Merge branch 'draw-instanced'

Conflicts:
	src/gallium/auxiliary/draw/draw_llvm.c
	src/gallium/drivers/llvmpipe/lp_state_fs.c
	src/glsl/ir_set_program_inouts.cpp
	src/mesa/tnl/t_vb_program.c

33 files changed, 347 insertions, 23 deletions

diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
index 943ec44bcde..41269ee869d 100644
--- a/src/gallium/auxiliary/draw/draw_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -437,6 +437,7 @@ generate_vs(struct draw_llvm *llvm,
             LLVMBuilderRef builder,
             LLVMValueRef (*outputs)[NUM_CHANNELS],
             const LLVMValueRef (*inputs)[NUM_CHANNELS],
+            LLVMValueRef system_values_array,
             LLVMValueRef context_ptr,
             struct lp_build_sampler_soa *draw_sampler)
 {
@@ -468,6 +469,7 @@ generate_vs(struct draw_llvm *llvm,
                      vs_type,
                      NULL /*struct lp_build_mask_context *mask*/,
                      consts_ptr,
+                     system_values_array,
                      NULL /*pos*/,
                      inputs,
                      outputs,
@@ -1118,7 +1120,9 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
    LLVMValueRef start, end, count, stride, step, io_itr;
    LLVMValueRef io_ptr, vbuffers_ptr, vb_ptr;
    LLVMValueRef instance_id;
+   LLVMValueRef system_values_array;
    struct draw_context *draw = llvm->draw;
+   const struct tgsi_shader_info *vs_info = &draw->vs.vertex_shader->info;
    unsigned i, j;
    struct lp_build_context bld;
    struct lp_build_loop_state lp_loop;
@@ -1179,6 +1183,9 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
 
    lp_build_context_init(&bld, llvm->gallivm, lp_type_int(32));
 
+   system_values_array = lp_build_system_values_array(gallivm, vs_info,
+                                                      instance_id, NULL);
+
    end = lp_build_add(&bld, start, count);
 
    step = lp_build_const_int32(gallivm, max_vertices);
@@ -1233,6 +1240,7 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
                   builder,
                   outputs,
                   ptr_aos,
+                  system_values_array,
                   context_ptr,
                   sampler);
 
@@ -1263,8 +1271,7 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
 
       /* store clipmask in vertex header and positions in data */
       convert_to_aos(gallivm, io, outputs, clipmask,
-                     draw->vs.vertex_shader->info.num_outputs,
-                     max_vertices);
+                     vs_info->num_outputs, max_vertices);
    }
 
    lp_build_loop_end_cond(&lp_loop, end, step, LLVMIntUGE);
@@ -1315,7 +1322,9 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
    LLVMValueRef fetch_elts, fetch_count, stride, step, io_itr;
    LLVMValueRef io_ptr, vbuffers_ptr, vb_ptr;
    LLVMValueRef instance_id;
+   LLVMValueRef system_values_array;
    struct draw_context *draw = llvm->draw;
+   const struct tgsi_shader_info *vs_info = &draw->vs.vertex_shader->info;
    unsigned i, j;
    struct lp_build_context bld;
    struct lp_build_loop_state lp_loop;
@@ -1376,6 +1385,10 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
 
    lp_build_context_init(&bld, gallivm, lp_type_int(32));
 
+   system_values_array = lp_build_system_values_array(gallivm, vs_info,
+                                                      instance_id, NULL);
+
+
    step = lp_build_const_int32(gallivm, max_vertices);
 
    /* code generated texture sampling */
@@ -1438,6 +1451,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
                   builder,
                   outputs,
                   ptr_aos,
+                  system_values_array,
                   context_ptr,
                   sampler);
 
@@ -1471,8 +1485,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
        * and transformed positions in data 
        */   
       convert_to_aos(gallivm, io, outputs, clipmask,
-                     draw->vs.vertex_shader->info.num_outputs,
-                     max_vertices);
+                     vs_info->num_outputs, max_vertices);
    }
 
    lp_build_loop_end_cond(&lp_loop, fetch_count, step, LLVMIntUGE);
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index 667eb507855..c41d7c42a01 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -99,6 +99,12 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
                                   constants, const_size);
 
+   if (shader->info.uses_instanceid) {
+      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
+      assert(i < Elements(machine->SystemValue));
+      machine->SystemValue[i][0] = shader->draw->instance_id;
+   }
+
    for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
       unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index dee7c0da9b6..d55b9b08070 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -71,6 +71,12 @@ vs_sse_prepare( struct draw_vertex_shader *base,
    struct tgsi_exec_machine *machine = shader->machine;
 
    machine->Samplers = draw->vs.samplers;
+
+   if (base->info.uses_instanceid) {
+      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
+      assert(i < Elements(machine->SystemValue));
+      machine->SystemValue[i][0] = base->draw->instance_id;
+   }
 }
 
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
index 40186befb9f..9713d100484 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
@@ -180,6 +180,7 @@ lp_build_tgsi_soa(struct gallivm_state *gallivm,
                   struct lp_type type,
                   struct lp_build_mask_context *mask,
                   LLVMValueRef consts_ptr,
+                  LLVMValueRef system_values_array,
                   const LLVMValueRef *pos,
                   const LLVMValueRef (*inputs)[4],
                   LLVMValueRef (*outputs)[4],
@@ -199,4 +200,11 @@ lp_build_tgsi_aos(struct gallivm_state *gallivm,
                   const struct tgsi_shader_info *info);
 
 
+LLVMValueRef
+lp_build_system_values_array(struct gallivm_state *gallivm,
+                             const struct tgsi_shader_info *info,
+                             LLVMValueRef instance_id,
+                             LLVMValueRef facing);
+
+
 #endif /* LP_BLD_TGSI_H */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index 1b5a8a5903b..d1585c8e2b7 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -157,6 +157,8 @@ struct lp_build_tgsi_soa_context
     */
    LLVMValueRef inputs_array;
 
+   LLVMValueRef system_values_array;
+
    const struct tgsi_shader_info *info;
    /** bitmask indicating which register files are accessed indirectly */
    unsigned indirect_files;
@@ -759,6 +761,23 @@ emit_fetch(
       }
       break;
 
+   case TGSI_FILE_SYSTEM_VALUE:
+      assert(!reg->Register.Indirect);
+      {
+         LLVMValueRef index;  /* index into the system value array */
+         LLVMValueRef scalar, scalar_ptr;
+
+         index = lp_build_const_int32(gallivm,
+                                      reg->Register.Index * 4 + swizzle);
+
+         scalar_ptr = LLVMBuildGEP(builder, bld->system_values_array,
+                                   &index, 1, "");
+         scalar = LLVMBuildLoad(builder, scalar_ptr, "");
+
+         res = lp_build_broadcast_scalar(&bld->base, scalar);
+      }
+      break;
+
    default:
       assert(0 && "invalid src register in emit_fetch()");
       return bld->base.undef;
@@ -2322,6 +2341,7 @@ lp_build_tgsi_soa(struct gallivm_state *gallivm,
                   struct lp_type type,
                   struct lp_build_mask_context *mask,
                   LLVMValueRef consts_ptr,
+                  LLVMValueRef system_values_array,
                   const LLVMValueRef *pos,
                   const LLVMValueRef (*inputs)[NUM_CHANNELS],
                   LLVMValueRef (*outputs)[NUM_CHANNELS],
@@ -2411,6 +2431,8 @@ lp_build_tgsi_soa(struct gallivm_state *gallivm,
       }
    }
 
+   bld.system_values_array = system_values_array;
+
    tgsi_parse_init( &parse, tokens );
 
    while( !tgsi_parse_end_of_tokens( &parse ) ) {
@@ -2512,3 +2534,54 @@ lp_build_tgsi_soa(struct gallivm_state *gallivm,
    FREE( bld.instructions );
 }
 
+
+/**
+ * Build up the system values array out of individual values such as
+ * the instance ID, front-face, primitive ID, etc.  The shader info is
+ * used to determine which system values are needed and where to put
+ * them in the system values array.
+ *
+ * XXX only instance ID is implemented at this time.
+ *
+ * The system values register file is similar to the constants buffer.
+ * Example declaration:
+ *    DCL SV[0], INSTANCEID
+ * Example instruction:
+ *    MOVE foo, SV[0].xxxx;
+ *
+ * \return  LLVM float array (interpreted as float [][4])
+ */
+LLVMValueRef
+lp_build_system_values_array(struct gallivm_state *gallivm,
+                             const struct tgsi_shader_info *info,
+                             LLVMValueRef instance_id,
+                             LLVMValueRef facing)
+{
+   LLVMValueRef size = lp_build_const_int32(gallivm, 4 * info->num_system_values);
+   LLVMTypeRef float_t = LLVMFloatTypeInContext(gallivm->context);
+   LLVMValueRef array = lp_build_array_alloca(gallivm, float_t,
+                                              size, "sysvals_array");
+   unsigned i;
+
+   for (i = 0; i < info->num_system_values; i++) {
+      LLVMValueRef index = lp_build_const_int32(gallivm, i * 4);
+      LLVMValueRef ptr, value;
+
+      switch (info->system_value_semantic_name[i]) {
+      case TGSI_SEMANTIC_INSTANCEID:
+         /* convert instance ID from int to float */
+         value = LLVMBuildSIToFP(gallivm->builder, instance_id, float_t,
+                                 "sysval_instanceid");
+         break;
+      case TGSI_SEMANTIC_FACE:
+         /* fall-through */
+      default:
+         assert(0 && "unexpected semantic in build_system_values_array()");
+      }
+
+      ptr = LLVMBuildGEP(gallivm->builder, array, &index, 1, "");
+      LLVMBuildStore(gallivm->builder, value, ptr);
+   }
+      
+   return array;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 7892a67f04c..35b27423513 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -1038,7 +1038,6 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
       break;
 
    case TGSI_FILE_INPUT:
-   case TGSI_FILE_SYSTEM_VALUE:
       for (i = 0; i < QUAD_SIZE; i++) {
          /*
          if (TGSI_PROCESSOR_GEOMETRY == mach->Processor) {
@@ -1053,6 +1052,15 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
       }
       break;
 
+   case TGSI_FILE_SYSTEM_VALUE:
+      /* XXX no swizzling at this point.  Will be needed if we put
+       * gl_FragCoord, for example, in a sys value register.
+       */
+      for (i = 0; i < QUAD_SIZE; i++) {
+         chan->f[i] = mach->SystemValue[index->i[i]][0];
+      }
+      break;
+
    case TGSI_FILE_TEMPORARY:
       for (i = 0; i < QUAD_SIZE; i++) {
          assert(index->i[i] < TGSI_EXEC_NUM_TEMPS);
@@ -1907,8 +1915,7 @@ exec_declaration(struct tgsi_exec_machine *mach,
                  const struct tgsi_full_declaration *decl)
 {
    if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) {
-      if (decl->Declaration.File == TGSI_FILE_INPUT ||
-          decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE) {
+      if (decl->Declaration.File == TGSI_FILE_INPUT) {
          uint first, last, mask;
 
          first = decl->Range.First;
@@ -1921,6 +1928,7 @@ exec_declaration(struct tgsi_exec_machine *mach,
           * ureg code to emit the right UsageMask value (WRITEMASK_X).
           * Then, we could remove the tgsi_exec_machine::Face field.
           */
+         /* XXX make FACE a system value */
          if (decl->Semantic.Name == TGSI_SEMANTIC_FACE) {
             uint i;
 
@@ -1962,8 +1970,13 @@ exec_declaration(struct tgsi_exec_machine *mach,
          }
       }
    }
+
+   if (decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE) {
+      mach->SysSemanticToIndex[decl->Declaration.Semantic] = decl->Range.First;
+   }
 }
 
+
 typedef void (* micro_op)(union tgsi_exec_channel *dst);
 
 static void
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index b5ebbfbfaab..6c204c73714 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -31,6 +31,7 @@
 
 #include "pipe/p_compiler.h"
 #include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
 
 #if defined __cplusplus
 extern "C" {
@@ -181,6 +182,8 @@ struct tgsi_sampler
 /* The maximum total number of vertices */
 #define TGSI_MAX_TOTAL_VERTICES (TGSI_MAX_PRIM_VERTICES * TGSI_MAX_PRIMITIVES * PIPE_MAX_ATTRIBS)
 
+#define TGSI_MAX_MISC_INPUTS 8
+
 /** function call/activation record */
 struct tgsi_call_record
 {
@@ -228,6 +231,10 @@ struct tgsi_exec_machine
    struct tgsi_exec_vector       Inputs[TGSI_MAX_PRIM_VERTICES * PIPE_MAX_ATTRIBS];
    struct tgsi_exec_vector       Outputs[TGSI_MAX_TOTAL_VERTICES];
 
+   /* System values */
+   unsigned                      SysSemanticToIndex[TGSI_SEMANTIC_COUNT];
+   float                         SystemValue[TGSI_MAX_MISC_INPUTS][4];
+
    struct tgsi_exec_vector       *Addrs;
    struct tgsi_exec_vector       *Predicates;
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
index 3521847b619..537a0f6c5e4 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
@@ -294,7 +294,6 @@ emit_fetch(struct gen_context *gen,
    case TGSI_SWIZZLE_W:
       switch (reg->Register.File) {
       case TGSI_FILE_INPUT:
-      case TGSI_FILE_SYSTEM_VALUE:
          {
             int offset = (reg->Register.Index * 4 + swizzle) * 16;
             int offset_reg = emit_li_offset(gen, offset);
@@ -302,6 +301,9 @@ emit_fetch(struct gen_context *gen,
             ppc_lvx(gen->f, dst_vec, gen->inputs_reg, offset_reg);
          }
          break;
+      case TGSI_FILE_SYSTEM_VALUE:
+         assert(!"unhandled system value in tgsi_ppc.c");
+         break;
       case TGSI_FILE_TEMPORARY:
          if (is_ppc_vec_temporary(reg)) {
             /* use PPC vec register */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
index 6585da3e838..83c6ac75e54 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -143,7 +143,7 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                info->file_count[file]++;
                info->file_max[file] = MAX2(info->file_max[file], (int)reg);
 
-               if (file == TGSI_FILE_INPUT || file == TGSI_FILE_SYSTEM_VALUE) {
+               if (file == TGSI_FILE_INPUT) {
                   info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.Name;
                   info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.Index;
                   info->input_interpolate[reg] = (ubyte)fulldecl->Declaration.Interpolate;
@@ -151,6 +151,23 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                   info->input_cylindrical_wrap[reg] = (ubyte)fulldecl->Declaration.CylindricalWrap;
                   info->num_inputs++;
                }
+               else if (file == TGSI_FILE_SYSTEM_VALUE) {
+                  unsigned index = fulldecl->Range.First;
+                  unsigned semName = fulldecl->Semantic.Name;
+
+                  info->system_value_semantic_name[index] = semName;
+                  info->num_system_values = MAX2(info->num_system_values,
+                                                 index + 1);
+
+                  /*
+                  info->system_value_semantic_name[info->num_system_values++] = 
+                     fulldecl->Semantic.Name;
+                  */
+
+                  if (fulldecl->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) {
+                     info->uses_instanceid = TRUE;
+                  }
+               }
                else if (file == TGSI_FILE_OUTPUT) {
                   info->output_semantic_name[reg] = (ubyte)fulldecl->Semantic.Name;
                   info->output_semantic_index[reg] = (ubyte)fulldecl->Semantic.Index;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h
index 104097fbc03..53ab3d509dd 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h
@@ -51,6 +51,9 @@ struct tgsi_shader_info
    ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */
    ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
 
+   ubyte num_system_values;
+   ubyte system_value_semantic_name[PIPE_MAX_SHADER_INPUTS];
+
    uint file_mask[TGSI_FILE_COUNT];  /**< bitmask of declared registers */
    uint file_count[TGSI_FILE_COUNT];  /**< number of declared registers */
    int file_max[TGSI_FILE_COUNT];  /**< highest index of declared registers */
@@ -64,6 +67,7 @@ struct tgsi_shader_info
    boolean writes_stencil; /**< does fragment shader write stencil value? */
    boolean writes_edgeflag; /**< vertex shader outputs edgeflag */
    boolean uses_kill;  /**< KIL or KILP instruction used? */
+   boolean uses_instanceid;
 
    /**
     * Bitmask indicating which register files are accessed with
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index 086d983a73a..3f2cda860e0 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -163,6 +163,14 @@ get_immediate_base( void )
       reg_DX );
 }
 
+static struct x86_reg
+get_system_value_base( void )
+{
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, SystemValue) );
+}
+
 
 /**
  * Data access helpers.
@@ -229,6 +237,16 @@ get_temp(
 }
 
 static struct x86_reg
+get_system_value(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_system_value_base(), /* base */
+      (vec * 4 + chan) * 4 );  /* byte offset from base */
+}
+
+static struct x86_reg
 get_coef(
    unsigned vec,
    unsigned chan,
@@ -423,6 +441,30 @@ emit_tempf(
 }
 
 /**
+ * Copy a system value to xmm register
+ * \param xmm  the destination xmm register
+ * \param vec  the source system value register
+ * \param chan  src channel to fetch (X, Y, Z or W)
+ */
+static void
+emit_system_value(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movss(
+      func,
+      make_xmm( xmm ),
+      get_system_value( vec, chan ) );
+   sse_shufps(
+      func,
+      make_xmm( xmm ),
+      make_xmm( xmm ),
+      SHUF( 0, 0, 0, 0 ) );
+}
+
+/**
  * Load an xmm register with an input attrib coefficient (a0, dadx or dady)
  * \param xmm  the destination xmm register
  * \param vec  the src input/attribute coefficient index
@@ -1281,8 +1323,15 @@ emit_fetch(
             swizzle );
          break;
 
-      case TGSI_FILE_INPUT:
       case TGSI_FILE_SYSTEM_VALUE:
+         emit_system_value(
+            func,
+            xmm,
+            reg->Register.Index,
+            swizzle );
+         break;
+
+      case TGSI_FILE_INPUT:
          emit_inputf(
             func,
             xmm,
@@ -2636,8 +2685,7 @@ emit_declaration(
    struct x86_function *func,
    struct tgsi_full_declaration *decl )
 {
-   if( decl->Declaration.File == TGSI_FILE_INPUT ||
-       decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE ) {
+   if( decl->Declaration.File == TGSI_FILE_INPUT ) {
       unsigned first, last, mask;
       unsigned i, j;
 
diff --git a/src/gallium/docs/source/screen.rst b/src/gallium/docs/source/screen.rst
index 09edbaa673d..ab90097add6 100644
--- a/src/gallium/docs/source/screen.rst
+++ b/src/gallium/docs/source/screen.rst
@@ -36,6 +36,7 @@ The integer capabilities:
   bound.
 * ``OCCLUSION_QUERY``: Whether occlusion queries are available.
 * ``TIMER_QUERY``: Whether timer queries are available.
+* ``INSTANCED_DRAWING``: indicates support for instanced drawing.
 * ``TEXTURE_SHADOW_MAP``: indicates whether the fragment shader hardware
   can do the depth texture / Z comparison operation in TEX instructions
   for shadow testing.
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
index 9459a3cd113..b6919a5c6d3 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -164,6 +164,8 @@ llvmpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
       return 1;
    case PIPE_CAP_DEPTH_CLAMP:
       return 0;
+   case PIPE_CAP_INSTANCED_DRAWING:
+      return 1;
    default:
       return 0;
    }
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index ae207617cc1..1b9119eda00 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -334,7 +334,8 @@ generate_fs(struct gallivm_state *gallivm,
    
    /* Build the actual shader */
    lp_build_tgsi_soa(gallivm, tokens, type, &mask,
-                     consts_ptr, interp->pos, interp->inputs,
+                     consts_ptr, NULL, /* sys values array */
+                     interp->pos, interp->inputs,
                      outputs, sampler, &shader->info.base);
 
    /* Alpha test */
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index cbdea19af43..e19f2e6fc7d 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -123,6 +123,8 @@ softpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
       return 0;
    case PIPE_CAP_SHADER_STENCIL_EXPORT:
       return 1;
+   case PIPE_CAP_INSTANCED_DRAWING:
+      return 1;
    default:
       return 0;
    }
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index f5af15ff9ff..2135c19dcfc 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -462,6 +462,7 @@ enum pipe_cap {
    PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER,
    PIPE_CAP_DEPTH_CLAMP,
    PIPE_CAP_SHADER_STENCIL_EXPORT,
+   PIPE_CAP_INSTANCED_DRAWING,
 };
 
 /* Shader caps not specific to any single stage */
diff --git a/src/glsl/glsl_parser_extras.cpp b/src/glsl/glsl_parser_extras.cpp
index 8dbe66927dc..cbeacd5633f 100644
--- a/src/glsl/glsl_parser_extras.cpp
+++ b/src/glsl/glsl_parser_extras.cpp
@@ -180,6 +180,15 @@ _mesa_glsl_process_extension(const char *name, YYLTYPE *name_locp,
 	 state->ARB_draw_buffers_enable = (ext_mode != extension_disable);
 	 state->ARB_draw_buffers_warn = (ext_mode == extension_warn);
       }
+   } else if (strcmp(name, "GL_ARB_draw_instanced") == 0) {
+      /* This extension is only supported in vertex shaders.
+       */
+      if (state->target != vertex_shader) {
+	 unsupported = true;
+      } else {
+	 state->ARB_draw_instanced_enable = (ext_mode != extension_disable);
+	 state->ARB_draw_instanced_warn = (ext_mode == extension_warn);
+      }
    } else if (strcmp(name, "GL_ARB_explicit_attrib_location") == 0) {
       state->ARB_explicit_attrib_location_enable =
 	 (ext_mode != extension_disable);
diff --git a/src/glsl/glsl_parser_extras.h b/src/glsl/glsl_parser_extras.h
index 39c10b99e83..030d27a2627 100644
--- a/src/glsl/glsl_parser_extras.h
+++ b/src/glsl/glsl_parser_extras.h
@@ -132,6 +132,8 @@ struct _mesa_glsl_parse_state {
    /*@{*/
    unsigned ARB_draw_buffers_enable:1;
    unsigned ARB_draw_buffers_warn:1;
+   unsigned ARB_draw_instanced_enable:1;
+   unsigned ARB_draw_instanced_warn:1;
    unsigned ARB_explicit_attrib_location_enable:1;
    unsigned ARB_explicit_attrib_location_warn:1;
    unsigned ARB_fragment_coord_conventions_enable:1;
diff --git a/src/glsl/ir.h b/src/glsl/ir.h
index 57e89f0f6bf..7b83c2836ee 100644
--- a/src/glsl/ir.h
+++ b/src/glsl/ir.h
@@ -225,6 +225,7 @@ enum ir_variable_mode {
    ir_var_in,
    ir_var_out,
    ir_var_inout,
+   ir_var_system_value, /**< Ex: front-face, instance-id, etc. */
    ir_var_temporary	/**< Temporary variable generated during compilation. */
 };
 
diff --git a/src/glsl/ir_set_program_inouts.cpp b/src/glsl/ir_set_program_inouts.cpp
index 714281539ac..085456533c5 100644
--- a/src/glsl/ir_set_program_inouts.cpp
+++ b/src/glsl/ir_set_program_inouts.cpp
@@ -79,6 +79,8 @@ mark(struct gl_program *prog, ir_variable *var, int offset, int len)
    for (int i = 0; i < len; i++) {
       if (var->mode == ir_var_in)
 	 prog->InputsRead |= BITFIELD64_BIT(var->location + offset + i);
+      else if (var->mode == ir_var_system_value)
+         prog->SystemValuesRead |= (1 << (var->location + offset + i));
       else
 	 prog->OutputsWritten |= BITFIELD64_BIT(var->location + offset + i);
    }
@@ -134,7 +136,8 @@ ir_visitor_status
 ir_set_program_inouts_visitor::visit(ir_variable *ir)
 {
    if (ir->mode == ir_var_in ||
-       ir->mode == ir_var_out) {
+       ir->mode == ir_var_out ||
+       ir->mode == ir_var_system_value) {
       hash_table_insert(this->ht, ir, ir);
    }
 
@@ -158,5 +161,6 @@ do_set_program_inouts(exec_list *instructions, struct gl_program *prog)
 
    prog->InputsRead = 0;
    prog->OutputsWritten = 0;
+   prog->SystemValuesRead = 0;
    visit_list_elements(&v, instructions);
 }
diff --git a/src/glsl/ir_variable.cpp b/src/glsl/ir_variable.cpp
index 5b8281e16e3..73da28faf4f 100644
--- a/src/glsl/ir_variable.cpp
+++ b/src/glsl/ir_variable.cpp
@@ -30,6 +30,11 @@ static void generate_ARB_draw_buffers_variables(exec_list *,
 						struct _mesa_glsl_parse_state *,
 						bool, _mesa_glsl_parser_targets);
 
+static void
+generate_ARB_draw_instanced_variables(exec_list *,
+                                      struct _mesa_glsl_parse_state *,
+                                      bool, _mesa_glsl_parser_targets);
+
 static ir_variable *
 add_variable(const char *name, enum ir_variable_mode mode, int slot,
 	     const glsl_type *type, exec_list *instructions,
@@ -41,6 +46,7 @@ add_variable(const char *name, enum ir_variable_mode mode, int slot,
    case ir_var_auto:
    case ir_var_in:
    case ir_var_uniform:
+   case ir_var_system_value:
       var->read_only = true;
       break;
    case ir_var_inout:
@@ -324,8 +330,13 @@ initialize_vs_variables(exec_list *instructions,
       generate_130_vs_variables(instructions, state);
       break;
    }
+
+   if (state->ARB_draw_instanced_enable)
+      generate_ARB_draw_instanced_variables(instructions, state, false,
+                                            vertex_shader);
 }
 
+
 /* This function should only be called for ES, not desktop GL. */
 static void
 generate_100ES_fs_variables(exec_list *instructions,
@@ -422,6 +433,27 @@ generate_ARB_draw_buffers_variables(exec_list *instructions,
    }
 }
 
+
+static void
+generate_ARB_draw_instanced_variables(exec_list *instructions,
+                                      struct _mesa_glsl_parse_state *state,
+                                      bool warn,
+                                      _mesa_glsl_parser_targets target)
+{
+   /* gl_InstanceIDARB is only available in the vertex shader.
+    */
+   if (target == vertex_shader) {
+      ir_variable *const inst =
+         add_variable("gl_InstanceIDARB", ir_var_system_value,
+                      SYSTEM_VALUE_INSTANCE_ID,
+                      glsl_type::int_type, instructions, state->symbols);
+
+      if (warn)
+         inst->warn_extension = "GL_ARB_draw_instanced";
+   }
+}
+
+
 static void
 generate_ARB_shader_stencil_export_variables(exec_list *instructions,
 					     struct _mesa_glsl_parse_state *state,
diff --git a/src/glsl/main.cpp b/src/glsl/main.cpp
index 2d146caca5b..9b041aafe42 100644
--- a/src/glsl/main.cpp
+++ b/src/glsl/main.cpp
@@ -73,6 +73,7 @@ initialize_context(struct gl_context *ctx, gl_api api)
    ctx->API = api;
 
    ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+   ctx->Extensions.ARB_draw_instanced = GL_TRUE;
    ctx->Extensions.ARB_fragment_coord_conventions = GL_TRUE;
    ctx->Extensions.EXT_texture_array = GL_TRUE;
    ctx->Extensions.NV_texture_rectangle = GL_TRUE;
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index c41ae4ad2d9..24404993c69 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -380,6 +380,7 @@ _mesa_enable_sw_extensions(struct gl_context *ctx)
    ctx->Extensions.ARB_depth_texture = GL_TRUE;
    /*ctx->Extensions.ARB_draw_buffers = GL_TRUE;*/
    ctx->Extensions.ARB_draw_elements_base_vertex = GL_TRUE;
+   ctx->Extensions.ARB_draw_instanced = GL_TRUE;
    ctx->Extensions.ARB_explicit_attrib_location = GL_TRUE;
    ctx->Extensions.ARB_fragment_coord_conventions = GL_TRUE;
 #if FEATURE_ARB_fragment_program
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index f76e1eb2ee0..1522332e8a1 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -1748,11 +1748,24 @@ typedef enum
    PROGRAM_WRITE_ONLY,  /**< A dummy, write-only register */
    PROGRAM_ADDRESS,     /**< machine->AddressReg */
    PROGRAM_SAMPLER,     /**< for shader samplers, compile-time only */
+   PROGRAM_SYSTEM_VALUE,/**< InstanceId, PrimitiveID, etc. */
    PROGRAM_UNDEFINED,   /**< Invalid/TBD value */
    PROGRAM_FILE_MAX
 } gl_register_file;
 
 
+/**
+ * If the register file is PROGRAM_SYSTEM_VALUE, the register index will be
+ * one of these values.
+ */
+typedef enum
+{
+   SYSTEM_VALUE_FRONT_FACE,  /**< Fragment shader only (not done yet) */
+   SYSTEM_VALUE_INSTANCE_ID, /**< Vertex shader only */
+   SYSTEM_VALUE_MAX          /**< Number of values */
+} gl_system_value;
+
+
 /** Vertex and fragment instructions */
 struct prog_instruction;
 struct gl_program_parameter_list;
@@ -1775,6 +1788,7 @@ struct gl_program
 
    GLbitfield InputsRead;     /**< Bitmask of which input regs are read */
    GLbitfield64 OutputsWritten; /**< Bitmask of which output regs are written */
+   GLbitfield SystemValuesRead;   /**< Bitmask of SYSTEM_VALUE_x inputs used */
    GLbitfield InputFlags[MAX_PROGRAM_INPUTS];   /**< PROG_PARAM_BIT_x flags */
    GLbitfield OutputFlags[MAX_PROGRAM_OUTPUTS]; /**< PROG_PARAM_BIT_x flags */
    GLbitfield TexturesUsed[MAX_TEXTURE_UNITS];  /**< TEXTURE_x_BIT bitmask */
diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp
index 0929b959505..786fdfbd231 100644
--- a/src/mesa/program/ir_to_mesa.cpp
+++ b/src/mesa/program/ir_to_mesa.cpp
@@ -1462,6 +1462,7 @@ ir_to_mesa_visitor::visit(ir_dereference_variable *ir)
       case ir_var_in:
       case ir_var_out:
       case ir_var_inout:
+      case ir_var_system_value:
 	 /* The linker assigns locations for varyings and attributes,
 	  * including deprecated builtins (like gl_Color), user-assign
 	  * generic attributes (glBindVertexLocation), and
@@ -1484,6 +1485,10 @@ ir_to_mesa_visitor::visit(ir_dereference_variable *ir)
 				   ir->var->type->gl_type,
 				   ir->var->location - VERT_ATTRIB_GENERIC0);
 	    }
+         } else if (ir->var->mode == ir_var_system_value) {
+	    entry = new(mem_ctx) variable_storage(ir->var,
+						  PROGRAM_SYSTEM_VALUE,
+						  ir->var->location);
 	 } else {
 	    entry = new(mem_ctx) variable_storage(ir->var,
 						  PROGRAM_OUTPUT,
diff --git a/src/mesa/program/prog_execute.c b/src/mesa/program/prog_execute.c
index dd15e9a1ccd..e7553c69dbe 100644
--- a/src/mesa/program/prog_execute.c
+++ b/src/mesa/program/prog_execute.c
@@ -159,6 +159,10 @@ get_src_register_pointer(const struct prog_src_register *source,
          return ZeroVec;
       return prog->Parameters->ParameterValues[reg];
 
+   case PROGRAM_SYSTEM_VALUE:
+      assert(reg < Elements(machine->SystemValues));
+      return machine->SystemValues[reg];
+
    default:
       _mesa_problem(NULL,
          "Invalid src register file %d in get_src_register_pointer()",
diff --git a/src/mesa/program/prog_execute.h b/src/mesa/program/prog_execute.h
index cefd468c36b..cdf37082a00 100644
--- a/src/mesa/program/prog_execute.h
+++ b/src/mesa/program/prog_execute.h
@@ -61,6 +61,7 @@ struct gl_program_machine
    GLfloat (*EnvParams)[4]; /**< Vertex or Fragment env parameters */
    GLuint CondCodes[4];  /**< COND_* value for x/y/z/w */
    GLint AddressReg[MAX_PROGRAM_ADDRESS_REGS][4];
+   GLfloat SystemValues[SYSTEM_VALUE_MAX][4];
 
    const GLubyte *Samplers;  /** Array mapping sampler var to tex unit */
 
diff --git a/src/mesa/program/prog_print.c b/src/mesa/program/prog_print.c
index abebf392c0a..484596af760 100644
--- a/src/mesa/program/prog_print.c
+++ b/src/mesa/program/prog_print.c
@@ -72,6 +72,8 @@ _mesa_register_file_name(gl_register_file f)
       return "ADDR";
    case PROGRAM_SAMPLER:
       return "SAMPLER";
+   case PROGRAM_SYSTEM_VALUE:
+      return "SYSVAL";
    case PROGRAM_UNDEFINED:
       return "UNDEFINED";
    default:
@@ -310,6 +312,9 @@ reg_string(gl_register_file f, GLint index, gl_prog_print_mode mode,
       case PROGRAM_UNIFORM: /* extension */
          sprintf(str, "uniform[%s%d]", addr, index);
          break;
+      case PROGRAM_SYSTEM_VALUE:
+         sprintf(str, "sysvalue[%s%d]", addr, index);
+         break;
       case PROGRAM_STATE_VAR:
          {
             struct gl_program_parameter *param
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index d240cab1a74..df4f5cf8b5c 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -461,4 +461,8 @@ void st_init_extensions(struct st_context *st)
    if (screen->get_param(screen, PIPE_CAP_SHADER_STENCIL_EXPORT)) {
       ctx->Extensions.ARB_shader_stencil_export = GL_TRUE;
    }
+
+   if (screen->get_param(screen, PIPE_CAP_INSTANCED_DRAWING)) {
+      ctx->Extensions.ARB_draw_instanced = GL_TRUE;
+   }
 }
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index f848462310e..42f1c2017f8 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -72,6 +72,7 @@ struct st_translate {
    struct ureg_src inputs[PIPE_MAX_SHADER_INPUTS];
    struct ureg_dst address[1];
    struct ureg_src samplers[PIPE_MAX_SAMPLERS];
+   struct ureg_src systemValues[SYSTEM_VALUE_MAX];
 
    /* Extra info for handling point size clamping in vertex shader */
    struct ureg_dst pointSizeResult; /**< Actual point size output register */
@@ -104,6 +105,13 @@ struct st_translate {
 };
 
 
+/** Map Mesa's SYSTEM_VALUE_x to TGSI_SEMANTIC_x */
+static unsigned mesa_sysval_to_semantic[SYSTEM_VALUE_MAX] = {
+   TGSI_SEMANTIC_FACE,
+   TGSI_SEMANTIC_INSTANCEID
+};
+
+
 /**
  * Make note of a branch to a label in the TGSI code.
  * After we've emitted all instructions, we'll go over the list
@@ -245,6 +253,10 @@ src_register( struct st_translate *t,
    case PROGRAM_ADDRESS:
       return ureg_src(t->address[index]);
 
+   case PROGRAM_SYSTEM_VALUE:
+      assert(index < Elements(t->systemValues));
+      return t->systemValues[index];
+
    default:
       debug_assert( 0 );
       return ureg_src_undef();
@@ -1089,6 +1101,21 @@ st_translate_mesa_program(
       t->address[0] = ureg_DECL_address( ureg );
    }
 
+   /* Declare misc input registers
+    */
+   {
+      GLbitfield sysInputs = program->SystemValuesRead;
+      unsigned numSys = 0;
+      for (i = 0; sysInputs; i++) {
+         if (sysInputs & (1 << i)) {
+            unsigned semName = mesa_sysval_to_semantic[i];
+            t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
+            numSys++;
+            sysInputs &= ~(1 << i);
+         }
+      }
+   }
+
    if (program->IndirectRegisterFiles & (1 << PROGRAM_TEMPORARY)) {
       /* If temps are accessed with indirect addressing, declare temporaries
        * in sequential order.  Else, we declare them on demand elsewhere.
diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h
index bc01646247c..6a9444216c0 100644
--- a/src/mesa/tnl/t_context.h
+++ b/src/mesa/tnl/t_context.h
@@ -527,6 +527,8 @@ typedef struct
    GLubyte *block[VERT_ATTRIB_MAX];
    GLuint nr_blocks;
 
+   GLuint CurInstance;
+
 } TNLcontext;
 
 
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
index 30f1bf323cb..bdb893eba22 100644
--- a/src/mesa/tnl/t_draw.c
+++ b/src/mesa/tnl/t_draw.c
@@ -453,6 +453,7 @@ void _tnl_draw_prims( struct gl_context *ctx,
        */
       struct gl_buffer_object *bo[VERT_ATTRIB_MAX + 1];
       GLuint nr_bo = 0;
+      GLuint inst;
 
       for (i = 0; i < nr_prims;) {
 	 GLuint this_nr_prims;
@@ -470,15 +471,19 @@ void _tnl_draw_prims( struct gl_context *ctx,
 	 /* Binding inputs may imply mapping some vertex buffer objects.
 	  * They will need to be unmapped below.
 	  */
-	 bind_prims(ctx, &prim[i], this_nr_prims);
-	 bind_inputs(ctx, arrays, max_index + prim[i].basevertex + 1,
-		     bo, &nr_bo);
-	 bind_indices(ctx, ib, bo, &nr_bo);
+         for (inst = 0; inst < prim[i].num_instances; inst++) {
 
-	 TNL_CONTEXT(ctx)->Driver.RunPipeline(ctx);
+            bind_prims(ctx, &prim[i], this_nr_prims);
+            bind_inputs(ctx, arrays, max_index + prim[i].basevertex + 1,
+                        bo, &nr_bo);
+            bind_indices(ctx, ib, bo, &nr_bo);
 
-	 unmap_vbos(ctx, bo, nr_bo);
-	 free_space(ctx);
+            tnl->CurInstance = inst;
+            TNL_CONTEXT(ctx)->Driver.RunPipeline(ctx);
+
+            unmap_vbos(ctx, bo, nr_bo);
+            free_space(ctx);
+         }
 
 	 i += this_nr_prims;
       }
diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index 94372bbafbc..f5d8f7477aa 100644
--- a/src/mesa/tnl/t_vb_program.c
+++ b/src/mesa/tnl/t_vb_program.c
@@ -220,7 +220,8 @@ _tnl_program_string(struct gl_context *ctx, GLenum target, struct gl_program *pr
  * Initialize virtual machine state prior to executing vertex program.
  */
 static void
-init_machine(struct gl_context *ctx, struct gl_program_machine *machine)
+init_machine(struct gl_context *ctx, struct gl_program_machine *machine,
+             GLuint instID)
 {
    /* Input registers get initialized from the current vertex attribs */
    memcpy(machine->VertAttribs, ctx->Current.Attrib,
@@ -256,6 +257,8 @@ init_machine(struct gl_context *ctx, struct gl_program_machine *machine)
    machine->FetchTexelDeriv = NULL; /* not used by vertex programs */
 
    machine->Samplers = ctx->VertexProgram._Current->Base.SamplerUnits;
+
+   machine->SystemValues[SYSTEM_VALUE_INSTANCE_ID][0] = (GLfloat) instID;
 }
 
 
@@ -341,7 +344,7 @@ run_vp( struct gl_context *ctx, struct tnl_pipeline_stage *stage )
    for (i = 0; i < VB->Count; i++) {
       GLuint attr;
 
-      init_machine(ctx, machine);
+      init_machine(ctx, machine, tnl->CurInstance);
 
 #if 0
       printf("Input  %d: %f, %f, %f, %f\n", i,