Merge branch 'instanced-arrays'

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

11 files changed, 106 insertions, 23 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;
       }