diff options
Diffstat (limited to 'arm_compute/runtime/NEON/functions')
54 files changed, 582 insertions, 81 deletions
diff --git a/arm_compute/runtime/NEON/functions/NEArgMinMaxLayer.h b/arm_compute/runtime/NEON/functions/NEArgMinMaxLayer.h index cbf1d5b444..4392de7b28 100644 --- a/arm_compute/runtime/NEON/functions/NEArgMinMaxLayer.h +++ b/arm_compute/runtime/NEON/functions/NEArgMinMaxLayer.h @@ -64,6 +64,18 @@ public: ~NEArgMinMaxLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:----------| + * |QASYMM8 |U32, S32 | + * |QASYMM8_SIGNED |U32, S32 | + * |S32 |U32, S32 | + * |F16 |U32, S32 | + * |F32 |U32, S32 | + * * @param[in] input Input source tensor. Data types supported: QASYMM8_SIGNED/QASYMM8/S32/F16/F32. * @param[in] axis Axis to find max/min index. * @param[out] output Output source tensor. Data types supported: U32/S32. diff --git a/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h b/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h index 6d56a267a7..ec00fbdbf2 100644 --- a/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NEBatchNormalizationLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -58,6 +58,16 @@ public: ~NEBatchNormalizationLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |F32 |F32 | + * |F16 |F16 | + * * @note If the output tensor is a nullptr or is equal to the input, the batch normalization function will be performed in-place * * @param[in, out] input Source tensor. In case of @p output tensor = nullptr, this tensor will store the result. diff --git a/arm_compute/runtime/NEON/functions/NEBatchToSpaceLayer.h b/arm_compute/runtime/NEON/functions/NEBatchToSpaceLayer.h index c2fd26d34c..810bf81a22 100644 --- a/arm_compute/runtime/NEON/functions/NEBatchToSpaceLayer.h +++ b/arm_compute/runtime/NEON/functions/NEBatchToSpaceLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -52,6 +52,15 @@ public: ~NEBatchToSpaceLayer() = default; /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:---------|:---------|:----------| + * |All |s32 |All | + * * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: All. * @param[in] block_shape 1-D tensor with shape [M]. Data types supported: S32 * @param[out] output Tensor output. Data types supported: same as @p input diff --git a/arm_compute/runtime/NEON/functions/NEBitwiseAnd.h b/arm_compute/runtime/NEON/functions/NEBitwiseAnd.h index 3203d2b9a7..1f95f193d3 100644 --- a/arm_compute/runtime/NEON/functions/NEBitwiseAnd.h +++ b/arm_compute/runtime/NEON/functions/NEBitwiseAnd.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2020 Arm Limited. + * Copyright (c) 2016-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -48,6 +48,14 @@ public: ~NEBitwiseAnd() = default; /** Initialise the kernel's inputs and output * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |U8 |U8 | + * * @param[in] input1 First tensor input. Data type supported: U8. * @param[in] input2 Second tensor input. Data type supported: U8. * @param[out] output Output tensor. Data type supported: U8. diff --git a/arm_compute/runtime/NEON/functions/NEBitwiseNot.h b/arm_compute/runtime/NEON/functions/NEBitwiseNot.h index 9fa0d38caf..c66bebf7cc 100644 --- a/arm_compute/runtime/NEON/functions/NEBitwiseNot.h +++ b/arm_compute/runtime/NEON/functions/NEBitwiseNot.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -36,6 +36,14 @@ class NEBitwiseNot : public INESimpleFunctionNoBorder public: /** Initialise the kernel's input and output * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |U8 |U8 | + * * @param[in] input Input tensor. Data type supported: U8. * @param[out] output Output tensor. Data type supported: U8. */ diff --git a/arm_compute/runtime/NEON/functions/NEBitwiseOr.h b/arm_compute/runtime/NEON/functions/NEBitwiseOr.h index fba6b784de..183df212e4 100644 --- a/arm_compute/runtime/NEON/functions/NEBitwiseOr.h +++ b/arm_compute/runtime/NEON/functions/NEBitwiseOr.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -36,6 +36,14 @@ class NEBitwiseOr : public INESimpleFunctionNoBorder public: /** Initialise the kernel's inputs and output * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |U8 |U8 | + * * @param[in] input1 First tensor input. Data type supported: U8. * @param[in] input2 Second tensor input. Data type supported: U8. * @param[out] output Output tensor. Data type supported: U8. diff --git a/arm_compute/runtime/NEON/functions/NEBitwiseXor.h b/arm_compute/runtime/NEON/functions/NEBitwiseXor.h index c6cb584284..126aaa6ddd 100644 --- a/arm_compute/runtime/NEON/functions/NEBitwiseXor.h +++ b/arm_compute/runtime/NEON/functions/NEBitwiseXor.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -36,6 +36,14 @@ class NEBitwiseXor : public INESimpleFunctionNoBorder public: /** Initialise the kernel's inputs and output * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |U8 |U8 | + * * @param[in] input1 First tensor input. Data type supported: U8. * @param[in] input2 Second tensor input. Data type supported: U8. * @param[out] output Output tensor. Data type supported: U8. diff --git a/arm_compute/runtime/NEON/functions/NEBoundingBoxTransform.h b/arm_compute/runtime/NEON/functions/NEBoundingBoxTransform.h index c377520a12..2a196a2de5 100644 --- a/arm_compute/runtime/NEON/functions/NEBoundingBoxTransform.h +++ b/arm_compute/runtime/NEON/functions/NEBoundingBoxTransform.h @@ -38,6 +38,17 @@ class NEBoundingBoxTransform : public INESimpleFunctionNoBorder public: /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:--------------|:--------------|:--------------| + * |QASYMM16 |QASYMM8 |QASYMM16 | + * |F16 |F16 |F16 | + * |F32 |F32 |F32 | + * * @param[in] boxes Source tensor. Bounding box proposals in pixel coordinates. Size(M, 4), format [x1, y1, x2, y2]. Data types supported: QASYMM16/F16/F32. * @param[out] pred_boxes Destination tensor. Pixel coordinates of the transformed bounding boxes. Size (M, 4*K), format [x1, y1, x2, y2]. Data types supported: Same as @p input * @param[in] deltas Bounding box translations and scales. Size (M, 4*K), format [dx, dy, dw, dh], K is the number of classes. diff --git a/arm_compute/runtime/NEON/functions/NECast.h b/arm_compute/runtime/NEON/functions/NECast.h index e536317660..eb7de1fadb 100644 --- a/arm_compute/runtime/NEON/functions/NECast.h +++ b/arm_compute/runtime/NEON/functions/NECast.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -40,18 +40,22 @@ class NECast : public INESimpleFunctionNoBorder public: /** Initialize the function's source, destination * - * Input data type must be different than output data type. + * Valid data layouts: + * - All * - * Valid conversions Input -> Output : + * Valid data type configurations: + * |src |dst | + * |:--------------|:-----------------------------------------------| + * |QASYMM8_SIGNED | S16, S32, F32, F16 | + * |QASYMM8 | U16, S16, S32, F32, F16 | + * |U8 | U16, S16, S32, F32, F16 | + * |U16 | U8, U32 | + * |S16 | QASYMM8_SIGNED, U8, S32 | + * |F16 | QASYMM8_SIGNED, QASYMM8, F32, S32, U8 | + * |S32 | QASYMM8_SIGNED, QASYMM8, F16, F32, U8 | + * |F32 | QASYMM8_SIGNED, QASYMM8, BFLOAT16, F16, S32, U8| * - * - QASYMM8_SIGNED -> S16, S32, F32, F16 - * - QASYMM8 -> U16, S16, S32, F32, F16 - * - U8 -> U16, S16, S32, F32, F16 - * - U16 -> U8, U32 - * - S16 -> QASYMM8_SIGNED, U8, S32 - * - F16 -> QASYMM8_SIGNED, QASYMM8, F32, S32, U8 - * - S32 -> QASYMM8_SIGNED, QASYMM8, F16, F32, U8 - * - F32 -> QASYMM8_SIGNED, QASYMM8, BFLOAT16, F16, S32, U8 + * Input data type must be different than output data type. * * @param[in] input The input tensor to convert. Data types supported: QASYMM8_SIGNED/QASYMM8/U8/U16/S16/F16/S32/F32. * @param[out] output The output tensor. Data types supported: QASYMM8_SIGNED/QASYMM8/U8/S8/U16/S16/U32/S32/BFLOAT16/F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NEChannelShuffleLayer.h b/arm_compute/runtime/NEON/functions/NEChannelShuffleLayer.h index aa11396c20..8888efec4f 100644 --- a/arm_compute/runtime/NEON/functions/NEChannelShuffleLayer.h +++ b/arm_compute/runtime/NEON/functions/NEChannelShuffleLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -44,6 +44,14 @@ class NEChannelShuffleLayer : public INESimpleFunctionNoBorder public: /** Initialize the function * + * Valid data layouts: + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input Input tensor. Data types supported: All * @param[out] output Output tensor. Data type supported: Same as @p input * @param[in] num_groups Number of groups. Must be greater than 1 and the number of channels of the tensors must be a multiple of the number of groups. diff --git a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h index b1e85523c5..f19aa8008b 100644 --- a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h @@ -86,12 +86,27 @@ public: ~NEConvolutionLayer() = default; /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | + * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. + * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. + * Data type supported: Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type. + * Data type supported: Same as @p input, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type. * @param[out] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. @@ -110,9 +125,10 @@ public: * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. + * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. + * Data type supported:Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type. + * Data type supported: Same as @p input, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type. * @param[in] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. @@ -134,7 +150,8 @@ public: * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. + * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. + * Data type supported:Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. * @param[in] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. diff --git a/arm_compute/runtime/NEON/functions/NECropResize.h b/arm_compute/runtime/NEON/functions/NECropResize.h index 7dcf925650..143bbbc6f1 100644 --- a/arm_compute/runtime/NEON/functions/NECropResize.h +++ b/arm_compute/runtime/NEON/functions/NECropResize.h @@ -54,6 +54,14 @@ public: /** Configure kernel * + * Valid data layouts: + * - NHWC + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------|:--------|:------|:--------| + * |All |F32 |F32 |F32 | + * * @note Supported tensor rank: up to 4 * @note Box indices may be outside of the bounds, in which case @p extrapolation_value is used. * @note Start and end indices of boxes are inclusive. diff --git a/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h index c16cf26095..34ab0707c2 100644 --- a/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h @@ -90,6 +90,20 @@ public: /** Set the input, weights, biases and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | + * * @param[in,out] input Input tensor. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: F32/F16/QASYMM8/QASYMM8_SIGNED. * @param[in] weights The 4d weights with dimensions [width, height, IFM, OFM]. Data type supported: Same as @p input, also could be QSYMM8_PER_CHANNEL if input is QASYMM8/QASYMM8_SIGNED. * @param[in] bias Optional, ignored if NULL. The biases have one dimension. Data type supported: Data types supported: S32 for QASYMM8/QASYMM8_SIGNED input, F32 for F32 input, F16 for F16 input. diff --git a/arm_compute/runtime/NEON/functions/NEDepthConvertLayer.h b/arm_compute/runtime/NEON/functions/NEDepthConvertLayer.h index c9817a63c1..17cf539717 100644 --- a/arm_compute/runtime/NEON/functions/NEDepthConvertLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDepthConvertLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2020 Arm Limited. + * Copyright (c) 2016-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -48,15 +48,21 @@ public: ~NEDepthConvertLayer() = default; /** Initialize the function's source, destination * - * Valid conversions Input -> Output : + * Valid data layouts: + * - All * - * - QASYMM8 -> F16, F32 - * - U8 -> U16, S16, S32 - * - U16 -> U8, U32 - * - S16 -> U8, S32 - * - BFLOAT16 -> F32 - * - F16 -> QASYMM8, F32 - * - F32 -> QASYMM8, F16, BFLOAT16 + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------------------| + * |QASYMM8 | F16, F32 | + * |U8 | U16, S16, S32 | + * |U16 | U8, U32 | + * |S16 | U8, S32 | + * |BFLOAT16 | F32 | + * |F16 | QASYMM8, F32 | + * |F32 | QASYMM8, F16, BFLOAT16 | + * + * Input data type must be different than output data type. * * @param[in] input The input tensor to convert. Data types supported: QASYMM8/U8/U16/S16/BFLOAT16/F16/F32. * @param[out] output The output tensor. Data types supported: QASYMM8/U8/U16/S16/U32/S32/BFLOAT16/F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h b/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h index 51f7ff7770..b9bdcd1f11 100644 --- a/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDepthToSpaceLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -53,6 +53,15 @@ public: ~NEDepthToSpaceLayer() = default; /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: All * @param[out] output Tensor output. Data types supported: same as @p input * @param[in] block_shape Block shape value. diff --git a/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h index c74b2a93ee..2f541758f4 100644 --- a/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h @@ -55,6 +55,20 @@ public: ~NEDepthwiseConvolutionLayer(); /** Initialize the function's source, destination, weights and convolution information. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | + * * @param[in, out] input Source tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32 * @param[out] output Destination tensor. Data type supported: same as @p input. * @param[in] weights Weights tensor. These are 3D tensors with shape [kernel_x, kernel_y, IFM]. diff --git a/arm_compute/runtime/NEON/functions/NEDequantizationLayer.h b/arm_compute/runtime/NEON/functions/NEDequantizationLayer.h index dfec835f45..91ed056cf3 100644 --- a/arm_compute/runtime/NEON/functions/NEDequantizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDequantizationLayer.h @@ -57,18 +57,13 @@ public: * - All * * Valid data type configurations: - * |src |dst | - * |:------------------|:--------------| - * |QASYMM8 |F16 | - * |QASYMM8 |F32 | - * |QASYMM8_SIGNED |F16 | - * |QASYMM8_SIGNED |F32 | - * |QSYMM8_PER_CHANNEL |F16 | - * |QSYMM8_PER_CHANNEL |F32 | - * |QSYMM8 |F16 | - * |QSYMM8 |F32 | - * |QSYMM16 |F16 | - * |QSYMM16 |F32 | + * |src |dst | + * |:------------------|:-----------| + * |QASYMM8 |F16, F32 | + * |QASYMM8_SIGNED |F16, F32 | + * |QSYMM8_PER_CHANNEL |F16, F32 | + * |QSYMM8 |F16, F32 | + * |QSYMM16 |F16, F32 | * * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/QSYMM8/QSYMM16. * @param[out] output Destination tensor with the same dimensions of input. Data type supported: F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NEFillBorder.h b/arm_compute/runtime/NEON/functions/NEFillBorder.h index 8a8a0c7dc2..ab77c28839 100644 --- a/arm_compute/runtime/NEON/functions/NEFillBorder.h +++ b/arm_compute/runtime/NEON/functions/NEFillBorder.h @@ -42,6 +42,14 @@ public: NEFillBorder(); /** Initialize the function's source, destination and border_mode. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @note This function fills the borders within the XY-planes. * * @param[in, out] input Source tensor. Data type supported: All diff --git a/arm_compute/runtime/NEON/functions/NEFlattenLayer.h b/arm_compute/runtime/NEON/functions/NEFlattenLayer.h index 1104aac77f..e688e918d9 100644 --- a/arm_compute/runtime/NEON/functions/NEFlattenLayer.h +++ b/arm_compute/runtime/NEON/functions/NEFlattenLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -39,6 +39,14 @@ class NEFlattenLayer : public IFunction public: /** Initialise the kernel's input and output. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input First input tensor to flatten with at least 3 dimensions. The dimensions over the third will be interpreted as batches. Data types supported: All * @param[out] output Output tensor with shape [w*h*d, input_batches] where: * w = width input tensor, h = height input tensor and d = depth input tensor. Data type supported: same as @p input diff --git a/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h b/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h index bc45e58b4b..9727e108a5 100644 --- a/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h +++ b/arm_compute/runtime/NEON/functions/NEFullyConnectedLayer.h @@ -36,7 +36,6 @@ namespace arm_compute { - namespace weights_transformations { /** Basic function to manage the reshape weights generated from @ref NETranspose */ @@ -102,6 +101,18 @@ public: ~NEFullyConnectedLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * * @param[in] input Source tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32. * @param[in] weights Weights tensor. The weights must be 2 dimensional. * If this function is called after a Convolution Layer, the (transposed) weights will have as many rows as the product of the first 3 input's dimensions. diff --git a/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h b/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h index 5dc804e240..3dd7f49044 100644 --- a/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h +++ b/arm_compute/runtime/NEON/functions/NEFuseBatchNormalization.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -52,6 +52,16 @@ public: ~NEFuseBatchNormalization(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |F32 |F32 | + * |F16 |F16 | + * * @param[in] input_weights Input weights tensor for convolution or depthwise convolution layer. Data type supported: F16/F32. Data layout supported: NCHW, NHWC * @param[in] bn_mean Batch normalization layer mean tensor. Same as @p input_weights * @param[in] bn_var Batch normalization layer variance tensor. Same as @p input_weights diff --git a/arm_compute/runtime/NEON/functions/NEGEMM.h b/arm_compute/runtime/NEON/functions/NEGEMM.h index a6c3436656..d4a9f68beb 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMM.h +++ b/arm_compute/runtime/NEON/functions/NEGEMM.h @@ -76,6 +76,16 @@ public: ~NEGEMM(); /** Initialise the kernel's inputs, output * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:------------|:-----------|:---------|:--------------| + * |F32 |F32 |F32 |F32 | + * |F16 |F16 |F16 |F16 | + * |BFLOAT16 |BFLOAT16 |BFLOAT16 |BFLOAT16 | + * * @note GEMM: General Matrix Multiply - [alpha * A * B + beta * C]. * @note GEMM: The tensors a, b, c, d must have the same data type. You should not mix data types when calling this function. * diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h index 9897bf1d4d..e89eae1d31 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h @@ -177,6 +177,21 @@ public: ~NEGEMMConvolutionLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:--------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |BFLOAT16 |BFLOAT16 |BFLOAT16 |BFLOAT16 | + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | + * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: QASYMM8/QASYMM8_SIGNED/BFLOAT16/F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h b/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h index b2b77bd406..780723e752 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h @@ -76,6 +76,26 @@ public: ~NEGEMMLowpMatrixMultiplyCore(); /** Initialise the kernel's inputs, output * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:--------|:--------------| + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |QASYMM8 | + * |QASYMM8 |QSYMM8 |S32 |QASYMM8 | + * |QASYMM8 |QASYMM8 |S32 |S32 | + * |QASYMM8 |QSYMM8_PER_CHANNEL |S32 |S32 | + * |QASYMM8 |QSYMM8 |S32 |S32 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QSYMM8 |S32 |QASYMM8_SIGNED | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |S32 | + * |QASYMM8_SIGNED |QSYMM8_PER_CHANNEL |S32 |S32 | + * |QASYMM8_SIGNED |QSYMM8 |S32 |S32 | + * * @note GEMM_LOWP: low precision GEMM kernel * This kernel performs the following computations: * diff --git a/arm_compute/runtime/NEON/functions/NEGather.h b/arm_compute/runtime/NEON/functions/NEGather.h index a5e0461227..393a38ee4d 100644 --- a/arm_compute/runtime/NEON/functions/NEGather.h +++ b/arm_compute/runtime/NEON/functions/NEGather.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -40,6 +40,14 @@ class NEGather : public INESimpleFunctionNoBorder public: /** Initialise the kernel's inputs and outputs * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input Source tensor. Supported tensor rank: up to 4. Data type supported: All * @param[in] indices Indices tensor. Supported tensor rank: up to 1. Must be one of the following type: U32/S32. Each value Must be in range [0, input.shape[@p axis]) * @param[out] output Destination tensor. Data type supported: Same as @p input diff --git a/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h b/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h index 22c6ba2ed6..3b683382ec 100644 --- a/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h +++ b/arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h @@ -72,6 +72,16 @@ public: /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:------------------|:--------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |QASYMM8 |QSYMM8 |QSYMM16 |QASYMM8 | + * * @param[in] scores Scores from convolution layer of size (W, H, A), where H and W are the height and width of the feature map, and A is the number of anchors. * Data types supported: QASYMM8/F16/F32 * @param[in] deltas Bounding box deltas from convolution layer of size (W, H, 4*A). Data types supported: Same as @p scores diff --git a/arm_compute/runtime/NEON/functions/NEInstanceNormalizationLayer.h b/arm_compute/runtime/NEON/functions/NEInstanceNormalizationLayer.h index 57165c94b4..bb0697072b 100644 --- a/arm_compute/runtime/NEON/functions/NEInstanceNormalizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NEInstanceNormalizationLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -60,6 +60,16 @@ public: ~NEInstanceNormalizationLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------|:---------| + * |F16 |F16 | + * |F32 |F32 | + * * @param[in, out] input Source tensor. In case of @p output tensor = nullptr this tensor will store the result of the normalization. * Data types supported: F16/F32. Data layout supported: NHWC, NCHW * @param[out] output Destination tensor. Data types and data layouts supported: same as @p input. diff --git a/arm_compute/runtime/NEON/functions/NEL2NormalizeLayer.h b/arm_compute/runtime/NEON/functions/NEL2NormalizeLayer.h index 173b9d2141..7f1a5e785e 100644 --- a/arm_compute/runtime/NEON/functions/NEL2NormalizeLayer.h +++ b/arm_compute/runtime/NEON/functions/NEL2NormalizeLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -60,6 +60,16 @@ public: ~NEL2NormalizeLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------|:---------| + * |F16 |F16 | + * |F32 |F32 | + * * @param[in, out] input Source tensor. Data types supported: F16/F32. (Written to only for border_size != 0) * @param[out] output Destination tensor. Data types and data layouts supported: same as @p input. * @param[in] axis Axis along which to reduce. Negative values wrap around. Maximum supported actual reduction axis : 2 diff --git a/arm_compute/runtime/NEON/functions/NELSTMLayer.h b/arm_compute/runtime/NEON/functions/NELSTMLayer.h index ef8defb827..075fb4530a 100644 --- a/arm_compute/runtime/NEON/functions/NELSTMLayer.h +++ b/arm_compute/runtime/NEON/functions/NELSTMLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -60,6 +60,15 @@ public: ~NELSTMLayer(); /** Initialize function's tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 - src13 | dst0 - dst3 | + * |:------------|:------------| + * |F16 |F16 | + * |F32 |F32 | + * * @param[in] input Source tensor. Input is a 2D tensor with dimensions [input_size, batch_size]. Data types supported: F16/F32. * @param[in] input_to_forget_weights 2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input. * @param[in] input_to_cell_weights 2D weights tensor with dimensions [input_size, num_units]. Data type supported: Same as @p input. diff --git a/arm_compute/runtime/NEON/functions/NELSTMLayerQuantized.h b/arm_compute/runtime/NEON/functions/NELSTMLayerQuantized.h index 53a024ae04..2f0c753691 100644 --- a/arm_compute/runtime/NEON/functions/NELSTMLayerQuantized.h +++ b/arm_compute/runtime/NEON/functions/NELSTMLayerQuantized.h @@ -77,6 +77,14 @@ public: ~NELSTMLayerQuantized(); /** Initialize function's tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 - src8 |src9 - src12 |src13 |src14 |dst0 |dst1 | + * |:-----------|:------------|:-------|:------|:------|:------| + * |QASYMM8 |S32 |QSYMM16 |QASYMM8|QSYMM16|QASYMM8| + * * @param[in] input Source tensor. Input is a 2D tensor with dimensions [input_size, batch_size]. Data types supported: QASYMM8. * @param[in] input_to_input_weights 2D weights tensor with dimensions [input_size, output_size]. Data type supported: Same as @p input. * @param[in] input_to_forget_weights 2D weights tensor with dimensions [input_size, output_size]. Data type supported: Same as @p input. diff --git a/arm_compute/runtime/NEON/functions/NEMaxUnpoolingLayer.h b/arm_compute/runtime/NEON/functions/NEMaxUnpoolingLayer.h index fae26b3c93..41ea040457 100644 --- a/arm_compute/runtime/NEON/functions/NEMaxUnpoolingLayer.h +++ b/arm_compute/runtime/NEON/functions/NEMaxUnpoolingLayer.h @@ -57,6 +57,18 @@ public: ~NEMaxUnpoolingLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |QASYMM8 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED | + * |F16 |F16 | + * |F32 |F32 | + * * @note Only supported pool size 2 * * @param[in, out] input Source tensor. (Written to only when padding != 0) Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.h b/arm_compute/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.h index 31e376191c..41aa81946b 100644 --- a/arm_compute/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2020 Arm Limited. + * Copyright (c) 2019-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -50,6 +50,16 @@ public: ~NEMeanStdDevNormalizationLayer(); /** Initialise the function's input and outputs. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------|:---------| + * |F32 |F32 | + * |F16 |F16 | + * * @note If the output tensor is a nullptr, the normalization will be performed in-place. * * @param[in, out] input Input tensor with 2 dimensions. Data types supported: F16/F32. diff --git a/arm_compute/runtime/NEON/functions/NENormalizationLayer.h b/arm_compute/runtime/NEON/functions/NENormalizationLayer.h index 8c4ad1516e..fbe000445c 100644 --- a/arm_compute/runtime/NEON/functions/NENormalizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NENormalizationLayer.h @@ -63,6 +63,16 @@ public: ~NENormalizationLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------|:---------| + * |F32 |F32 | + * |F16 |F16 | + * * @param[in] input Source tensor. 3 lower dims represent a single input with dimensions [width, height, IFM], * and an optional 4th dimension for batch of inputs. Data type supported: F16/F32. Data layouts supported: NCHW/NHWC. * @param[out] output Destination with the same dimensions, data type, data layout and number of channels of @p input diff --git a/arm_compute/runtime/NEON/functions/NEPadLayer.h b/arm_compute/runtime/NEON/functions/NEPadLayer.h index 76ff0643a0..4aa6725496 100644 --- a/arm_compute/runtime/NEON/functions/NEPadLayer.h +++ b/arm_compute/runtime/NEON/functions/NEPadLayer.h @@ -65,6 +65,15 @@ public: ~NEPadLayer(); /** Initialize the function * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------|:---------| + * |All |All | + * * @param[in] input Source tensor. Data types supported: All. * @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] diff --git a/arm_compute/runtime/NEON/functions/NEPriorBoxLayer.h b/arm_compute/runtime/NEON/functions/NEPriorBoxLayer.h index 3cc79fa28e..38e0c9f3ad 100644 --- a/arm_compute/runtime/NEON/functions/NEPriorBoxLayer.h +++ b/arm_compute/runtime/NEON/functions/NEPriorBoxLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -38,6 +38,15 @@ class NEPriorBoxLayer : public INESimpleFunctionNoBorder public: /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:--------|:--------|:--------| + * |F32 |F32 |F32 | + * * @param[in] input1 First source tensor. Data types supported: F32. Data layouts supported: NCHW/NHWC. * @param[in] input2 Second source tensor. Data types and layouts supported: same as @p input1 * @param[out] output Destination tensor. Output dimensions are [W * H * num_priors * 4, 2]. Data type supported: same as @p input diff --git a/arm_compute/runtime/NEON/functions/NEQLSTMLayer.h b/arm_compute/runtime/NEON/functions/NEQLSTMLayer.h index e706179415..7c2e9bc5a1 100644 --- a/arm_compute/runtime/NEON/functions/NEQLSTMLayer.h +++ b/arm_compute/runtime/NEON/functions/NEQLSTMLayer.h @@ -76,6 +76,14 @@ public: ~NEQLSTMLayer(); /** Initialize function's tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 - src6 |src7 -src9 |src10 |src11 |dst0 |dst1 - dst2 | + * |:-------------|:------------|:------------|:------|:-------------|:------|:-----------------| + * |QASYMM8_SIGNED|QASYMM8 |S32 |QSYMM16|QASYMM8_SIGNED|QSYMM16|QASYMM8_SIGNED | + * * @param[in] input Source tensor. Input is a 2D tensor with dimensions [input_size, batch_size]. Data types supported: QASYMM8_SIGNED. * @param[in] input_to_forget_weights 2D weights tensor with dimensions [input_size, num_units]. Data type supported: QSYMM8. * @param[in] input_to_cell_weights 2D weights tensor with dimensions [input_size, num_units]. Data type supported: QSYMM8. diff --git a/arm_compute/runtime/NEON/functions/NEQuantizationLayer.h b/arm_compute/runtime/NEON/functions/NEQuantizationLayer.h index a7fadfc7cd..eeca2bb1db 100644 --- a/arm_compute/runtime/NEON/functions/NEQuantizationLayer.h +++ b/arm_compute/runtime/NEON/functions/NEQuantizationLayer.h @@ -56,20 +56,12 @@ public: * - All * * Valid data type configurations: - * |src |dst | - * |:------------------|:--------------| - * |QASYMM8 |QASYMM8 | - * |QASYMM8 |QASYMM8_SIGNED | - * |QASYMM8 |QASYMM16 | - * |QASYMM8_SIGNED |QASYMM8 | - * |QASYMM8_SIGNED |QASYMM8_SIGNED | - * |QASYMM8_SIGNED |QASYMM16 | - * |F16 |QASYMM8 | - * |F16 |QASYMM8_SIGNED | - * |F16 |QASYMM16 | - * |F32 |QASYMM8 | - * |F32 |QASYMM8_SIGNED | - * |F32 |QASYMM16 | + * |src |dst | + * |:------------------|:--------------------------------------| + * |QASYMM8 |QASYMM8, QASYMM8_SIGNED, QASYMM16 | + * |QASYMM8_SIGNED |QASYMM8, QASYMM8_SIGNED, QASYMM16 | + * |F16 |QASYMM8, QASYMM8_SIGNED, QASYMM16 | + * |F32 |QASYMM8, QASYMM8_SIGNED, QASYMM16 | * * @param[in] input Source tensor. The dimensions over the third will be interpreted as batches. Data types supported: QASYMM8/QASYMM8_SIGNED/F32/F16. * @param[out] output Destination tensor with the same dimensions of input. Data types supported: QASYMM8/QASYMM8_SIGNED/QASYMM16 diff --git a/arm_compute/runtime/NEON/functions/NERNNLayer.h b/arm_compute/runtime/NEON/functions/NERNNLayer.h index 66f7f2ea3f..667d3144ac 100644 --- a/arm_compute/runtime/NEON/functions/NERNNLayer.h +++ b/arm_compute/runtime/NEON/functions/NERNNLayer.h @@ -54,6 +54,16 @@ public: ~NERNNLayer(); /** Initialize the function * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |src3 |dst0 |dst1 | + * |:------|:------|:------|:------|:------|:------| + * |F16 |F16 |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 |F32 |F32 | + * * @param[in] input Input is a 2-D tensor of shape [input_size, batch_size]. Data types supported: F16/F32 * @param[in] weights Weights tensor of shape [input_size, num_units] that multiplies the input. Data types supported: Same as @p input * @param[in] recurrent_weights Weights tensor of shape [num_units, num_units] that multiplies the current 'state'. Data types supported: Same as @p input diff --git a/arm_compute/runtime/NEON/functions/NEROIAlignLayer.h b/arm_compute/runtime/NEON/functions/NEROIAlignLayer.h index c72cd494d2..ea1af4daea 100644 --- a/arm_compute/runtime/NEON/functions/NEROIAlignLayer.h +++ b/arm_compute/runtime/NEON/functions/NEROIAlignLayer.h @@ -38,10 +38,21 @@ class NEROIAlignLayer : public INESimpleFunctionNoBorder public: /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:--------------|:--------------|:--------------| + * |F16 |F16 |F16 | + * |F32 |F32 |F32 | + * |QASYMM8 |QASYMM16 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM16 |QASYMM8_SIGNED | + * * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. * @param[in] rois ROIs tensor, it is a 2D tensor of size [5, N] (where N is the number of ROIs) containing top left and bottom right corner * as coordinate of an image and batch_id of ROI [ batch_id, x1, y1, x2, y2 ]. - * Data types supported: QASYMM16 with scale of 0.125 and 0 offset if @p input is QASYMM8, otherwise same as @p input + * Data types supported: QASYMM16 with scale of 0.125 and 0 offset if @p input is QASYMM8/QASYMM8_SIGNED, otherwise same as @p input * @param[out] output Destination tensor. Data types supported: Same as @p input. * @param[in] pool_info Contains pooling operation information described in @ref ROIPoolingLayerInfo. * @@ -54,7 +65,7 @@ public: /** Static function to check if given info will lead to a valid configuration of @ref NEROIAlignLayerKernel * * @param[in] input Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] rois ROIs tensor info. Data types supported: QASYMM16 with scale of 0.125 and 0 offset if @p input is QASYMM8, + * @param[in] rois ROIs tensor info. Data types supported: QASYMM16 with scale of 0.125 and 0 offset if @p input is QASYMM8/QASYMM8_SIGNED, * otherwise same as @p input * @param[in] output Destination tensor info. Data types supported: Same as @p input. * @param[in] pool_info Contains pooling operation information described in @ref ROIPoolingLayerInfo. diff --git a/arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h b/arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h index 214dd43402..2992b3eb95 100644 --- a/arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h +++ b/arm_compute/runtime/NEON/functions/NEROIPoolingLayer.h @@ -53,6 +53,15 @@ public: ~NEROIPoolingLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:--------------|:--------------|:--------------| + * |F32 |U16 |F32 | + * |QASYMM8 |U16 |QASYMM8 | + * * @param[in] input Source tensor. Data types supported: QASYMM8/F32 * @param[in] rois ROIs tensor, it is a 2D tensor of size [5, N] (where N is the number of ROIs) containing top left and bottom right corner * as coordinate of an image and batch_id of ROI [ batch_id, x1, y1, x2, y2 ]. Data types supported: U16 diff --git a/arm_compute/runtime/NEON/functions/NERange.h b/arm_compute/runtime/NEON/functions/NERange.h index 28976001d7..cb14c8fdde 100644 --- a/arm_compute/runtime/NEON/functions/NERange.h +++ b/arm_compute/runtime/NEON/functions/NERange.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -57,6 +57,21 @@ public: ~NERange(); /** Initialize the kernel's start, end, step and output tensor. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |dst | + * |:---------| + * |U8 | + * |S8 | + * |U16 | + * |S16 | + * |U32 | + * |S32 | + * |F16 | + * |F32 | + * * @param[out] output Output tensor. Data types supported: U8/S8/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. diff --git a/arm_compute/runtime/NEON/functions/NEReduceMean.h b/arm_compute/runtime/NEON/functions/NEReduceMean.h index 89cd09812b..7512115a3f 100644 --- a/arm_compute/runtime/NEON/functions/NEReduceMean.h +++ b/arm_compute/runtime/NEON/functions/NEReduceMean.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -54,6 +54,17 @@ public: ~NEReduceMean(); /** Configure kernel * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |QASYMM8 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED | + * |F16 |F16 | + * |F32 |F32 | + * * @note Supported tensor rank: up to 4 * * @param[in] input Source tensor. Data type supported: QASYMM8_SIGNED/QASYMM8/F16/F32 diff --git a/arm_compute/runtime/NEON/functions/NEReductionOperation.h b/arm_compute/runtime/NEON/functions/NEReductionOperation.h index b96b70926c..533c10adcf 100644 --- a/arm_compute/runtime/NEON/functions/NEReductionOperation.h +++ b/arm_compute/runtime/NEON/functions/NEReductionOperation.h @@ -58,7 +58,19 @@ public: ~NEReductionOperation(); /** Set the input and output tensors. * - * @param[in, out] input Source tensor. Data type supported: QASYMM8_SIGNED/QASYMM8/F16/F32/S32. Data layouts supported: NCHW. (Written to only for border_size != 0) + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |QASYMM8 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED | + * |F16 |F16 | + * |F32 |F32 | + * |S32 |S32 | + * + * @param[in, out] input Source tensor. Data type supported: QASYMM8_SIGNED/QASYMM8/F16/F32/S32. (Written to only for border_size != 0) * @param[out] output Destination tensor. Data types and data layouts supported: same as @p input. * @param[in] axis Dimension along which to reduce. Supported reduction axis : 0 * @param[in] op Reduction operation to perform. @@ -68,7 +80,7 @@ public: /** Static function to check if given info will lead to a valid configuration of @ref NEReductionOperation. * - * @param[in] input Source tensor info. Data type supported: QASYMM8_SIGNED/QASYMM8/F16/F32/S32. Data layouts supported: NCHW. + * @param[in] input Source tensor info. Data type supported: QASYMM8_SIGNED/QASYMM8/F16/F32/S32. * @param[in] output Destination tensor info. Data types and data layouts supported: same as @p input. * @param[in] axis Dimension along which to reduce. Supported reduction axis : 0 * @param[in] op Reduction operation to perform. diff --git a/arm_compute/runtime/NEON/functions/NERemap.h b/arm_compute/runtime/NEON/functions/NERemap.h index 835ebfab7e..1693078f66 100644 --- a/arm_compute/runtime/NEON/functions/NERemap.h +++ b/arm_compute/runtime/NEON/functions/NERemap.h @@ -44,6 +44,14 @@ class NERemap : public INESimpleFunction public: /** Initialise the function's sources, destination, interpolation policy and border mode. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:------|:------|:------|:------| + * |U8 |F32 |F32 |U 8 | + * * @param[in, out] input Source tensor. Data type supported: U8. (Written to only for @p border_mode != UNDEFINED) * @param[in] map_x Map for X coordinates. Data type supported: F32. * @param[in] map_y Map for Y coordinates. Data type supported: F32. diff --git a/arm_compute/runtime/NEON/functions/NEReorgLayer.h b/arm_compute/runtime/NEON/functions/NEReorgLayer.h index f76d1d252c..0a7d824d10 100644 --- a/arm_compute/runtime/NEON/functions/NEReorgLayer.h +++ b/arm_compute/runtime/NEON/functions/NEReorgLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -39,6 +39,15 @@ class NEReorgLayer : public INESimpleFunctionNoBorder public: /** Initialise the kernel's inputs and outputs * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input First tensor input. Data type supported: All * @param[out] output Output tensor. Data type supported: Same as @p input * @param[in] stride Stride to be used during data re-organization diff --git a/arm_compute/runtime/NEON/functions/NEReverse.h b/arm_compute/runtime/NEON/functions/NEReverse.h index 2048dafcb5..c02fff54a5 100644 --- a/arm_compute/runtime/NEON/functions/NEReverse.h +++ b/arm_compute/runtime/NEON/functions/NEReverse.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -39,6 +39,14 @@ class NEReverse : public INESimpleFunctionNoBorder public: /** Initialize the function * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |dst | + * |:--------------|:--------------|:--------------| + * |All |U32 |All | + * * @param[in] input Input tensor. Data types supported: All * @param[out] output Output tensor. Data type supported: Same as @p input * @param[in] axis Axis tensor. Contains the indices of the dimensions to reverse. Data type supported: U32 diff --git a/arm_compute/runtime/NEON/functions/NESelect.h b/arm_compute/runtime/NEON/functions/NESelect.h index c66fbfa7d4..c8e5a204dd 100644 --- a/arm_compute/runtime/NEON/functions/NESelect.h +++ b/arm_compute/runtime/NEON/functions/NESelect.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -39,6 +39,14 @@ class NESelect : public INESimpleFunctionNoBorder public: /** Initialise the kernel's inputs and output. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:--------------|:------|:--------------| + * |U8 |All |All |All | + * * @param[in] c Condition input tensor. Data types supported: U8. * @param[in] x First input tensor. Data types supported: All. * @param[in] y Second input tensor. Data types supported: Same as @p x diff --git a/arm_compute/runtime/NEON/functions/NESpaceToBatchLayer.h b/arm_compute/runtime/NEON/functions/NESpaceToBatchLayer.h index 27c1ddf8e3..ad8c1467d0 100644 --- a/arm_compute/runtime/NEON/functions/NESpaceToBatchLayer.h +++ b/arm_compute/runtime/NEON/functions/NESpaceToBatchLayer.h @@ -58,6 +58,15 @@ public: ~NESpaceToBatchLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:---------|:---------|:---------|:---------| + * |All |S32 |S32 |All | + * * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: All. * @param[in] block_shape 1-D tensor with shape [M]. Supported M: 2. Data types supported: S32 * @param[in] paddings 2-D tensor with shape [2, M] (First dimension is the fastest-changing dimension). Supported M: 2. Data types supported: S32 diff --git a/arm_compute/runtime/NEON/functions/NESpaceToDepthLayer.h b/arm_compute/runtime/NEON/functions/NESpaceToDepthLayer.h index 73c228d8ee..1820cb8f6b 100644 --- a/arm_compute/runtime/NEON/functions/NESpaceToDepthLayer.h +++ b/arm_compute/runtime/NEON/functions/NESpaceToDepthLayer.h @@ -53,6 +53,15 @@ public: ~NESpaceToDepthLayer(); /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: All. * @param[out] output Tensor output. Data types supported: same as @p input * @param[in] block_shape Block shape value diff --git a/arm_compute/runtime/NEON/functions/NESplit.h b/arm_compute/runtime/NEON/functions/NESplit.h index ede5ecf65a..206f299c06 100644 --- a/arm_compute/runtime/NEON/functions/NESplit.h +++ b/arm_compute/runtime/NEON/functions/NESplit.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -40,6 +40,18 @@ namespace arm_compute class NESplit : public CPPSplit<NESlice> { public: + /** NESplit + * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * + */ + // Inherited methods overridden: void run() override; }; diff --git a/arm_compute/runtime/NEON/functions/NEStackLayer.h b/arm_compute/runtime/NEON/functions/NEStackLayer.h index f6fa4f2eb3..ae4e468f21 100644 --- a/arm_compute/runtime/NEON/functions/NEStackLayer.h +++ b/arm_compute/runtime/NEON/functions/NEStackLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -58,6 +58,14 @@ public: ~NEStackLayer(); /** Initialise the kernel's inputs vector and output. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @note Supported input tensor rank: up to 4 * * @param[in] input The vectors containing all the tensors with the same shape to stack. Data types supported: All diff --git a/arm_compute/runtime/NEON/functions/NETile.h b/arm_compute/runtime/NEON/functions/NETile.h index d5ce76c9cf..915e5aa1da 100644 --- a/arm_compute/runtime/NEON/functions/NETile.h +++ b/arm_compute/runtime/NEON/functions/NETile.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -39,6 +39,14 @@ class NETile : public INESimpleFunctionNoBorder public: /** Set the source, destination of the kernel * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input Source tensor. Data type supported: All. * @param[out] output Destination tensor. Same as @p input * @param[in] multiples Contains the number of times the input tensor should be replicated on the given dimension. diff --git a/arm_compute/runtime/NEON/functions/NEUnstack.h b/arm_compute/runtime/NEON/functions/NEUnstack.h index c8e85115f7..079fee5b9e 100644 --- a/arm_compute/runtime/NEON/functions/NEUnstack.h +++ b/arm_compute/runtime/NEON/functions/NEUnstack.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -57,6 +57,14 @@ public: ~NEUnstack() = default; /** Set the input, output and unstacking axis. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |All |All | + * * @param[in] input A tensor to be unstacked. Data type supported: All. * @param[in,out] output_vector A vector of tensors. Data types supported: same as @p input. * Note: The number of elements of the vector will be used as the number of slices to be taken from the axis. diff --git a/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h index 3367b10a96..77f9093ed4 100644 --- a/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h @@ -66,6 +66,16 @@ public: /** Set the input and output tensors. * + * Valid data layouts: + * - NHWC + * - NCHW + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:--------------|:------|:--------------| + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: F16/F32. |