benibus commented on code in PR #36073: URL: https://github.com/apache/arrow/pull/36073#discussion_r1234470603
########## cpp/src/arrow/util/float16.h: ########## @@ -0,0 +1,179 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +#pragma once + +#include <array> +#include <cmath> +#include <cstdint> +#include <cstring> +#include <limits> +#include <type_traits> + +#include "arrow/util/bit_util.h" +#include "arrow/util/ubsan.h" +#include "arrow/util/visibility.h" + +namespace arrow { +namespace util { + +/// \brief Base class for an IEEE half-precision float, encoded as a `uint16_t` +/// +/// The exact format is as follows (from MSB to LSB): +/// - bit 0: sign +/// - bits 1-5: exponent +/// - bits 6-15: mantissa +/// +/// NOTE: Methods in the class should not mutate the unerlying value or produce copies. +/// Such functionality is delegated to subclasses. +class Float16Base { + public: + Float16Base() = default; + constexpr explicit Float16Base(uint16_t value) : value_(value) {} + + constexpr uint16_t bits() const { return value_; } + constexpr explicit operator uint16_t() const { return bits(); } + + constexpr bool signbit() const { return (value_ & 0x8000) != 0; } + + constexpr bool is_nan() const { + return (value_ & 0x7c00) == 0x7c00 && (value_ & 0x03ff) != 0; + } + constexpr bool is_infinity() const { return (value_ & 0x7fff) == 0x7c00; } + constexpr bool is_zero() const { return (value_ & 0x7fff) == 0; } + + /// \brief Copy the value's bytes in native-endian byte order + void ToBytes(uint8_t* dest) const { std::memcpy(dest, &value_, sizeof(value_)); } + /// \brief Return the value's bytes in native-endian byte order + std::array<uint8_t, 2> ToBytes() const { + std::array<uint8_t, 2> bytes; + ToBytes(bytes.data()); + return bytes; + } + + void ToLittleEndian(uint8_t* dest) const { + Float16Base{bit_util::ToLittleEndian(value_)}.ToBytes(dest); + } + std::array<uint8_t, 2> ToLittleEndian() const { + return Float16Base{bit_util::ToLittleEndian(value_)}.ToBytes(); + } + + void ToBigEndian(uint8_t* dest) const { Review Comment: Not sure how generally useful it would be either. I only really added it for symmetry given that the little-endian requirement is Parquet-specific. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
