Make Gpu Elementwise Operations kernels and functions state-less

Partially resolves COMPMID-4004

Change-Id: Ie99b6c676af5be4c2a67affc06fec378f3dff8ec
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4918
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 282 insertions, 0 deletions

diff --git a/src/runtime/gpu/cl/operators/ClElementwiseOperations.cpp b/src/runtime/gpu/cl/operators/ClElementwiseOperations.cpp
new file mode 100644
index 0000000000..e5b836a0d8
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClElementwiseOperations.cpp
@@ -0,0 +1,92 @@
+/*
+ * Copyright (c) 2021 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 "src/runtime/gpu/cl/operators/ClElementwiseOperations.h"
+
+#include "src/core/gpu/cl/kernels/ClElementwiseKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+void ClElementwiseDivision::configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClArithmeticKernel>();
+    k->configure(compile_context, ArithmeticOperation::DIV, src1, src2, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClElementwiseDivision::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClArithmeticKernel::validate(ArithmeticOperation::DIV, src1, src2, dst, act_info);
+}
+
+void ClElementwiseMax::configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClArithmeticKernel>();
+    k->configure(compile_context, ArithmeticOperation::MAX, src1, src2, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClElementwiseMax::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClArithmeticKernel::validate(ArithmeticOperation::MAX, src1, src2, dst, act_info);
+}
+
+void ClElementwiseMin::configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClArithmeticKernel>();
+    k->configure(compile_context, ArithmeticOperation::MIN, src1, src2, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClElementwiseMin::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClArithmeticKernel::validate(ArithmeticOperation::MIN, src1, src2, dst, act_info);
+}
+
+void ClElementwiseSquaredDiff::configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClArithmeticKernel>();
+    k->configure(compile_context, ArithmeticOperation::SQUARED_DIFF, src1, src2, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClElementwiseSquaredDiff::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClArithmeticKernel::validate(ArithmeticOperation::SQUARED_DIFF, src1, src2, dst, act_info);
+}
+
+void ClElementwisePower::configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    auto k = std::make_unique<kernels::ClArithmeticKernel>();
+    k->configure(compile_context, ArithmeticOperation::POWER, src1, src2, dst, act_info);
+    _kernel = std::move(k);
+}
+
+Status ClElementwisePower::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info)
+{
+    return kernels::ClArithmeticKernel::validate(ArithmeticOperation::POWER, src1, src2, dst, act_info);
+}
+} // namespace opencl
+} // namespace arm_compute
diff --git a/src/runtime/gpu/cl/operators/ClElementwiseOperations.h b/src/runtime/gpu/cl/operators/ClElementwiseOperations.h
new file mode 100644
index 0000000000..b9ab1405c8
--- /dev/null
+++ b/src/runtime/gpu/cl/operators/ClElementwiseOperations.h
@@ -0,0 +1,190 @@
+/*
+ * Copyright (c) 2021 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_CL_ELEMENTWISE_OPERATIONS_H
+#define ARM_COMPUTE_CL_ELEMENTWISE_OPERATIONS_H
+
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/runtime/gpu/cl/IClOperator.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+/** Basic function to run @ref opencl::kernels::ClArithmeticKernel for division
+ *
+ * @note The tensor data type for the inputs must be F16/F32.
+ * @note The function performs an arithmetic division between two tensors.
+ */
+class ClElementwiseDivision : public IClOperator
+{
+public:
+    /** Default Constructor */
+    ClElementwiseDivision() = default;
+    /** Configure function for a given list of arguments.
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor info. Data types supported: F16/F32.
+     * @param[in]  src2            Second source tensor info. same as @p src1.
+     * @param[out] dst             Destination tensor info. Data types supported: same as @p src1.
+     * @param[in]  act_info        (Optional) Activation layer information in case of a fused activation.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref ClElementwiseDivision
+     *
+     * @param[in] src1     First source tensor info. Data types supported: F16/F32.
+     * @param[in] src2     Second source tensor info. Data types supported: same as @p src1.
+     * @param[in] dst      Destination tensor info. Data types supported: same as @p src1.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+};
+
+/** Basic function to run @ref opencl::kernels::ClArithmeticKernel for max
+ *
+ * @note The tensor data type for the inputs must be U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32.
+ * @note The function performs a max operation between two tensors.
+ */
+class ClElementwiseMax : public IClOperator
+{
+public:
+    /** Default Constructor */
+    ClElementwiseMax() = default;
+    /** Configure function for a given list of arguments.
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32.
+     * @param[in]  src2            Second source tensor info. Data types supported: same as @p src1.
+     * @param[out] dst             Destination tensor info. Data types supported: same as @p src1.
+     * @param[in]  act_info        (Optional) Activation layer information in case of a fused activation.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref opencl::kernels::ClArithmeticKernel for max
+     *
+     * @param[in] src1     First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32.
+     * @param[in] src2     Second source tensor info. Data types supported: same as @p src1.
+     * @param[in] dst      Destination tensor info. Data types supported: same as @p src1.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+};
+
+/** Basic function to run @ref opencl::kernels::ClArithmeticKernel for min
+ *
+ * @note The tensor data type for the inputs must be U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32.
+ * @note The function performs a max operation between two tensors.
+ */
+class ClElementwiseMin : public IClOperator
+{
+public:
+    /** Default Constructor */
+    ClElementwiseMin() = default;
+    /** Configure function for a given list of arguments.
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32.
+     * @param[in]  src2            Second source tensor info. Data types supported: same as @p src1.
+     * @param[out] dst             Destination tensor info. Data types supported: same as @p src1.
+     * @param[in]  act_info        (Optional) Activation layer information in case of a fused activation.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref opencl::kernels::ClArithmeticKernel for min
+     *
+     * @param[in] src1     First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32.
+     * @param[in] src2     Second source tensor info. Data types supported: same as @p src1.
+     * @param[in] dst      Destination tensor info. Data types supported: same as @p src1.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+};
+
+/** Basic function to run @ref opencl::kernels::ClArithmeticKernel for squared difference
+ *
+ * @note The tensor data type for the inputs must be QASYMM8/U8/S16/QSYMM16/F16/F32.
+ * @note The function performs a squared different operation between two tensors (i.e., out[i] = (in1[i] - in2[i])^2
+ */
+class ClElementwiseSquaredDiff : public IClOperator
+{
+public:
+    /** Default Constructor */
+    ClElementwiseSquaredDiff() = default;
+    /** Configure function for a given list of arguments.
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32.
+     * @param[in]  src2            Second source tensor info. Data types supported: same as @p src1.
+     * @param[out] dst             Destination tensor info. Data types supported: same as @p src1.
+     * @param[in]  act_info        (Optional) Activation layer information in case of a fused activation.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref opencl::kernels::ClArithmeticKernel for squared difference
+     *
+     * @param[in] src1     First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32.
+     * @param[in] src2     Second source tensor info. Data types supported: same as @p src1.
+     * @param[in] dst      Destination tensor info. Data types supported: same as @p src1.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+};
+
+/** Basic function to run @ref opencl::kernels::ClArithmeticKernel for power
+ *
+ * @note The tensor data type for the inputs must be F16/F32.
+ * @note The function performs an elementwise power of in1 to in2 (i.e., out[i] = in1[i] ^ in2[i])
+ */
+class ClElementwisePower : public IClOperator
+{
+public:
+    /** Default Constructor */
+    ClElementwisePower() = default;
+    /** Configure function for a given list of arguments.
+     *
+     * @param[in]  compile_context The compile context to be used.
+     * @param[in]  src1            First source tensor info. Data types supported: F16/F32.
+     * @param[in]  src2            Second source tensor info. Data types supported: F16/F32.
+     * @param[out] dst             Destination tensor info. Data types supported:F16/F32.
+     * @param[in]  act_info        (Optional) Activation layer information in case of a fused activation.
+     */
+    void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref opencl::kernels::ClArithmeticKernel for power
+     *
+     * @param[in] src1     First source tensor info. Data types supported: F16/F32.
+     * @param[in] src2     Second source tensor info. Data types supported: F16/F32.
+     * @param[in] dst      Destination tensor info. Data types supported: F16/F32.
+     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());
+};
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_ELEMENTWISE_OPERATIONS_H */