COMPMID-1226 Extend CLMeanStdDev to support FP32 / FP16

- Extend support for FP16 in CLReduction.
- For F16/F32 MeanStdDev we perform one reduction operation for mean
and one for stddev and we calculate the final result in the host CPU.

Change-Id: Iad2099f26c0ba7969737d22f00c6c275634d875c
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/135870
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

1 files changed, 17 insertions, 6 deletions

diff --git a/src/core/CL/kernels/CLMeanStdDevKernel.cpp b/src/core/CL/kernels/CLMeanStdDevKernel.cpp
index fc8764dbfe..bd31131fe5 100644
--- a/src/core/CL/kernels/CLMeanStdDevKernel.cpp
+++ b/src/core/CL/kernels/CLMeanStdDevKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016, 2017 ARM Limited.
+ * Copyright (c) 2016-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -23,6 +23,7 @@
  */
 #include "arm_compute/core/CL/kernels/CLMeanStdDevKernel.h"
 
+#include "arm_compute/core/CL/CLHelpers.h"
 #include "arm_compute/core/CL/CLKernelLibrary.h"
 #include "arm_compute/core/CL/CLValidate.h"
 #include "arm_compute/core/CL/ICLTensor.h"
@@ -49,14 +50,24 @@ BorderSize CLMeanStdDevKernel::border_size() const
     return _border_size;
 }
 
+Status CLMeanStdDevKernel::validate(const ITensorInfo *input, float *mean, cl::Buffer *global_sum, float *stddev, cl::Buffer *global_sum_squared)
+{
+    ARM_COMPUTE_UNUSED(mean);
+    ARM_COMPUTE_UNUSED(stddev);
+    ARM_COMPUTE_UNUSED(global_sum);
+    ARM_COMPUTE_UNUSED(global_sum_squared);
+    ARM_COMPUTE_RETURN_ERROR_ON_INT64_BASE_ATOMICS_UNSUPPORTED();
+    ARM_COMPUTE_RETURN_ERROR_ON_TENSOR_NOT_2D(input);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
+
+    return Status{};
+}
+
 void CLMeanStdDevKernel::configure(const ICLImage *input, float *mean, cl::Buffer *global_sum, float *stddev, cl::Buffer *global_sum_squared)
 {
-    ARM_COMPUTE_ERROR_ON_INT64_BASE_ATOMICS_UNSUPPORTED();
-    ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
-    ARM_COMPUTE_ERROR_ON(nullptr == mean);
-    ARM_COMPUTE_ERROR_ON(nullptr == global_sum);
+    ARM_COMPUTE_ERROR_ON_NULLPTR(input, mean, global_sum);
     ARM_COMPUTE_ERROR_ON(stddev && nullptr == global_sum_squared);
+    ARM_COMPUTE_ERROR_THROW_ON(CLMeanStdDevKernel::validate(input->info(), mean, global_sum, stddev, global_sum_squared));
 
     _input              = input;
     _mean               = mean;