COMPMID-1722 : CL: Implement Range

Change-Id: I88da6eb5289c303b1dc91606c1560ce629746058
Reviewed-on: https://review.mlplatform.org/381
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

16 files changed, 1018 insertions, 2 deletions

diff --git a/arm_compute/core/CL/CLHelpers.h b/arm_compute/core/CL/CLHelpers.h
index a86870a250..78427c3738 100644
--- a/arm_compute/core/CL/CLHelpers.h
+++ b/arm_compute/core/CL/CLHelpers.h
@@ -135,5 +135,14 @@ bool dot8_acc_supported(const cl::Device &device);
  * @return True if the configuration is supported
  */
 bool cl_winograd_convolution_layer_supported(const Size2D &output_tile, const Size2D &kernel_size, DataLayout data_layout);
+
+/** Helper function to get the preferred native vector width size for built-in scalar types that can be put into vectors
+ *
+ * @param[in] device A CL device
+ * @param[in] dt     data type
+ *
+ * @return preferred vector width
+ */
+size_t preferred_vector_width(const cl::Device &device, DataType dt);
 }
 #endif /* __ARM_COMPUTE_CLHELPERS_H__ */
diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index d89426dd32..cfcfb7400b 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -113,6 +113,7 @@
 #include "arm_compute/core/CL/kernels/CLQuantizationLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLROIAlignLayerKernel.h"
 #include "arm_compute/core/CL/kernels/CLROIPoolingLayerKernel.h"
+#include "arm_compute/core/CL/kernels/CLRangeKernel.h"
 #include "arm_compute/core/CL/kernels/CLReductionOperationKernel.h"
 #include "arm_compute/core/CL/kernels/CLRemapKernel.h"
 #include "arm_compute/core/CL/kernels/CLReorgLayerKernel.h"
diff --git a/arm_compute/core/CL/kernels/CLRangeKernel.h b/arm_compute/core/CL/kernels/CLRangeKernel.h
new file mode 100644
index 0000000000..2da21175ce
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLRangeKernel.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_CLRANGEKERNEL_H__
+#define __ARM_COMPUTE_CLRANGEKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Kernel class for Range()
+ *
+ * range generates a 1-D tensor containing a sequence of numbers that begins at 'start' and extends by increments
+ * of 'step' up to but not including 'end'.
+ */
+class CLRangeKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    CLRangeKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLRangeKernel(const CLRangeKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLRangeKernel &operator=(const CLRangeKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CLRangeKernel(CLRangeKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CLRangeKernel &operator=(CLRangeKernel &&) = default;
+    /** Default destructor */
+    ~CLRangeKernel() = default;
+    /** Initialise the kernel's output tensor, start, end and step of the sequence.
+     *
+     * @param[out] output Output tensor. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32.
+     * @param[in]  start  The starting value of the sequence.
+     * @param[in]  end    The ending (not including) value of the sequence.
+     * @param[in]  step   The gap between each pair of values in the sequence.
+     */
+    void configure(ICLTensor *output, float start, float end, float step);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLRangeKernel
+     *
+     * @param[in] output Output tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32.
+     * @param[in] start  The starting value of the sequence.
+     * @param[in] end    The ending (not including) value of the sequence.
+     * @param[in] step   The gap between each pair of values in the sequence.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *output, float start, float end, float step);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+    float      _start;  /**< Start of sequence */
+    float      _end;    /**< End of sequence */
+    float      _step;   /**< Increment/step value */
+    ICLTensor *_output; /**< Destination tensor */
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CLRANGEKERNEL_H__ */
diff --git a/arm_compute/core/utils/misc/Utility.h b/arm_compute/core/utils/misc/Utility.h
index 37c8b66e06..8dd9afd5cd 100644
--- a/arm_compute/core/utils/misc/Utility.h
+++ b/arm_compute/core/utils/misc/Utility.h
@@ -80,8 +80,10 @@ std::array<typename std::iterator_traits<Iterator>::value_type, N> make_array(It
  *
  *  @return Clamped value.
  */
-template <typename T>
-inline T clamp(const T &n, const T &lower, const T &upper)
+template <typename DataType, typename RangeType = DataType>
+inline DataType clamp(const DataType &n,
+                      const DataType &lower = std::numeric_limits<RangeType>::lowest(),
+                      const DataType &upper = std::numeric_limits<RangeType>::max())
 {
     return std::max(lower, std::min(n, upper));
 }
diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h
index d4827af88a..303a8989a9 100644
--- a/arm_compute/runtime/CL/CLFunctions.h
+++ b/arm_compute/runtime/CL/CLFunctions.h
@@ -111,6 +111,7 @@
 #include "arm_compute/runtime/CL/functions/CLRNNLayer.h"
 #include "arm_compute/runtime/CL/functions/CLROIAlignLayer.h"
 #include "arm_compute/runtime/CL/functions/CLROIPoolingLayer.h"
+#include "arm_compute/runtime/CL/functions/CLRange.h"
 #include "arm_compute/runtime/CL/functions/CLReduceMean.h"
 #include "arm_compute/runtime/CL/functions/CLReductionOperation.h"
 #include "arm_compute/runtime/CL/functions/CLRemap.h"
diff --git a/arm_compute/runtime/CL/functions/CLRange.h b/arm_compute/runtime/CL/functions/CLRange.h
new file mode 100644
index 0000000000..2614534f14
--- /dev/null
+++ b/arm_compute/runtime/CL/functions/CLRange.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_CLRANGE_H__
+#define __ARM_COMPUTE_CLRANGE_H__
+
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/ICLSimpleFunction.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Basic function to run @ref CLRangeKernel
+ *
+ * @note The tensor data type for the output must be U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32.
+ * @note The function performs generates a sequence with the given start, end and step.
+ */
+class CLRange : public ICLSimpleFunction
+{
+public:
+    /** Initialise the kernel's start, end, step and output tensor.
+     *
+     * @param[out] output Output tensor. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32.
+     * @param[in]  start  The starting value of the sequence.
+     * @param[in]  end    The ending (not including) value of the sequence.
+     * @param[in]  step   The gap between each pair of values in the sequence. Default is 1.
+     */
+    void configure(ICLTensor *output, float start, float end, float step = 1.f);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLRange
+     *
+     * @param[in] output Output tensor info. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32.
+     * @param[in] start  The starting value of the sequence.
+     * @param[in] end    The ending (not including) value of the sequence.
+     * @param[in] step   The gap between each pair of values in the sequence. Default is 1.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *output, float start, float end, float step = 1.f);
+};
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_CLRANGE_H__ */
diff --git a/src/core/CL/CLHelpers.cpp b/src/core/CL/CLHelpers.cpp
index 0947d58973..924fb1d322 100644
--- a/src/core/CL/CLHelpers.cpp
+++ b/src/core/CL/CLHelpers.cpp
@@ -230,4 +230,30 @@ bool cl_winograd_convolution_layer_supported(const Size2D &output_tile, const Si
         return (std::find(winograd_configs_nhwc.begin(), winograd_configs_nhwc.end(), p) != winograd_configs_nhwc.end());
     }
 }
+
+size_t preferred_vector_width(const cl::Device &device, const DataType dt)
+{
+    switch(dt)
+    {
+        case DataType::U8:
+        case DataType::S8:
+        case DataType::QASYMM8:
+            return device.getInfo<CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR>();
+        case DataType::U16:
+        case DataType::S16:
+            return device.getInfo<CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT>();
+        case DataType::U32:
+        case DataType::S32:
+            return device.getInfo<CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT>();
+        case DataType::F16:
+        case DataType::F32:
+            return device.getInfo<CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT>();
+        case DataType::U64:
+        case DataType::S64:
+            return device.getInfo<CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG>();
+        default:
+            return 1;
+    }
+}
+
 } // namespace arm_compute
diff --git a/src/core/CL/CLKernelLibrary.cpp b/src/core/CL/CLKernelLibrary.cpp
index 03bc8d15db..ea5342c19d 100644
--- a/src/core/CL/CLKernelLibrary.cpp
+++ b/src/core/CL/CLKernelLibrary.cpp
@@ -382,6 +382,8 @@ const std::map<std::string, std::string> CLKernelLibrary::_kernel_program_map =
     { "pooling_layer_MxN_quantized_nchw", "pooling_layer_quantized.cl" },
     { "prior_box_layer_nchw", "prior_box_layer.cl" },
     { "quantization_layer", "quantization_layer.cl" },
+    { "range", "range.cl" },
+    { "range_quantized", "range.cl" },
     { "reduction_operation_x", "reduction_operation.cl" },
     { "reduction_operation_non_parallel_x", "reduction_operation.cl" },
     { "reduction_operation_y", "reduction_operation.cl" },
@@ -817,6 +819,10 @@ const std::map<std::string, std::string> CLKernelLibrary::_program_source_map =
 #include "./cl_kernels/quantization_layer.clembed"
     },
     {
+        "range.cl",
+#include "./cl_kernels/range.clembed"
+    },
+    {
         "reduction_operation.cl",
 #include "./cl_kernels/reduction_operation.clembed"
     },
diff --git a/src/core/CL/cl_kernels/range.cl b/src/core/CL/cl_kernels/range.cl
new file mode 100644
index 0000000000..d122c9ac89
--- /dev/null
+++ b/src/core/CL/cl_kernels/range.cl
@@ -0,0 +1,145 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "helpers.h"
+
+#if defined(VECTOR_SIZE) && defined(START) && defined(STEP) && defined(DATA_TYPE)
+/** Generates a sequence of numbers starting from START and extends by increments of 'STEP' up to but not including 'END'.
+ *
+ * @note starting value of the sequence must be given as a preprocessor argument using -DSTART=value. e.g. -DSTART=0
+ * @note difference between consequtive elements of the sequence must be given as a preprocessor argument using -DSTEP=value. e.g. -DSTEP=1
+ * @note Datatype must be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=short
+ * @note vector size supported by the device must be given as a preprocessor argument using -DVECTOR_SIZE=value. e.g. -DDATA_TYPE=4
+ *
+ * @param[out] out_ptr                           Pointer to the destination tensor. Supported data types: U8/S8/U16/S16/U32/S32/F16/F32.
+ * @param[in]  out_stride_x                      Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  out_step_x                        dst_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  out_offset_first_element_in_bytes The offset of the first element in the destination tensor
+ */
+__kernel void range(
+    VECTOR_DECLARATION(out))
+{
+    uint           id      = get_global_id(0) * VECTOR_SIZE;
+    __global void *dst_ptr = out_ptr + out_offset_first_element_in_bytes + id * sizeof(DATA_TYPE);
+#if VECTOR_SIZE == 1
+    DATA_TYPE seq;
+    seq = (DATA_TYPE)START + (DATA_TYPE)id * (DATA_TYPE)STEP;
+
+    *((__global DATA_TYPE *)dst_ptr) = seq;
+#else // VECTOR_SIZE == 1
+    VEC_DATA_TYPE(DATA_TYPE, VECTOR_SIZE)
+    seq;
+
+    seq.s0 = ((DATA_TYPE)START + (DATA_TYPE)id * (DATA_TYPE)STEP);
+#if VECTOR_SIZE > 1
+    seq.s1 = seq.s0 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 2
+    seq.s2 = seq.s1 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 3
+    seq.s3 = seq.s2 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 4
+    seq.s4 = seq.s3 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 5
+    seq.s5 = seq.s4 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 6
+    seq.s6 = seq.s5 + (DATA_TYPE)STEP;
+#if VECTOR_SIZE > 7
+    seq.s7 = seq.s6 + (DATA_TYPE)STEP;
+#endif // VECTOR_SIZE > 7
+#endif // VECTOR_SIZE > 6
+#endif // VECTOR_SIZE > 5
+#endif // VECTOR_SIZE > 4
+#endif // VECTOR_SIZE > 3
+#endif // VECTOR_SIZE > 2
+#endif // VECTOR_SIZE > 1
+    VSTORE(VECTOR_SIZE)
+    (seq, 0, ((__global DATA_TYPE *)dst_ptr));
+#endif //VECTOR_SIZE == 1
+}
+
+#if defined(OFFSET_OUT) && defined(SCALE_OUT)
+
+#define CONVERT_RTE(x, type) (convert_##type##_rte((x)))
+#define CONVERT_DOWN(x, type) CONVERT_RTE(x, type)
+
+/** Generates a sequence of numbers starting from START and extends by increments of 'STEP' up to but not including 'END'.
+ *
+ * @note starting value of the sequence must be given as a preprocessor argument using -DSTART=value. e.g. -DSTART=0
+ * @note difference between consequtive elements of the sequence must be given as a preprocessor argument using -DSTEP=value. e.g. -DSTEP=1
+ * @note Datatype must be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=short
+ * @note vector size supported by the device must be given as a preprocessor argument using -DVECTOR_SIZE=vector_size. e.g. -DDATA_TYPE=4
+ * @note The quantization offset of the output must be passed at compile time using -DOFFSET_OUT, i.e. -DOFFSET_OUT=10
+ * @note The quantization scale of the output must be passed at compile time using -DSCALE_OUT, i.e. -DSCALE_OUT=10
+ *
+ * @param[out] out_ptr                           Pointer to the destination tensor. Supported data types: QASYMM8.
+ * @param[in]  out_stride_x                      Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  out_step_x                        dst_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  out_offset_first_element_in_bytes The offset of the first element in the destination tensor
+ */
+__kernel void range_quantized(
+    VECTOR_DECLARATION(out))
+{
+    size_t         id      = get_global_id(0) * VECTOR_SIZE;
+    __global void *dst_ptr = out_ptr + out_offset_first_element_in_bytes + id * sizeof(DATA_TYPE);
+#if VECTOR_SIZE == 1
+    float seq;
+    seq                          = (float)START + (float)id * (float)STEP;
+    seq                          = (DATA_TYPE)(int)(seq / ((float)SCALE_OUT) + (float)OFFSET_OUT);
+    seq                          = max(0.0f, min(seq, 255.0f));
+    *((__global uchar *)dst_ptr) = CONVERT_SAT(CONVERT_DOWN(seq, int), uchar);
+#else // VECTOR_SIZE == 1
+    VEC_DATA_TYPE(float, VECTOR_SIZE)
+    seq;
+    seq.s0 = (float)START + id * (float)STEP;
+#if VECTOR_SIZE > 1
+    seq.s1 = seq.s0 + (float)STEP;
+#if VECTOR_SIZE > 2
+    seq.s2 = seq.s1 + (float)STEP;
+#if VECTOR_SIZE > 3
+    seq.s3 = seq.s2 + (float)STEP;
+#if VECTOR_SIZE > 4
+    seq.s4 = seq.s3 + (float)STEP;
+#if VECTOR_SIZE > 5
+    seq.s5 = seq.s4 + (float)STEP;
+#if VECTOR_SIZE > 6
+    seq.s6 = seq.s5 + (float)STEP;
+#if VECTOR_SIZE > 7
+    seq.s7 = seq.s6 + (float)STEP;
+#endif // VECTOR_SIZE > 7
+#endif // VECTOR_SIZE > 6
+#endif // VECTOR_SIZE > 5
+#endif // VECTOR_SIZE > 4
+#endif // VECTOR_SIZE > 3
+#endif // VECTOR_SIZE > 2
+#endif // VECTOR_SIZE > 1
+    seq    = seq / ((VEC_DATA_TYPE(float, VECTOR_SIZE))((float)SCALE_OUT)) + ((VEC_DATA_TYPE(float, VECTOR_SIZE))((float)OFFSET_OUT));
+    seq    = max((VEC_DATA_TYPE(float, VECTOR_SIZE))(0.0f), min(seq, (VEC_DATA_TYPE(float, VECTOR_SIZE))(255.0f)));
+    VEC_DATA_TYPE(uchar, VECTOR_SIZE)
+    res = CONVERT_SAT(CONVERT_DOWN(seq, VEC_DATA_TYPE(int, VECTOR_SIZE)), VEC_DATA_TYPE(uchar, VECTOR_SIZE));
+    VSTORE(VECTOR_SIZE)
+    (res, 0, ((__global DATA_TYPE *)dst_ptr));
+#endif // VECTOR_SIZE == 1
+}
+#endif // defined(OFFSET_OUT) && defined(SCALE_OUT)
+
+#endif // defined(VECTOR_SIZE) && defined(START) && defined(STEP) && defined(DATA_TYPE)
diff --git a/src/core/CL/kernels/CLRangeKernel.cpp b/src/core/CL/kernels/CLRangeKernel.cpp
new file mode 100644
index 0000000000..f53fe87a70
--- /dev/null
+++ b/src/core/CL/kernels/CLRangeKernel.cpp
@@ -0,0 +1,159 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/core/CL/kernels/CLRangeKernel.h"
+
+#include "arm_compute/core/CL/CLHelpers.h"
+#include "arm_compute/core/CL/CLValidate.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "utils/Utils.h"
+
+using namespace arm_compute;
+
+namespace
+{
+size_t num_of_elements_in_range(const float start, const float end, const float step)
+{
+    ARM_COMPUTE_ERROR_ON_MSG(step == 0, "CLRange Step cannot be 0");
+    return size_t(std::ceil((end - start) / step));
+}
+
+unsigned int get_num_elems_processed_per_iteration(const DataType dt)
+{
+    unsigned int num_elems_processed_per_iteration = preferred_vector_width(CLKernelLibrary::get().get_device(), dt);
+    if(num_elems_processed_per_iteration > 8)
+    {
+        num_elems_processed_per_iteration = 8; //kernel uses only 8 lanes.
+    }
+    return num_elems_processed_per_iteration;
+}
+
+Status validate_arguments(const ITensorInfo &output, const float start, const float end, const float step)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&output,
+                                                         1,
+                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
+                                                         DataType::U16, DataType::S16,
+                                                         DataType::U32, DataType::S32,
+                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&output);
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG((start == end), "start of the requested sequence must not be equal to the end");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((start < end) && (step <= 0)), "step must be greater than 0 when start < end");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((start > end) && (step >= 0)), "step must be less than 0 when start > end");
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(!utils::check_value_range(start, output.data_type(), output.quantization_info()), "start value is outside the range of the data type");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(!utils::check_value_range(end, output.data_type(), output.quantization_info()), "end value is outside the range of the data type");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(!utils::check_value_range(step, output.data_type(), output.quantization_info()), "step value is outside the range of the data type");
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG((start == end), "start of the requested sequence must not be equal to the end");
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output.num_dimensions() != 1, "Output has to be a 1-D tensor");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(output.tensor_shape().total_size() < num_of_elements_in_range(start, end, step), "Output tensor size is incorrect");
+
+    return Status{};
+}
+
+std::pair<Status, Window> validate_and_configure_window(ITensorInfo &output, const float start, const float end, const float step)
+{
+    unsigned int num_elems_processed_per_iteration = get_num_elems_processed_per_iteration(output.data_type());
+    // Auto initialize output if not initialized
+    auto_init_if_empty(output, TensorShape(num_of_elements_in_range(start, end, step)), 1, output.data_type(), output.quantization_info());
+
+    // Configure kernel window
+    Window win = calculate_max_window(output, Steps(num_elems_processed_per_iteration));
+
+    AccessWindowHorizontal output_access(&output, 0, num_elems_processed_per_iteration);
+    bool                   window_changed = update_window_and_padding(win, output_access);
+    output_access.set_valid_region(win, ValidRegion(Coordinates(), TensorShape(num_of_elements_in_range(start, end, step))));
+    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
+    return std::make_pair(err, win);
+}
+} // namespace
+
+CLRangeKernel::CLRangeKernel()
+    : _start(0), _end(1), _step(1), _output(nullptr)
+{
+}
+
+void CLRangeKernel::configure(ICLTensor *output, const float start, const float end, const float step)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(output);
+
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*(output->info()), start, end, step));
+
+    // Configure kernel window
+    auto win_config = validate_and_configure_window(*(output->info()), start, end, step);
+    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
+
+    _start  = start;
+    _end    = end;
+    _step   = step;
+    _output = output;
+
+    std::string kernel_name = "range";
+
+    unsigned int num_elems_processed_per_iteration = get_num_elems_processed_per_iteration(output->info()->data_type());
+    // Set build options
+    CLBuildOptions build_opts;
+    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(output->info()->data_type()));
+    build_opts.add_option("-DVECTOR_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
+    build_opts.add_option("-DSTART=" + support::cpp11::to_string(start));
+    build_opts.add_option("-DSTEP=" + support::cpp11::to_string(step));
+    if(is_data_type_quantized_asymmetric(output->info()->data_type()))
+    {
+        build_opts.add_option("-DOFFSET_OUT=" + support::cpp11::to_string(output->info()->quantization_info().offset));
+        build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(output->info()->quantization_info().scale));
+        kernel_name += "_quantized";
+    }
+    // Create kernel
+    _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel(kernel_name, build_opts.options()));
+    ICLKernel::configure_internal(win_config.second);
+
+    // Set config_id for enabling LWS tuning
+    _config_id = kernel_name;
+    _config_id += "_";
+    _config_id += lower_string(string_from_data_type(output->info()->data_type()));
+    _config_id += "_";
+    _config_id += support::cpp11::to_string(output->info()->dimension(0));
+}
+
+Status CLRangeKernel::validate(const ITensorInfo *output, const float start, const float end, const float step)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
+
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(*output, start, end, step));
+    ARM_COMPUTE_RETURN_ON_ERROR((validate_and_configure_window(*(output->clone()), start, end, step)).first);
+
+    return Status{};
+}
+
+void CLRangeKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
+    unsigned int idx = 0;
+    add_1D_tensor_argument(idx, _output, window);
+
+    enqueue(queue, *this, window, lws_hint());
+}
diff --git a/src/runtime/CL/functions/CLRange.cpp b/src/runtime/CL/functions/CLRange.cpp
new file mode 100644
index 0000000000..b2cd472d02
--- /dev/null
+++ b/src/runtime/CL/functions/CLRange.cpp
@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/runtime/CL/functions/CLRange.h"
+
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/CL/kernels/CLRangeKernel.h"
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/runtime/CL/CLScheduler.h"
+#include "support/ToolchainSupport.h"
+
+using namespace arm_compute;
+
+void CLRange::configure(ICLTensor *output, const float start, const float end, const float step)
+{
+    auto k = arm_compute::support::cpp14::make_unique<CLRangeKernel>();
+    k->set_target(CLScheduler::get().target());
+    k->configure(output, start, end, step);
+    _kernel = std::move(k);
+
+    // Tune kernels
+    CLScheduler::get().tune_kernel_static(*_kernel);
+}
+
+Status CLRange::validate(const ITensorInfo *output, const float start, const float end, const float step)
+{
+    return CLRangeKernel::validate(output, start, end, step);
+}
diff --git a/tests/validation/CL/Range.cpp b/tests/validation/CL/Range.cpp
new file mode 100644
index 0000000000..b1804f2184
--- /dev/null
+++ b/tests/validation/CL/Range.cpp
@@ -0,0 +1,162 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLRange.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/ConvertPolicyDataset.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/RangeFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+constexpr RelativeTolerance<float> tolerance(0.01f);
+constexpr AbsoluteTolerance<float> abs_tolerance(0.02f);
+const auto                         start_dataset          = framework::dataset::make("Start", { float(3), float(-17), float(16) });
+const auto                         unsigned_start_dataset = framework::dataset::make("Start", { float(3), float(16) });
+const auto                         float_step_dataset     = framework::dataset::make("Step", { float(1), float(-0.2f), float(0.2), float(12.2), float(-12.2), float(-1.2), float(-3), float(3) });
+const auto                         step_dataset           = framework::dataset::make("Step", { float(1), float(12), float(-12), float(-1), float(-3), float(3) });
+const auto                         unsigned_step_dataset  = framework::dataset::make("Step", { float(1), float(12), float(3) });
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(Range)
+
+// *INDENT-OFF*
+// clang-format off
+
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
+               framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(32U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(27U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(32U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(32U), 1, DataType::F32),
+                                                        TensorInfo(TensorShape(27U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(27U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(10U), 1, DataType::U8),
+                                                      }),
+               framework::dataset::make("Start",{ 0.0f, 15.0f, 1500.0f, 100.0f, -15.0f, 0.2f , 2.0f , 10.0f})),
+               framework::dataset::make("End",{ 100.0f, 15.0f, 2500.0f, -1000.0f, 15.0f, 10.0f, 10.0f,100.0f })),
+               framework::dataset::make("Step",{ 100.0f, 15.0f, 10.0f, 100.0f, -15.0f, 1.0f, 0.0f, 10.0f })),
+               framework::dataset::make("Expected", { false, //1-D tensor expected
+                                                    false, //start == end
+                                                    false, //output vector size insufficient
+                                                    false, //sign of step incorrect
+                                                    false, //sign of step incorrect
+                                                    false, //data type incompatible
+                                                    false, //step = 0
+                                                    true,
+                                                    })),
+               output_info, start, end, step, expected)
+{
+    ARM_COMPUTE_EXPECT(bool(CLRange::validate(&output_info, start, end, step)) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLRangeFixture = RangeFixture<CLTensor, CLAccessor, CLRange, T>;
+
+TEST_SUITE(U8)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLRangeFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(
+                                                                                                                 framework::dataset::make("DataType", DataType::U8),
+                                                                                                                 unsigned_start_dataset),
+                                                                                                             unsigned_step_dataset),
+                                                                                                     framework::dataset::make("QuantizationInfo", { QuantizationInfo() })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance, 0.f, abs_tolerance);
+}
+TEST_SUITE_END() //U8
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLRangeFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(
+                                                                                                                 framework::dataset::make("DataType", DataType::QASYMM8),
+                                                                                                                 start_dataset),
+                                                                                                             step_dataset),
+                                                                                                     framework::dataset::make("QuantizationInfo", { QuantizationInfo(0.3457f, 120.0f) })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance, 0.f, abs_tolerance);
+}
+TEST_SUITE_END() //QASYMM8
+TEST_SUITE_END() //Quantized
+
+TEST_SUITE(S16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLRangeFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(
+                                                                                                                 framework::dataset::make("DataType", DataType::S16),
+                                                                                                                 start_dataset),
+                                                                                                             step_dataset),
+                                                                                                     framework::dataset::make("QuantizationInfo", { QuantizationInfo() })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance, 0.f, abs_tolerance);
+}
+TEST_SUITE_END() //S16
+
+TEST_SUITE(Float)
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLRangeFixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(
+                                                                                                              framework::dataset::make("DataType", DataType::F16),
+                                                                                                              start_dataset),
+                                                                                                          float_step_dataset),
+                                                                                                  framework::dataset::make("QuantizationInfo", { QuantizationInfo() })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance, 0.f, abs_tolerance);
+}
+TEST_SUITE_END() //FP16
+
+TEST_SUITE(FP32)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLRangeFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(
+                                                                                                               framework::dataset::make("DataType", DataType::F32),
+                                                                                                               start_dataset),
+                                                                                                           float_step_dataset),
+                                                                                                   framework::dataset::make("QuantizationInfo", { QuantizationInfo() })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance, 0.f, abs_tolerance);
+}
+TEST_SUITE_END() //FP32
+TEST_SUITE_END() //Float
+
+TEST_SUITE_END() //Range
+TEST_SUITE_END() //CL
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/RangeFixture.h b/tests/validation/fixtures/RangeFixture.h
new file mode 100644
index 0000000000..c192eee14f
--- /dev/null
+++ b/tests/validation/fixtures/RangeFixture.h
@@ -0,0 +1,140 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_TEST_RANGE_FIXTURE
+#define ARM_COMPUTE_TEST_RANGE_FIXTURE
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+#include "tests/AssetsLibrary.h"
+#include "tests/Globals.h"
+#include "tests/IAccessor.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Fixture.h"
+#include "tests/validation/Helpers.h"
+#include "tests/validation/reference/Range.h"
+
+#include <algorithm>
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+size_t num_of_elements_in_range(float start, float end, float step)
+{
+    ARM_COMPUTE_ERROR_ON(step == 0);
+    return size_t(std::ceil((end - start) / step));
+}
+}
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class RangeFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(const DataType data_type0, float start, float step, const QuantizationInfo qinfo0 = QuantizationInfo())
+    {
+        _target    = compute_target(data_type0, qinfo0, start, step);
+        _reference = compute_reference(data_type0, qinfo0, start, step);
+    }
+
+protected:
+    float get_random_end(const DataType output_data_type, const QuantizationInfo qinfo_out, float start, float step)
+    {
+        std::uniform_real_distribution<> distribution(1, 100);
+        std::mt19937                     gen(library->seed());
+        float                            end = start;
+        switch(output_data_type)
+        {
+            case DataType::U8:
+                end += std::max((uint8_t)1, static_cast<uint8_t>(distribution(gen))) * step;
+                return utility::clamp<float, uint8_t>(end);
+            case DataType::U16:
+                end += std::max((uint16_t)1, static_cast<uint16_t>(distribution(gen))) * step;
+                return utility::clamp<float, uint16_t>(end);
+            case DataType::U32:
+                end += std::max((uint32_t)1, static_cast<uint32_t>(distribution(gen))) * step;
+                return utility::clamp<float, uint32_t>(end);
+            case DataType::S8:
+                end += std::max((int8_t)1, static_cast<int8_t>(distribution(gen))) * step;
+                return utility::clamp<float, int8_t>(end);
+            case DataType::S16:
+                end += std::max((int16_t)1, static_cast<int16_t>(distribution(gen))) * step;
+                return utility::clamp<float, int16_t>(end);
+            case DataType::S32:
+                end += std::max((int32_t)1, static_cast<int32_t>(distribution(gen))) * step;
+                return utility::clamp<float, int32_t>(end);
+            case DataType::F32:
+                end += std::max(1.0f, static_cast<float>(distribution(gen))) * step;
+                return end;
+            case DataType::F16:
+                end += std::max(half(1.0f), static_cast<half>(distribution(gen))) * step;
+                return utility::clamp<float, half>(end);
+            case DataType::QASYMM8:
+                return utility::clamp<float, uint8_t>(end + (float)distribution(gen) * step, qinfo_out.dequantize(0), qinfo_out.dequantize(std::numeric_limits<uint8_t>::max()));
+            default:
+                return 0;
+        }
+    }
+
+    TensorType compute_target(const DataType output_data_type, const QuantizationInfo qinfo_out, float start, float step)
+    {
+        float  end             = get_random_end(output_data_type, qinfo_out, start, step);
+        size_t num_of_elements = num_of_elements_in_range(start, end, step);
+        // Create tensor
+        TensorType dst = create_tensor<TensorType>(TensorShape(num_of_elements), output_data_type, 1, qinfo_out);
+        // Create and configure function
+        FunctionType range_func;
+        range_func.configure(&dst, start, end, step);
+
+        ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+        // Allocate tensors
+        dst.allocator()->allocate();
+
+        ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        // Compute function
+        range_func.run();
+        return dst;
+    }
+
+    SimpleTensor<T> compute_reference(const DataType output_data_type, const QuantizationInfo qinfo_out, float start, float step)
+    {
+        // Create tensor
+        const float     end             = get_random_end(output_data_type, qinfo_out, start, step);
+        size_t          num_of_elements = num_of_elements_in_range(start, end, step);
+        SimpleTensor<T> ref_dst{ TensorShape(num_of_elements ? num_of_elements : 1), output_data_type, 1, qinfo_out };
+        return reference::range<T>(ref_dst, start, num_of_elements, step);
+    }
+
+    TensorType      _target{};
+    SimpleTensor<T> _reference{};
+};
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_RANGE_FIXTURE */
diff --git a/tests/validation/reference/Range.cpp b/tests/validation/reference/Range.cpp
new file mode 100644
index 0000000000..c24512fa9d
--- /dev/null
+++ b/tests/validation/reference/Range.cpp
@@ -0,0 +1,77 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/core/Types.h"
+#include "tests/validation/Helpers.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+namespace
+{
+template <typename T>
+void generate_range(SimpleTensor<T> &dst, float start, const size_t num_of_elements, float step)
+{
+    float val = start;
+    for(size_t index = 0; index < num_of_elements; index++)
+    {
+        dst[index] = static_cast<T>(val);
+        val += step;
+    }
+}
+} // namespace
+
+template <typename T>
+SimpleTensor<T> range(SimpleTensor<T> &dst, float start, const size_t num_of_elements, float step)
+{
+    generate_range(dst, start, num_of_elements, step);
+    return dst;
+}
+
+template <>
+SimpleTensor<uint8_t> range(SimpleTensor<uint8_t> &dst, float start, const size_t num_of_elements, float step)
+{
+    if(dst.data_type() == DataType::QASYMM8)
+    {
+        SimpleTensor<float> dst_tmp{ dst.shape(), DataType::F32, 1 };
+        generate_range(dst_tmp, start, num_of_elements, step);
+        return convert_to_asymmetric(dst_tmp, dst.quantization_info());
+    }
+    generate_range(dst, start, num_of_elements, step);
+    return dst;
+}
+template SimpleTensor<float> range(SimpleTensor<float> &dst, float start, const size_t num_of_elements, float step);
+template SimpleTensor<half> range(SimpleTensor<half> &dst, float start, const size_t num_of_elements, float step);
+template SimpleTensor<int8_t> range(SimpleTensor<int8_t> &dst, float start, const size_t num_of_elements, float step);
+template SimpleTensor<uint16_t> range(SimpleTensor<uint16_t> &dst, float start, const size_t num_of_elements, float step);
+template SimpleTensor<int16_t> range(SimpleTensor<int16_t> &dst, float start, const size_t num_of_elements, float step);
+
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/reference/Range.h b/tests/validation/reference/Range.h
new file mode 100644
index 0000000000..d31166bb86
--- /dev/null
+++ b/tests/validation/reference/Range.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_TEST_RANGE_H__
+#define __ARM_COMPUTE_TEST_RANGE_H__
+
+#include "tests/SimpleTensor.h"
+#include "tests/validation/Helpers.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+template <typename T>
+SimpleTensor<T> range(SimpleTensor<T> &dst, float start, size_t num_elements, float step);
+
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_TEST_RANGE_H__ */
diff --git a/utils/Utils.h b/utils/Utils.h
index 8cac857178..031a3726a1 100644
--- a/utils/Utils.h
+++ b/utils/Utils.h
@@ -194,6 +194,55 @@ inline std::string get_typestring(DataType data_type)
     }
 }
 
+/** Returns true if the value can be represented by the given data type
+ *
+ * @param[in] val        value to be checked
+ * @param[in] dt         data type that is checked
+ * @param[in] quant_info quantization info if the data type is QASYMM8
+ *
+ * @return true if the data type can hold the value.
+ */
+template <typename T>
+bool check_value_range(T val, DataType dt, QuantizationInfo quant_info = QuantizationInfo())
+{
+    switch(dt)
+    {
+        case DataType::U8:
+            return ((static_cast<uint8_t>(val) == val) && val >= std::numeric_limits<uint8_t>::lowest() && val <= std::numeric_limits<uint8_t>::max());
+        case DataType::QASYMM8:
+        {
+            double min = static_cast<double>(quant_info.dequantize(0));
+            double max = static_cast<double>(quant_info.dequantize(std::numeric_limits<uint8_t>::max()));
+            return ((double)val >= min && (double)val <= max);
+        }
+        case DataType::S8:
+            return ((static_cast<int8_t>(val) == val) && val >= std::numeric_limits<int8_t>::lowest() && val <= std::numeric_limits<int8_t>::max());
+        case DataType::U16:
+            return ((static_cast<uint16_t>(val) == val) && val >= std::numeric_limits<uint16_t>::lowest() && val <= std::numeric_limits<uint16_t>::max());
+        case DataType::S16:
+            return ((static_cast<int16_t>(val) == val) && val >= std::numeric_limits<int16_t>::lowest() && val <= std::numeric_limits<int16_t>::max());
+        case DataType::U32:
+            return ((static_cast<uint32_t>(val) == val) && val >= std::numeric_limits<uint32_t>::lowest() && val <= std::numeric_limits<uint32_t>::max());
+        case DataType::S32:
+            return ((static_cast<int32_t>(val) == val) && val >= std::numeric_limits<int32_t>::lowest() && val <= std::numeric_limits<int32_t>::max());
+        case DataType::U64:
+            return (val >= std::numeric_limits<uint64_t>::lowest() && val <= std::numeric_limits<uint64_t>::max());
+        case DataType::S64:
+            return (val >= std::numeric_limits<int64_t>::lowest() && val <= std::numeric_limits<int64_t>::max());
+        case DataType::F16:
+            return (val >= std::numeric_limits<half>::lowest() && val <= std::numeric_limits<half>::max());
+        case DataType::F32:
+            return (val >= std::numeric_limits<float>::lowest() && val <= std::numeric_limits<float>::max());
+        case DataType::F64:
+            return (val >= std::numeric_limits<double>::lowest() && val <= std::numeric_limits<double>::max());
+        case DataType::SIZET:
+            return ((static_cast<size_t>(val) == val) && val >= std::numeric_limits<size_t>::lowest() && val <= std::numeric_limits<size_t>::max());
+        default:
+            ARM_COMPUTE_ERROR("Data type not supported");
+            return false;
+    }
+}
+
 /** Maps a tensor if needed
  *
  * @param[in] tensor   Tensor to be mapped