iris: Use streaming loads to read from tiled surfaces

Always use the streaming load (since we know we have Broadwell+, all of
our target CPU support sse41) for reading back form the tiled surface
for mapping the resource. This means we hit the fast WC handling paths
on Atoms (without LLC), and for big Core (with LLC) using the streaming
load is no less efficient as we do not require the tiled buffer to be
pulled into the CPU cache.

Reviewed-by: Kenneth Graunke <[email protected]>

2 files changed, 5 insertions, 2 deletions

diff --git a/src/gallium/drivers/iris/iris_bufmgr.c b/src/gallium/drivers/iris/iris_bufmgr.c
index b20297fd94a..ae335163d45 100644
--- a/src/gallium/drivers/iris/iris_bufmgr.c
+++ b/src/gallium/drivers/iris/iris_bufmgr.c
@@ -1186,8 +1186,11 @@ can_map_cpu(struct iris_bo *bo, unsigned flags)
     * most drawing while non-persistent mappings are active, we may still use
     * the GPU for blits or other operations, causing batches to happen at
     * inconvenient times.
+    *
+    * If RAW is set, we expect the caller to be able to handle a WC buffer
+    * more efficiently than the involuntary clflushes.
     */
-   if (flags & (MAP_PERSISTENT | MAP_COHERENT | MAP_ASYNC))
+   if (flags & (MAP_PERSISTENT | MAP_COHERENT | MAP_ASYNC | MAP_RAW))
       return false;
 
    return !(flags & MAP_WRITE);
diff --git a/src/gallium/drivers/iris/iris_resource.c b/src/gallium/drivers/iris/iris_resource.c
index 155c8786d27..acb19364a3b 100644
--- a/src/gallium/drivers/iris/iris_resource.c
+++ b/src/gallium/drivers/iris/iris_resource.c
@@ -1143,7 +1143,7 @@ iris_map_tiled_memcpy(struct iris_transfer *map)
 
          isl_memcpy_tiled_to_linear(x1, x2, y1, y2, ptr, src, xfer->stride,
                                     surf->row_pitch_B, has_swizzling,
-                                    surf->tiling, ISL_MEMCPY);
+                                    surf->tiling, ISL_MEMCPY_STREAMING_LOAD);
          box.z++;
       }
    }