intel: Emit 3DSTATE_VF_STATISTICS dynamically

Pipeline statistics queries should not count BLORP's rectangles.

    (23) How do operations like Clear, TexSubImage, etc. affect the
         results of the newly introduced queries?

      DISCUSSION: Implementations might require "helper" rendering
      commands be issued to implement certain operations like Clear,
      TexSubImage, etc.

      RESOLVED: They don't. Only application submitted rendering
      commands should have an effect on the results of the queries.

Piglit's arb_pipeline_statistics_query-vert_adj exposes this bug when
the driver is hacked to always perform glBufferData via a GPU staging
copy (for debugging purposes).

Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Lionel Landwerlin <[email protected]>

6 files changed, 35 insertions, 11 deletions

diff --git a/src/gallium/drivers/iris/iris_context.h b/src/gallium/drivers/iris/iris_context.h
index 4ec73973dc5..4465eef40dd 100644
--- a/src/gallium/drivers/iris/iris_context.h
+++ b/src/gallium/drivers/iris/iris_context.h
@@ -127,6 +127,7 @@ enum iris_param_domain {
 #define IRIS_DIRTY_VF_TOPOLOGY              (1ull << 54)
 #define IRIS_DIRTY_RENDER_RESOLVES_AND_FLUSHES  (1ull << 55)
 #define IRIS_DIRTY_COMPUTE_RESOLVES_AND_FLUSHES (1ull << 56)
+#define IRIS_DIRTY_VF_STATISTICS            (1ull << 57)
 
 #define IRIS_ALL_DIRTY_FOR_COMPUTE (IRIS_DIRTY_CS | \
                                     IRIS_DIRTY_SAMPLER_STATES_CS | \
diff --git a/src/gallium/drivers/iris/iris_state.c b/src/gallium/drivers/iris/iris_state.c
index bf4fdcae1c9..8d5383598d2 100644
--- a/src/gallium/drivers/iris/iris_state.c
+++ b/src/gallium/drivers/iris/iris_state.c
@@ -5055,6 +5055,12 @@ iris_upload_dirty_render_state(struct iris_context *ice,
       }
    }
 
+   if (dirty & IRIS_DIRTY_VF_STATISTICS) {
+      iris_emit_cmd(batch, GENX(3DSTATE_VF_STATISTICS), vf) {
+         vf.StatisticsEnable = true;
+      }
+   }
+
    /* TODO: Gen8 PMA fix */
 }
 
diff --git a/src/intel/blorp/blorp_genX_exec.h b/src/intel/blorp/blorp_genX_exec.h
index 152f40d9338..c3bf92c9148 100644
--- a/src/intel/blorp/blorp_genX_exec.h
+++ b/src/intel/blorp/blorp_genX_exec.h
@@ -513,6 +513,10 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
       dw += GENX(VERTEX_ELEMENT_STATE_length);
    }
 
+   blorp_emit(batch, GENX(3DSTATE_VF_STATISTICS), vf) {
+      vf.StatisticsEnable = false;
+   }
+
 #if GEN_GEN >= 8
    /* Overwrite Render Target Array Index (2nd dword) in the VUE header with
     * primitive instance identifier. This is used for layered clears.
diff --git a/src/intel/vulkan/genX_blorp_exec.c b/src/intel/vulkan/genX_blorp_exec.c
index e9c85d56d5f..1592e7f7e3d 100644
--- a/src/intel/vulkan/genX_blorp_exec.c
+++ b/src/intel/vulkan/genX_blorp_exec.c
@@ -258,11 +258,6 @@ genX(blorp_exec)(struct blorp_batch *batch,
     */
    genX(cmd_buffer_enable_pma_fix)(cmd_buffer, false);
 
-   /* Disable VF statistics */
-   blorp_emit(batch, GENX(3DSTATE_VF_STATISTICS), vf) {
-      vf.StatisticsEnable = false;
-   }
-
    blorp_exec(batch, params);
 
    cmd_buffer->state.gfx.vb_dirty = ~0;
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 9bff2c8ac92..b752294250e 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -614,12 +614,6 @@ brw_upload_invariant_state(struct brw_context *brw)
       OUT_BATCH(0);
       ADVANCE_BATCH();
    }
-
-   const uint32_t _3DSTATE_VF_STATISTICS =
-      is_965 ? GEN4_3DSTATE_VF_STATISTICS : GM45_3DSTATE_VF_STATISTICS;
-   BEGIN_BATCH(1);
-   OUT_BATCH(_3DSTATE_VF_STATISTICS << 16 | 1);
-   ADVANCE_BATCH();
 }
 
 /**
diff --git a/src/mesa/drivers/dri/i965/genX_state_upload.c b/src/mesa/drivers/dri/i965/genX_state_upload.c
index e904c4c0cf9..ecffa2e8e86 100644
--- a/src/mesa/drivers/dri/i965/genX_state_upload.c
+++ b/src/mesa/drivers/dri/i965/genX_state_upload.c
@@ -928,6 +928,22 @@ const struct brw_tracked_state genX(cut_index) = {
 };
 #endif
 
+static void
+genX(upload_vf_statistics)(struct brw_context *brw)
+{
+   brw_batch_emit(brw, GENX(3DSTATE_VF_STATISTICS), vf) {
+      vf.StatisticsEnable = true;
+   }
+}
+
+const struct brw_tracked_state genX(vf_statistics) = {
+   .dirty = {
+      .mesa  = 0,
+      .brw   = BRW_NEW_BLORP | BRW_NEW_CONTEXT,
+   },
+   .emit = genX(upload_vf_statistics),
+};
+
 #if GEN_GEN >= 6
 /**
  * Determine the appropriate attribute override value to store into the
@@ -5640,6 +5656,8 @@ genX(init_atoms)(struct brw_context *brw)
 #if GEN_GEN < 6
    static const struct brw_tracked_state *render_atoms[] =
    {
+      &genX(vf_statistics),
+
       /* Once all the programs are done, we know how large urb entry
        * sizes need to be and can decide if we need to change the urb
        * layout.
@@ -5696,6 +5714,8 @@ genX(init_atoms)(struct brw_context *brw)
 #elif GEN_GEN == 6
    static const struct brw_tracked_state *render_atoms[] =
    {
+      &genX(vf_statistics),
+
       &genX(sf_clip_viewport),
 
       /* Command packets: */
@@ -5760,6 +5780,8 @@ genX(init_atoms)(struct brw_context *brw)
 #elif GEN_GEN == 7
    static const struct brw_tracked_state *render_atoms[] =
    {
+      &genX(vf_statistics),
+
       /* Command packets: */
 
       &genX(cc_vp),
@@ -5850,6 +5872,8 @@ genX(init_atoms)(struct brw_context *brw)
 #elif GEN_GEN >= 8
    static const struct brw_tracked_state *render_atoms[] =
    {
+      &genX(vf_statistics),
+
       &genX(cc_vp),
       &genX(sf_clip_viewport),