Make CL Elementwise Unary kernels and functions state-less
Resolves COMPMID-4004
Change-Id: I1dfe8bc52c1ff394ea208ba98b51033c738746a4
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4922
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
diff --git a/src/core/CL/CLKernels.h b/src/core/CL/CLKernels.h
index 916cf45..d31c876 100644
--- a/src/core/CL/CLKernels.h
+++ b/src/core/CL/CLKernels.h
@@ -56,7 +56,6 @@
#include "src/core/CL/kernels/CLDerivativeKernel.h"
#include "src/core/CL/kernels/CLDilateKernel.h"
#include "src/core/CL/kernels/CLDirectConvolutionLayerKernel.h"
-#include "src/core/CL/kernels/CLElementWiseUnaryLayerKernel.h"
#include "src/core/CL/kernels/CLErodeKernel.h"
#include "src/core/CL/kernels/CLFFTDigitReverseKernel.h"
#include "src/core/CL/kernels/CLFFTRadixStageKernel.h"
diff --git a/src/core/CL/kernels/CLElementWiseUnaryLayerKernel.h b/src/core/CL/kernels/CLElementWiseUnaryLayerKernel.h
deleted file mode 100644
index 95b5872..0000000
--- a/src/core/CL/kernels/CLElementWiseUnaryLayerKernel.h
+++ /dev/null
@@ -1,66 +0,0 @@
-/*
- * Copyright (c) 2018-2020 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_CLELEMENTWISEUNARYLAYERKERNEL_H
-#define ARM_COMPUTE_CLELEMENTWISEUNARYLAYERKERNEL_H
-
-#include "arm_compute/core/Types.h"
-#include "src/core/CL/ICLKernel.h"
-#include "src/core/CL/ICLSimpleKernel.h"
-
-namespace arm_compute
-{
-/** Interface for the elementwise unary operator */
-class CLElementWiseUnaryLayerKernel : public ICLKernel
-{
-public:
- /** Initialise the kernel's inputs, output.
- *
- * @param[in] input First tensor input info. Data types supported: F16/F32.
- * @param[out] output Output tensor info. Data types supported: Same as @p input.
- * @param[in] op Element wise unary operation to perform.
- */
- void configure(const ITensorInfo *input, ITensorInfo *output, const ElementWiseUnary &op);
- /** Initialise the kernel's inputs, output.
- *
- * @param[in] compile_context The compile context to be used.
- * @param[in] input First tensor input info. Data types supported: F16/F32.
- * @param[out] output Output tensor info. Data types supported: Same as @p input.
- * @param[in] op Element wise unary operation to perform.
- */
- void configure(const CLCompileContext &compile_context, const ITensorInfo *input, ITensorInfo *output, const ElementWiseUnary &op);
- /** Static function to check if given info will lead to a valid configuration of @ref CLElementWiseUnaryLayerKernel
- *
- * @param[in] input First tensor input info. Data types supported: F16/F32.
- * @param[in] output Output tensor info. Data types supported: Same as @p input.
- * @param[in] op Element wise unary operation to perform.
- *
- * @return a status
- */
- static Status validate(const ITensorInfo *input, const ITensorInfo *output, const ElementWiseUnary &op);
-
- // Inherited methods overridden:
- void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
-};
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_CLELEMENTWISEUNARYLAYERKERNEL_H */
diff --git a/src/core/CL/kernels/CLElementWiseUnaryLayerKernel.cpp b/src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.cpp
similarity index 66%
rename from src/core/CL/kernels/CLElementWiseUnaryLayerKernel.cpp
rename to src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.cpp
index e7c95ba..8e9698c 100644
--- a/src/core/CL/kernels/CLElementWiseUnaryLayerKernel.cpp
+++ b/src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -21,7 +21,7 @@
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
-#include "src/core/CL/kernels/CLElementWiseUnaryLayerKernel.h"
+#include "src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.h"
#include "arm_compute/core/CL/CLHelpers.h"
#include "arm_compute/core/CL/ICLTensor.h"
@@ -32,55 +32,54 @@
namespace arm_compute
{
+namespace opencl
+{
+namespace kernels
+{
namespace
{
-Status validate_arguments(const ITensorInfo &input, const ITensorInfo &output, const ElementWiseUnary op)
+Status validate_arguments(const ITensorInfo &src, const ITensorInfo &dst, const ElementWiseUnary op)
{
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&input);
+ ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&src);
if(op == ElementWiseUnary::LOGICAL_NOT)
{
- ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input, 1, DataType::U8);
+ ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&src, 1, DataType::U8);
}
else
{
- ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input, 1, DataType::F16, DataType::F32);
+ ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&src, 1, DataType::F16, DataType::F32);
}
- // Validate in case of configured output
- if(output.total_size() > 0)
+ // Validate in case of configured dst
+ if(dst.total_size() > 0)
{
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&output);
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(&input, &output);
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(&input, &output);
+ ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&dst);
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(&src, &dst);
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(&src, &dst);
}
return Status{};
}
} // namespace
-void CLElementWiseUnaryLayerKernel::configure(const ITensorInfo *input, ITensorInfo *output, const ElementWiseUnary &op)
+void ClElementWiseUnaryKernel::configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst, const ElementWiseUnary &op)
{
- configure(CLKernelLibrary::get().get_compile_context(), input, output, op);
-}
+ ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
-void CLElementWiseUnaryLayerKernel::configure(const CLCompileContext &compile_context, const ITensorInfo *input, ITensorInfo *output, const ElementWiseUnary &op)
-{
- ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+ auto padding_info = get_padding_info({ src, dst });
- auto padding_info = get_padding_info({ input, output });
-
- ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*input, *output, op));
+ ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*src, *dst, op));
const std::string kernel_name = "elementwise_unary";
- const int vec_size_x = 16 / output->element_size();
- const int output_width_x = output->tensor_shape().x();
- const bool multi_access_x = (output_width_x / vec_size_x > 0);
+ const int vec_size_x = 16 / dst->element_size();
+ const int dst_width_x = dst->tensor_shape().x();
+ const bool multi_access_x = (dst_width_x / vec_size_x > 0);
// Set kernel build options
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->data_type()));
build_opts.add_option_if(multi_access_x, "-DVEC_SIZE=" + support::cpp11::to_string(vec_size_x));
- build_opts.add_option_if(multi_access_x, "-DLAST_ACCESSED_X=" + support::cpp11::to_string(std::max<int>(output_width_x - vec_size_x, 0)));
+ build_opts.add_option_if(multi_access_x, "-DLAST_ACCESSED_X=" + support::cpp11::to_string(std::max<int>(dst_width_x - vec_size_x, 0)));
switch(op)
{
case ElementWiseUnary::RSQRT:
@@ -115,7 +114,7 @@
_kernel = create_kernel(compile_context, kernel_name, build_opts.options());
// Configure kernel window
- Window win = calculate_max_window(*output);
+ Window win = calculate_max_window(*dst);
if(multi_access_x)
{
win.set(Window::DimX,
@@ -126,16 +125,16 @@
ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
}
-Status CLElementWiseUnaryLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const ElementWiseUnary &op)
+Status ClElementWiseUnaryKernel::validate(const ITensorInfo *src, const ITensorInfo *dst, const ElementWiseUnary &op)
{
ARM_COMPUTE_UNUSED(op);
- ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
- ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(*input, *output, op));
+ ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, dst);
+ ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(*src, *dst, op));
return Status{};
}
-void CLElementWiseUnaryLayerKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
+void ClElementWiseUnaryKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
{
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
@@ -155,4 +154,6 @@
}
while(collapsed.slide_window_slice_3D(slice));
}
+} // namespace kernels
+} // namespace opencl
} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.h b/src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.h
new file mode 100644
index 0000000..7e5edef
--- /dev/null
+++ b/src/core/gpu/cl/kernels/ClElementwiseUnaryKernel.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2018-2021 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_UNARY_KERNEL_H
+#define ARM_COMPUTE_CL_ELEMENTWISE_UNARY_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/core/gpu/cl/ClCompileContext.h"
+#include "src/core/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** Interface for the elementwise unary operator */
+class ClElementWiseUnaryKernel : public IClKernel
+{
+public:
+ ClElementWiseUnaryKernel() = default;
+ ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClElementWiseUnaryKernel);
+ /** Initialise the kernel's srcs, dst.
+ *
+ * @param[in] compile_context The compile context to be used.
+ * @param[in] src First source tensor info. Data types supported: F16/F32.
+ * @param[out] dst Destination tensor info. Data types supported: same as @p src.
+ * @param[in] op Element wise unary operation to perform.
+ */
+ void configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst, const ElementWiseUnary &op);
+ /** Static function to check if given info will lead to a valid configuration of @ref ClElementWiseUnaryKernel
+ *
+ * @param[in] src First source tensor info. Data types supported: F16/F32.
+ * @param[in] dst Destination tensor info. Data types supported: same as @p src.
+ * @param[in] op Element wise unary operation to perform.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const ElementWiseUnary &op);
+
+ // Inherited methods overridden:
+ void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_CL_ELEMENTWISE_UNARY_KERNEL_H */