Revert "i965: Implement GL_PRIMITIVES_GENERATED with non-zero streams."

This reverts commit 3178d2474ae5bdd1102fb3d76a60d1d63c961ff5.

This caused GPU hangs on Ivybridge for some users and huge (80%)
performance regressions across the board on multiple platforms.

We need to find a better solution.  I've made several attempts, but none
of them have worked yet.  In the meantime, we should revert this.

Reverting it breaks GL_PRIMITIVES_GENERATED for non-zero streams, but
that's okay, since we don't expose GL_ARB_gpu_shader5 yet.

Fixes Piglit's EXT_transform_feedback/generatemipmap prims_generated
test case on Haswell.

diff --git a/src/mesa/drivers/dri/i965/gen6_queryobj.c b/src/mesa/drivers/dri/i965/gen6_queryobj.c
index 720dcbd..3aec99c 100644 (file)
--- a/src/mesa/drivers/dri/i965/gen6_queryobj.c
+++ b/src/mesa/drivers/dri/i965/gen6_queryobj.c
@@ -84,16 +84,11 @@ brw_store_register_mem64(struct brw_context *brw,
 
 static void
 write_primitives_generated(struct brw_context *brw,
-                           drm_intel_bo *query_bo, int stream, int idx)
+                           drm_intel_bo *query_bo, int idx)
 {
    intel_batchbuffer_emit_mi_flush(brw);
 
-   if (brw->gen >= 7) {
-      brw_store_register_mem64(brw, query_bo,
-                               GEN7_SO_PRIM_STORAGE_NEEDED(stream), idx);
-   } else {
-      brw_store_register_mem64(brw, query_bo, CL_INVOCATION_COUNT, idx);
-   }
+   brw_store_register_mem64(brw, query_bo, CL_INVOCATION_COUNT, idx);
 }
 
 static void
@@ -244,7 +239,7 @@ gen6_begin_query(struct gl_context *ctx, struct gl_query_object *q)
       break;
 
    case GL_PRIMITIVES_GENERATED:
-      write_primitives_generated(brw, query->bo, query->Base.Stream, 0);
+      write_primitives_generated(brw, query->bo, 0);
       break;
 
    case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:
@@ -282,7 +277,7 @@ gen6_end_query(struct gl_context *ctx, struct gl_query_object *q)
       break;
 
    case GL_PRIMITIVES_GENERATED:
-      write_primitives_generated(brw, query->bo, query->Base.Stream, 1);
+      write_primitives_generated(brw, query->bo, 1);
       break;
 
    case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:

diff --git a/src/mesa/drivers/dri/i965/gen7_sol_state.c b/src/mesa/drivers/dri/i965/gen7_sol_state.c
index faac716..166720e 100644 (file)
--- a/src/mesa/drivers/dri/i965/gen7_sol_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_sol_state.c
@@ -223,28 +223,14 @@ upload_3dstate_streamout(struct brw_context *brw, bool active,
    uint32_t dw1 = 0, dw2 = 0;
    int i;
 
-   /*
-    * From ARB_transform_feedback3:
-    *
-    *   "When a generated primitive query for a vertex stream is active, the
-    *   primitives-generated count is incremented every time a primitive
-    *   emitted to that stream reaches the Discarding Rasterization stage
-    *   (see Section 3.x) right before rasterization. This counter is
-    *   incremented whether or not transform feedback is active."
-    *
-    * Since we can only keep track of generated primitives for each stream
-    * in the SOL stage we need to make sure it is always active even if
-    * transform beedback is not. This way we can track primitives generated
-    * in each stream via SO_PRIMITIVE_STORAGE_NEEDED.
-    */
-   dw1 |= SO_FUNCTION_ENABLE;
-   dw1 |= SO_STATISTICS_ENABLE;
-
    if (active) {
       int urb_entry_read_offset = 0;
       int urb_entry_read_length = (vue_map->num_slots + 1) / 2 -
         urb_entry_read_offset;
 
+      dw1 |= SO_FUNCTION_ENABLE;
+      dw1 |= SO_STATISTICS_ENABLE;
+
       /* _NEW_LIGHT */
       if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION)
         dw1 |= SO_REORDER_TRAILING;