PIPE_MAX_ATTRIBS];
+   unsigned instance_divisors[PIPE_MAX_ATTRIBS];
+   struct ilo_state_vf_info vf_info;
    struct ilo_ve_state *ve;
+   unsigned i;
 
-   ve = MALLOC_STRUCT(ilo_ve_state);
+   ve = CALLOC_STRUCT(ilo_ve_state);
    assert(ve);
 
-   ilo_gpe_init_ve(ilo->dev, num_elements, elements, ve);
+   for (i = 0; i < num_elements; i++) {
+      const struct pipe_vertex_element *elem = &elements[i];
+      struct ilo_state_vf_element_info *attr = &vf_elements[i];
+      unsigned hw_idx;
+
+      /*
+       * map the pipe vb to the hardware vb, which has a fixed instance
+       * divisor
+       */
+      for (hw_idx = 0; hw_idx < ve->vb_count; hw_idx++) {
+         if (ve->vb_mapping[hw_idx] == elem->vertex_buffer_index &&
+             instance_divisors[hw_idx] == elem->instance_divisor)
+            break;
+      }
+
+      /* create one if there is no matching hardware vb */
+      if (hw_idx >= ve->vb_count) {
+         hw_idx = ve->vb_count++;
+
+         ve->vb_mapping[hw_idx] = elem->vertex_buffer_index;
+         instance_divisors[hw_idx] = elem->instance_divisor;
+      }
+
+      attr->buffer = hw_idx;
+      attr->vertex_offset = elem->src_offset;
+      attr->format = ilo_format_translate_vertex(dev, elem->src_format);
+      attr->format_size = util_format_get_blocksize(elem->src_format);
+      attr->component_count = util_format_get_nr_components(elem->src_format);
+      attr->is_integer = util_format_is_pure_integer(elem->src_format);
+
+      attr->instancing_enable = (elem->instance_divisor != 0);
+      attr->instancing_step_rate = elem->instance_divisor;
+   }
+
+   memset(&vf_info, 0, sizeof(vf_info));
+   vf_info.data = ve->vf_data;
+   vf_info.data_size = sizeof(ve->vf_data);
+   vf_info.elements = vf_elements;
+   vf_info.element_count = num_elements;
+   /* vf_info.params and ve->vf_params are both zeroed */
+
+   if (!ilo_state_vf_init(&ve->vf, dev, &vf_info)) {
+      FREE(ve);
+      return NULL;
+   }
 
    return ve;
 }
@@ -582,11 +1463,11 @@ ilo_create_vertex_elements_state(struct pipe_context *pipe,
 static void
 ilo_bind_vertex_elements_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->ve = state;
+   vec->ve = state;
 
-   ilo->dirty |= ILO_DIRTY_VE;
+   vec->dirty |= ILO_DIRTY_VE;
 }
 
 static void
@@ -601,66 +1482,70 @@ static void
 ilo_set_blend_color(struct pipe_context *pipe,
                     const struct pipe_blend_color *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->blend_color = *state;
+   memcpy(vec->cc_params.blend_rgba, state->color, sizeof(state->color));
 
-   ilo->dirty |= ILO_DIRTY_BLEND_COLOR;
+   vec->dirty |= ILO_DIRTY_BLEND_COLOR;
 }
 
 static void
 ilo_set_stencil_ref(struct pipe_context *pipe,
                     const struct pipe_stencil_ref *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
    /* util_blitter may set this unnecessarily */
-   if (!memcmp(&ilo->stencil_ref, state, sizeof(*state)))
+   if (!memcmp(&vec->stencil_ref, state, sizeof(*state)))
       return;
 
-   ilo->stencil_ref = *state;
+   vec->stencil_ref = *state;
+
+   vec->cc_params.stencil_front.test_ref = state->ref_value[0];
+   vec->cc_params.stencil_back.test_ref = state->ref_value[1];
 
-   ilo->dirty |= ILO_DIRTY_STENCIL_REF;
+   vec->dirty |= ILO_DIRTY_STENCIL_REF;
 }
 
 static void
 ilo_set_sample_mask(struct pipe_context *pipe,
                     unsigned sample_mask)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
    /* util_blitter may set this unnecessarily */
-   if (ilo->sample_mask == sample_mask)
+   if (vec->sample_mask == sample_mask)
       return;
 
-   ilo->sample_mask = sample_mask;
+   vec->sample_mask = sample_mask;
 
-   ilo->dirty |= ILO_DIRTY_SAMPLE_MASK;
+   vec->dirty |= ILO_DIRTY_SAMPLE_MASK;
 }
 
 static void
 ilo_set_clip_state(struct pipe_context *pipe,
                    const struct pipe_clip_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->clip = *state;
+   vec->clip = *state;
 
-   ilo->dirty |= ILO_DIRTY_CLIP;
+   vec->dirty |= ILO_DIRTY_CLIP;
 }
 
 static void
 ilo_set_constant_buffer(struct pipe_context *pipe,
                         uint shader, uint index,
-                        struct pipe_constant_buffer *buf)
+                        const struct pipe_constant_buffer *buf)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_cbuf_state *cbuf = &ilo->cbuf[shader];
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_cbuf_state *cbuf = &vec->cbuf[shader];
    const unsigned count = 1;
    unsigned i;
 
-   assert(shader < Elements(ilo->cbuf));
-   assert(index + count <= Elements(ilo->cbuf[shader].cso));
+   assert(shader < ARRAY_SIZE(vec->cbuf));
+   assert(index + count <= ARRAY_SIZE(vec->cbuf[shader].cso));
 
    if (buf) {
       for (i = 0; i < count; i++) {
@@ -668,85 +1553,185 @@ ilo_set_constant_buffer(struct pipe_context *pipe,
 
          pipe_resource_reference(&cso->resource, buf[i].buffer);
 
+         cso->info.access = ILO_STATE_SURFACE_ACCESS_DP_DATA;
+         cso->info.format = GEN6_FORMAT_R32G32B32A32_FLOAT;
+         cso->info.format_size = 16;
+         cso->info.struct_size = 16;
+         cso->info.readonly = true;
+         cso->info.size = buf[i].buffer_size;
+
          if (buf[i].buffer) {
-            const enum pipe_format elem_format =
-               PIPE_FORMAT_R32G32B32A32_FLOAT;
+            cso->info.vma = ilo_resource_get_vma(buf[i].buffer);
+            cso->info.offset = buf[i].buffer_offset;
 
-            ilo_gpe_init_view_surface_for_buffer(ilo->dev,
-                  ilo_buffer(buf[i].buffer),
-                  buf[i].buffer_offset, buf[i].buffer_size,
-                  util_format_get_blocksize(elem_format), elem_format,
-                  false, false, &cso->surface);
+            memset(&cso->surface, 0, sizeof(cso->surface));
+            ilo_state_surface_init_for_buffer(&cso->surface, dev, &cso->info);
 
             cso->user_buffer = NULL;
-            cso->user_buffer_size = 0;
 
             cbuf->enabled_mask |= 1 << (index + i);
-         }
-         else if (buf[i].user_buffer) {
-            cso->surface.bo = NULL;
-
+         } else if (buf[i].user_buffer) {
+            cso->info.vma = NULL;
             /* buffer_offset does not apply for user buffer */
             cso->user_buffer = buf[i].user_buffer;
-            cso->user_buffer_size = buf[i].buffer_size;
 
             cbuf->enabled_mask |= 1 << (index + i);
-         }
-         else {
-            cso->surface.bo = NULL;
+         } else {
+            cso->info.vma = NULL;
+            cso->info.size = 0;
             cso->user_buffer = NULL;
-            cso->user_buffer_size = 0;
 
             cbuf->enabled_mask &= ~(1 << (index + i));
          }
       }
-   }
-   else {
+   } else {
       for (i = 0; i < count; i++) {
          struct ilo_cbuf_cso *cso = &cbuf->cso[index + i];
 
          pipe_resource_reference(&cso->resource, NULL);
-         cso->surface.bo = NULL;
+
+         cso->info.vma = NULL;
+         cso->info.size = 0;
          cso->user_buffer = NULL;
-         cso->user_buffer_size = 0;
 
          cbuf->enabled_mask &= ~(1 << (index + i));
       }
    }
 
-   ilo->dirty |= ILO_DIRTY_CBUF;
+   vec->dirty |= ILO_DIRTY_CBUF;
+}
+
+static void
+fb_set_blend_caps(const struct ilo_dev *dev,
+                  enum pipe_format format,
+                  struct ilo_fb_blend_caps *caps)
+{
+   const struct util_format_description *desc =
+      util_format_description(format);
+   const int ch = util_format_get_first_non_void_channel(format);
+
+   memset(caps, 0, sizeof(*caps));
+
+   if (format == PIPE_FORMAT_NONE || desc->is_mixed)
+      return;
+
+   caps->is_unorm = (ch >= 0 && desc->channel[ch].normalized &&
+         desc->channel[ch].type == UTIL_FORMAT_TYPE_UNSIGNED &&
+         desc->colorspace == UTIL_FORMAT_COLORSPACE_RGB);
+   caps->is_integer = util_format_is_pure_integer(format);
+
+   /*
+    * From the Sandy Bridge PRM, volume 2 part 1, page 365:
+    *
+    *     "Logic Ops are only supported on *_UNORM surfaces (excluding _SRGB
+    *      variants), otherwise Logic Ops must be DISABLED."
+    *
+    * According to the classic driver, this is lifted on Gen8+.
+    */
+   caps->can_logicop = (ilo_dev_gen(dev) >= ILO_GEN(8) || caps->is_unorm);
+
+   /* no blending for pure integer formats */
+   caps->can_blend = !caps->is_integer;
+
+   /*
+    * From the Sandy Bridge PRM, volume 2 part 1, page 382:
+    *
+    *     "Alpha Test can only be enabled if Pixel Shader outputs a float
+    *      alpha value."
+    */
+   caps->can_alpha_test = !caps->is_integer;
+
+   caps->force_dst_alpha_one =
+      (ilo_format_translate_render(dev, format) !=
+       ilo_format_translate_color(dev, format));
+
+   /* sanity check */
+   if (caps->force_dst_alpha_one) {
+      enum pipe_format render_format;
+
+      switch (format) {
+      case PIPE_FORMAT_B8G8R8X8_UNORM:
+         render_format = PIPE_FORMAT_B8G8R8A8_UNORM;
+         break;
+      default:
+         render_format = PIPE_FORMAT_NONE;
+         break;
+      }
+
+      assert(ilo_format_translate_render(dev, format) ==
+             ilo_format_translate_color(dev, render_format));
+   }
 }
 
 static void
 ilo_set_framebuffer_state(struct pipe_context *pipe,
                           const struct pipe_framebuffer_state *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_fb_state *fb = &vec->fb;
+   const struct pipe_surface *first_surf = NULL;
+   int i;
 
-   util_copy_framebuffer_state(&ilo->fb.state, state);
+   util_copy_framebuffer_state(&fb->state, state);
 
-   if (state->nr_cbufs)
-      ilo->fb.num_samples = state->cbufs[0]->texture->nr_samples;
-   else if (state->zsbuf)
-      ilo->fb.num_samples = state->zsbuf->texture->nr_samples;
-   else
-      ilo->fb.num_samples = 1;
+   fb->has_integer_rt = false;
+   for (i = 0; i < state->nr_cbufs; i++) {
+      if (state->cbufs[i]) {
+         fb_set_blend_caps(dev, state->cbufs[i]->format, &fb->blend_caps[i]);
+
+         fb->has_integer_rt |= fb->blend_caps[i].is_integer;
+
+         if (!first_surf)
+            first_surf = state->cbufs[i];
+      } else {
+         fb_set_blend_caps(dev, PIPE_FORMAT_NONE, &fb->blend_caps[i]);
+      }
+   }
+
+   if (!first_surf && state->zsbuf)
+      first_surf = state->zsbuf;
 
-   if (!ilo->fb.num_samples)
-      ilo->fb.num_samples = 1;
+   fb->num_samples = (first_surf) ? first_surf->texture->nr_samples : 1;
+   if (!fb->num_samples)
+      fb->num_samples = 1;
 
-   ilo->dirty |= ILO_DIRTY_FB;
+   if (state->zsbuf) {
+      const struct ilo_surface_cso *cso =
+         (const struct ilo_surface_cso *) state->zsbuf;
+      const struct ilo_texture *tex = ilo_texture(cso->base.texture);
+
+      fb->has_hiz = cso->u.zs.hiz_vma;
+      fb->depth_offset_format =
+         ilo_format_translate_depth(dev, tex->image_format);
+   } else {
+      fb->has_hiz = false;
+      fb->depth_offset_format = GEN6_ZFORMAT_D32_FLOAT;
+   }
+
+   /*
+    * The PRMs list several restrictions when the framebuffer has more than
+    * one surface.  It seems they are actually lifted on GEN6+.
+    */
+
+   vec->dirty |= ILO_DIRTY_FB;
 }
 
 static void
 ilo_set_polygon_stipple(struct pipe_context *pipe,
                         const struct pipe_poly_stipple *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_state_poly_stipple_info info;
+   int i;
 
-   ilo->poly_stipple = *state;
+   for (i = 0; i < 32; i++)
+      info.pattern[i] = state->stipple[i];
 
-   ilo->dirty |= ILO_DIRTY_POLY_STIPPLE;
+   ilo_state_poly_stipple_set_info(&vec->poly_stipple, dev, &info);
+
+   vec->dirty |= ILO_DIRTY_POLY_STIPPLE;
 }
 
 static void
@@ -755,12 +1740,28 @@ ilo_set_scissor_states(struct pipe_context *pipe,
                        unsigned num_scissors,
                        const struct pipe_scissor_state *scissors)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   unsigned i;
 
-   ilo_gpe_set_scissor(ilo->dev, start_slot, num_scissors,
-         scissors, &ilo->scissor);
+   for (i = 0; i < num_scissors; i++) {
+      struct ilo_state_viewport_scissor_info *info =
+         &vec->viewport.scissors[start_slot + i];
+
+      if (scissors[i].minx < scissors[i].maxx &&
+          scissors[i].miny < scissors[i].maxy) {
+         info->min_x = scissors[i].minx;
+         info->min_y = scissors[i].miny;
+         info->max_x = scissors[i].maxx - 1;
+         info->max_y = scissors[i].maxy - 1;
+      } else {
+         info->min_x = 1;
+         info->min_y = 1;
+         info->max_x = 0;
+         info->max_y = 0;
+      }
+   }
 
-   ilo->dirty |= ILO_DIRTY_SCISSOR;
+   vec->dirty |= ILO_DIRTY_SCISSOR;
 }
 
 static void
@@ -769,30 +1770,34 @@ ilo_set_viewport_states(struct pipe_context *pipe,
                         unsigned num_viewports,
                         const struct pipe_viewport_state *viewports)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
    if (viewports) {
       unsigned i;
 
       for (i = 0; i < num_viewports; i++) {
-         ilo_gpe_set_viewport_cso(ilo->dev, &viewports[i],
-               &ilo->viewport.cso[start_slot + i]);
+         struct ilo_state_viewport_matrix_info *info =
+            &vec->viewport.matrices[start_slot + i];
+
+         memcpy(info->scale, viewports[i].scale, sizeof(info->scale));
+         memcpy(info->translate, viewports[i].translate,
+               sizeof(info->translate));
       }
 
-      if (ilo->viewport.count < start_slot + num_viewports)
-         ilo->viewport.count = start_slot + num_viewports;
+      if (vec->viewport.params.count < start_slot + num_viewports)
+         vec->viewport.params.count = start_slot + num_viewports;
 
       /* need to save viewport 0 for util_blitter */
       if (!start_slot && num_viewports)
-         ilo->viewport.viewport0 = viewports[0];
+         vec->viewport.viewport0 = viewports[0];
    }
    else {
-      if (ilo->viewport.count <= start_slot + num_viewports &&
-          ilo->viewport.count > start_slot)
-         ilo->viewport.count = start_slot;
+      if (vec->viewport.params.count <= start_slot + num_viewports &&
+          vec->viewport.params.count > start_slot)
+         vec->viewport.params.count = start_slot;
    }
 
-   ilo->dirty |= ILO_DIRTY_VIEWPORT;
+   vec->dirty |= ILO_DIRTY_VIEWPORT;
 }
 
 static void
@@ -800,11 +1805,11 @@ ilo_set_sampler_views(struct pipe_context *pipe, unsigned shader,
                       unsigned start, unsigned count,
                       struct pipe_sampler_view **views)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_view_state *dst = &ilo->view[shader];
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_view_state *dst = &vec->view[shader];
    unsigned i;
 
-   assert(start + count <= Elements(dst->states));
+   assert(start + count <= ARRAY_SIZE(dst->states));
 
    if (views) {
       for (i = 0; i < count; i++)
@@ -826,82 +1831,34 @@ ilo_set_sampler_views(struct pipe_context *pipe, unsigned shader,
 
       dst->count = count;
    }
-}
-
-static void
-ilo_set_fragment_sampler_views(struct pipe_context *pipe,
-                               unsigned num_views,
-                               struct pipe_sampler_view **views)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0, num_views, views);
-
-   if (ilo->view[PIPE_SHADER_FRAGMENT].count > num_views) {
-      ilo_set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, num_views,
-            ilo->view[PIPE_SHADER_FRAGMENT].count - num_views, NULL);
-   }
-
-   ilo->dirty |= ILO_DIRTY_VIEW_FS;
-}
-
-static void
-ilo_set_vertex_sampler_views(struct pipe_context *pipe,
-                             unsigned num_views,
-                             struct pipe_sampler_view **views)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
 
-   ilo_set_sampler_views(pipe, PIPE_SHADER_VERTEX, 0, num_views, views);
-
-   if (ilo->view[PIPE_SHADER_VERTEX].count > num_views) {
-      ilo_set_sampler_views(pipe, PIPE_SHADER_VERTEX, num_views,
-            ilo->view[PIPE_SHADER_VERTEX].count - num_views, NULL);
-   }
-
-   ilo->dirty |= ILO_DIRTY_VIEW_VS;
-}
-
-static void
-ilo_set_geometry_sampler_views(struct pipe_context *pipe,
-                               unsigned num_views,
-                               struct pipe_sampler_view **views)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_set_sampler_views(pipe, PIPE_SHADER_GEOMETRY, 0, num_views, views);
-
-   if (ilo->view[PIPE_SHADER_GEOMETRY].count > num_views) {
-      ilo_set_sampler_views(pipe, PIPE_SHADER_GEOMETRY, num_views,
-            ilo->view[PIPE_SHADER_GEOMETRY].count - num_views, NULL);
+   switch (shader) {
+   case PIPE_SHADER_VERTEX:
+      vec->dirty |= ILO_DIRTY_VIEW_VS;
+      break;
+   case PIPE_SHADER_GEOMETRY:
+      vec->dirty |= ILO_DIRTY_VIEW_GS;
+      break;
+   case PIPE_SHADER_FRAGMENT:
+      vec->dirty |= ILO_DIRTY_VIEW_FS;
+      break;
+   case PIPE_SHADER_COMPUTE:
+      vec->dirty |= ILO_DIRTY_VIEW_CS;
+      break;
    }
-
-   ilo->dirty |= ILO_DIRTY_VIEW_GS;
-}
-
-static void
-ilo_set_compute_sampler_views(struct pipe_context *pipe,
-                              unsigned start_slot, unsigned num_views,
-                              struct pipe_sampler_view **views)
-{
-   struct ilo_context *ilo = ilo_context(pipe);
-
-   ilo_set_sampler_views(pipe, PIPE_SHADER_COMPUTE,
-         start_slot, num_views, views);
-
-   ilo->dirty |= ILO_DIRTY_VIEW_CS;
 }
 
 static void
-ilo_set_shader_resources(struct pipe_context *pipe,
-                         unsigned start, unsigned count,
-                         struct pipe_surface **surfaces)
+ilo_set_shader_images(struct pipe_context *pipe, unsigned shader,
+                      unsigned start, unsigned count,
+                      const struct pipe_image_view *views)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_resource_state *dst = &ilo->resource;
+#if 0
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_resource_state *dst = &vec->resource;
    unsigned i;
 
-   assert(start + count <= Elements(dst->states));
+   assert(start + count <= ARRAY_SIZE(dst->states));
 
    if (surfaces) {
       for (i = 0; i < count; i++)
@@ -924,7 +1881,8 @@ ilo_set_shader_resources(struct pipe_context *pipe,
       dst->count = count;
    }
 
-   ilo->dirty |= ILO_DIRTY_RESOURCE;
+   vec->dirty |= ILO_DIRTY_RESOURCE;
+#endif
 }
 
 static void
@@ -932,7 +1890,7 @@ ilo_set_vertex_buffers(struct pipe_context *pipe,
                        unsigned start_slot, unsigned num_buffers,
                        const struct pipe_vertex_buffer *buffers)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
    unsigned i;
 
    /* no PIPE_CAP_USER_VERTEX_BUFFERS */
@@ -941,32 +1899,27 @@ ilo_set_vertex_buffers(struct pipe_context *pipe,
          assert(!buffers[i].user_buffer);
    }
 
-   util_set_vertex_buffers_mask(ilo->vb.states,
-         &ilo->vb.enabled_mask, buffers, start_slot, num_buffers);
+   util_set_vertex_buffers_mask(vec->vb.states,
+         &vec->vb.enabled_mask, buffers, start_slot, num_buffers);
 
-   ilo->dirty |= ILO_DIRTY_VB;
+   vec->dirty |= ILO_DIRTY_VB;
 }
 
 static void
 ilo_set_index_buffer(struct pipe_context *pipe,
                      const struct pipe_index_buffer *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
    if (state) {
-      pipe_resource_reference(&ilo->ib.buffer, state->buffer);
-      ilo->ib.user_buffer = state->user_buffer;
-      ilo->ib.offset = state->offset;
-      ilo->ib.index_size = state->index_size;
-   }
-   else {
-      pipe_resource_reference(&ilo->ib.buffer, NULL);
-      ilo->ib.user_buffer = NULL;
-      ilo->ib.offset = 0;
-      ilo->ib.index_size = 0;
+      pipe_resource_reference(&vec->ib.state.buffer, state->buffer);
+      vec->ib.state = *state;
+   } else {
+      pipe_resource_reference(&vec->ib.state.buffer, NULL);
+      memset(&vec->ib.state, 0, sizeof(vec->ib.state));
    }
 
-   ilo->dirty |= ILO_DIRTY_IB;
+   vec->dirty |= ILO_DIRTY_IB;
 }
 
 static struct pipe_stream_output_target *
@@ -975,49 +1928,61 @@ ilo_create_stream_output_target(struct pipe_context *pipe,
                                 unsigned buffer_offset,
                                 unsigned buffer_size)
 {
-   struct pipe_stream_output_target *target;
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_stream_output_target *target;
+   struct ilo_state_sol_buffer_info info;
 
-   target = MALLOC_STRUCT(pipe_stream_output_target);
+   target = CALLOC_STRUCT(ilo_stream_output_target);
    assert(target);
 
-   pipe_reference_init(&target->reference, 1);
-   target->buffer = NULL;
-   pipe_resource_reference(&target->buffer, res);
-   target->context = pipe;
-   target->buffer_offset = buffer_offset;
-   target->buffer_size = buffer_size;
+   pipe_reference_init(&target->base.reference, 1);
+   pipe_resource_reference(&target->base.buffer, res);
+   target->base.context = pipe;
+   target->base.buffer_offset = buffer_offset;
+   target->base.buffer_size = buffer_size;
+
+   memset(&info, 0, sizeof(info));
+   info.vma = ilo_resource_get_vma(res);
+   info.offset = buffer_offset;
+   info.size = buffer_size;
+
+   ilo_state_sol_buffer_init(&target->sb, dev, &info);
 
-   return target;
+   return &target->base;
 }
 
 static void
 ilo_set_stream_output_targets(struct pipe_context *pipe,
                               unsigned num_targets,
                               struct pipe_stream_output_target **targets,
-                              unsigned append_bitmask)
+                              const unsigned *offset)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
    unsigned i;
+   unsigned append_bitmask = 0;
 
    if (!targets)
       num_targets = 0;
 
    /* util_blitter may set this unnecessarily */
-   if (!ilo->so.count && !num_targets)
+   if (!vec->so.count && !num_targets)
       return;
 
-   for (i = 0; i < num_targets; i++)
-      pipe_so_target_reference(&ilo->so.states[i], targets[i]);
+   for (i = 0; i < num_targets; i++) {
+      pipe_so_target_reference(&vec->so.states[i], targets[i]);
+      if (offset[i] == (unsigned)-1)
+         append_bitmask |= 1 << i;
+   }
 
-   for (; i < ilo->so.count; i++)
-      pipe_so_target_reference(&ilo->so.states[i], NULL);
+   for (; i < vec->so.count; i++)
+      pipe_so_target_reference(&vec->so.states[i], NULL);
 
-   ilo->so.count = num_targets;
-   ilo->so.append_bitmask = append_bitmask;
+   vec->so.count = num_targets;
+   vec->so.append_bitmask = append_bitmask;
 
-   ilo->so.enabled = (ilo->so.count > 0);
+   vec->so.enabled = (vec->so.count > 0);
 
-   ilo->dirty |= ILO_DIRTY_SO;
+   vec->dirty |= ILO_DIRTY_SO;
 }
 
 static void
@@ -1033,10 +1998,10 @@ ilo_create_sampler_view(struct pipe_context *pipe,
                         struct pipe_resource *res,
                         const struct pipe_sampler_view *templ)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
    struct ilo_view_cso *view;
 
-   view = MALLOC_STRUCT(ilo_view_cso);
+   view = CALLOC_STRUCT(ilo_view_cso);
    assert(view);
 
    view->base = *templ;
@@ -1046,31 +2011,56 @@ ilo_create_sampler_view(struct pipe_context *pipe,
    view->base.context = pipe;
 
    if (res->target == PIPE_BUFFER) {
-      const unsigned elem_size = util_format_get_blocksize(templ->format);
-      const unsigned first_elem = templ->u.buf.first_element;
-      const unsigned num_elems = templ->u.buf.last_element - first_elem + 1;
-
-      ilo_gpe_init_view_surface_for_buffer(ilo->dev, ilo_buffer(res),
-            first_elem * elem_size, num_elems * elem_size,
-            elem_size, templ->format, false, false, &view->surface);
-   }
-   else {
+      struct ilo_state_surface_buffer_info info;
+
+      memset(&info, 0, sizeof(info));
+      info.vma = ilo_resource_get_vma(res);
+      info.offset = templ->u.buf.offset;
+      info.size = templ->u.buf.size;
+      info.access = ILO_STATE_SURFACE_ACCESS_SAMPLER;
+      info.format = ilo_format_translate_color(dev, templ->format);
+      info.format_size = util_format_get_blocksize(templ->format);
+      info.struct_size = info.format_size;
+      info.readonly = true;
+
+      ilo_state_surface_init_for_buffer(&view->surface, dev, &info);
+   } else {
       struct ilo_texture *tex = ilo_texture(res);
+      struct ilo_state_surface_image_info info;
 
       /* warn about degraded performance because of a missing binding flag */
-      if (tex->tiling == INTEL_TILING_NONE &&
+      if (tex->image.tiling == GEN6_TILING_NONE &&
           !(tex->base.bind & PIPE_BIND_SAMPLER_VIEW)) {
          ilo_warn("creating sampler view for a resource "
                   "not created for sampling\n");
       }
 
-      ilo_gpe_init_view_surface_for_texture(ilo->dev, tex,
-            templ->format,
-            templ->u.tex.first_level,
-            templ->u.tex.last_level - templ->u.tex.first_level + 1,
-            templ->u.tex.first_layer,
-            templ->u.tex.last_layer - templ->u.tex.first_layer + 1,
-            false, false, &view->surface);
+      memset(&info, 0, sizeof(info));
+
+      info.img = &tex->image;
+      info.level_base = templ->u.tex.first_level;
+      info.level_count = templ->u.tex.last_level -
+         templ->u.tex.first_level + 1;
+      info.slice_base = templ->u.tex.first_layer;
+      info.slice_count = templ->u.tex.last_layer -
+         templ->u.tex.first_layer + 1;
+
+      info.vma = &tex->vma;
+      info.access = ILO_STATE_SURFACE_ACCESS_SAMPLER;
+      info.type = tex->image.type;
+
+      if (templ->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT &&
+          tex->separate_s8) {
+         info.format = ilo_format_translate_texture(dev,
+               PIPE_FORMAT_Z32_FLOAT);
+      } else {
+         info.format = ilo_format_translate_texture(dev, templ->format);
+      }
+
+      info.is_array = util_resource_is_array_texture(&tex->base);
+      info.readonly = true;
+
+      ilo_state_surface_init_for_image(&view->surface, dev, &info);
    }
 
    return &view->base;
@@ -1089,45 +2079,89 @@ ilo_create_surface(struct pipe_context *pipe,
                    struct pipe_resource *res,
                    const struct pipe_surface *templ)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   const struct ilo_dev *dev = ilo_context(pipe)->dev;
+   struct ilo_texture *tex = ilo_texture(res);
    struct ilo_surface_cso *surf;
 
-   surf = MALLOC_STRUCT(ilo_surface_cso);
+   surf = CALLOC_STRUCT(ilo_surface_cso);
    assert(surf);
 
    surf->base = *templ;
    pipe_reference_init(&surf->base.reference, 1);
    surf->base.texture = NULL;
-   pipe_resource_reference(&surf->base.texture, res);
+   pipe_resource_reference(&surf->base.texture, &tex->base);
 
    surf->base.context = pipe;
-   surf->base.width = u_minify(res->width0, templ->u.tex.level);
-   surf->base.height = u_minify(res->height0, templ->u.tex.level);
+   surf->base.width = u_minify(tex->base.width0, templ->u.tex.level);
+   surf->base.height = u_minify(tex->base.height0, templ->u.tex.level);
 
    surf->is_rt = !util_format_is_depth_or_stencil(templ->format);
 
    if (surf->is_rt) {
+      struct ilo_state_surface_image_info info;
+
       /* relax this? */
-      assert(res->target != PIPE_BUFFER);
+      assert(tex->base.target != PIPE_BUFFER);
+
+      memset(&info, 0, sizeof(info));
+
+      info.img = &tex->image;
+      info.level_base = templ->u.tex.level;
+      info.level_count = 1;
+      info.slice_base = templ->u.tex.first_layer;
+      info.slice_count = templ->u.tex.last_layer -
+         templ->u.tex.first_layer + 1;
+
+      info.vma = &tex->vma;
+      if (ilo_image_can_enable_aux(&tex->image, templ->u.tex.level))
+         info.aux_vma = &tex->aux_vma;
+
+      info.access = ILO_STATE_SURFACE_ACCESS_DP_RENDER;
+
+      info.type = (tex->image.type == GEN6_SURFTYPE_CUBE) ?
+         GEN6_SURFTYPE_2D : tex->image.type;
+
+      info.format = ilo_format_translate_render(dev, templ->format);
+      info.is_array = util_resource_is_array_texture(&tex->base);
+
+      ilo_state_surface_init_for_image(&surf->u.rt, dev, &info);
+   } else {
+      struct ilo_state_zs_info info;
 
-      /*
-       * classic i965 sets render_cache_rw for constant buffers and sol
-       * surfaces but not render buffers.  Why?
-       */
-      ilo_gpe_init_view_surface_for_texture(ilo->dev, ilo_texture(res),
-            templ->format, templ->u.tex.level, 1,
-            templ->u.tex.first_layer,
-            templ->u.tex.last_layer - templ->u.tex.first_layer + 1,
-            true, true, &surf->u.rt);
-   }
-   else {
       assert(res->target != PIPE_BUFFER);
 
-      ilo_gpe_init_zs_surface(ilo->dev, ilo_texture(res),
-            templ->format, templ->u.tex.level,
-            templ->u.tex.first_layer,
-            templ->u.tex.last_layer - templ->u.tex.first_layer + 1,
-            &surf->u.zs);
+      memset(&info, 0, sizeof(info));
+
+      if (templ->format == PIPE_FORMAT_S8_UINT) {
+         info.s_vma = &tex->vma;
+         info.s_img = &tex->image;
+      } else {
+         info.z_vma = &tex->vma;
+         info.z_img = &tex->image;
+
+         if (tex->separate_s8) {
+            info.s_vma = &tex->separate_s8->vma;
+            info.s_img = &tex->separate_s8->image;
+         }
+
+         if (ilo_image_can_enable_aux(&tex->image, templ->u.tex.level))
+            info.hiz_vma = &tex->aux_vma;
+      }
+
+      info.level = templ->u.tex.level;
+      info.slice_base = templ->u.tex.first_layer;
+      info.slice_count = templ->u.tex.last_layer -
+         templ->u.tex.first_layer + 1;
+
+      info.type = (tex->image.type == GEN6_SURFTYPE_CUBE) ?
+         GEN6_SURFTYPE_2D : tex->image.type;
+
+      info.format = ilo_format_translate_depth(dev, tex->image_format);
+      if (ilo_dev_gen(dev) == ILO_GEN(6) && !info.hiz_vma &&
+          tex->image_format == PIPE_FORMAT_Z24X8_UNORM)
+         info.format = GEN6_ZFORMAT_D24_UNORM_S8_UINT;
+
+      ilo_state_zs_init(&surf->u.zs, dev, &info);
    }
 
    return &surf->base;
@@ -1148,7 +2182,7 @@ ilo_create_compute_state(struct pipe_context *pipe,
    struct ilo_context *ilo = ilo_context(pipe);
    struct ilo_shader_state *shader;
 
-   shader = ilo_shader_create_cs(ilo->dev, state, ilo);
+   shader = ilo_shader_create_cs(ilo->dev, state, &ilo->state_vector);
    assert(shader);
 
    ilo_shader_cache_add(ilo->shader_cache, shader);
@@ -1159,11 +2193,11 @@ ilo_create_compute_state(struct pipe_context *pipe,
 static void
 ilo_bind_compute_state(struct pipe_context *pipe, void *state)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
 
-   ilo->cs = state;
+   vec->cs = state;
 
-   ilo->dirty |= ILO_DIRTY_CS;
+   vec->dirty |= ILO_DIRTY_CS;
 }
 
 static void
@@ -1181,11 +2215,11 @@ ilo_set_compute_resources(struct pipe_context *pipe,
                           unsigned start, unsigned count,
                           struct pipe_surface **surfaces)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_resource_state *dst = &ilo->cs_resource;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_resource_state *dst = &vec->cs_resource;
    unsigned i;
 
-   assert(start + count <= Elements(dst->states));
+   assert(start + count <= ARRAY_SIZE(dst->states));
 
    if (surfaces) {
       for (i = 0; i < count; i++)
@@ -1208,7 +2242,7 @@ ilo_set_compute_resources(struct pipe_context *pipe,
       dst->count = count;
    }
 
-   ilo->dirty |= ILO_DIRTY_CS_RESOURCE;
+   vec->dirty |= ILO_DIRTY_CS_RESOURCE;
 }
 
 static void
@@ -1217,34 +2251,56 @@ ilo_set_global_binding(struct pipe_context *pipe,
                        struct pipe_resource **resources,
                        uint32_t **handles)
 {
-   struct ilo_context *ilo = ilo_context(pipe);
-   struct ilo_global_binding *dst = &ilo->global_binding;
+   struct ilo_state_vector *vec = &ilo_context(pipe)->state_vector;
+   struct ilo_global_binding_cso *dst;
    unsigned i;
 
-   assert(start + count <= Elements(dst->resources));
+   /* make room */
+   if (vec->global_binding.count < start + count) {
+      if (resources) {
+         const unsigned old_size = vec->global_binding.bindings.size;
+         const unsigned new_size = sizeof(*dst) * (start + count);
 
-   if (resources) {
-      for (i = 0; i < count; i++)
-         pipe_resource_reference(&dst->resources[start + i], resources[i]);
+         if (old_size < new_size) {
+            util_dynarray_resize(&vec->global_binding.bindings, new_size);
+            memset(vec->global_binding.bindings.data + old_size, 0,
+                  new_size - old_size);
+         }
+      } else {
+         count = vec->global_binding.count - start;
+      }
    }
-   else {
-      for (i = 0; i < count; i++)
-         pipe_resource_reference(&dst->resources[start + i], NULL);
+
+   dst = util_dynarray_element(&vec->global_binding.bindings,
+         struct ilo_global_binding_cso, start);
+
+   if (resources) {
+      for (i = 0; i < count; i++) {
+         pipe_resource_reference(&dst[i].resource, resources[i]);
+         dst[i].handle = handles[i];
+      }
+   } else {
+      for (i = 0; i < count; i++) {
+         pipe_resource_reference(&dst[i].resource, NULL);
+         dst[i].handle = NULL;
+      }
    }
 
-   if (dst->count <= start + count) {
+   if (vec->global_binding.count <= start + count) {
+      dst = util_dynarray_begin(&vec->global_binding.bindings);
+
       if (resources)
          count += start;
       else
          count = start;
 
-      while (count > 0 && !dst->resources[count - 1])
+      while (count > 0 && !dst[count - 1].resource)
          count--;
 
-      dst->count = count;
+      vec->global_binding.count = count;
    }
 
-   ilo->dirty |= ILO_DIRTY_GLOBAL_BINDING;
+   vec->dirty |= ILO_DIRTY_GLOBAL_BINDING;
 }
 
 /**
@@ -1259,7 +2315,7 @@ ilo_init_state_functions(struct ilo_context *ilo)
    ilo->base.bind_blend_state = ilo_bind_blend_state;
    ilo->base.delete_blend_state = ilo_delete_blend_state;
    ilo->base.create_sampler_state = ilo_create_sampler_state;
-   ilo->base.bind_sampler_states = ilo_bind_sampler_states2;
+   ilo->base.bind_sampler_states = ilo_bind_sampler_states;
    ilo->base.delete_sampler_state = ilo_delete_sampler_state;
    ilo->base.create_rasterizer_state = ilo_create_rasterizer_state;
    ilo->base.bind_rasterizer_state = ilo_bind_rasterizer_state;
@@ -1289,11 +2345,8 @@ ilo_init_state_functions(struct ilo_context *ilo)
    ilo->base.set_polygon_stipple = ilo_set_polygon_stipple;
    ilo->base.set_scissor_states = ilo_set_scissor_states;
    ilo->base.set_viewport_states = ilo_set_viewport_states;
-   ilo->base.set_fragment_sampler_views = ilo_set_fragment_sampler_views;
-   ilo->base.set_vertex_sampler_views = ilo_set_vertex_sampler_views;
-   ilo->base.set_geometry_sampler_views = ilo_set_geometry_sampler_views;
-   ilo->base.set_compute_sampler_views = ilo_set_compute_sampler_views;
-   ilo->base.set_shader_resources = ilo_set_shader_resources;
+   ilo->base.set_sampler_views = ilo_set_sampler_views;
+   ilo->base.set_shader_images = ilo_set_shader_images;
    ilo->base.set_vertex_buffers = ilo_set_vertex_buffers;
    ilo->base.set_index_buffer = ilo_set_index_buffer;
 
@@ -1315,83 +2368,112 @@ ilo_init_state_functions(struct ilo_context *ilo)
 }
 
 void
-ilo_init_states(struct ilo_context *ilo)
+ilo_state_vector_init(const struct ilo_dev *dev,
+                      struct ilo_state_vector *vec)
 {
-   ilo_gpe_set_scissor_null(ilo->dev, &ilo->scissor);
+   struct ilo_state_urb_info urb_info;
+
+   vec->sample_mask = ~0u;
+
+   ilo_state_viewport_init_data_only(&vec->viewport.vp, dev,
+         vec->viewport.vp_data, sizeof(vec->viewport.vp_data));
+   assert(vec->viewport.vp.array_size >= ILO_MAX_VIEWPORTS);
+
+   vec->viewport.params.matrices = vec->viewport.matrices;
+   vec->viewport.params.scissors = vec->viewport.scissors;
+
+   ilo_state_hs_init_disabled(&vec->disabled_hs, dev);
+   ilo_state_ds_init_disabled(&vec->disabled_ds, dev);
+   ilo_state_gs_init_disabled(&vec->disabled_gs, dev);
+
+   ilo_state_sol_buffer_init_disabled(&vec->so.dummy_sb, dev);
 
-   ilo_gpe_init_zs_surface(ilo->dev, NULL,
-         PIPE_FORMAT_NONE, 0, 0, 1, &ilo->fb.null_zs);
+   ilo_state_surface_init_for_null(&vec->fb.null_rt, dev);
+   ilo_state_zs_init_for_null(&vec->fb.null_zs, dev);
 
-   ilo->dirty = ILO_DIRTY_ALL;
+   ilo_state_sampler_init_disabled(&vec->disabled_sampler, dev);
+
+   memset(&urb_info, 0, sizeof(urb_info));
+   ilo_state_urb_init(&vec->urb, dev, &urb_info);
+
+   util_dynarray_init(&vec->global_binding.bindings);
+
+   vec->dirty = ILO_DIRTY_ALL;
 }
 
 void
-ilo_cleanup_states(struct ilo_context *ilo)
+ilo_state_vector_cleanup(struct ilo_state_vector *vec)
 {
    unsigned i, sh;
 
-   for (i = 0; i < Elements(ilo->vb.states); i++) {
-      if (ilo->vb.enabled_mask & (1 << i))
-         pipe_resource_reference(&ilo->vb.states[i].buffer, NULL);
+   for (i = 0; i < ARRAY_SIZE(vec->vb.states); i++) {
+      if (vec->vb.enabled_mask & (1 << i))
+         pipe_resource_reference(&vec->vb.states[i].buffer, NULL);
    }
 
-   pipe_resource_reference(&ilo->ib.buffer, NULL);
-   pipe_resource_reference(&ilo->ib.hw_resource, NULL);
+   pipe_resource_reference(&vec->ib.state.buffer, NULL);
+   pipe_resource_reference(&vec->ib.hw_resource, NULL);
 
-   for (i = 0; i < ilo->so.count; i++)
-      pipe_so_target_reference(&ilo->so.states[i], NULL);
+   for (i = 0; i < vec->so.count; i++)
+      pipe_so_target_reference(&vec->so.states[i], NULL);
 
    for (sh = 0; sh < PIPE_SHADER_TYPES; sh++) {
-      for (i = 0; i < ilo->view[sh].count; i++) {
-         struct pipe_sampler_view *view = ilo->view[sh].states[i];
+      for (i = 0; i < vec->view[sh].count; i++) {
+         struct pipe_sampler_view *view = vec->view[sh].states[i];
          pipe_sampler_view_reference(&view, NULL);
       }
 
-      for (i = 0; i < Elements(ilo->cbuf[sh].cso); i++) {
-         struct ilo_cbuf_cso *cbuf = &ilo->cbuf[sh].cso[i];
+      for (i = 0; i < ARRAY_SIZE(vec->cbuf[sh].cso); i++) {
+         struct ilo_cbuf_cso *cbuf = &vec->cbuf[sh].cso[i];
          pipe_resource_reference(&cbuf->resource, NULL);
       }
    }
 
-   for (i = 0; i < ilo->resource.count; i++)
-      pipe_surface_reference(&ilo->resource.states[i], NULL);
+   for (i = 0; i < vec->resource.count; i++)
+      pipe_surface_reference(&vec->resource.states[i], NULL);
 
-   for (i = 0; i < ilo->fb.state.nr_cbufs; i++)
-      pipe_surface_reference(&ilo->fb.state.cbufs[i], NULL);
+   for (i = 0; i < vec->fb.state.nr_cbufs; i++)
+      pipe_surface_reference(&vec->fb.state.cbufs[i], NULL);
 
-   if (ilo->fb.state.zsbuf)
-      pipe_surface_reference(&ilo->fb.state.zsbuf, NULL);
+   if (vec->fb.state.zsbuf)
+      pipe_surface_reference(&vec->fb.state.zsbuf, NULL);
 
-   for (i = 0; i < ilo->cs_resource.count; i++)
-      pipe_surface_reference(&ilo->cs_resource.states[i], NULL);
+   for (i = 0; i < vec->cs_resource.count; i++)
+      pipe_surface_reference(&vec->cs_resource.states[i], NULL);
 
-   for (i = 0; i < ilo->global_binding.count; i++)
-      pipe_resource_reference(&ilo->global_binding.resources[i], NULL);
+   for (i = 0; i < vec->global_binding.count; i++) {
+      struct ilo_global_binding_cso *cso =
+         util_dynarray_element(&vec->global_binding.bindings,
+               struct ilo_global_binding_cso, i);
+      pipe_resource_reference(&cso->resource, NULL);
+   }
+
+   util_dynarray_fini(&vec->global_binding.bindings);
 }
 
 /**
  * Mark all states that have the resource dirty.
  */
 void
-ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
-                                    const struct pipe_resource *res)
+ilo_state_vector_resource_renamed(struct ilo_state_vector *vec,
+                                  struct pipe_resource *res)
 {
    uint32_t states = 0;
    unsigned sh, i;
 
    if (res->target == PIPE_BUFFER) {
-      uint32_t vb_mask = ilo->vb.enabled_mask;
+      uint32_t vb_mask = vec->vb.enabled_mask;
 
       while (vb_mask) {
          const unsigned idx = u_bit_scan(&vb_mask);
 
-         if (ilo->vb.states[idx].buffer == res) {
+         if (vec->vb.states[idx].buffer == res) {
             states |= ILO_DIRTY_VB;
             break;
          }
       }
 
-      if (ilo->ib.buffer == res) {
+      if (vec->ib.state.buffer == res) {
          states |= ILO_DIRTY_IB;
 
          /*
@@ -1401,11 +2483,11 @@ ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
           * contents of the IB has changed.  Here, we set the index size to an
           * invalid value to avoid the optimization.
           */
-         ilo->ib.hw_index_size = 0;
+         vec->ib.hw_index_size = 0;
       }
 
-      for (i = 0; i < ilo->so.count; i++) {
-         if (ilo->so.states[i]->buffer == res) {
+      for (i = 0; i < vec->so.count; i++) {
+         if (vec->so.states[i]->buffer == res) {
             states |= ILO_DIRTY_SO;
             break;
          }
@@ -1413,10 +2495,10 @@ ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
    }
 
    for (sh = 0; sh < PIPE_SHADER_TYPES; sh++) {
-      for (i = 0; i < ilo->view[sh].count; i++) {
-         struct pipe_sampler_view *view = ilo->view[sh].states[i];
+      for (i = 0; i < vec->view[sh].count; i++) {
+         struct ilo_view_cso *cso = (struct ilo_view_cso *) vec->view[sh].states[i];
 
-         if (view->texture == res) {
+         if (cso->base.texture == res) {
             static const unsigned view_dirty_bits[PIPE_SHADER_TYPES] = {
                [PIPE_SHADER_VERTEX]    = ILO_DIRTY_VIEW_VS,
                [PIPE_SHADER_FRAGMENT]  = ILO_DIRTY_VIEW_FS,
@@ -1430,8 +2512,8 @@ ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
       }
 
       if (res->target == PIPE_BUFFER) {
-         for (i = 0; i < Elements(ilo->cbuf[sh].cso); i++) {
-            struct ilo_cbuf_cso *cbuf = &ilo->cbuf[sh].cso[i];
+         for (i = 0; i < ARRAY_SIZE(vec->cbuf[sh].cso); i++) {
+            struct ilo_cbuf_cso *cbuf = &vec->cbuf[sh].cso[i];
 
             if (cbuf->resource == res) {
                states |= ILO_DIRTY_CBUF;
@@ -1441,8 +2523,11 @@ ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
       }
    }
 
-   for (i = 0; i < ilo->resource.count; i++) {
-      if (ilo->resource.states[i]->texture == res) {
+   for (i = 0; i < vec->resource.count; i++) {
+      struct ilo_surface_cso *cso =
+         (struct ilo_surface_cso *) vec->resource.states[i];
+
+      if (cso->base.texture == res) {
          states |= ILO_DIRTY_RESOURCE;
          break;
       }
@@ -1450,37 +2535,44 @@ ilo_mark_states_with_resource_dirty(struct ilo_context *ilo,
 
    /* for now? */
    if (res->target != PIPE_BUFFER) {
-      for (i = 0; i < ilo->fb.state.nr_cbufs; i++) {
-         if (ilo->fb.state.cbufs[i]->texture == res) {
+      for (i = 0; i < vec->fb.state.nr_cbufs; i++) {
+         struct ilo_surface_cso *cso =
+            (struct ilo_surface_cso *) vec->fb.state.cbufs[i];
+         if (cso && cso->base.texture == res) {
             states |= ILO_DIRTY_FB;
             break;
          }
       }
 
-      if (ilo->fb.state.zsbuf && ilo->fb.state.zsbuf->texture == res)
+      if (vec->fb.state.zsbuf && vec->fb.state.zsbuf->texture == res)
          states |= ILO_DIRTY_FB;
    }
 
-   for (i = 0; i < ilo->cs_resource.count; i++) {
-      pipe_surface_reference(&ilo->cs_resource.states[i], NULL);
-      if (ilo->cs_resource.states[i]->texture == res) {
+   for (i = 0; i < vec->cs_resource.count; i++) {
+      struct ilo_surface_cso *cso =
+         (struct ilo_surface_cso *) vec->cs_resource.states[i];
+      if (cso->base.texture == res) {
          states |= ILO_DIRTY_CS_RESOURCE;
          break;
       }
    }
 
-   for (i = 0; i < ilo->global_binding.count; i++) {
-      if (ilo->global_binding.resources[i] == res) {
+   for (i = 0; i < vec->global_binding.count; i++) {
+      struct ilo_global_binding_cso *cso =
+         util_dynarray_element(&vec->global_binding.bindings,
+               struct ilo_global_binding_cso, i);
+
+      if (cso->resource == res) {
          states |= ILO_DIRTY_GLOBAL_BINDING;
          break;
       }
    }
 
-   ilo->dirty |= states;
+   vec->dirty |= states;
 }
 
 void
-ilo_dump_dirty_flags(uint32_t dirty)
+ilo_state_vector_dump_dirty(const struct ilo_state_vector *vec)
 {
    static const char *state_names[ILO_STATE_COUNT] = {
       [ILO_STATE_VB]              = "VB",
@@ -1515,6 +2607,7 @@ ilo_dump_dirty_flags(uint32_t dirty)
       [ILO_STATE_CS_RESOURCE]     = "CS_RESOURCE",
       [ILO_STATE_GLOBAL_BINDING]  = "GLOBAL_BINDING",
    };
+   uint32_t dirty = vec->dirty;
 
    if (!dirty) {
       ilo_printf("no state is dirty\n");