COMPMID-443: Use 3D tensors for fill_border_image

2x performance improvement on some GoogLeNet Pooling tests

Change-Id: If75336aa6308731a06462a73cd9209d24574509e
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/80342
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Steven Niu <steven.niu@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

1 files changed, 4 insertions, 4 deletions

diff --git a/src/core/CL/kernels/CLFillBorderKernel.cpp b/src/core/CL/kernels/CLFillBorderKernel.cpp
index 7683ff9a49..2c751a489c 100644
--- a/src/core/CL/kernels/CLFillBorderKernel.cpp
+++ b/src/core/CL/kernels/CLFillBorderKernel.cpp
@@ -108,7 +108,7 @@ void CLFillBorderKernel::configure(ICLTensor *tensor, BorderSize border_size, Bo
     const unsigned int total_valid_width = border_size.left + valid_width + border_size.right;
 
     // Set static kernel arguments
-    unsigned int idx = num_arguments_per_2D_tensor(); //Skip the tensor parameters
+    unsigned int idx = num_arguments_per_3D_tensor(); //Skip the tensor parameters
     ICLKernel::add_argument<cl_uint>(idx, valid_width);
     ICLKernel::add_argument<cl_uint>(idx, valid_height);
     ICLKernel::add_argument<cl_int2>(idx, valid_region_coords);
@@ -163,13 +163,13 @@ void CLFillBorderKernel::run(const Window &window, cl::CommandQueue &queue)
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
 
-    Window slice = window.first_slice_window_2D();
+    Window slice = window.first_slice_window_3D();
 
     do
     {
         unsigned int idx = 0;
-        add_2D_tensor_argument(idx, _tensor, slice);
+        add_3D_tensor_argument(idx, _tensor, slice);
         enqueue(queue, *this, slice, cl::NullRange);
     }
-    while(window.slide_window_slice_2D(slice));
+    while(window.slide_window_slice_3D(slice));
 }