| /* |
| * Copyright (c) 2016-2020 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_UTILS_H |
| #define ARM_COMPUTE_UTILS_H |
| |
| #include "arm_compute/core/Error.h" |
| #include "arm_compute/core/PixelValue.h" |
| #include "arm_compute/core/Rounding.h" |
| #include "arm_compute/core/Types.h" |
| #include "arm_compute/core/Version.h" |
| |
| #include <algorithm> |
| #include <cstdint> |
| #include <cstdlib> |
| #include <iomanip> |
| #include <numeric> |
| #include <sstream> |
| #include <string> |
| #include <type_traits> |
| #include <unordered_map> |
| #include <utility> |
| #include <vector> |
| |
| namespace arm_compute |
| { |
| class ITensor; |
| class ITensorInfo; |
| |
| /** Calculate the rounded up quotient of val / m. |
| * |
| * @param[in] val Value to divide and round up. |
| * @param[in] m Value to divide by. |
| * |
| * @return the result. |
| */ |
| template <typename S, typename T> |
| constexpr auto DIV_CEIL(S val, T m) -> decltype((val + m - 1) / m) |
| { |
| return (val + m - 1) / m; |
| } |
| |
| /** Computes the smallest number larger or equal to value that is a multiple of divisor. |
| * |
| * @param[in] value Lower bound value |
| * @param[in] divisor Value to compute multiple of. |
| * |
| * @return the result. |
| */ |
| template <typename S, typename T> |
| inline auto ceil_to_multiple(S value, T divisor) -> decltype(((value + divisor - 1) / divisor) * divisor) |
| { |
| ARM_COMPUTE_ERROR_ON(value < 0 || divisor <= 0); |
| return DIV_CEIL(value, divisor) * divisor; |
| } |
| |
| /** Computes the largest number smaller or equal to value that is a multiple of divisor. |
| * |
| * @param[in] value Upper bound value |
| * @param[in] divisor Value to compute multiple of. |
| * |
| * @return the result. |
| */ |
| template <typename S, typename T> |
| inline auto floor_to_multiple(S value, T divisor) -> decltype((value / divisor) * divisor) |
| { |
| ARM_COMPUTE_ERROR_ON(value < 0 || divisor <= 0); |
| return (value / divisor) * divisor; |
| } |
| |
| /** Load an entire file in memory |
| * |
| * @param[in] filename Name of the file to read. |
| * @param[in] binary Is it a binary file ? |
| * |
| * @return The content of the file. |
| */ |
| std::string read_file(const std::string &filename, bool binary); |
| |
| /** The size in bytes of the data type |
| * |
| * @param[in] data_type Input data type |
| * |
| * @return The size in bytes of the data type |
| */ |
| inline size_t data_size_from_type(DataType data_type) |
| { |
| switch(data_type) |
| { |
| case DataType::U8: |
| case DataType::S8: |
| case DataType::QSYMM8: |
| case DataType::QASYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QSYMM8_PER_CHANNEL: |
| return 1; |
| case DataType::U16: |
| case DataType::S16: |
| case DataType::QSYMM16: |
| case DataType::QASYMM16: |
| case DataType::BFLOAT16: |
| case DataType::F16: |
| return 2; |
| case DataType::F32: |
| case DataType::U32: |
| case DataType::S32: |
| return 4; |
| case DataType::F64: |
| case DataType::U64: |
| case DataType::S64: |
| return 8; |
| case DataType::SIZET: |
| return sizeof(size_t); |
| default: |
| ARM_COMPUTE_ERROR("Invalid data type"); |
| return 0; |
| } |
| } |
| |
| /** The size in bytes of the pixel format |
| * |
| * @param[in] format Input format |
| * |
| * @return The size in bytes of the pixel format |
| */ |
| inline size_t pixel_size_from_format(Format format) |
| { |
| switch(format) |
| { |
| case Format::U8: |
| return 1; |
| case Format::U16: |
| case Format::S16: |
| case Format::BFLOAT16: |
| case Format::F16: |
| case Format::UV88: |
| case Format::YUYV422: |
| case Format::UYVY422: |
| return 2; |
| case Format::RGB888: |
| return 3; |
| case Format::RGBA8888: |
| return 4; |
| case Format::U32: |
| case Format::S32: |
| case Format::F32: |
| return 4; |
| //Doesn't make sense for planar formats: |
| case Format::NV12: |
| case Format::NV21: |
| case Format::IYUV: |
| case Format::YUV444: |
| default: |
| ARM_COMPUTE_ERROR("Undefined pixel size for given format"); |
| return 0; |
| } |
| } |
| |
| /** The size in bytes of the data type |
| * |
| * @param[in] dt Input data type |
| * |
| * @return The size in bytes of the data type |
| */ |
| inline size_t element_size_from_data_type(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::S8: |
| case DataType::U8: |
| case DataType::QSYMM8: |
| case DataType::QASYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QSYMM8_PER_CHANNEL: |
| return 1; |
| case DataType::U16: |
| case DataType::S16: |
| case DataType::QSYMM16: |
| case DataType::QASYMM16: |
| case DataType::BFLOAT16: |
| case DataType::F16: |
| return 2; |
| case DataType::U32: |
| case DataType::S32: |
| case DataType::F32: |
| return 4; |
| default: |
| ARM_COMPUTE_ERROR("Undefined element size for given data type"); |
| return 0; |
| } |
| } |
| |
| /** Return the data type used by a given single-planar pixel format |
| * |
| * @param[in] format Input format |
| * |
| * @return The size in bytes of the pixel format |
| */ |
| inline DataType data_type_from_format(Format format) |
| { |
| switch(format) |
| { |
| case Format::U8: |
| case Format::UV88: |
| case Format::RGB888: |
| case Format::RGBA8888: |
| case Format::YUYV422: |
| case Format::UYVY422: |
| return DataType::U8; |
| case Format::U16: |
| return DataType::U16; |
| case Format::S16: |
| return DataType::S16; |
| case Format::U32: |
| return DataType::U32; |
| case Format::S32: |
| return DataType::S32; |
| case Format::BFLOAT16: |
| return DataType::BFLOAT16; |
| case Format::F16: |
| return DataType::F16; |
| case Format::F32: |
| return DataType::F32; |
| //Doesn't make sense for planar formats: |
| case Format::NV12: |
| case Format::NV21: |
| case Format::IYUV: |
| case Format::YUV444: |
| default: |
| ARM_COMPUTE_ERROR("Not supported data_type for given format"); |
| return DataType::UNKNOWN; |
| } |
| } |
| |
| /** Return the plane index of a given channel given an input format. |
| * |
| * @param[in] format Input format |
| * @param[in] channel Input channel |
| * |
| * @return The plane index of the specific channel of the specific format |
| */ |
| inline int plane_idx_from_channel(Format format, Channel channel) |
| { |
| switch(format) |
| { |
| // Single planar formats have a single plane |
| case Format::U8: |
| case Format::U16: |
| case Format::S16: |
| case Format::U32: |
| case Format::S32: |
| case Format::BFLOAT16: |
| case Format::F16: |
| case Format::F32: |
| case Format::UV88: |
| case Format::RGB888: |
| case Format::RGBA8888: |
| case Format::YUYV422: |
| case Format::UYVY422: |
| return 0; |
| // Multi planar formats |
| case Format::NV12: |
| case Format::NV21: |
| { |
| // Channel U and V share the same plane of format UV88 |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| case Channel::V: |
| return 1; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::IYUV: |
| case Format::YUV444: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| return 1; |
| case Channel::V: |
| return 2; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| default: |
| ARM_COMPUTE_ERROR("Not supported format"); |
| return 0; |
| } |
| } |
| |
| /** Return the channel index of a given channel given an input format. |
| * |
| * @param[in] format Input format |
| * @param[in] channel Input channel |
| * |
| * @return The channel index of the specific channel of the specific format |
| */ |
| inline int channel_idx_from_format(Format format, Channel channel) |
| { |
| switch(format) |
| { |
| case Format::RGB888: |
| { |
| switch(channel) |
| { |
| case Channel::R: |
| return 0; |
| case Channel::G: |
| return 1; |
| case Channel::B: |
| return 2; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::RGBA8888: |
| { |
| switch(channel) |
| { |
| case Channel::R: |
| return 0; |
| case Channel::G: |
| return 1; |
| case Channel::B: |
| return 2; |
| case Channel::A: |
| return 3; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::YUYV422: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| return 1; |
| case Channel::V: |
| return 3; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::UYVY422: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 1; |
| case Channel::U: |
| return 0; |
| case Channel::V: |
| return 2; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::NV12: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| return 0; |
| case Channel::V: |
| return 1; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::NV21: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| return 1; |
| case Channel::V: |
| return 0; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| case Format::YUV444: |
| case Format::IYUV: |
| { |
| switch(channel) |
| { |
| case Channel::Y: |
| return 0; |
| case Channel::U: |
| return 0; |
| case Channel::V: |
| return 0; |
| default: |
| ARM_COMPUTE_ERROR("Not supported channel"); |
| return 0; |
| } |
| } |
| default: |
| ARM_COMPUTE_ERROR("Not supported format"); |
| return 0; |
| } |
| } |
| |
| /** Return the number of planes for a given format |
| * |
| * @param[in] format Input format |
| * |
| * @return The number of planes for a given image format. |
| */ |
| inline size_t num_planes_from_format(Format format) |
| { |
| switch(format) |
| { |
| case Format::U8: |
| case Format::S16: |
| case Format::U16: |
| case Format::S32: |
| case Format::U32: |
| case Format::BFLOAT16: |
| case Format::F16: |
| case Format::F32: |
| case Format::RGB888: |
| case Format::RGBA8888: |
| case Format::YUYV422: |
| case Format::UYVY422: |
| return 1; |
| case Format::NV12: |
| case Format::NV21: |
| return 2; |
| case Format::IYUV: |
| case Format::YUV444: |
| return 3; |
| default: |
| ARM_COMPUTE_ERROR("Not supported format"); |
| return 0; |
| } |
| } |
| |
| /** Return the number of channels for a given single-planar pixel format |
| * |
| * @param[in] format Input format |
| * |
| * @return The number of channels for a given image format. |
| */ |
| inline size_t num_channels_from_format(Format format) |
| { |
| switch(format) |
| { |
| case Format::U8: |
| case Format::U16: |
| case Format::S16: |
| case Format::U32: |
| case Format::S32: |
| case Format::BFLOAT16: |
| case Format::F16: |
| case Format::F32: |
| return 1; |
| // Because the U and V channels are subsampled |
| // these formats appear like having only 2 channels: |
| case Format::YUYV422: |
| case Format::UYVY422: |
| return 2; |
| case Format::UV88: |
| return 2; |
| case Format::RGB888: |
| return 3; |
| case Format::RGBA8888: |
| return 4; |
| //Doesn't make sense for planar formats: |
| case Format::NV12: |
| case Format::NV21: |
| case Format::IYUV: |
| case Format::YUV444: |
| default: |
| return 0; |
| } |
| } |
| |
| /** Return the promoted data type of a given data type. |
| * |
| * @note If promoted data type is not supported an error will be thrown |
| * |
| * @param[in] dt Data type to get the promoted type of. |
| * |
| * @return Promoted data type |
| */ |
| inline DataType get_promoted_data_type(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::U8: |
| return DataType::U16; |
| case DataType::S8: |
| return DataType::S16; |
| case DataType::U16: |
| return DataType::U32; |
| case DataType::S16: |
| return DataType::S32; |
| case DataType::QSYMM8: |
| case DataType::QASYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QSYMM8_PER_CHANNEL: |
| case DataType::QSYMM16: |
| case DataType::QASYMM16: |
| case DataType::BFLOAT16: |
| case DataType::F16: |
| case DataType::U32: |
| case DataType::S32: |
| case DataType::F32: |
| ARM_COMPUTE_ERROR("Unsupported data type promotions!"); |
| default: |
| ARM_COMPUTE_ERROR("Undefined data type!"); |
| } |
| return DataType::UNKNOWN; |
| } |
| |
| /** Compute the mininum and maximum values a data type can take |
| * |
| * @param[in] dt Data type to get the min/max bounds of |
| * |
| * @return A tuple (min,max) with the minimum and maximum values respectively wrapped in PixelValue. |
| */ |
| inline std::tuple<PixelValue, PixelValue> get_min_max(DataType dt) |
| { |
| PixelValue min{}; |
| PixelValue max{}; |
| switch(dt) |
| { |
| case DataType::U8: |
| case DataType::QASYMM8: |
| { |
| min = PixelValue(static_cast<int32_t>(std::numeric_limits<uint8_t>::lowest())); |
| max = PixelValue(static_cast<int32_t>(std::numeric_limits<uint8_t>::max())); |
| break; |
| } |
| case DataType::S8: |
| case DataType::QSYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QSYMM8_PER_CHANNEL: |
| { |
| min = PixelValue(static_cast<int32_t>(std::numeric_limits<int8_t>::lowest())); |
| max = PixelValue(static_cast<int32_t>(std::numeric_limits<int8_t>::max())); |
| break; |
| } |
| case DataType::U16: |
| case DataType::QASYMM16: |
| { |
| min = PixelValue(static_cast<int32_t>(std::numeric_limits<uint16_t>::lowest())); |
| max = PixelValue(static_cast<int32_t>(std::numeric_limits<uint16_t>::max())); |
| break; |
| } |
| case DataType::S16: |
| case DataType::QSYMM16: |
| { |
| min = PixelValue(static_cast<int32_t>(std::numeric_limits<int16_t>::lowest())); |
| max = PixelValue(static_cast<int32_t>(std::numeric_limits<int16_t>::max())); |
| break; |
| } |
| case DataType::U32: |
| { |
| min = PixelValue(std::numeric_limits<uint32_t>::lowest()); |
| max = PixelValue(std::numeric_limits<uint32_t>::max()); |
| break; |
| } |
| case DataType::S32: |
| { |
| min = PixelValue(std::numeric_limits<int32_t>::lowest()); |
| max = PixelValue(std::numeric_limits<int32_t>::max()); |
| break; |
| } |
| case DataType::BFLOAT16: |
| { |
| min = PixelValue(bfloat16::lowest()); |
| max = PixelValue(bfloat16::max()); |
| break; |
| } |
| case DataType::F16: |
| { |
| min = PixelValue(std::numeric_limits<half>::lowest()); |
| max = PixelValue(std::numeric_limits<half>::max()); |
| break; |
| } |
| case DataType::F32: |
| { |
| min = PixelValue(std::numeric_limits<float>::lowest()); |
| max = PixelValue(std::numeric_limits<float>::max()); |
| break; |
| } |
| default: |
| ARM_COMPUTE_ERROR("Undefined data type!"); |
| } |
| return std::make_tuple(min, max); |
| } |
| |
| /** Return true if the given format has horizontal subsampling. |
| * |
| * @param[in] format Format to determine subsampling. |
| * |
| * @return True if the format can be subsampled horizontaly. |
| */ |
| inline bool has_format_horizontal_subsampling(Format format) |
| { |
| return (format == Format::YUYV422 || format == Format::UYVY422 || format == Format::NV12 || format == Format::NV21 || format == Format::IYUV || format == Format::UV88) ? true : false; |
| } |
| |
| /** Return true if the given format has vertical subsampling. |
| * |
| * @param[in] format Format to determine subsampling. |
| * |
| * @return True if the format can be subsampled verticaly. |
| */ |
| inline bool has_format_vertical_subsampling(Format format) |
| { |
| return (format == Format::NV12 || format == Format::NV21 || format == Format::IYUV || format == Format::UV88) ? true : false; |
| } |
| |
| /** Separate a 2D convolution into two 1D convolutions |
| * |
| * @param[in] conv 2D convolution |
| * @param[out] conv_col 1D vertical convolution |
| * @param[out] conv_row 1D horizontal convolution |
| * @param[in] size Size of the 2D convolution |
| * |
| * @return true if the separation was successful |
| */ |
| inline bool separate_matrix(const int16_t *conv, int16_t *conv_col, int16_t *conv_row, uint8_t size) |
| { |
| int32_t min_col = -1; |
| int16_t min_col_val = -1; |
| |
| for(int32_t i = 0; i < size; ++i) |
| { |
| if(conv[i] != 0 && (min_col < 0 || abs(min_col_val) > abs(conv[i]))) |
| { |
| min_col = i; |
| min_col_val = conv[i]; |
| } |
| } |
| |
| if(min_col < 0) |
| { |
| return false; |
| } |
| |
| for(uint32_t j = 0; j < size; ++j) |
| { |
| conv_col[j] = conv[min_col + j * size]; |
| } |
| |
| for(uint32_t i = 0; i < size; i++) |
| { |
| if(static_cast<int>(i) == min_col) |
| { |
| conv_row[i] = 1; |
| } |
| else |
| { |
| int16_t coeff = conv[i] / conv[min_col]; |
| |
| for(uint32_t j = 1; j < size; ++j) |
| { |
| if(conv[i + j * size] != (conv_col[j] * coeff)) |
| { |
| return false; |
| } |
| } |
| |
| conv_row[i] = coeff; |
| } |
| } |
| |
| return true; |
| } |
| |
| /** Calculate the scale of the given square matrix |
| * |
| * The scale is the absolute value of the sum of all the coefficients in the matrix. |
| * |
| * @note If the coefficients add up to 0 then the scale is set to 1. |
| * |
| * @param[in] matrix Matrix coefficients |
| * @param[in] matrix_size Number of elements per side of the square matrix. (Number of coefficients = matrix_size * matrix_size). |
| * |
| * @return The absolute value of the sum of the coefficients if they don't add up to 0, otherwise 1. |
| */ |
| inline uint32_t calculate_matrix_scale(const int16_t *matrix, unsigned int matrix_size) |
| { |
| const size_t size = matrix_size * matrix_size; |
| |
| return std::max(1, std::abs(std::accumulate(matrix, matrix + size, 0))); |
| } |
| |
| /** Adjust tensor shape size if width or height are odd for a given multi-planar format. No modification is done for other formats. |
| * |
| * @note Adding here a few links discussing the issue of odd size and sharing the same solution: |
| * <a href="https://android.googlesource.com/platform/frameworks/base/+/refs/heads/master/graphics/java/android/graphics/YuvImage.java">Android Source</a> |
| * <a href="https://groups.google.com/a/webmproject.org/forum/#!topic/webm-discuss/LaCKpqiDTXM">WebM</a> |
| * <a href="https://bugs.chromium.org/p/libyuv/issues/detail?id=198&can=1&q=odd%20width">libYUV</a> |
| * <a href="https://sourceforge.net/p/raw-yuvplayer/bugs/1/">YUVPlayer</a> * |
| * |
| * @param[in, out] shape Tensor shape of 2D size |
| * @param[in] format Format of the tensor |
| * |
| * @return The adjusted tensor shape. |
| */ |
| inline TensorShape adjust_odd_shape(const TensorShape &shape, Format format) |
| { |
| TensorShape output{ shape }; |
| |
| // Force width to be even for formats which require subsampling of the U and V channels |
| if(has_format_horizontal_subsampling(format)) |
| { |
| output.set(0, (output.x() + 1) & ~1U); |
| } |
| |
| // Force height to be even for formats which require subsampling of the U and V channels |
| if(has_format_vertical_subsampling(format)) |
| { |
| output.set(1, (output.y() + 1) & ~1U); |
| } |
| |
| return output; |
| } |
| |
| /** Calculate subsampled shape for a given format and channel |
| * |
| * @param[in] shape Shape of the tensor to calculate the extracted channel. |
| * @param[in] format Format of the tensor. |
| * @param[in] channel Channel to create tensor shape to be extracted. |
| * |
| * @return The subsampled tensor shape. |
| */ |
| inline TensorShape calculate_subsampled_shape(const TensorShape &shape, Format format, Channel channel = Channel::UNKNOWN) |
| { |
| TensorShape output{ shape }; |
| |
| // Subsample shape only for U or V channel |
| if(Channel::U == channel || Channel::V == channel || Channel::UNKNOWN == channel) |
| { |
| // Subsample width for the tensor shape when channel is U or V |
| if(has_format_horizontal_subsampling(format)) |
| { |
| output.set(0, output.x() / 2U); |
| } |
| |
| // Subsample height for the tensor shape when channel is U or V |
| if(has_format_vertical_subsampling(format)) |
| { |
| output.set(1, output.y() / 2U); |
| } |
| } |
| |
| return output; |
| } |
| |
| /** Calculate accurary required by the horizontal and vertical convolution computations |
| * |
| * @param[in] conv_col Pointer to the vertical vector of the separated convolution filter |
| * @param[in] conv_row Pointer to the horizontal vector of the convolution filter |
| * @param[in] size Number of elements per vector of the separated matrix |
| * |
| * @return The return type is a pair. The first element of the pair is the biggest data type needed for the first stage. The second |
| * element of the pair is the biggest data type needed for the second stage. |
| */ |
| inline std::pair<DataType, DataType> data_type_for_convolution(const int16_t *conv_col, const int16_t *conv_row, size_t size) |
| { |
| DataType first_stage = DataType::UNKNOWN; |
| DataType second_stage = DataType::UNKNOWN; |
| |
| auto gez = [](const int16_t &v) |
| { |
| return v >= 0; |
| }; |
| |
| auto accu_neg = [](const int &first, const int &second) |
| { |
| return first + (second < 0 ? second : 0); |
| }; |
| |
| auto accu_pos = [](const int &first, const int &second) |
| { |
| return first + (second > 0 ? second : 0); |
| }; |
| |
| const bool only_positive_coefficients = std::all_of(conv_row, conv_row + size, gez) && std::all_of(conv_col, conv_col + size, gez); |
| |
| if(only_positive_coefficients) |
| { |
| const int max_row_value = std::accumulate(conv_row, conv_row + size, 0) * UINT8_MAX; |
| const int max_value = std::accumulate(conv_col, conv_col + size, 0) * max_row_value; |
| |
| first_stage = (max_row_value <= UINT16_MAX) ? DataType::U16 : DataType::S32; |
| |
| second_stage = (max_value <= UINT16_MAX) ? DataType::U16 : DataType::S32; |
| } |
| else |
| { |
| const int min_row_value = std::accumulate(conv_row, conv_row + size, 0, accu_neg) * UINT8_MAX; |
| const int max_row_value = std::accumulate(conv_row, conv_row + size, 0, accu_pos) * UINT8_MAX; |
| const int neg_coeffs_sum = std::accumulate(conv_col, conv_col + size, 0, accu_neg); |
| const int pos_coeffs_sum = std::accumulate(conv_col, conv_col + size, 0, accu_pos); |
| const int min_value = neg_coeffs_sum * max_row_value + pos_coeffs_sum * min_row_value; |
| const int max_value = neg_coeffs_sum * min_row_value + pos_coeffs_sum * max_row_value; |
| |
| first_stage = ((INT16_MIN <= min_row_value) && (max_row_value <= INT16_MAX)) ? DataType::S16 : DataType::S32; |
| |
| second_stage = ((INT16_MIN <= min_value) && (max_value <= INT16_MAX)) ? DataType::S16 : DataType::S32; |
| } |
| |
| return std::make_pair(first_stage, second_stage); |
| } |
| |
| /** Calculate the accuracy required by the squared convolution calculation. |
| * |
| * |
| * @param[in] conv Pointer to the squared convolution matrix |
| * @param[in] size The total size of the convolution matrix |
| * |
| * @return The return is the biggest data type needed to do the convolution |
| */ |
| inline DataType data_type_for_convolution_matrix(const int16_t *conv, size_t size) |
| { |
| auto gez = [](const int16_t v) |
| { |
| return v >= 0; |
| }; |
| |
| const bool only_positive_coefficients = std::all_of(conv, conv + size, gez); |
| |
| if(only_positive_coefficients) |
| { |
| const int max_conv_value = std::accumulate(conv, conv + size, 0) * UINT8_MAX; |
| if(max_conv_value <= UINT16_MAX) |
| { |
| return DataType::U16; |
| } |
| else |
| { |
| return DataType::S32; |
| } |
| } |
| else |
| { |
| const int min_value = std::accumulate(conv, conv + size, 0, [](int a, int b) |
| { |
| return b < 0 ? a + b : a; |
| }) |
| * UINT8_MAX; |
| |
| const int max_value = std::accumulate(conv, conv + size, 0, [](int a, int b) |
| { |
| return b > 0 ? a + b : a; |
| }) |
| * UINT8_MAX; |
| |
| if((INT16_MIN <= min_value) && (INT16_MAX >= max_value)) |
| { |
| return DataType::S16; |
| } |
| else |
| { |
| return DataType::S32; |
| } |
| } |
| } |
| |
| /** Permutes the given dimensions according the permutation vector |
| * |
| * @param[in,out] dimensions Dimensions to be permuted. |
| * @param[in] perm Vector describing the permutation. |
| * |
| */ |
| template <typename T> |
| inline void permute_strides(Dimensions<T> &dimensions, const PermutationVector &perm) |
| { |
| const auto old_dim = utility::make_array<Dimensions<T>::num_max_dimensions>(dimensions.begin(), dimensions.end()); |
| for(unsigned int i = 0; i < perm.num_dimensions(); ++i) |
| { |
| T dimension_val = old_dim[i]; |
| dimensions.set(perm[i], dimension_val); |
| } |
| } |
| |
| /** Calculate padding requirements in case of SAME padding |
| * |
| * @param[in] input_shape Input shape |
| * @param[in] weights_shape Weights shape |
| * @param[in] conv_info Convolution information (containing strides) |
| * @param[in] data_layout (Optional) Data layout of the input and weights tensor |
| * @param[in] dilation (Optional) Dilation factor used in the convolution. |
| * @param[in] rounding_type (Optional) Dimension rounding type when down-scaling. |
| * |
| * @return PadStrideInfo for SAME padding |
| */ |
| PadStrideInfo calculate_same_pad(TensorShape input_shape, TensorShape weights_shape, PadStrideInfo conv_info, DataLayout data_layout = DataLayout::NCHW, const Size2D &dilation = Size2D(1u, 1u), |
| const DimensionRoundingType &rounding_type = DimensionRoundingType::FLOOR); |
| |
| /** Returns expected width and height of the deconvolution's output tensor. |
| * |
| * @param[in] in_width Width of input tensor (Number of columns) |
| * @param[in] in_height Height of input tensor (Number of rows) |
| * @param[in] kernel_width Kernel width. |
| * @param[in] kernel_height Kernel height. |
| * @param[in] pad_stride_info Pad and stride information. |
| * |
| * @return A pair with the new width in the first position and the new height in the second. |
| */ |
| std::pair<unsigned int, unsigned int> deconvolution_output_dimensions(unsigned int in_width, unsigned int in_height, |
| unsigned int kernel_width, unsigned int kernel_height, |
| const PadStrideInfo &pad_stride_info); |
| |
| /** Returns expected width and height of output scaled tensor depending on dimensions rounding mode. |
| * |
| * @param[in] width Width of input tensor (Number of columns) |
| * @param[in] height Height of input tensor (Number of rows) |
| * @param[in] kernel_width Kernel width. |
| * @param[in] kernel_height Kernel height. |
| * @param[in] pad_stride_info Pad and stride information. |
| * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1). |
| * |
| * @return A pair with the new width in the first position and the new height in the second. |
| */ |
| std::pair<unsigned int, unsigned int> scaled_dimensions(int width, int height, |
| int kernel_width, int kernel_height, |
| const PadStrideInfo &pad_stride_info, |
| const Size2D &dilation = Size2D(1U, 1U)); |
| |
| /** Check if the given reduction operation should be handled in a serial way. |
| * |
| * @param[in] op Reduction operation to perform |
| * @param[in] dt Data type |
| * @param[in] axis Axis along which to reduce |
| * |
| * @return True if the given reduction operation should be handled in a serial way. |
| */ |
| bool needs_serialized_reduction(ReductionOperation op, DataType dt, unsigned int axis); |
| |
| /** Returns output quantization information for softmax layer |
| * |
| * @param[in] input_type The data type of the input tensor |
| * @param[in] is_log True for log softmax |
| * |
| * @return Quantization information for the output tensor |
| */ |
| QuantizationInfo get_softmax_output_quantization_info(DataType input_type, bool is_log); |
| |
| /** Returns a pair of minimum and maximum values for a quantized activation |
| * |
| * @param[in] act_info The information for activation |
| * @param[in] data_type The used data type |
| * @param[in] oq_info The output quantization information |
| * |
| * @return The pair with minimum and maximum values |
| */ |
| std::pair<int32_t, int32_t> get_quantized_activation_min_max(ActivationLayerInfo act_info, DataType data_type, UniformQuantizationInfo oq_info); |
| |
| /** Convert a tensor format into a string. |
| * |
| * @param[in] format @ref Format to be translated to string. |
| * |
| * @return The string describing the format. |
| */ |
| const std::string &string_from_format(Format format); |
| |
| /** Convert a channel identity into a string. |
| * |
| * @param[in] channel @ref Channel to be translated to string. |
| * |
| * @return The string describing the channel. |
| */ |
| const std::string &string_from_channel(Channel channel); |
| /** Convert a data layout identity into a string. |
| * |
| * @param[in] dl @ref DataLayout to be translated to string. |
| * |
| * @return The string describing the data layout. |
| */ |
| const std::string &string_from_data_layout(DataLayout dl); |
| /** Convert a data type identity into a string. |
| * |
| * @param[in] dt @ref DataType to be translated to string. |
| * |
| * @return The string describing the data type. |
| */ |
| const std::string &string_from_data_type(DataType dt); |
| /** Convert a matrix pattern into a string. |
| * |
| * @param[in] pattern @ref MatrixPattern to be translated to string. |
| * |
| * @return The string describing the matrix pattern. |
| */ |
| const std::string &string_from_matrix_pattern(MatrixPattern pattern); |
| /** Translates a given activation function to a string. |
| * |
| * @param[in] act @ref ActivationLayerInfo::ActivationFunction to be translated to string. |
| * |
| * @return The string describing the activation function. |
| */ |
| const std::string &string_from_activation_func(ActivationLayerInfo::ActivationFunction act); |
| /** Translates a given non linear function to a string. |
| * |
| * @param[in] function @ref NonLinearFilterFunction to be translated to string. |
| * |
| * @return The string describing the non linear function. |
| */ |
| const std::string &string_from_non_linear_filter_function(NonLinearFilterFunction function); |
| /** Translates a given interpolation policy to a string. |
| * |
| * @param[in] policy @ref InterpolationPolicy to be translated to string. |
| * |
| * @return The string describing the interpolation policy. |
| */ |
| const std::string &string_from_interpolation_policy(InterpolationPolicy policy); |
| /** Translates a given border mode policy to a string. |
| * |
| * @param[in] border_mode @ref BorderMode to be translated to string. |
| * |
| * @return The string describing the border mode. |
| */ |
| const std::string &string_from_border_mode(BorderMode border_mode); |
| /** Translates a given normalization type to a string. |
| * |
| * @param[in] type @ref NormType to be translated to string. |
| * |
| * @return The string describing the normalization type. |
| */ |
| const std::string &string_from_norm_type(NormType type); |
| /** Translates a given pooling type to a string. |
| * |
| * @param[in] type @ref PoolingType to be translated to string. |
| * |
| * @return The string describing the pooling type. |
| */ |
| const std::string &string_from_pooling_type(PoolingType type); |
| /** Translates a given GEMMLowp output stage to a string. |
| * |
| * @param[in] output_stage @ref GEMMLowpOutputStageInfo to be translated to string. |
| * |
| * @return The string describing the GEMMLowp output stage |
| */ |
| const std::string &string_from_gemmlowp_output_stage(GEMMLowpOutputStageType output_stage); |
| /** Convert a PixelValue to a string, represented through the specific data type |
| * |
| * @param[in] value The PixelValue to convert |
| * @param[in] data_type The type to be used to convert the @p value |
| * |
| * @return String representation of the PixelValue through the given data type. |
| */ |
| std::string string_from_pixel_value(const PixelValue &value, const DataType data_type); |
| /** Convert a string to DataType |
| * |
| * @param[in] name The name of the data type |
| * |
| * @return DataType |
| */ |
| DataType data_type_from_name(const std::string &name); |
| /** Stores padding information before configuring a kernel |
| * |
| * @param[in] infos list of tensor infos to store the padding info for |
| * |
| * @return An unordered map where each tensor info pointer is paired with its original padding info |
| */ |
| std::unordered_map<const ITensorInfo *, PaddingSize> get_padding_info(std::initializer_list<const ITensorInfo *> infos); |
| /** Stores padding information before configuring a kernel |
| * |
| * @param[in] tensors list of tensors to store the padding info for |
| * |
| * @return An unordered map where each tensor info pointer is paired with its original padding info |
| */ |
| std::unordered_map<const ITensorInfo *, PaddingSize> get_padding_info(std::initializer_list<const ITensor *> tensors); |
| /** Check if the previously stored padding info has changed after configuring a kernel |
| * |
| * @param[in] padding_map an unordered map where each tensor info pointer is paired with its original padding info |
| * |
| * @return true if any of the tensor infos has changed its paddings |
| */ |
| bool has_padding_changed(const std::unordered_map<const ITensorInfo *, PaddingSize> &padding_map); |
| |
| /** Input Stream operator for @ref DataType |
| * |
| * @param[in] stream Stream to parse |
| * @param[out] data_type Output data type |
| * |
| * @return Updated stream |
| */ |
| inline ::std::istream &operator>>(::std::istream &stream, DataType &data_type) |
| { |
| std::string value; |
| stream >> value; |
| data_type = data_type_from_name(value); |
| return stream; |
| } |
| /** Lower a given string. |
| * |
| * @param[in] val Given string to lower. |
| * |
| * @return The lowered string |
| */ |
| std::string lower_string(const std::string &val); |
| |
| /** Check if a given data type is of floating point type |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of floating point type, else false. |
| */ |
| inline bool is_data_type_float(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::F16: |
| case DataType::F32: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Check if a given data type is of quantized type |
| * |
| * @note Quantized is considered a super-set of fixed-point and asymmetric data types. |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of quantized type, else false. |
| */ |
| inline bool is_data_type_quantized(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::QSYMM8: |
| case DataType::QASYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QSYMM8_PER_CHANNEL: |
| case DataType::QSYMM16: |
| case DataType::QASYMM16: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Check if a given data type is of asymmetric quantized type |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of asymmetric quantized type, else false. |
| */ |
| inline bool is_data_type_quantized_asymmetric(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::QASYMM8: |
| case DataType::QASYMM8_SIGNED: |
| case DataType::QASYMM16: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Check if a given data type is of asymmetric quantized signed type |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of asymmetric quantized signed type, else false. |
| */ |
| inline bool is_data_type_quantized_asymmetric_signed(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::QASYMM8_SIGNED: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Check if a given data type is of symmetric quantized type |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of symmetric quantized type, else false. |
| */ |
| inline bool is_data_type_quantized_symmetric(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::QSYMM8: |
| case DataType::QSYMM8_PER_CHANNEL: |
| case DataType::QSYMM16: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Check if a given data type is of per channel type |
| * |
| * @param[in] dt Input data type. |
| * |
| * @return True if data type is of per channel type, else false. |
| */ |
| inline bool is_data_type_quantized_per_channel(DataType dt) |
| { |
| switch(dt) |
| { |
| case DataType::QSYMM8_PER_CHANNEL: |
| return true; |
| default: |
| return false; |
| } |
| } |
| |
| /** Create a string with the float in full precision. |
| * |
| * @param val Floating point value |
| * |
| * @return String with the floating point value. |
| */ |
| inline std::string float_to_string_with_full_precision(float val) |
| { |
| std::stringstream ss; |
| ss.precision(std::numeric_limits<float>::max_digits10); |
| ss << val; |
| |
| if(val != static_cast<int>(val)) |
| { |
| ss << "f"; |
| } |
| |
| return ss.str(); |
| } |
| |
| /** Returns the number of elements required to go from start to end with the wanted step |
| * |
| * @param[in] start start value |
| * @param[in] end end value |
| * @param[in] step step value between each number in the wanted sequence |
| * |
| * @return number of elements to go from start value to end value using the wanted step |
| */ |
| inline size_t num_of_elements_in_range(const float start, const float end, const float step) |
| { |
| ARM_COMPUTE_ERROR_ON_MSG(step == 0, "Range Step cannot be 0"); |
| return size_t(std::ceil((end - start) / step)); |
| } |
| |
| /** Returns true if the value can be represented by the given data type |
| * |
| * @param[in] val value to be checked |
| * @param[in] dt data type that is checked |
| * @param[in] qinfo (Optional) quantization info if the data type is QASYMM8 |
| * |
| * @return true if the data type can hold the value. |
| */ |
| template <typename T> |
| bool check_value_range(T val, DataType dt, QuantizationInfo qinfo = QuantizationInfo()) |
| { |
| switch(dt) |
| { |
| case DataType::U8: |
| { |
| const auto val_u8 = static_cast<uint8_t>(val); |
| return ((val_u8 == val) && val_u8 >= std::numeric_limits<uint8_t>::lowest() && val_u8 <= std::numeric_limits<uint8_t>::max()); |
| } |
| case DataType::QASYMM8: |
| { |
| double min = static_cast<double>(dequantize_qasymm8(0, qinfo)); |
| double max = static_cast<double>(dequantize_qasymm8(std::numeric_limits<uint8_t>::max(), qinfo)); |
| return ((double)val >= min && (double)val <= max); |
| } |
| case DataType::S8: |
| { |
| const auto val_s8 = static_cast<int8_t>(val); |
| return ((val_s8 == val) && val_s8 >= std::numeric_limits<int8_t>::lowest() && val_s8 <= std::numeric_limits<int8_t>::max()); |
| } |
| case DataType::U16: |
| { |
| const auto val_u16 = static_cast<uint16_t>(val); |
| return ((val_u16 == val) && val_u16 >= std::numeric_limits<uint16_t>::lowest() && val_u16 <= std::numeric_limits<uint16_t>::max()); |
| } |
| case DataType::S16: |
| { |
| const auto val_s16 = static_cast<int16_t>(val); |
| return ((val_s16 == val) && val_s16 >= std::numeric_limits<int16_t>::lowest() && val_s16 <= std::numeric_limits<int16_t>::max()); |
| } |
| case DataType::U32: |
| { |
| const auto val_u32 = static_cast<uint32_t>(val); |
| return ((val_u32 == val) && val_u32 >= std::numeric_limits<uint32_t>::lowest() && val_u32 <= std::numeric_limits<uint32_t>::max()); |
| } |
| case DataType::S32: |
| { |
| const auto val_s32 = static_cast<int32_t>(val); |
| return ((val_s32 == val) && val_s32 >= std::numeric_limits<int32_t>::lowest() && val_s32 <= std::numeric_limits<int32_t>::max()); |
| } |
| case DataType::BFLOAT16: |
| return (val >= bfloat16::lowest() && val <= bfloat16::max()); |
| case DataType::F16: |
| return (val >= std::numeric_limits<half>::lowest() && val <= std::numeric_limits<half>::max()); |
| case DataType::F32: |
| return (val >= std::numeric_limits<float>::lowest() && val <= std::numeric_limits<float>::max()); |
| default: |
| ARM_COMPUTE_ERROR("Data type not supported"); |
| return false; |
| } |
| } |
| |
| /** Returns the adjusted vector size in case it is less than the input's first dimension, getting rounded down to its closest valid vector size |
| * |
| * @param[in] vec_size vector size to be adjusted |
| * @param[in] dim0 size of the first dimension |
| * |
| * @return the number of element processed along the X axis per thread |
| */ |
| inline unsigned int adjust_vec_size(unsigned int vec_size, size_t dim0) |
| { |
| ARM_COMPUTE_ERROR_ON(vec_size > 16); |
| |
| if((vec_size >= dim0) && (dim0 == 3)) |
| { |
| return dim0; |
| } |
| |
| while(vec_size > dim0) |
| { |
| vec_size >>= 1; |
| } |
| |
| return vec_size; |
| } |
| |
| #ifdef ARM_COMPUTE_ASSERTS_ENABLED |
| /** Print consecutive elements to an output stream. |
| * |
| * @param[out] s Output stream to print the elements to. |
| * @param[in] ptr Pointer to print the elements from. |
| * @param[in] n Number of elements to print. |
| * @param[in] stream_width (Optional) Width of the stream. If set to 0 the element's width is used. Defaults to 0. |
| * @param[in] element_delim (Optional) Delimeter among the consecutive elements. Defaults to space delimeter |
| */ |
| template <typename T> |
| void print_consecutive_elements_impl(std::ostream &s, const T *ptr, unsigned int n, int stream_width = 0, const std::string &element_delim = " ") |
| { |
| using print_type = typename std::conditional<std::is_floating_point<T>::value, T, int>::type; |
| std::ios stream_status(nullptr); |
| stream_status.copyfmt(s); |
| |
| for(unsigned int i = 0; i < n; ++i) |
| { |
| // Set stream width as it is not a "sticky" stream manipulator |
| if(stream_width != 0) |
| { |
| s.width(stream_width); |
| } |
| |
| if(std::is_same<typename std::decay<T>::type, half>::value) |
| { |
| // We use T instead of print_type here is because the std::is_floating_point<half> returns false and then the print_type becomes int. |
| s << std::right << static_cast<T>(ptr[i]) << element_delim; |
| } |
| else if(std::is_same<typename std::decay<T>::type, bfloat16>::value) |
| { |
| // We use T instead of print_type here is because the std::is_floating_point<bfloat16> returns false and then the print_type becomes int. |
| s << std::right << float(ptr[i]) << element_delim; |
| } |
| else |
| { |
| s << std::right << static_cast<print_type>(ptr[i]) << element_delim; |
| } |
| } |
| |
| // Restore output stream flags |
| s.copyfmt(stream_status); |
| } |
| |
| /** Identify the maximum width of n consecutive elements. |
| * |
| * @param[in] s The output stream which will be used to print the elements. Used to extract the stream format. |
| * @param[in] ptr Pointer to the elements. |
| * @param[in] n Number of elements. |
| * |
| * @return The maximum width of the elements. |
| */ |
| template <typename T> |
| int max_consecutive_elements_display_width_impl(std::ostream &s, const T *ptr, unsigned int n) |
| { |
| using print_type = typename std::conditional<std::is_floating_point<T>::value, T, int>::type; |
| |
| int max_width = -1; |
| for(unsigned int i = 0; i < n; ++i) |
| { |
| std::stringstream ss; |
| ss.copyfmt(s); |
| |
| if(std::is_same<typename std::decay<T>::type, half>::value) |
| { |
| // We use T instead of print_type here is because the std::is_floating_point<half> returns false and then the print_type becomes int. |
| ss << static_cast<T>(ptr[i]); |
| } |
| else if(std::is_same<typename std::decay<T>::type, bfloat16>::value) |
| { |
| // We use T instead of print_type here is because the std::is_floating_point<bfloat> returns false and then the print_type becomes int. |
| ss << float(ptr[i]); |
| } |
| else |
| { |
| ss << static_cast<print_type>(ptr[i]); |
| } |
| |
| max_width = std::max<int>(max_width, ss.str().size()); |
| } |
| return max_width; |
| } |
| |
| /** Print consecutive elements to an output stream. |
| * |
| * @param[out] s Output stream to print the elements to. |
| * @param[in] dt Data type of the elements |
| * @param[in] ptr Pointer to print the elements from. |
| * @param[in] n Number of elements to print. |
| * @param[in] stream_width (Optional) Width of the stream. If set to 0 the element's width is used. Defaults to 0. |
| * @param[in] element_delim (Optional) Delimeter among the consecutive elements. Defaults to space delimeter |
| */ |
| void print_consecutive_elements(std::ostream &s, DataType dt, const uint8_t *ptr, unsigned int n, int stream_width, const std::string &element_delim = " "); |
| |
| /** Identify the maximum width of n consecutive elements. |
| * |
| * @param[in] s Output stream to print the elements to. |
| * @param[in] dt Data type of the elements |
| * @param[in] ptr Pointer to print the elements from. |
| * @param[in] n Number of elements to print. |
| * |
| * @return The maximum width of the elements. |
| */ |
| int max_consecutive_elements_display_width(std::ostream &s, DataType dt, const uint8_t *ptr, unsigned int n); |
| #endif /* ARM_COMPUTE_ASSERTS_ENABLED */ |
| } |
| #endif /*ARM_COMPUTE_UTILS_H */ |