~CLPadLayerKernel() = default;
/** Set the input and output tensor.
*
- * @param[in] input Source tensor. Data types supported: All.
+ * @param[in] input Source tensor. Data types supported: U8, S8, QASYMM8, QASYMM8_SIGNED, U16, S16, U32, S32, F16, F32.
* @param[out] output Output tensor. Data type supported: same as @p input
* @param[in] padding The padding for each spatial dimension of the input tensor. The pair padding[i]
* specifies the front and the end padding in the i-th dimension.
PaddingMode mode = PaddingMode::CONSTANT);
/** Static function to check if given info will lead to a valid configuration of @ref CLPadLayerKernel
*
- * @param[in] input Source tensor info. Data types supported: All.
+ * @param[in] input Source tensor info. Data types supported: U8, S8, QASYMM8, QASYMM8_SIGNED, U16, S16, U32, S32, F16, F32.
* @param[in] output Output tensor info. Data type supported: same as @p input
* @param[in] padding The padding for each spatial dimension of the input tensor. The pair padding[i]
* specifies the front and the end padding in the i-th dimension.
/*
- * Copyright (c) 2018 ARM Limited.
+ * 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 withoutput restriction, including withoutput limitation the
+ * 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:
* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
* @note The input tensor batch size must be passed at compile time using -DBATCH_SIZE. e.g. -DBATCH_SIZE=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
* @note The input tensor batch size must be passed at compile time using -DBATCH_SIZE. e.g. -DBATCH_SIZE=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The block shape x must be passed at compile time using -DBLOCK_SHAPE_X. e.g. -DBLOCK_SHAPE_X=2
* @note The block shape y must be passed at compile time using -DBLOCK_SHAPE_Y. e.g. -DBLOCK_SHAPE_Y=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The block shape x must be passed at compile time using -DBLOCK_SHAPE_X. e.g. -DBLOCK_SHAPE_X=2
* @note The block shape y must be passed at compile time using -DBLOCK_SHAPE_Y. e.g. -DBLOCK_SHAPE_Y=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
/*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @note The element size must be passed at compile time using -DELEMENT_SIZE: e.g. -DELEMENT_SIZE=4
* @note The number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/QASYMM8_SIGNED/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @note The width, height and depth of the input tensor must be passed at compile time using -DSRC_WIDTH, -DSRC_HEIGHT and -DSRC_DEPTH. e.g. -DSRC_WIDTH=24, -DSRC_HEIGHT=24, -DSRC_DEPTH=16
* @note If the output has 3 dimensions, the 2nd dimension of the output tensor must be passed at compile time using -DDST_DIM1. e.g -DDST_DIM1=3
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: All
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
#if defined(DST_DIM1)
uint b_tmp = b0;
- b0 = b_tmp % DST_DIM1; // batch id0
- b1 = b_tmp / DST_DIM1; // batch id1
-#endif // defined(DST_DIM1)
+ b0 = b_tmp % DST_DIM1; // batch id0
+ b1 = b_tmp / DST_DIM1; // batch id1
+#endif // defined(DST_DIM1)
- __global uchar *output_ptr = dst_ptr + dst_offset_first_element_in_bytes +
- (get_global_id(0) + get_global_id(1) * (uint)SRC_WIDTH + c * (uint)(SRC_WIDTH * SRC_HEIGHT)) * sizeof(DATA_TYPE) +
- b0 * dst_stride_y +
- b1 * dst_stride_z;
+ __global uchar *output_ptr = dst_ptr + dst_offset_first_element_in_bytes + (get_global_id(0) + get_global_id(1) * (uint)SRC_WIDTH + c * (uint)(SRC_WIDTH * SRC_HEIGHT)) * sizeof(
+ DATA_TYPE) + b0 * dst_stride_y + b1 * dst_stride_z;
*((__global DATA_TYPE *)output_ptr) = *((__global DATA_TYPE *)src.ptr);
}
/*
- * Copyright (c) 2018-2019 ARM Limited.
+ * Copyright (c) 2018-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @attention Input tensor depth should be given as a preprocessor argument using -DINPUT_DIM_Z=size. e.g. -DINPUT_DIM_Z=16
*
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/U16/S16/U32/S32/F16/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per work item (in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
* @note In case grouping is performed, the number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
* @note In case grouping is performed, the number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The stride along the X and Y directions must be passed at compile time using -DSTRIDE_X and -DSTRIDE_Y: e.g. -DSTRIDE_X=1 and -DSTRIDE_Y=1
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
* @note In case grouping is performed, the number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
* @note In case grouping is performed, the number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
* @note In case grouping is performed, the number of groups must be passed at compile time using -DNUM_GROUPS: e.g. -DNUM_GROUPS=4
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The stride along the Y direction must be passed at compile time using -DSTRIDE_Y: e.g. -DSTRIDE_Y=1
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The stride along the Y direction must be passed at compile time using -DSTRIDE_Y: e.g. -DSTRIDE_Y=1
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The dilation_x and dilation_y must be passed at compile time using -DDILATION_X and -DDILATION_Y: e.g. -DDILATION_X=1, -DDILATION_Y=1
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8/F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QASYMM8_SIGNED/QASYMM8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
/*
- * Copyright (c) 2019 ARM Limited.
+ * Copyright (c) 2019-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* -# -DPAD_W_BEFORE: Pad to add before the first batch of the input tensor (e.g. -DPAD_W_BEFORE=3)
* -# -DSRC_BATCH: Input tensor's batch size (e.g. -DSRC_BATCH=4)
*
- * @param[in] src_ptr Pointer to the source image. Supported data types: U8, S8, QASYMM8, U16, S16, U32, S32, F16, F32
+ * @param[in] src_ptr Pointer to the source image. Supported data types: U8, S8, QASYMM8, QASYMM8_SIGNED, U16, S16, U32, S32, F16, F32
* @param[in] src_stride_x Stride of the source image in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source image in Y dimension (in bytes)
* @note If the starting point to read backward from is less than the output's last element accessed in the X, the following compile flags must be passed at compile time to avoid negative offsets:
* -# -DAFTER_PAD_REM: Defines how much to rotate the vector if the backward calculation attempted to read from a negative offset (e.g. -DAFTER_PAD_REM=3)
*
- * @param[in] src_ptr Pointer to the source image. Supported data types: U8, S8, QASYMM8, U16, S16, U32, S32, F16, F32
+ * @param[in] src_ptr Pointer to the source image. Supported data types: U8, S8, QASYMM8, QASYMM8_SIGNED, U16, S16, U32, S32, F16, F32
* @param[in] src_stride_x Stride of the source image in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source image in Y dimension (in bytes)
/*
- * Copyright (c) 2018 ARM Limited.
+ * 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 withoutput restriction, including withoutput limitation the
+ * 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 KOUTD, EXPRESS OR
- * IMPLIED, OUTCLUDOUTG BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONOUTFROUTGEMENT. OUT NO EVENT SHALL THE
+ * 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 OUT AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISOUTG FROM,
- * OUT OF OR OUT CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALOUTGS OUT THE
+ * 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"
* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
* @note The block shape tensor rank must be passed at compile time using -DBLOCK_SHAPE_DIM. e.g. -DBLOCK_SHAPE_DIM=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)
* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
* @note The block shape tensor rank must be passed at compile time using -DBLOCK_SHAPE_DIM. e.g. -DBLOCK_SHAPE_DIM=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)
* @note The starting pad value of y must be passed at compile time using -DPAD_LEFT_Y. e.g. -DPAD_LEFT_Y=2
* @note The ending pad value of y must be passed at compile time using -DPAD_RIGHT_Y. e.g. -DPAD_RIGHT_X=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)
* @note The starting pad value of y must be passed at compile time using -DPAD_LEFT_Y. e.g. -DPAD_LEFT_Y=2
* @note The ending pad value of y must be passed at compile time using -DPAD_RIGHT_Y. e.g. -DPAD_RIGHT_X=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)
/*
- * Copyright (c) 2019 ARM Limited.
+ * Copyright (c) 2019-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @note The input tensor batch size must be passed at compile time using -DCHANNEL_SIZE. e.g. -DCHANNEL_SIZE=2
* @note The block shape must be passed at compile time using -DBLOCK_SHAPE. e.g. -DBLOCK_SHAPE=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
* @note The input tensor batch size must be passed at compile time using -DCHANNEL_SIZE. e.g. -DCHANNEL_SIZE=2
* @note The block shape must be passed at compile time using -DBLOCK_SHAPE. e.g. -DBLOCK_SHAPE=2
*
- * @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+ * @param[in] input_ptr Pointer to the source tensor. Supported data types: All
* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source tensor in Y dimension (in bytes)
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* @attention Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16
* @note Can only take floating point data types.
*
- * @param[in] input_ptr Pointer to the source image. Supported data types: F16/F32
+ * @param[in] input_ptr Pointer to the source image. Supported data types: All
* @param[in] input_stride_x Stride of the source image in X dimension (in bytes)
* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(input->info()->dimension(3)));
build_opts.add_option("-DBLOCK_SHAPE_X=" + support::cpp11::to_string(block_shape_x));
build_opts.add_option("-DBLOCK_SHAPE_Y=" + support::cpp11::to_string(block_shape_y));
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLFlattenLayerKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Validate.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/StringSupport.h"
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
// Checks performed when output is configured
ICLKernel::configure_internal(win_config.second);
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(input->info()->dimension(0)));
build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(input->info()->dimension(1)));
build_opts.add_option("-DSRC_DEPTH=" + support::cpp11::to_string(input->info()->dimension(2)));
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLGatherKernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Window.h"
+#include "arm_compute/core/Error.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/StringSupport.h"
ARM_COMPUTE_RETURN_ERROR_ON(indices->num_dimensions() > 1);
ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
ARM_COMPUTE_RETURN_ERROR_ON(actual_axis >= input->num_dimensions());
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
if(output->total_size() != 0)
// Set build options
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DOUTPUT_DIM_Z=" + support::cpp11::to_string(output->info()->dimension(2)));
build_opts.add_option("-DINPUT_DIM_Z=" + support::cpp11::to_string(input->info()->dimension(2)));
build_opts.add_option("-DAXIS=" + support::cpp11::to_string(_axis));
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLMemsetKernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/StringSupport.h"
namespace arm_compute
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLPermuteKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/CL/OpenCL.h"
#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/Types.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/StringSupport.h"
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
ARM_COMPUTE_RETURN_ERROR_ON_MSG(input->num_dimensions() < 1 || input->num_dimensions() > 4,
"Permutation upto 4-D input tensor is supported");
ARM_COMPUTE_RETURN_ERROR_ON_MSG(perm.num_dimensions() < 1 || perm.num_dimensions() > 4,
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DDEPTH_IN=" + support::cpp11::to_string(input->info()->dimension(2)));
// New positions of width(W), height(H), channel(C) and batch(D) based on permutation vector
build_opts.add_option("-DP1=" + support::cpp11::to_string((_perm.num_dimensions() >= 1) ? perm[0] : 0));
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(output->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(output->info()->data_type())));
build_opts.add_option("-DCHANNEL_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_channel)));
build_opts.add_option("-DBLOCK_SHAPE=" + support::cpp11::to_string(block_shape));
build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLStridedSliceKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/IAccessWindow.h"
#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Window.h"
-
-#include "arm_compute/core/Types.h"
#include "arm_compute/core/utils/helpers/bit_ops.h"
#include "arm_compute/core/utils/helpers/tensor_transform.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
- ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape().num_dimensions() > 4);
// Create build options
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
for(unsigned int i = 0; i < input_shape.num_dimensions(); ++i)
{
const bool is_shrink = arm_compute::helpers::bit_ops::is_bit_set(shrink_axis_mask, i);
* SOFTWARE.
*/
#include "arm_compute/core/CL/kernels/CLTileKernel.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/CL/CLValidate.h"
#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
+#include "arm_compute/core/Error.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "support/StringSupport.h"
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
+ build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(data_type)));
build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(input_width_x));
build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(input->info()->dimension(1)));
build_opts.add_option("-DSRC_DEPTH=" + support::cpp11::to_string(input->info()->dimension(2)));
/*
- * Copyright (c) 2018-2019 ARM Limited.
+ * Copyright (c) 2018-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
namespace
{
const auto PaddingSizesDataset3D = framework::dataset::make("PaddingSize",
-{ PaddingList{ { 0, 0 } },
+{
+ PaddingList{ { 0, 0 } },
PaddingList{ { 1, 1 } },
PaddingList{ { 33, 33 } },
PaddingList{ { 1, 1 }, { 5, 5 } },
PaddingList{ { 0, 0 }, { 0, 0 }, { 0, 0 } }
});
const auto PaddingSizesDataset4D = framework::dataset::make("PaddingSize",
-{ PaddingList{ { 1, 1 }, { 1, 0 }, { 1, 1 }, { 0, 0 } },
+{
+ PaddingList{ { 1, 1 }, { 1, 0 }, { 1, 1 }, { 0, 0 } },
PaddingList{ { 0, 0 }, { 0, 0 }, { 0, 0 }, { 1, 1 } },
PaddingList{ { 0, 1 }, { 1, 0 }, { 2, 2 }, { 1, 0 } },
PaddingList{ { 1, 1 }, { 1, 1 }, { 1, 1 }, { 3, 3 } }
validate(CLAccessor(_target), _reference);
}
TEST_SUITE_END() // QASYMM8
+
+TEST_SUITE(QASYMM8_SIGNED)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLPaddingFixture<int8_t>, framework::DatasetMode::PRECOMMIT,
+ combine(combine(combine(datasets::Small3DShapes(), framework::dataset::make("DataType", { DataType::QASYMM8_SIGNED })), PaddingSizesDataset3D),
+ framework::dataset::make("PaddingMode", { PaddingMode::CONSTANT, PaddingMode::REFLECT })))
+{
+ // Validate output
+ validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // QASYMM8_SIGNED
+
TEST_SUITE_END() // Quantized
TEST_SUITE_END() // PadLayer
projects / platform / upstream / armcl.git / commitdiff