anv/query: flush render target before copying results

This change tracks render target writes in the pipeline and applies a
render target flush before copying the query results to make sure the
preceding operations have landed in memory before the command streamer
initiates the copy.

v2: Simplify logic in CopyQueryResults (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108909
Fixes: 37f9788e9a ("anv: flush pipeline before query result copies")
Cc: mesa-stable@lists.freedesktop.org

src/intel/vulkan/anv_private.h

@@ -1747,6 +1747,13 @@ enum anv_pipe_bits {
     * we would have to CS stall on every flush which could be bad.
     */
    ANV_PIPE_NEEDS_CS_STALL_BIT               = (1 << 21),
+
+   /* This bit does not exist directly in PIPE_CONTROL. It means that render
+    * target operations are ongoing. Some operations like copies on the
+    * command streamer might need to be aware of this to trigger the
+    * appropriate stall before they can proceed with the copy.
+    */
+   ANV_PIPE_RENDER_TARGET_WRITES              = (1 << 22),
 };
 
 #define ANV_PIPE_FLUSH_BITS ( \

src/intel/vulkan/genX_blorp_exec.c

@@ -263,4 +263,5 @@ genX(blorp_exec)(struct blorp_batch *batch,
    cmd_buffer->state.gfx.vb_dirty = ~0;
    cmd_buffer->state.gfx.dirty = ~0;
    cmd_buffer->state.push_constants_dirty = ~0;
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }

src/intel/vulkan/genX_cmd_buffer.c

@@ -1766,6 +1766,12 @@ genX(cmd_buffer_apply_pipe_flushes)(struct anv_cmd_buffer *cmd_buffer)
             pipe.StallAtPixelScoreboard = true;
       }
 
+      /* If a render target flush was emitted, then we can toggle off the bit
+       * saying that render target writes are ongoing.
+       */
+      if (bits & ANV_PIPE_RENDER_TARGET_CACHE_FLUSH_BIT)
+         bits &= ~(ANV_PIPE_RENDER_TARGET_WRITES);
+
       bits &= ~(ANV_PIPE_FLUSH_BITS | ANV_PIPE_CS_STALL_BIT);
    }
 
@@ -2777,6 +2783,8 @@ void genX(CmdDraw)(
       prim.StartInstanceLocation    = firstInstance;
       prim.BaseVertexLocation       = 0;
    }
+
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }
 
 void genX(CmdDrawIndexed)(
@@ -2816,6 +2824,8 @@ void genX(CmdDrawIndexed)(
       prim.StartInstanceLocation    = firstInstance;
       prim.BaseVertexLocation       = vertexOffset;
    }
+
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }
 
 /* Auto-Draw / Indirect Registers */
@@ -2949,6 +2959,8 @@ void genX(CmdDrawIndirect)(
 
       offset += stride;
    }
+
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }
 
 void genX(CmdDrawIndexedIndirect)(
@@ -2988,6 +3000,8 @@ void genX(CmdDrawIndexedIndirect)(
 
       offset += stride;
    }
+
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }
 
 static VkResult

src/intel/vulkan/genX_gpu_memcpy.c

@@ -302,4 +302,5 @@ genX(cmd_buffer_so_memcpy)(struct anv_cmd_buffer *cmd_buffer,
    }
 
    cmd_buffer->state.gfx.dirty |= ANV_CMD_DIRTY_PIPELINE;
+   cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
 }

src/intel/vulkan/genX_query.c

@@ -729,6 +729,15 @@ void genX(CmdCopyQueryPoolResults)(
    ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
    ANV_FROM_HANDLE(anv_buffer, buffer, destBuffer);
 
+   /* If render target writes are ongoing, request a render target cache flush
+    * to ensure proper ordering of the commands from the 3d pipe and the
+    * command streamer.
+    */
+   if (cmd_buffer->state.pending_pipe_bits & ANV_PIPE_RENDER_TARGET_WRITES) {
+      cmd_buffer->state.pending_pipe_bits |=
+         ANV_PIPE_RENDER_TARGET_CACHE_FLUSH_BIT;
+   }
+
    if ((flags & VK_QUERY_RESULT_WAIT_BIT) ||
        (cmd_buffer->state.pending_pipe_bits & ANV_PIPE_FLUSH_BITS)) {
       cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_CS_STALL_BIT;