Add dynamic tensor support to CpuElementwise
The kernels and operators for binary and unary operations
are now capable of being configured with dynamic shapes and
computing windows at run-time.
Additionally, changing arguments' names is done
for consistency.
Partially Implements: COMPMID-4127
Change-Id: I48e5038692db667dec7cb2b2906fe5683214fe19
Signed-off-by: Sang-Hoon Park <sang-hoon.park@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4973
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/runtime/NEON/INEOperator.cpp b/src/runtime/NEON/INEOperator.cpp
index ccee8ff..a5fc0a2 100644
--- a/src/runtime/NEON/INEOperator.cpp
+++ b/src/runtime/NEON/INEOperator.cpp
@@ -44,7 +44,12 @@
ARM_COMPUTE_ERROR("No inputs provided");
}
- NEScheduler::get().schedule_op(_kernel.get(), Window::DimY, _kernel->window(), tensors);
+ run(tensors, _kernel->window());
+}
+
+void INEOperator::run(ITensorPack &tensors, const Window &window)
+{
+ NEScheduler::get().schedule_op(_kernel.get(), Window::DimY, window, tensors);
}
void INEOperator::prepare(ITensorPack &constants)
diff --git a/src/runtime/cpu/operators/CpuElementwise.cpp b/src/runtime/cpu/operators/CpuElementwise.cpp
index 322bd09..b5c8dde 100644
--- a/src/runtime/cpu/operators/CpuElementwise.cpp
+++ b/src/runtime/cpu/operators/CpuElementwise.cpp
@@ -23,95 +23,111 @@
*/
#include "src/runtime/cpu/operators/CpuElementwise.h"
#include "src/core/cpu/kernels/CpuElementwiseKernel.h"
+#include "src/core/helpers/WindowHelpers.h"
namespace arm_compute
{
namespace cpu
{
-void CpuElementwiseMax::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwiseBase::run(ITensorPack &tensors)
+{
+ // If the kernel has been configured, use the window from the kernel.
+ if(_kernel->is_window_configured())
+ {
+ ICpuOperator::run(tensors);
+ return;
+ }
+
+ auto src0_info = tensors.get_const_tensor(TensorType::ACL_SRC_0)->info();
+ auto src1_info = tensors.get_const_tensor(TensorType::ACL_SRC_1)->info();
+ auto shape_and_window = compute_output_shape_and_window(*src0_info, *src1_info);
+ ICpuOperator::run(tensors, shape_and_window.second);
+}
+
+void CpuElementwiseMax::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuArithmeticKernel>();
- k->configure(ArithmeticOperation::MAX, input1, input2, output);
+ k->configure(ArithmeticOperation::MAX, src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwiseMax::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwiseMax::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::MAX, input1, input2, output);
+ return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::MAX, src0, src1, dst);
}
-void CpuElementwiseMin::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwiseMin::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuArithmeticKernel>();
- k->configure(ArithmeticOperation::MIN, input1, input2, output);
+ k->configure(ArithmeticOperation::MIN, src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwiseMin::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwiseMin::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::MIN, input1, input2, output);
+ return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::MIN, src0, src1, dst);
}
-void CpuElementwiseSquaredDiff::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwiseSquaredDiff::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuArithmeticKernel>();
- k->configure(ArithmeticOperation::SQUARED_DIFF, input1, input2, output);
+ k->configure(ArithmeticOperation::SQUARED_DIFF, src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwiseSquaredDiff::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwiseSquaredDiff::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::SQUARED_DIFF, input1, input2, output);
+ return kernels::CpuArithmeticKernel::validate(ArithmeticOperation::SQUARED_DIFF, src0, src1, dst);
}
-void CpuElementwiseDivision::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwiseDivision::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuDivisionKernel>();
- k->configure(input1, input2, output);
+ k->configure(src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwiseDivision::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwiseDivision::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuDivisionKernel::validate(input1, input2, output);
+ return kernels::CpuDivisionKernel::validate(src0, src1, dst);
}
-void CpuElementwisePower::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwisePower::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuPowerKernel>();
- k->configure(input1, input2, output);
+ k->configure(src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwisePower::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwisePower::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuPowerKernel::validate(input1, input2, output);
+ return kernels::CpuPowerKernel::validate(src0, src1, dst);
}
template <ComparisonOperation COP>
-void CpuElementwiseComparisonStatic<COP>::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output)
+void CpuElementwiseComparisonStatic<COP>::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst)
{
auto k = std::make_unique<kernels::CpuComparisonKernel>();
- k->configure(COP, input1, input2, output);
+ k->configure(COP, src0, src1, dst);
_kernel = std::move(k);
}
template <ComparisonOperation COP>
-Status CpuElementwiseComparisonStatic<COP>::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+Status CpuElementwiseComparisonStatic<COP>::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst)
{
- return kernels::CpuComparisonKernel::validate(COP, input1, input2, output);
+ return kernels::CpuComparisonKernel::validate(COP, src0, src1, dst);
}
-void CpuElementwiseComparison::configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output, ComparisonOperation op)
+void CpuElementwiseComparison::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst, ComparisonOperation op)
{
auto k = std::make_unique<kernels::CpuComparisonKernel>();
- k->configure(op, input1, input2, output);
+ k->configure(op, src0, src1, dst);
_kernel = std::move(k);
}
-Status CpuElementwiseComparison::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ComparisonOperation op)
+Status CpuElementwiseComparison::validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst, ComparisonOperation op)
{
- return kernels::CpuComparisonKernel::validate(op, input1, input2, output);
+ return kernels::CpuComparisonKernel::validate(op, src0, src1, dst);
}
// Supported Specializations
diff --git a/src/runtime/cpu/operators/CpuElementwise.h b/src/runtime/cpu/operators/CpuElementwise.h
index 611a374..4b350d5 100644
--- a/src/runtime/cpu/operators/CpuElementwise.h
+++ b/src/runtime/cpu/operators/CpuElementwise.h
@@ -30,30 +30,36 @@
{
namespace cpu
{
+class CpuElementwiseBase : public ICpuOperator
+{
+public:
+ // Inherited methods overridden:
+ void run(ITensorPack &tensors) override;
+};
/** Basic function to run @ref cpu::kernels::CpuArithmeticKernel for max
*
* @note The tensor data type for the inputs must be QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
* @note The function performs a max operation between two tensors.
*/
-class CpuElementwiseMax : public ICpuOperator
+class CpuElementwiseMax : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in, out] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: Same as @p src0.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuArithmeticKernel for max
*
- * @param[in] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: Same as @p src0.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run @ref cpu::kernels::CpuArithmeticKernel for min
@@ -61,25 +67,25 @@
* @note The tensor data type for the inputs must be QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
* @note The function performs a min operation between two tensors.
*/
-class CpuElementwiseMin : public ICpuOperator
+class CpuElementwiseMin : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in, out] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: Same as @p src0.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuArithmeticKernel for min
*
- * @param[in] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: Same as @p src0.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run @ref cpu::kernels::CpuArithmeticKernel for squared difference
@@ -87,25 +93,25 @@
* @note The tensor data type for the inputs must be QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
* @note The function performs a squared different operation between two tensors (i.e., out[i] = (in1[i] - in2[i])^2
*/
-class CpuElementwiseSquaredDiff : public ICpuOperator
+class CpuElementwiseSquaredDiff : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in, out] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: Same as @p src0.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuArithmeticKernel for squared difference
*
- * @param[in] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: Same as @p src0.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run @ref cpu::kernels::CpuArithmeticKernel for division
@@ -113,25 +119,25 @@
* @note The tensor data type for the inputs must be S32/F16/F32.
* @note The function performs a division operation between two tensors (i.e., out[i] = in1[i] / in2[i])
*/
-class CpuElementwiseDivision : public ICpuOperator
+class CpuElementwiseDivision : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: S32/F16/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in, out] src0 First tensor input info. Data types supported: S32/F16/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: Same as @p src0.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuArithmeticKernel for division
*
- * @param[in] input1 First tensor input info. Data types supported: S32/F16/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in] src0 First tensor input info. Data types supported: S32/F16/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: Same as @p src0.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run @ref cpu::kernels::CpuArithmeticKernel for power
@@ -140,25 +146,25 @@
* @note The function performs a elementwise power of in1 to in2 (i.e., out[i] = in1[i] ^ in2[i])
* @note For an exponent that is a float, this function will only work with a positive base.
*/
-class CpuElementwisePower : public ICpuOperator
+class CpuElementwisePower : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: F16/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in, out] src0 First tensor input info. Data types supported: F16/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: Same as @p src0.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuArithmeticKernel for power
*
- * @param[in] input1 First tensor input info. Data types supported: F16/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ * @param[in] src0 First tensor input info. Data types supported: F16/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: Same as @p src0.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run @ref cpu::kernels::CpuComparisonKernel.
@@ -166,27 +172,27 @@
* @note The tensor data type for the inputs must be QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
* @note The function performs a comparison operation between two tensors.
*/
-class CpuElementwiseComparison : public ICpuOperator
+class CpuElementwiseComparison : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: U16/U32.
- * @param[in] op Comparison Operation to be performed.
+ * @param[in, out] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: U16/U32.
+ * @param[in] op Comparison Operation to be performed.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output, ComparisonOperation op);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst, ComparisonOperation op);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuComparisonKernel
*
- * @param[in] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: U16/U32.
- * @param[in] op Comparison Operation to be performed.
+ * @param[in] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: U16/U32.
+ * @param[in] op Comparison Operation to be performed.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ComparisonOperation op);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst, ComparisonOperation op);
};
/** Basic function to run @ref cpu::kernels::CpuComparisonKernel
@@ -195,25 +201,25 @@
* @note The function performs a comparison operation between two tensors.
*/
template <ComparisonOperation op>
-class CpuElementwiseComparisonStatic : public ICpuOperator
+class CpuElementwiseComparisonStatic : public CpuElementwiseBase
{
public:
- /** Initialise the kernel's inputs, output and conversion policy.
+ /** Initialise the kernel's inputs, dst and conversion policy.
*
- * @param[in, out] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in, out] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[out] output Output tensor info. Data types supported: U16/U32.
+ * @param[in, out] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in, out] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[out] dst Output tensor info. Data types supported: U16/U32.
*/
- void configure(const ITensorInfo *input1, const ITensorInfo *input2, ITensorInfo *output);
+ void configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst);
/** Static function to check if given info will lead to a valid configuration of @ref cpu::kernels::CpuComparisonKernel
*
- * @param[in] input1 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
- * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
- * @param[in] output Output tensor info. Data types supported: U16/U32.
+ * @param[in] src0 First tensor input info. Data types supported: QASYMM8/QASYMM8_SIGNED/S16/F16/S32/F32.
+ * @param[in] src1 Second tensor input info. Data types supported: Same as @p src0.
+ * @param[in] dst Output tensor info. Data types supported: U16/U32.
*
* @return a status
*/
- static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+ static Status validate(const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *dst);
};
/** Basic function to run equal comparison. */
diff --git a/src/runtime/cpu/operators/CpuElementwiseUnary.cpp b/src/runtime/cpu/operators/CpuElementwiseUnary.cpp
index d1b1700..2140c5c 100644
--- a/src/runtime/cpu/operators/CpuElementwiseUnary.cpp
+++ b/src/runtime/cpu/operators/CpuElementwiseUnary.cpp
@@ -23,6 +23,7 @@
*/
#include "src/runtime/cpu/operators/CpuElementwiseUnary.h"
#include "src/core/cpu/kernels/CpuElementwiseUnaryKernel.h"
+#include "src/core/helpers/WindowHelpers.h"
namespace arm_compute
{
@@ -41,5 +42,17 @@
{
return KernelType::validate(op, src, dst);
}
+
+void CpuElementwiseUnary::run(ITensorPack &tensors)
+{
+ if(_kernel->is_window_configured())
+ {
+ ICpuOperator::run(tensors);
+ return;
+ }
+
+ auto src_info = tensors.get_const_tensor(TensorType::ACL_SRC)->info();
+ ICpuOperator::run(tensors, compute_output_shape_and_window(*src_info).second);
+}
} // namespace cpu
} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/cpu/operators/CpuElementwiseUnary.h b/src/runtime/cpu/operators/CpuElementwiseUnary.h
index 0b2a9e7..721ba2a 100644
--- a/src/runtime/cpu/operators/CpuElementwiseUnary.h
+++ b/src/runtime/cpu/operators/CpuElementwiseUnary.h
@@ -50,6 +50,9 @@
* @return a status
*/
static Status validate(ElementWiseUnary op, const ITensorInfo &src, const ITensorInfo &dst);
+
+ // Inherited methods overridden:
+ void run(ITensorPack &tensors) override;
};
} // namespace cpu