draw: implement pipeline statistics in the draw module

This is a basic implementation of the pipeline statistics in the
draw module. The interface is similar to the stream output statistics
and also requires that the callers explicitly enable it.
Included is the implementation of the interface in llvmpipe and
softpipe. Only softpipe enables the pipeline statistics capability
though because llvmpipe is lacking gathering of the fragment shading
and rasterization statistics.

Signed-off-by: Zack Rusin <[email protected]>
Reviewed-by: Jose Fonseca <[email protected]>

13 files changed, 146 insertions, 2 deletions

diff --git a/src/gallium/drivers/llvmpipe/lp_context.h b/src/gallium/drivers/llvmpipe/lp_context.h
index 6ee7b99f478..d605dba2260 100644
--- a/src/gallium/drivers/llvmpipe/lp_context.h
+++ b/src/gallium/drivers/llvmpipe/lp_context.h
@@ -94,6 +94,9 @@ struct llvmpipe_context {
    struct pipe_query_data_so_statistics so_stats;
    unsigned num_primitives_generated;
 
+   struct pipe_query_data_pipeline_statistics pipeline_statistics;
+   unsigned active_statistics_queries;
+
    unsigned dirty; /**< Mask of LP_NEW_x flags */
 
    unsigned active_occlusion_query;
diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
index efeca255755..e8a5880deef 100644
--- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
+++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
@@ -108,6 +108,8 @@ llvmpipe_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
          draw_vs_attach_so(lp->vs->draw_data, &lp->gs->shader.stream_output);
       }
    }
+   draw_collect_pipeline_statistics(draw,
+                                    lp->active_statistics_queries > 0);
 
    /* draw! */
    draw_vbo(draw, info);
diff --git a/src/gallium/drivers/llvmpipe/lp_query.c b/src/gallium/drivers/llvmpipe/lp_query.c
index 013d1928c11..96e1e3ff1f3 100644
--- a/src/gallium/drivers/llvmpipe/lp_query.c
+++ b/src/gallium/drivers/llvmpipe/lp_query.c
@@ -144,6 +144,12 @@ llvmpipe_get_query_result(struct pipe_context *pipe,
       stats->primitives_storage_needed = pq->num_primitives_generated;
    }
       break;
+   case PIPE_QUERY_PIPELINE_STATISTICS: {
+      struct pipe_query_data_pipeline_statistics *stats =
+         (struct pipe_query_data_pipeline_statistics *)vresult;
+      *stats = pq->stats;
+   }
+      break;
    default:
       assert(0);
       break;
@@ -188,6 +194,16 @@ llvmpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
       llvmpipe->num_primitives_generated = 0;
    }
 
+   if (pq->type == PIPE_QUERY_PIPELINE_STATISTICS) {
+      /* reset our cache */
+      if (llvmpipe->active_statistics_queries == 0) {
+         memset(&llvmpipe->pipeline_statistics, 0,
+                sizeof(llvmpipe->pipeline_statistics));
+      }
+      memcpy(&pq->stats, &llvmpipe->pipeline_statistics, sizeof(pq->stats));
+      llvmpipe->active_statistics_queries++;
+   }
+
    if (pq->type == PIPE_QUERY_OCCLUSION_COUNTER) {
       llvmpipe->active_occlusion_query = TRUE;
       llvmpipe->dirty |= LP_NEW_OCCLUSION_QUERY;
@@ -216,6 +232,27 @@ llvmpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
       pq->num_primitives_generated = llvmpipe->num_primitives_generated;
    }
 
+   if (pq->type == PIPE_QUERY_PIPELINE_STATISTICS) {
+      pq->stats.ia_vertices =
+         llvmpipe->pipeline_statistics.ia_vertices - pq->stats.ia_vertices;
+      pq->stats.ia_primitives =
+         llvmpipe->pipeline_statistics.ia_primitives - pq->stats.ia_primitives;
+      pq->stats.vs_invocations =
+         llvmpipe->pipeline_statistics.vs_invocations - pq->stats.vs_invocations;
+      pq->stats.gs_invocations =
+         llvmpipe->pipeline_statistics.gs_invocations - pq->stats.gs_invocations;
+      pq->stats.gs_primitives =
+         llvmpipe->pipeline_statistics.gs_primitives - pq->stats.gs_primitives;
+      pq->stats.c_invocations =
+         llvmpipe->pipeline_statistics.c_invocations - pq->stats.c_invocations;
+      pq->stats.c_primitives =
+         llvmpipe->pipeline_statistics.c_primitives - pq->stats.c_primitives;
+      pq->stats.ps_invocations =
+         llvmpipe->pipeline_statistics.ps_invocations - pq->stats.ps_invocations;
+
+      llvmpipe->active_statistics_queries--;
+   }
+
    if (pq->type == PIPE_QUERY_OCCLUSION_COUNTER) {
       assert(llvmpipe->active_occlusion_query);
       llvmpipe->active_occlusion_query = FALSE;
diff --git a/src/gallium/drivers/llvmpipe/lp_query.h b/src/gallium/drivers/llvmpipe/lp_query.h
index 19d977fd2e4..68dfe99f428 100644
--- a/src/gallium/drivers/llvmpipe/lp_query.h
+++ b/src/gallium/drivers/llvmpipe/lp_query.h
@@ -47,6 +47,8 @@ struct llvmpipe_query {
    unsigned type;                   /* PIPE_QUERY_* */
    unsigned num_primitives_generated;
    unsigned num_primitives_written;
+
+   struct pipe_query_data_pipeline_statistics stats;
 };
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 6cb27020566..29dd933c644 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -477,6 +477,7 @@ lp_rast_begin_query(struct lp_rasterizer_task *task,
    case PIPE_QUERY_PRIMITIVES_GENERATED:
    case PIPE_QUERY_PRIMITIVES_EMITTED:
    case PIPE_QUERY_SO_STATISTICS:
+   case PIPE_QUERY_PIPELINE_STATISTICS:
       break;
    default:
       assert(0);
@@ -509,6 +510,7 @@ lp_rast_end_query(struct lp_rasterizer_task *task,
    case PIPE_QUERY_PRIMITIVES_GENERATED:
    case PIPE_QUERY_PRIMITIVES_EMITTED:
    case PIPE_QUERY_SO_STATISTICS:
+   case PIPE_QUERY_PIPELINE_STATISTICS:
       break;
    default:
       assert(0);
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c b/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
index 9e0a5b39954..8173994e5c1 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
@@ -547,6 +547,28 @@ lp_setup_so_info(struct vbuf_render *vbr, uint primitives, uint vertices,
    lp->num_primitives_generated += prim_generated;
 }
 
+static void
+lp_setup_pipeline_statistics(
+   struct vbuf_render *vbr,
+   const struct pipe_query_data_pipeline_statistics *stats)
+{
+   struct lp_setup_context *setup = lp_setup_context(vbr);
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(setup->pipe);
+
+   llvmpipe->pipeline_statistics.ia_vertices +=
+      stats->ia_vertices;
+   llvmpipe->pipeline_statistics.ia_primitives +=
+      stats->ia_primitives;
+   llvmpipe->pipeline_statistics.vs_invocations +=
+      stats->vs_invocations;
+   llvmpipe->pipeline_statistics.gs_invocations +=
+      stats->gs_invocations;
+   llvmpipe->pipeline_statistics.gs_primitives +=
+      stats->gs_primitives;
+   llvmpipe->pipeline_statistics.c_invocations +=
+      stats->c_invocations;
+}
+
 /**
  * Create the post-transform vertex handler for the given context.
  */
@@ -566,4 +588,5 @@ lp_setup_init_vbuf(struct lp_setup_context *setup)
    setup->base.release_vertices = lp_setup_release_vertices;
    setup->base.destroy = lp_setup_vbuf_destroy;
    setup->base.set_stream_output_info = lp_setup_so_info;
+   setup->base.pipeline_statistics = lp_setup_pipeline_statistics;
 }
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index d78352ab630..431864ae087 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -92,6 +92,9 @@ struct softpipe_context {
    struct pipe_query_data_so_statistics so_stats;
    unsigned num_primitives_generated;
 
+   struct pipe_query_data_pipeline_statistics pipeline_statistics;
+   unsigned active_statistics_queries;
+
    unsigned num_samplers[PIPE_SHADER_TYPES];
    unsigned num_sampler_views[PIPE_SHADER_TYPES];
 
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index ba9e7817ceb..4378312bc36 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -105,6 +105,9 @@ softpipe_draw_vbo(struct pipe_context *pipe,
    draw_set_mapped_so_targets(draw, sp->num_so_targets,
                               sp->so_targets);
 
+   draw_collect_pipeline_statistics(draw,
+                                    sp->active_statistics_queries > 0);
+
    /* draw! */
    draw_vbo(draw, info);
 
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index bb5ebcc7aee..5d0b5e1c588 100644
--- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -600,6 +600,28 @@ sp_vbuf_so_info(struct vbuf_render *vbr, uint primitives, uint vertices,
    softpipe->num_primitives_generated += prim_generated;
 }
 
+static void
+sp_vbuf_pipeline_statistics(
+   struct vbuf_render *vbr, 
+   const struct pipe_query_data_pipeline_statistics *stats)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   struct softpipe_context *softpipe = cvbr->softpipe;
+
+   softpipe->pipeline_statistics.ia_vertices +=
+      stats->ia_vertices;
+   softpipe->pipeline_statistics.ia_primitives +=
+      stats->ia_primitives;
+   softpipe->pipeline_statistics.vs_invocations +=
+      stats->vs_invocations;
+   softpipe->pipeline_statistics.gs_invocations +=
+      stats->gs_invocations;
+   softpipe->pipeline_statistics.gs_primitives +=
+      stats->gs_primitives;
+   softpipe->pipeline_statistics.c_invocations +=
+      stats->c_invocations;
+}
+
 
 static void
 sp_vbuf_destroy(struct vbuf_render *vbr)
@@ -634,6 +656,7 @@ sp_create_vbuf_backend(struct softpipe_context *sp)
    cvbr->base.draw_arrays = sp_vbuf_draw_arrays;
    cvbr->base.release_vertices = sp_vbuf_release_vertices;
    cvbr->base.set_stream_output_info = sp_vbuf_so_info;
+   cvbr->base.pipeline_statistics = sp_vbuf_pipeline_statistics;
    cvbr->base.destroy = sp_vbuf_destroy;
 
    cvbr->softpipe = sp;
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index f32fe5f4068..d379f6457e3 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -73,6 +73,11 @@ shade_quad(struct quad_stage *qs, struct quad_header *quad)
    struct softpipe_context *softpipe = qs->softpipe;
    struct tgsi_exec_machine *machine = softpipe->fs_machine;
 
+   if (softpipe->active_statistics_queries) {
+      softpipe->pipeline_statistics.ps_invocations +=
+         util_bitcount(quad->inout.mask);         
+   }
+
    /* run shader */
    machine->flatshade_color = softpipe->rasterizer->flatshade ? TRUE : FALSE;
    return softpipe->fs_variant->run( softpipe->fs_variant, machine, quad );
diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
index 2dd82c81d6f..62411de5036 100644
--- a/src/gallium/drivers/softpipe/sp_query.c
+++ b/src/gallium/drivers/softpipe/sp_query.c
@@ -43,6 +43,8 @@ struct softpipe_query {
    uint64_t end;
    struct pipe_query_data_so_statistics so;
    unsigned num_primitives_generated;
+
+   struct pipe_query_data_pipeline_statistics stats;
 };
 
 
@@ -61,7 +63,8 @@ softpipe_create_query(struct pipe_context *pipe,
           type == PIPE_QUERY_TIME_ELAPSED ||
           type == PIPE_QUERY_SO_STATISTICS ||
           type == PIPE_QUERY_PRIMITIVES_EMITTED ||
-          type == PIPE_QUERY_PRIMITIVES_GENERATED ||
+          type == PIPE_QUERY_PRIMITIVES_GENERATED || 
+          type == PIPE_QUERY_PIPELINE_STATISTICS ||
           type == PIPE_QUERY_GPU_FINISHED ||
           type == PIPE_QUERY_TIMESTAMP ||
           type == PIPE_QUERY_TIMESTAMP_DISJOINT);
@@ -106,6 +109,16 @@ softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
    case PIPE_QUERY_TIMESTAMP:
    case PIPE_QUERY_GPU_FINISHED:
       break;
+   case PIPE_QUERY_PIPELINE_STATISTICS:
+      /* reset our cache */
+      if (softpipe->active_statistics_queries == 0) {
+         memset(&softpipe->pipeline_statistics, 0,
+                sizeof(softpipe->pipeline_statistics));
+      }
+      memcpy(&sq->stats, &softpipe->pipeline_statistics,
+             sizeof(sq->stats));
+      softpipe->active_statistics_queries++;
+      break;
    default:
       assert(0);
       break;
@@ -145,6 +158,26 @@ softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
       break;
    case PIPE_QUERY_GPU_FINISHED:
       break;
+   case PIPE_QUERY_PIPELINE_STATISTICS:
+      sq->stats.ia_vertices =
+         softpipe->pipeline_statistics.ia_vertices - sq->stats.ia_vertices;
+      sq->stats.ia_primitives =
+         softpipe->pipeline_statistics.ia_primitives - sq->stats.ia_primitives;
+      sq->stats.vs_invocations =
+         softpipe->pipeline_statistics.vs_invocations - sq->stats.vs_invocations;
+      sq->stats.gs_invocations =
+         softpipe->pipeline_statistics.gs_invocations - sq->stats.gs_invocations;
+      sq->stats.gs_primitives =
+         softpipe->pipeline_statistics.gs_primitives - sq->stats.gs_primitives;
+      sq->stats.c_invocations =
+         softpipe->pipeline_statistics.c_invocations - sq->stats.c_invocations;
+      sq->stats.c_primitives =
+         softpipe->pipeline_statistics.c_primitives - sq->stats.c_primitives;
+      sq->stats.ps_invocations =
+         softpipe->pipeline_statistics.ps_invocations - sq->stats.ps_invocations;
+
+      softpipe->active_statistics_queries--;
+      break;
    default:
       assert(0);
       break;
@@ -167,6 +200,10 @@ softpipe_get_query_result(struct pipe_context *pipe,
       memcpy(vresult, &sq->so,
              sizeof(struct pipe_query_data_so_statistics));
       break;
+   case PIPE_QUERY_PIPELINE_STATISTICS:
+      memcpy(vresult, &sq->stats,
+             sizeof(struct pipe_query_data_pipeline_statistics));;
+      break;
    case PIPE_QUERY_GPU_FINISHED:
       *result = TRUE;
       break;
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index 6915f918007..b2b7f2e3208 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -87,7 +87,7 @@ softpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_QUERY_TIME_ELAPSED:
       return 1;
    case PIPE_CAP_QUERY_PIPELINE_STATISTICS:
-      return 0;
+      return 1;
    case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
       return 1;
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 3dad131d881..08cd4da5bd0 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -857,6 +857,10 @@ sp_setup_tri(struct setup_context *setup,
 
    flush_spans( setup );
 
+   if (setup->softpipe->active_statistics_queries) {
+      setup->softpipe->pipeline_statistics.c_primitives++;
+   }
+
 #if DEBUG_FRAGS
    printf("Tri: %u frags emitted, %u written\n",
           setup->numFragsEmitted,