Merge branch 'instanced-arrays'

Conflicts:
	src/gallium/drivers/softpipe/sp_draw_arrays.c

26 files changed, 547 insertions, 129 deletions

diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index b716209df2..8a64c06efc 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -164,6 +164,14 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw,
 void draw_arrays(struct draw_context *draw, unsigned prim,
 		 unsigned start, unsigned count);
 
+void
+draw_arrays_instanced(struct draw_context *draw,
+                      unsigned mode,
+                      unsigned start,
+                      unsigned count,
+                      unsigned startInstance,
+                      unsigned instanceCount);
+
 void draw_flush(struct draw_context *draw);
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index 1a5269c0de..d40c035240 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -138,7 +138,7 @@ emit_vertex( struct vbuf_stage *vbuf,
       /* Note: we really do want data[0] here, not data[pos]: 
        */
       vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0);
-      vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr);
+      vbuf->translate->run(vbuf->translate, 0, 1, 0, vbuf->vertex_ptr);
 
       if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr);
       
@@ -271,10 +271,12 @@ vbuf_start_prim( struct vbuf_stage *vbuf, uint prim )
 	 emit_sz = 0;
 	 break;
       }
-      
+
+      hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
       hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
       hw_key.element[i].input_buffer = src_buffer;
       hw_key.element[i].input_offset = src_offset;
+      hw_key.element[i].instance_divisor = 0;
       hw_key.element[i].output_format = output_format;
       hw_key.element[i].output_offset = dst_offset;
 
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index e49041556b..2a11b8c545 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -239,6 +239,8 @@ struct draw_context
 
    unsigned reduced_prim;
 
+   unsigned instance_id;
+
    void *driver_private;
 };
 
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 2801dbafe4..a5ddec5286 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -280,20 +280,33 @@ void
 draw_arrays(struct draw_context *draw, unsigned prim,
             unsigned start, unsigned count)
 {
-   unsigned reduced_prim = u_reduced_prim(prim);
+   draw_arrays_instanced(draw, prim, start, count, 0, 1);
+}
+
+void
+draw_arrays_instanced(struct draw_context *draw,
+                      unsigned mode,
+                      unsigned start,
+                      unsigned count,
+                      unsigned startInstance,
+                      unsigned instanceCount)
+{
+   unsigned reduced_prim = u_reduced_prim(mode);
+   unsigned instance;
+
    if (reduced_prim != draw->reduced_prim) {
-      draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+      draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
       draw->reduced_prim = reduced_prim;
    }
 
    if (0)
-      draw_print_arrays(draw, prim, start, MIN2(count, 20));
+      draw_print_arrays(draw, mode, start, MIN2(count, 20));
 
 #if 0
    {
       int i;
-      debug_printf("draw_arrays(prim=%u start=%u count=%u):\n",
-                   prim, start, count);
+      debug_printf("draw_arrays(mode=%u start=%u count=%u):\n",
+                   mode, start, count);
       tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
       debug_printf("Elements:\n");
       for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
@@ -311,6 +324,8 @@ draw_arrays(struct draw_context *draw, unsigned prim,
    }
 #endif
 
-   /* drawing done here: */
-   draw_pt_arrays(draw, prim, start, count);
+   for (instance = 0; instance < instanceCount; instance++) {
+      draw->instance_id = instance + startInstance;
+      draw_pt_arrays(draw, mode, start, count);
+   }
 }
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 20edf7a227..d5e0d92a60 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -183,7 +183,8 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw );
 struct pt_fetch;
 void draw_pt_fetch_prepare( struct pt_fetch *fetch,
                             unsigned vertex_input_count,
-			    unsigned vertex_size );
+                            unsigned vertex_size,
+                            unsigned instance_id_index );
 
 void draw_pt_fetch_run( struct pt_fetch *fetch,
 			const unsigned *elts,
diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c
index 064e16c295..4fb53276bb 100644
--- a/src/gallium/auxiliary/draw/draw_pt_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_emit.c
@@ -121,10 +121,12 @@ void draw_pt_emit_prepare( struct pt_emit *emit,
 	 emit_sz = 0;
 	 break;
       }
-      
+
+      hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
       hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
       hw_key.element[i].input_buffer = src_buffer;
       hw_key.element[i].input_offset = src_offset;
+      hw_key.element[i].instance_divisor = 0;
       hw_key.element[i].output_format = output_format;
       hw_key.element[i].output_offset = dst_offset;
 
@@ -204,6 +206,7 @@ void draw_pt_emit( struct pt_emit *emit,
    translate->run( translate,
 		   0, 
 		   vertex_count,
+                   draw->instance_id,
 		   hw_verts );
 
    render->unmap_vertices( render, 
@@ -263,6 +266,7 @@ void draw_pt_emit_linear(struct pt_emit *emit,
    translate->run(translate,
                   0,
                   count,
+                  draw->instance_id,
                   hw_verts);
 
    if (0) {
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 305bfef435..55e7a7b81a 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -58,12 +58,14 @@ struct pt_fetch {
  */
 void draw_pt_fetch_prepare( struct pt_fetch *fetch,
                             unsigned vs_input_count,
-			    unsigned vertex_size )
+                            unsigned vertex_size,
+                            unsigned instance_id_index )
 {
    struct draw_context *draw = fetch->draw;
    unsigned nr_inputs;
-   unsigned i, nr = 0;
+   unsigned i, nr = 0, ei = 0;
    unsigned dst_offset = 0;
+   unsigned num_extra_inputs = 0;
    struct translate_key key;
 
    fetch->vertex_size = vertex_size;
@@ -78,9 +80,11 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
    {
       /* Need to set header->vertex_id = 0xffff somehow.
        */
+      key.element[nr].type = TRANSLATE_ELEMENT_NORMAL;
       key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT;
       key.element[nr].input_buffer = draw->pt.nr_vertex_buffers;
       key.element[nr].input_offset = 0;
+      key.element[nr].instance_divisor = 0;
       key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT;
       key.element[nr].output_offset = dst_offset;
       dst_offset += 1 * sizeof(float);
@@ -91,19 +95,36 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
        */
       dst_offset += 4 * sizeof(float);
    }
-      
-   assert( draw->pt.nr_vertex_elements >= vs_input_count );
 
-   nr_inputs = MIN2( vs_input_count, draw->pt.nr_vertex_elements );
+   if (instance_id_index != ~0) {
+      num_extra_inputs++;
+   }
+
+   assert(draw->pt.nr_vertex_elements + num_extra_inputs >= vs_input_count);
+
+   nr_inputs = MIN2(vs_input_count, draw->pt.nr_vertex_elements + num_extra_inputs);
 
    for (i = 0; i < nr_inputs; i++) {
-      key.element[nr].input_format = draw->pt.vertex_element[i].src_format;
-      key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index;
-      key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset;
-      key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-      key.element[nr].output_offset = dst_offset;
+      if (i == instance_id_index) {
+         key.element[nr].type = TRANSLATE_ELEMENT_INSTANCE_ID;
+         key.element[nr].input_format = PIPE_FORMAT_R32_USCALED;
+         key.element[nr].output_format = PIPE_FORMAT_R32_USCALED;
+         key.element[nr].output_offset = dst_offset;
+
+         dst_offset += sizeof(uint);
+      } else {
+         key.element[nr].type = TRANSLATE_ELEMENT_NORMAL;
+         key.element[nr].input_format = draw->pt.vertex_element[ei].src_format;
+         key.element[nr].input_buffer = draw->pt.vertex_element[ei].vertex_buffer_index;
+         key.element[nr].input_offset = draw->pt.vertex_element[ei].src_offset;
+         key.element[nr].instance_divisor = draw->pt.vertex_element[ei].instance_divisor;
+         key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+         key.element[nr].output_offset = dst_offset;
+
+         ei++;
+         dst_offset += 4 * sizeof(float);
+      }
 
-      dst_offset += 4 * sizeof(float);
       nr++;
    }
 
@@ -158,6 +179,7 @@ void draw_pt_fetch_run( struct pt_fetch *fetch,
    translate->run_elts( translate,
 			elts, 
 			count,
+                        draw->instance_id,
 			verts );
 
 }
@@ -183,6 +205,7 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch,
    translate->run( translate,
                    start,
                    count,
+                   draw->instance_id,
                    verts );
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
index e7fe6b3b76..2a604470e9 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -166,9 +166,11 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
 	 continue;
       }
 
+      key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
       key.element[i].input_format = input_format;
       key.element[i].input_buffer = input_buffer;
       key.element[i].input_offset = input_offset;
+      key.element[i].instance_divisor = src->instance_divisor;
       key.element[i].output_format = output_format;
       key.element[i].output_offset = dst_offset;
       
@@ -256,6 +258,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle,
    feme->translate->run_elts( feme->translate, 
 			      fetch_elts,
 			      fetch_count,
+                              draw->instance_id,
 			      hw_verts );
 
    if (0) {
@@ -314,6 +317,7 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
    feme->translate->run( feme->translate,
                          start,
                          count,
+                         draw->instance_id,
                          hw_verts );
 
    if (0) {
@@ -374,6 +378,7 @@ static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
    feme->translate->run( feme->translate,
                          start,
                          count,
+                         draw->instance_id,
                          hw_verts );
 
    draw->render->unmap_vertices( draw->render, 0, (ushort)(count - 1) );
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index 1a9df4cac5..279f4eec63 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -59,6 +59,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+   unsigned i;
+   boolean instance_id_index = ~0;
 
    /* Add one to num_outputs because the pipeline occasionally tags on
     * an additional texcoord, eg for AA lines.
@@ -66,6 +68,15 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
    unsigned nr = MAX2( vs->info.num_inputs,
 		       vs->info.num_outputs + 1 );
 
+   /* Scan for instanceID system value.
+    */
+   for (i = 0; i < vs->info.num_inputs; i++) {
+      if (vs->info.input_semantic_name[i] == TGSI_SEMANTIC_INSTANCEID) {
+         instance_id_index = i;
+         break;
+      }
+   }
+
    fpme->prim = prim;
    fpme->opt = opt;
 
@@ -79,7 +90,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
 
    draw_pt_fetch_prepare( fpme->fetch, 
                           vs->info.num_inputs,
-			  fpme->vertex_size );
+                          fpme->vertex_size,
+                          instance_id_index );
    /* XXX: it's not really gl rasterization rules we care about here,
     * but gl vs dx9 clip spaces.
     */
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
index e3b807ebd0..00036cfe68 100644
--- a/src/gallium/auxiliary/draw/draw_vs.h
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -43,6 +43,7 @@ struct draw_varient_input
    enum pipe_format format;
    unsigned buffer;
    unsigned offset; 
+   unsigned instance_divisor;
 };
 
 struct draw_varient_output
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index d16692584e..9f40030f39 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -142,6 +142,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->fetch->run_elts( vsvg->fetch, 
                           elts,
                           count,
+                          vsvg->draw->instance_id,
                           temp_buffer );
 
    vsvg->base.vs->run_linear( vsvg->base.vs, 
@@ -181,6 +182,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
 
    vsvg->emit->run( vsvg->emit,
                     0, count,
+                    vsvg->draw->instance_id,
                     output_buffer );
 
    FREE(temp_buffer);
@@ -203,6 +205,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->fetch->run( vsvg->fetch, 
                      start,
                      count,
+                     vsvg->draw->instance_id,
                      temp_buffer );
 
    vsvg->base.vs->run_linear( vsvg->base.vs, 
@@ -239,6 +242,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    
    vsvg->emit->run( vsvg->emit,
                     0, count,
+                    vsvg->draw->instance_id,
                     output_buffer );
 
    FREE(temp_buffer);
@@ -281,9 +285,11 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
    fetch.nr_elements = key->nr_inputs;
    fetch.output_stride = vsvg->temp_vertex_stride;
    for (i = 0; i < key->nr_inputs; i++) {
+      fetch.element[i].type = TRANSLATE_ELEMENT_NORMAL;
       fetch.element[i].input_format = key->element[i].in.format;
       fetch.element[i].input_buffer = key->element[i].in.buffer;
       fetch.element[i].input_offset = key->element[i].in.offset;
+      fetch.element[i].instance_divisor = 0;
       fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
       fetch.element[i].output_offset = i * 4 * sizeof(float);
       assert(fetch.element[i].output_offset < fetch.output_stride);
@@ -295,17 +301,21 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
    for (i = 0; i < key->nr_outputs; i++) {
       if (key->element[i].out.format != EMIT_1F_PSIZE)
       {      
+         emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
          emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
          emit.element[i].input_buffer = 0;
          emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float);
+         emit.element[i].instance_divisor = 0;
          emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format);
          emit.element[i].output_offset = key->element[i].out.offset;
          assert(emit.element[i].input_offset <= fetch.output_stride);
       }
       else {
+         emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
          emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT;
          emit.element[i].input_buffer = 1;
          emit.element[i].input_offset = 0;
+         emit.element[i].instance_divisor = 0;
          emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT;
          emit.element[i].output_offset = key->element[i].out.offset;
       }
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
index 1acf3c373e..f675427d98 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
@@ -673,6 +673,13 @@ void x86_and( struct x86_function *p,
    emit_op_modrm( p, 0x23, 0x21, dst, src );
 }
 
+void x86_div( struct x86_function *p,
+              struct x86_reg src )
+{
+   assert(src.file == file_REG32 && src.mod == mod_REG);
+   emit_op_modrm(p, 0xf7, 0, x86_make_reg(file_REG32, 6), src);
+}
+
 
 
 /***********************************************************************
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
index 731a651796..f7612d416a 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
@@ -244,6 +244,7 @@ void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_sahf( struct x86_function *p );
+void x86_div( struct x86_function *p, struct x86_reg src );
 
 
 void x86_cdecl_caller_push_regs( struct x86_function *p );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index e2e5394f86..d7ff262f30 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -123,7 +123,8 @@ static const char *semantic_names[] =
    "NORMAL",
    "FACE",
    "EDGEFLAG",
-   "PRIM_ID"
+   "PRIM_ID",
+   "INSTANCEID"
 };
 
 static const char *immediate_type_names[] =
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
index e64e2b731d..2b51672b8e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
@@ -64,6 +64,7 @@ struct ureg_tokens {
 };
 
 #define UREG_MAX_INPUT PIPE_MAX_ATTRIBS
+#define UREG_MAX_SYSTEM_VALUE PIPE_MAX_ATTRIBS
 #define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS
 #define UREG_MAX_CONSTANT_RANGE 32
 #define UREG_MAX_IMMEDIATE 32
@@ -95,6 +96,13 @@ struct ureg_program
    unsigned nr_gs_inputs;
 
    struct {
+      unsigned index;
+      unsigned semantic_name;
+      unsigned semantic_index;
+   } system_value[UREG_MAX_SYSTEM_VALUE];
+   unsigned nr_system_values;
+
+   struct {
       unsigned semantic_name;
       unsigned semantic_index;
    } output[UREG_MAX_OUTPUT];
@@ -304,6 +312,25 @@ ureg_DECL_gs_input(struct ureg_program *ureg,
 }
 
 
+struct ureg_src
+ureg_DECL_system_value(struct ureg_program *ureg,
+                       unsigned index,
+                       unsigned semantic_name,
+                       unsigned semantic_index)
+{
+   if (ureg->nr_system_values < UREG_MAX_SYSTEM_VALUE) {
+      ureg->system_value[ureg->nr_system_values].index = index;
+      ureg->system_value[ureg->nr_system_values].semantic_name = semantic_name;
+      ureg->system_value[ureg->nr_system_values].semantic_index = semantic_index;
+      ureg->nr_system_values++;
+   } else {
+      set_bad(ureg);
+   }
+
+   return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, index);
+}
+
+
 struct ureg_dst 
 ureg_DECL_output( struct ureg_program *ureg,
                   unsigned name,
@@ -1058,6 +1085,15 @@ static void emit_decls( struct ureg_program *ureg )
       }
    }
 
+   for (i = 0; i < ureg->nr_system_values; i++) {
+      emit_decl(ureg,
+                TGSI_FILE_SYSTEM_VALUE,
+                ureg->system_value[i].index,
+                ureg->system_value[i].semantic_name,
+                ureg->system_value[i].semantic_index,
+                TGSI_INTERPOLATE_CONSTANT);
+   }
+
    for (i = 0; i < ureg->nr_outputs; i++) {
       emit_decl( ureg, 
                  TGSI_FILE_OUTPUT, 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
index 6f11273320..38e2fd8d0a 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
@@ -137,6 +137,12 @@ struct ureg_src
 ureg_DECL_gs_input(struct ureg_program *,
                    unsigned index);
 
+struct ureg_src
+ureg_DECL_system_value(struct ureg_program *,
+                       unsigned index,
+                       unsigned semantic_name,
+                       unsigned semantic_index);
+
 struct ureg_dst
 ureg_DECL_output( struct ureg_program *,
                   unsigned semantic_name,
diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h
index 34526eb061..54ed2c1a4b 100644
--- a/src/gallium/auxiliary/translate/translate.h
+++ b/src/gallium/auxiliary/translate/translate.h
@@ -44,12 +44,19 @@
 #include "pipe/p_format.h"
 #include "pipe/p_state.h"
 
+enum translate_element_type {
+   TRANSLATE_ELEMENT_NORMAL,
+   TRANSLATE_ELEMENT_INSTANCE_ID
+};
+
 struct translate_element 
 {
+   enum translate_element_type type;
    enum pipe_format input_format;
    enum pipe_format output_format;
    unsigned input_buffer:8;
    unsigned input_offset:24;
+   unsigned instance_divisor;
    unsigned output_offset;
 };
 
@@ -74,11 +81,13 @@ struct translate {
    void (PIPE_CDECL *run_elts)( struct translate *,
                                 const unsigned *elts,
                                 unsigned count,
+                                unsigned instance_id,
                                 void *output_buffer);
 
    void (PIPE_CDECL *run)( struct translate *,
                            unsigned start,
                            unsigned count,
+                           unsigned instance_id,
                            void *output_buffer);
 };
 
@@ -103,8 +112,13 @@ static INLINE int translate_keysize( const struct translate_key *key )
 static INLINE int translate_key_compare( const struct translate_key *a,
                                          const struct translate_key *b )
 {
-   int keysize = translate_keysize(a);
-   return memcmp(a, b, keysize);
+   int keysize_a = translate_keysize(a);
+   int keysize_b = translate_keysize(b);
+
+   if (keysize_a != keysize_b) {
+      return keysize_a - keysize_b;
+   }
+   return memcmp(a, b, keysize_a);
 }
 
 
diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c
index 266e7ee81e..24727d4988 100644
--- a/src/gallium/auxiliary/translate/translate_generic.c
+++ b/src/gallium/auxiliary/translate/translate_generic.c
@@ -46,9 +46,12 @@ struct translate_generic {
    struct translate translate;
 
    struct {
+      enum translate_element_type type;
+
       fetch_func fetch;
       unsigned buffer;
       unsigned input_offset;
+      unsigned instance_divisor;
 
       emit_func emit;
       unsigned output_offset;
@@ -568,6 +571,7 @@ static emit_func get_emit_func( enum pipe_format format )
 static void PIPE_CDECL generic_run_elts( struct translate *translate,
                                          const unsigned *elts,
                                          unsigned count,
+                                         unsigned instance_id,
                                          void *output_buffer )
 {
    struct translate_generic *tg = translate_generic(translate);
@@ -583,13 +587,20 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-
-	 const char *src = (tg->attrib[attr].input_ptr + 
-			    tg->attrib[attr].input_stride * elt);
+         const char *src;
 
 	 char *dst = (vert + 
 		      tg->attrib[attr].output_offset);
 
+         if (tg->attrib[attr].instance_divisor) {
+            src = tg->attrib[attr].input_ptr +
+                  tg->attrib[attr].input_stride *
+                  (instance_id / tg->attrib[attr].instance_divisor);
+         } else {
+            src = tg->attrib[attr].input_ptr +
+                  tg->attrib[attr].input_stride * elt;
+         }
+
 	 tg->attrib[attr].fetch( src, data );
 
          if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
@@ -607,6 +618,7 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
 static void PIPE_CDECL generic_run( struct translate *translate,
                                     unsigned start,
                                     unsigned count,
+                                    unsigned instance_id,
                                     void *output_buffer )
 {
    struct translate_generic *tg = translate_generic(translate);
@@ -623,13 +635,25 @@ static void PIPE_CDECL generic_run( struct translate *translate,
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
 
-	 const char *src = (tg->attrib[attr].input_ptr + 
-			    tg->attrib[attr].input_stride * elt);
-
 	 char *dst = (vert + 
 		      tg->attrib[attr].output_offset);
 
-	 tg->attrib[attr].fetch( src, data );
+         if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
+            const char *src;
+
+            if (tg->attrib[attr].instance_divisor) {
+               src = tg->attrib[attr].input_ptr +
+                     tg->attrib[attr].input_stride *
+                     (instance_id / tg->attrib[attr].instance_divisor);
+            } else {
+               src = tg->attrib[attr].input_ptr +
+                     tg->attrib[attr].input_stride * elt;
+            }
+
+            tg->attrib[attr].fetch( src, data );
+         } else {
+            data[0] = (float)instance_id;
+         }
 
          if (0) debug_printf("vert %d attr %d: %f %f %f %f\n",
                              i, attr, data[0], data[1], data[2], data[3]);
@@ -683,10 +707,12 @@ struct translate *translate_generic_create( const struct translate_key *key )
    tg->translate.run = generic_run;
 
    for (i = 0; i < key->nr_elements; i++) {
+      tg->attrib[i].type = key->element[i].type;
 
       tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format);
       tg->attrib[i].buffer = key->element[i].input_buffer;
       tg->attrib[i].input_offset = key->element[i].input_offset;
+      tg->attrib[i].instance_divisor = key->element[i].instance_divisor;
 
       tg->attrib[i].emit = get_emit_func(key->element[i].output_format);
       tg->attrib[i].output_offset = key->element[i].output_offset;
diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c
index b62db8d8f3..c13e742738 100644
--- a/src/gallium/auxiliary/translate/translate_sse.c
+++ b/src/gallium/auxiliary/translate/translate_sse.c
@@ -49,19 +49,29 @@
 typedef void (PIPE_CDECL *run_func)( struct translate *translate,
                                      unsigned start,
                                      unsigned count,
-                                     void *output_buffer );
+                                     unsigned instance_id,
+                                     void *output_buffer);
 
 typedef void (PIPE_CDECL *run_elts_func)( struct translate *translate,
                                           const unsigned *elts,
                                           unsigned count,
-                                          void *output_buffer );
+                                          unsigned instance_id,
+                                          void *output_buffer);
 
 struct translate_buffer {
    const void *base_ptr;
    unsigned stride;
-   void *ptr;                   /* updated per vertex */
 };
 
+struct translate_buffer_varient {
+   unsigned buffer_index;
+   unsigned instance_divisor;
+   void *ptr;                    /* updated either per vertex or per instance */
+};
+
+
+#define ELEMENT_BUFFER_INSTANCE_ID  1001
+
 
 struct translate_sse {
    struct translate translate;
@@ -81,6 +91,16 @@ struct translate_sse {
    struct translate_buffer buffer[PIPE_MAX_ATTRIBS];
    unsigned nr_buffers;
 
+   /* Multiple buffer varients can map to a single buffer. */
+   struct translate_buffer_varient buffer_varient[PIPE_MAX_ATTRIBS];
+   unsigned nr_buffer_varients;
+
+   /* Multiple elements can map to a single buffer varient. */
+   unsigned element_to_buffer_varient[PIPE_MAX_ATTRIBS];
+
+   boolean use_instancing;
+   unsigned instance_id;
+
    run_func      gen_run;
    run_elts_func gen_run_elts;
 
@@ -359,32 +379,61 @@ static boolean init_inputs( struct translate_sse *p,
                             boolean linear )
 {
    unsigned i;
-   if (linear) {
-      for (i = 0; i < p->nr_buffers; i++) {
+   struct x86_reg instance_id = x86_make_disp(p->machine_EDX,
+                                              get_offset(p, &p->instance_id));
+
+   for (i = 0; i < p->nr_buffer_varients; i++) {
+      struct translate_buffer_varient *varient = &p->buffer_varient[i];
+      struct translate_buffer *buffer = &p->buffer[varient->buffer_index];
+
+      if (linear || varient->instance_divisor) {
          struct x86_reg buf_stride   = x86_make_disp(p->machine_EDX,
-                                                     get_offset(p, &p->buffer[i].stride));
+                                                     get_offset(p, &buffer->stride));
          struct x86_reg buf_ptr      = x86_make_disp(p->machine_EDX,
-                                                     get_offset(p, &p->buffer[i].ptr));
+                                                     get_offset(p, &varient->ptr));
          struct x86_reg buf_base_ptr = x86_make_disp(p->machine_EDX,
-                                                     get_offset(p, &p->buffer[i].base_ptr));
+                                                     get_offset(p, &buffer->base_ptr));
          struct x86_reg elt = p->idx_EBX;
-         struct x86_reg tmp = p->tmp_EAX;
-
+         struct x86_reg tmp_EAX = p->tmp_EAX;
 
          /* Calculate pointer to first attrib:
+          *   base_ptr + stride * index, where index depends on instance divisor
           */
-         x86_mov(p->func, tmp, buf_stride);
-         x86_imul(p->func, tmp, elt);
-         x86_add(p->func, tmp, buf_base_ptr);
+         if (varient->instance_divisor) {
+            /* Our index is instance ID divided by instance divisor.
+             */
+            x86_mov(p->func, tmp_EAX, instance_id);
+
+            if (varient->instance_divisor != 1) {
+               struct x86_reg tmp_EDX = p->machine_EDX;
+               struct x86_reg tmp_ECX = p->outbuf_ECX;
+
+               /* TODO: Add x86_shr() to rtasm and use it whenever
+                *       instance divisor is power of two.
+                */
+
+               x86_push(p->func, tmp_EDX);
+               x86_push(p->func, tmp_ECX);
+               x86_xor(p->func, tmp_EDX, tmp_EDX);
+               x86_mov_reg_imm(p->func, tmp_ECX, varient->instance_divisor);
+               x86_div(p->func, tmp_ECX);    /* EAX = EDX:EAX / ECX */
+               x86_pop(p->func, tmp_ECX);
+               x86_pop(p->func, tmp_EDX);
+            }
+         } else {
+            x86_mov(p->func, tmp_EAX, elt);
+         }
+         x86_imul(p->func, tmp_EAX, buf_stride);
+         x86_add(p->func, tmp_EAX, buf_base_ptr);
 
 
          /* In the linear case, keep the buffer pointer instead of the
           * index number.
           */
-         if (p->nr_buffers == 1) 
-            x86_mov( p->func, elt, tmp );
+         if (linear && p->nr_buffer_varients == 1)
+            x86_mov(p->func, elt, tmp_EAX);
          else
-            x86_mov( p->func, buf_ptr, tmp );
+            x86_mov(p->func, buf_ptr, tmp_EAX);
       }
    }
 
@@ -394,31 +443,36 @@ static boolean init_inputs( struct translate_sse *p,
 
 static struct x86_reg get_buffer_ptr( struct translate_sse *p,
                                       boolean linear,
-                                      unsigned buf_idx,
+                                      unsigned var_idx,
                                       struct x86_reg elt )
 {
-   if (linear && p->nr_buffers == 1) {
+   if (var_idx == ELEMENT_BUFFER_INSTANCE_ID) {
+      return x86_make_disp(p->machine_EDX,
+                           get_offset(p, &p->instance_id));
+   }
+   if (linear && p->nr_buffer_varients == 1) {
       return p->idx_EBX;
    }
-   else if (linear) {
+   else if (linear || p->buffer_varient[var_idx].instance_divisor) {
       struct x86_reg ptr = p->tmp_EAX;
       struct x86_reg buf_ptr = 
          x86_make_disp(p->machine_EDX, 
-                       get_offset(p, &p->buffer[buf_idx].ptr));
+                       get_offset(p, &p->buffer_varient[var_idx].ptr));
       
       x86_mov(p->func, ptr, buf_ptr);
       return ptr;
    }
    else {
       struct x86_reg ptr = p->tmp_EAX;
+      const struct translate_buffer_varient *varient = &p->buffer_varient[var_idx];
 
       struct x86_reg buf_stride = 
          x86_make_disp(p->machine_EDX, 
-                       get_offset(p, &p->buffer[buf_idx].stride));
+                       get_offset(p, &p->buffer[varient->buffer_index].stride));
 
       struct x86_reg buf_base_ptr = 
          x86_make_disp(p->machine_EDX, 
-                       get_offset(p, &p->buffer[buf_idx].base_ptr));
+                       get_offset(p, &p->buffer[varient->buffer_index].base_ptr));
 
 
 
@@ -436,28 +490,33 @@ static struct x86_reg get_buffer_ptr( struct translate_sse *p,
 static boolean incr_inputs( struct translate_sse *p, 
                             boolean linear )
 {
-   if (linear && p->nr_buffers == 1) {
+   if (linear && p->nr_buffer_varients == 1) {
       struct x86_reg stride = x86_make_disp(p->machine_EDX,
                                             get_offset(p, &p->buffer[0].stride));
 
-      x86_add(p->func, p->idx_EBX, stride);
-      sse_prefetchnta(p->func, x86_make_disp(p->idx_EBX, 192));
+      if (p->buffer_varient[0].instance_divisor == 0) {
+         x86_add(p->func, p->idx_EBX, stride);
+         sse_prefetchnta(p->func, x86_make_disp(p->idx_EBX, 192));
+      }
    }
    else if (linear) {
       unsigned i;
 
       /* Is this worthwhile??
        */
-      for (i = 0; i < p->nr_buffers; i++) {
+      for (i = 0; i < p->nr_buffer_varients; i++) {
+         struct translate_buffer_varient *varient = &p->buffer_varient[i];
          struct x86_reg buf_ptr = x86_make_disp(p->machine_EDX,
-                                                get_offset(p, &p->buffer[i].ptr));
+                                                get_offset(p, &varient->ptr));
          struct x86_reg buf_stride = x86_make_disp(p->machine_EDX,
-                                                   get_offset(p, &p->buffer[i].stride));
+                                                   get_offset(p, &p->buffer[varient->buffer_index].stride));
 
-         x86_mov(p->func, p->tmp_EAX, buf_ptr);
-         x86_add(p->func, p->tmp_EAX, buf_stride);
-         if (i == 0) sse_prefetchnta(p->func, x86_make_disp(p->tmp_EAX, 192));
-         x86_mov(p->func, buf_ptr, p->tmp_EAX);
+         if (varient->instance_divisor == 0) {
+            x86_mov(p->func, p->tmp_EAX, buf_ptr);
+            x86_add(p->func, p->tmp_EAX, buf_stride);
+            if (i == 0) sse_prefetchnta(p->func, x86_make_disp(p->tmp_EAX, 192));
+            x86_mov(p->func, buf_ptr, p->tmp_EAX);
+         }
       }
    } 
    else {
@@ -514,7 +573,18 @@ static boolean build_vertex_emit( struct translate_sse *p,
    x86_mov(p->func, p->machine_EDX, x86_fn_arg(p->func, 1));
    x86_mov(p->func, p->idx_EBX, x86_fn_arg(p->func, 2));
    x86_mov(p->func, p->count_ESI, x86_fn_arg(p->func, 3));
-   x86_mov(p->func, p->outbuf_ECX, x86_fn_arg(p->func, 4));
+   x86_mov(p->func, p->outbuf_ECX, x86_fn_arg(p->func, 5));
+
+   /* Load instance ID.
+    */
+   if (p->use_instancing) {
+      x86_mov(p->func,
+              p->tmp_EAX,
+              x86_fn_arg(p->func, 4));
+      x86_mov(p->func,
+              x86_make_disp(p->machine_EDX, get_offset(p, &p->instance_id)),
+              p->tmp_EAX);
+   }
 
    /* Get vertex count, compare to zero
     */
@@ -531,17 +601,18 @@ static boolean build_vertex_emit( struct translate_sse *p,
    label = x86_get_label(p->func);
    {
       struct x86_reg elt = linear ? p->idx_EBX : x86_deref(p->idx_EBX);
-      int last_vb = -1;
+      int last_varient = -1;
       struct x86_reg vb;
 
       for (j = 0; j < p->translate.key.nr_elements; j++) {
          const struct translate_element *a = &p->translate.key.element[j];
+         unsigned varient = p->element_to_buffer_varient[j];
 
          /* Figure out source pointer address:
           */
-         if (a->input_buffer != last_vb) {
-            last_vb = a->input_buffer;
-            vb = get_buffer_ptr(p, linear, a->input_buffer, elt);
+         if (varient != last_varient) {
+            last_varient = varient;
+            vb = get_buffer_ptr(p, linear, varient, elt);
          }
          
          if (!translate_attr( p, a, 
@@ -624,6 +695,7 @@ static void translate_sse_release( struct translate *translate )
 static void PIPE_CDECL translate_sse_run_elts( struct translate *translate,
 			      const unsigned *elts,
 			      unsigned count,
+                              unsigned instance_id,
 			      void *output_buffer )
 {
    struct translate_sse *p = (struct translate_sse *)translate;
@@ -631,12 +703,14 @@ static void PIPE_CDECL translate_sse_run_elts( struct translate *translate,
    p->gen_run_elts( translate,
 		    elts,
 		    count,
-		    output_buffer );
+                    instance_id,
+                    output_buffer);
 }
 
 static void PIPE_CDECL translate_sse_run( struct translate *translate,
 			 unsigned start,
 			 unsigned count,
+                         unsigned instance_id,
 			 void *output_buffer )
 {
    struct translate_sse *p = (struct translate_sse *)translate;
@@ -644,7 +718,8 @@ static void PIPE_CDECL translate_sse_run( struct translate *translate,
    p->gen_run( translate,
 	       start,
 	       count,
-	       output_buffer );
+               instance_id,
+               output_buffer);
 }
 
 
@@ -666,8 +741,37 @@ struct translate *translate_sse2_create( const struct translate_key *key )
    p->translate.run_elts = translate_sse_run_elts;
    p->translate.run = translate_sse_run;
 
-   for (i = 0; i < key->nr_elements; i++) 
-      p->nr_buffers = MAX2( p->nr_buffers, key->element[i].input_buffer + 1 );
+   for (i = 0; i < key->nr_elements; i++) {
+      if (key->element[i].type == TRANSLATE_ELEMENT_NORMAL) {
+         unsigned j;
+
+         p->nr_buffers = MAX2(p->nr_buffers, key->element[i].input_buffer + 1);
+
+         if (key->element[i].instance_divisor) {
+            p->use_instancing = TRUE;
+         }
+
+         /*
+          * Map vertex element to vertex buffer varient.
+          */
+         for (j = 0; j < p->nr_buffer_varients; j++) {
+            if (p->buffer_varient[j].buffer_index == key->element[i].input_buffer &&
+                p->buffer_varient[j].instance_divisor == key->element[i].instance_divisor) {
+               break;
+            }
+         }
+         if (j == p->nr_buffer_varients) {
+            p->buffer_varient[j].buffer_index = key->element[i].input_buffer;
+            p->buffer_varient[j].instance_divisor = key->element[i].instance_divisor;
+            p->nr_buffer_varients++;
+         }
+         p->element_to_buffer_varient[i] = j;
+      } else {
+         assert(key->element[i].type == TRANSLATE_ELEMENT_INSTANCE_ID);
+
+         p->element_to_buffer_varient[i] = ELEMENT_BUFFER_INSTANCE_ID;
+      }
+   }
 
    if (0) debug_printf("nr_buffers: %d\n", p->nr_buffers);
 
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index f3ac6760db..3ac807d4b5 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -256,6 +256,8 @@ softpipe_create( struct pipe_screen *screen )
    softpipe->pipe.draw_arrays = softpipe_draw_arrays;
    softpipe->pipe.draw_elements = softpipe_draw_elements;
    softpipe->pipe.draw_range_elements = softpipe_draw_range_elements;
+   softpipe->pipe.draw_arrays_instanced = softpipe_draw_arrays_instanced;
+   softpipe->pipe.draw_elements_instanced = softpipe_draw_elements_instanced;
 
    softpipe->pipe.clear = softpipe_clear;
    softpipe->pipe.flush = softpipe_flush;
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index 03d35fb3cb..2a27e5ce64 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -98,19 +98,41 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp)
 }
 
 
+/**
+ * Draw vertex arrays, with optional indexing.
+ * Basically, map the vertex buffers (and drawing surfaces), then hand off
+ * the drawing to the 'draw' module.
+ */
+static void
+softpipe_draw_range_elements_instanced(struct pipe_context *pipe,
+                                       struct pipe_buffer *indexBuffer,
+                                       unsigned indexSize,
+                                       unsigned minIndex,
+                                       unsigned maxIndex,
+                                       unsigned mode,
+                                       unsigned start,
+                                       unsigned count,
+                                       unsigned startInstance,
+                                       unsigned instanceCount);
+
+
 void
 softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
                      unsigned start, unsigned count)
 {
-   softpipe_draw_elements(pipe, NULL, 0, mode, start, count);
+   softpipe_draw_range_elements_instanced(pipe,
+                                          NULL,
+                                          0,
+                                          0,
+                                          0xffffffff,
+                                          mode,
+                                          start,
+                                          count,
+                                          0,
+                                          1);
 }
 
 
-/**
- * Draw vertex arrays, with optional indexing.
- * Basically, map the vertex buffers (and drawing surfaces), then hand off
- * the drawing to the 'draw' module.
- */
 void
 softpipe_draw_range_elements(struct pipe_context *pipe,
                              struct pipe_buffer *indexBuffer,
@@ -119,6 +141,91 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
                              unsigned max_index,
                              unsigned mode, unsigned start, unsigned count)
 {
+   softpipe_draw_range_elements_instanced(pipe,
+                                          indexBuffer,
+                                          indexSize,
+                                          min_index,
+                                          max_index,
+                                          mode,
+                                          start,
+                                          count,
+                                          0,
+                                          1);
+}
+
+
+void
+softpipe_draw_elements(struct pipe_context *pipe,
+                       struct pipe_buffer *indexBuffer,
+                       unsigned indexSize,
+                       unsigned mode, unsigned start, unsigned count)
+{
+   softpipe_draw_range_elements_instanced(pipe,
+                                          indexBuffer,
+                                          indexSize,
+                                          0,
+                                          0xffffffff,
+                                          mode,
+                                          start,
+                                          count,
+                                          0,
+                                          1);
+}
+
+void
+softpipe_draw_arrays_instanced(struct pipe_context *pipe,
+                               unsigned mode,
+                               unsigned start,
+                               unsigned count,
+                               unsigned startInstance,
+                               unsigned instanceCount)
+{
+   softpipe_draw_range_elements_instanced(pipe,
+                                          NULL,
+                                          0,
+                                          0,
+                                          0xffffffff,
+                                          mode,
+                                          start,
+                                          count,
+                                          startInstance,
+                                          instanceCount);
+}
+
+void
+softpipe_draw_elements_instanced(struct pipe_context *pipe,
+                                 struct pipe_buffer *indexBuffer,
+                                 unsigned indexSize,
+                                 unsigned mode,
+                                 unsigned start,
+                                 unsigned count,
+                                 unsigned startInstance,
+                                 unsigned instanceCount)
+{
+   softpipe_draw_range_elements_instanced(pipe,
+                                          indexBuffer,
+                                          indexSize,
+                                          0,
+                                          0xffffffff,
+                                          mode,
+                                          start,
+                                          count,
+                                          startInstance,
+                                          instanceCount);
+}
+
+static void
+softpipe_draw_range_elements_instanced(struct pipe_context *pipe,
+                                       struct pipe_buffer *indexBuffer,
+                                       unsigned indexSize,
+                                       unsigned minIndex,
+                                       unsigned maxIndex,
+                                       unsigned mode,
+                                       unsigned start,
+                                       unsigned count,
+                                       unsigned startInstance,
+                                       unsigned instanceCount)
+{
    struct softpipe_context *sp = softpipe_context(pipe);
    struct draw_context *draw = sp->draw;
    unsigned i;
@@ -128,45 +235,48 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
 
    sp->reduced_api_prim = u_reduced_prim(mode);
 
-   if (sp->dirty)
-      softpipe_update_derived( sp );
+   if (sp->dirty) {
+      softpipe_update_derived(sp);
+   }
 
    softpipe_map_transfers(sp);
    softpipe_map_constant_buffers(sp);
 
-   /*
-    * Map vertex buffers
-    */
+   /* Map vertex buffers */
    for (i = 0; i < sp->num_vertex_buffers; i++) {
-      void *buf
-         = pipe_buffer_map(pipe->screen,
-                                    sp->vertex_buffer[i].buffer,
-                                    PIPE_BUFFER_USAGE_CPU_READ);
+      void *buf;
+
+      buf = pipe_buffer_map(pipe->screen,
+                            sp->vertex_buffer[i].buffer,
+                            PIPE_BUFFER_USAGE_CPU_READ);
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
 
    /* Map index buffer, if present */
    if (indexBuffer) {
-      void *mapped_indexes
-         = pipe_buffer_map(pipe->screen, indexBuffer,
-                                    PIPE_BUFFER_USAGE_CPU_READ);
-      draw_set_mapped_element_buffer_range(draw, indexSize,
-                                           min_index,
-                                           max_index,
+      void *mapped_indexes;
+
+      mapped_indexes = pipe_buffer_map(pipe->screen,
+                                       indexBuffer,
+                                       PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer_range(draw,
+                                           indexSize,
+                                           minIndex,
+                                           maxIndex,
                                            mapped_indexes);
-   }
-   else {
+   } else {
       /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw, 0, start,
-                                           start + count - 1, NULL);
+      draw_set_mapped_element_buffer_range(draw,
+                                           0,
+                                           start,
+                                           start + count - 1,
+                                           NULL);
    }
 
    /* draw! */
-   draw_arrays(draw, mode, start, count);
+   draw_arrays_instanced(draw, mode, start, count, startInstance, instanceCount);
 
-   /*
-    * unmap vertex/index buffers - will cause draw module to flush
-    */
+   /* unmap vertex/index buffers - will cause draw module to flush */
    for (i = 0; i < sp->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
       pipe_buffer_unmap(pipe->screen, sp->vertex_buffer[i].buffer);
@@ -176,22 +286,8 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
       pipe_buffer_unmap(pipe->screen, indexBuffer);
    }
 
-
    /* Note: leave drawing surfaces mapped */
    softpipe_unmap_constant_buffers(sp);
 
    sp->dirty_render_cache = TRUE;
 }
-
-
-void
-softpipe_draw_elements(struct pipe_context *pipe,
-                       struct pipe_buffer *indexBuffer,
-                       unsigned indexSize,
-                       unsigned mode, unsigned start, unsigned count)
-{
-   softpipe_draw_range_elements( pipe, indexBuffer,
-                                 indexSize,
-                                 0, 0xffffffff,
-                                 mode, start, count );
-}
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 9b18dac67b..0f9b1546df 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -200,6 +200,24 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
                              unsigned mode, unsigned start, unsigned count);
 
 void
+softpipe_draw_arrays_instanced(struct pipe_context *pipe,
+                               unsigned mode,
+                               unsigned start,
+                               unsigned count,
+                               unsigned startInstance,
+                               unsigned instanceCount);
+
+void
+softpipe_draw_elements_instanced(struct pipe_context *pipe,
+                                 struct pipe_buffer *indexBuffer,
+                                 unsigned indexSize,
+                                 unsigned mode,
+                                 unsigned start,
+                                 unsigned count,
+                                 unsigned startInstance,
+                                 unsigned instanceCount);
+
+void
 softpipe_map_transfers(struct softpipe_context *sp);
 
 void
diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c
index ae1e77e7d4..fef652c0c0 100644
--- a/src/gallium/drivers/svga/svga_state_vs.c
+++ b/src/gallium/drivers/svga/svga_state_vs.c
@@ -200,10 +200,12 @@ static int update_zero_stride( struct svga_context *svga,
 
          key.output_stride = 4 * sizeof(float);
          key.nr_elements = 1;
+         key.element[0].type = TRANSLATE_ELEMENT_NORMAL;
          key.element[0].input_format = vel->src_format;
          key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
          key.element[0].input_buffer = vel->vertex_buffer_index;
          key.element[0].input_offset = vel->src_offset;
+         key.element[0].instance_divisor = vel->instance_divisor;
          key.element[0].output_offset = const_idx * 4 * sizeof(float);
 
          translate_key_sanitize(&key);
@@ -222,7 +224,7 @@ static int update_zero_stride( struct svga_context *svga,
          translate->set_buffer(translate, vel->vertex_buffer_index,
                                mapped_buffer,
                                vbuffer->stride);
-         translate->run(translate, 0, 1,
+         translate->run(translate, 0, 1, 0,
                         svga->curr.zero_stride_constants);
 
          pipe_buffer_unmap(svga->pipe.screen,
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index d2f8085b42..497b35a680 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -69,6 +69,22 @@ struct pipe_context {
                           unsigned indexSize,
                           unsigned mode, unsigned start, unsigned count);
 
+   void (*draw_arrays_instanced)(struct pipe_context *pipe,
+                                 unsigned mode,
+                                 unsigned start,
+                                 unsigned count,
+                                 unsigned startInstance,
+                                 unsigned instanceCount);
+
+   void (*draw_elements_instanced)(struct pipe_context *pipe,
+                                   struct pipe_buffer *indexBuffer,
+                                   unsigned indexSize,
+                                   unsigned mode,
+                                   unsigned start,
+                                   unsigned count,
+                                   unsigned startInstance,
+                                   unsigned instanceCount);
+
    /* XXX: this is (probably) a temporary entrypoint, as the range
     * information should be available from the vertex_buffer state.
     * Using this to quickly evaluate a specialized path in the draw
diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
index 550e2abc32..b489b04466 100644
--- a/src/gallium/include/pipe/p_shader_tokens.h
+++ b/src/gallium/include/pipe/p_shader_tokens.h
@@ -121,17 +121,18 @@ struct tgsi_declaration_range
    unsigned Last    : 16; /**< UINT */
 };
 
-#define TGSI_SEMANTIC_POSITION  0
-#define TGSI_SEMANTIC_COLOR     1
-#define TGSI_SEMANTIC_BCOLOR    2 /**< back-face color */
-#define TGSI_SEMANTIC_FOG       3
-#define TGSI_SEMANTIC_PSIZE     4
-#define TGSI_SEMANTIC_GENERIC   5
-#define TGSI_SEMANTIC_NORMAL    6
-#define TGSI_SEMANTIC_FACE      7
-#define TGSI_SEMANTIC_EDGEFLAG  8
-#define TGSI_SEMANTIC_PRIMID    9
-#define TGSI_SEMANTIC_COUNT    10 /**< number of semantic values */
+#define TGSI_SEMANTIC_POSITION   0
+#define TGSI_SEMANTIC_COLOR      1
+#define TGSI_SEMANTIC_BCOLOR     2  /**< back-face color */
+#define TGSI_SEMANTIC_FOG        3
+#define TGSI_SEMANTIC_PSIZE      4
+#define TGSI_SEMANTIC_GENERIC    5
+#define TGSI_SEMANTIC_NORMAL     6
+#define TGSI_SEMANTIC_FACE       7
+#define TGSI_SEMANTIC_EDGEFLAG   8
+#define TGSI_SEMANTIC_PRIMID     9
+#define TGSI_SEMANTIC_INSTANCEID 10
+#define TGSI_SEMANTIC_COUNT      11 /**< number of semantic values */
 
 struct tgsi_declaration_semantic
 {
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 66d6cfa11e..e2885eb2b8 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -372,6 +372,11 @@ struct pipe_vertex_element
    /** Offset of this attribute, in bytes, from the start of the vertex */
    unsigned src_offset;
 
+   /** Instance data rate divisor. 0 means this is per-vertex data,
+    *  n means per-instance data used for n consecutive instances (n > 0).
+    */
+   unsigned instance_divisor;
+
    /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does
     * this attribute live in?
     */