COMPMID-802 Add NHWC data format support for NEON im2col.
Change-Id: I86e678179106a2b83d1c6a7cfe562df91b0f9eb2
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/124000
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
diff --git a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
index 676706f..5dd1f00 100644
--- a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
+++ b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
@@ -114,7 +114,7 @@
// If the fully connected layer is called after a convolution layer, the input tensor must be linearized
// Initialize output tensor for im2col
- TensorShape shape_im2col = compute_im2col_shape(input->info());
+ TensorShape shape_im2col = compute_im2col_fc_shape(input->info());
_im2col_output.allocator()->init(input->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_im2col));
// Configure im2col kernel
@@ -244,7 +244,7 @@
bool is_quantized = is_data_type_quantized_asymmetric(input->data_type());
const GPUTarget gpu_target = CLScheduler::get().target();
- const ITensorInfo &im2col_input = TensorInfo(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_im2col_shape(input)));
+ const ITensorInfo &im2col_input = TensorInfo(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_im2col_fc_shape(input)));
const ITensorInfo &reshaped_weights = TensorInfo(weights->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_transposed_shape(*weights)));
const ITensorInfo &gemmlowp_output = TensorInfo(output->clone()->set_is_resizable(true).reset_padding().set_data_type(DataType::S32));