//===- Endian.h - Utilities for IO with endian specific data ----*- C++ -*-===//
 
//
 
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 
// See https://llvm.org/LICENSE.txt for license information.
 
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
//
 
//===----------------------------------------------------------------------===//
 
//
 
// This file declares generic functions to read and write endian specific data.
 
//
 
//===----------------------------------------------------------------------===//
 
 
 
#ifndef LLVM_SUPPORT_ENDIAN_H
 
#define LLVM_SUPPORT_ENDIAN_H
 
 
 
#include "llvm/Support/Compiler.h"
 
#include "llvm/Support/SwapByteOrder.h"
 
#include <cassert>
 
#include <cstddef>
 
#include <cstdint>
 
#include <cstring>
 
#include <type_traits>
 
 
 
namespace llvm {
 
namespace support {
 
 
 
enum endianness {big, little, native};
 
 
 
// These are named values for common alignments.
 
enum {aligned = 0, unaligned = 1};
 
 
 
namespace detail {
 
 
 
/// ::value is either alignment, or alignof(T) if alignment is 0.
 
template<class T, int alignment>
 
struct PickAlignment {
 
 enum { value = alignment == 0 ? alignof(T) : alignment };
 
};
 
 
 
} // end namespace detail
 
 
 
namespace endian {
 
 
 
constexpr endianness system_endianness() {
 
  return sys::IsBigEndianHost ? big : little;
 
}
 
 
 
template <typename value_type>
 
inline value_type byte_swap(value_type value, endianness endian) {
 
  if ((endian != native) && (endian != system_endianness()))
 
    sys::swapByteOrder(value);
 
  return value;
 
}
 
 
 
/// Swap the bytes of value to match the given endianness.
 
template<typename value_type, endianness endian>
 
inline value_type byte_swap(value_type value) {
 
  return byte_swap(value, endian);
 
}
 
 
 
/// Read a value of a particular endianness from memory.
 
template <typename value_type, std::size_t alignment>
 
inline value_type read(const void *memory, endianness endian) {
 
  value_type ret;
 
 
 
  memcpy(&ret,
 
         LLVM_ASSUME_ALIGNED(
 
             memory, (detail::PickAlignment<value_type, alignment>::value)),
 
         sizeof(value_type));
 
  return byte_swap<value_type>(ret, endian);
 
}
 
 
 
template<typename value_type,
 
         endianness endian,
 
         std::size_t alignment>
 
inline value_type read(const void *memory) {
 
  return read<value_type, alignment>(memory, endian);
 
}
 
 
 
/// Read a value of a particular endianness from a buffer, and increment the
 
/// buffer past that value.
 
template <typename value_type, std::size_t alignment, typename CharT>
 
inline value_type readNext(const CharT *&memory, endianness endian) {
 
  value_type ret = read<value_type, alignment>(memory, endian);
 
  memory += sizeof(value_type);
 
  return ret;
 
}
 
 
 
template<typename value_type, endianness endian, std::size_t alignment,
 
         typename CharT>
 
inline value_type readNext(const CharT *&memory) {
 
  return readNext<value_type, alignment, CharT>(memory, endian);
 
}
 
 
 
/// Write a value to memory with a particular endianness.
 
template <typename value_type, std::size_t alignment>
 
inline void write(void *memory, value_type value, endianness endian) {
 
  value = byte_swap<value_type>(value, endian);
 
  memcpy(LLVM_ASSUME_ALIGNED(
 
             memory, (detail::PickAlignment<value_type, alignment>::value)),
 
         &value, sizeof(value_type));
 
}
 
 
 
template<typename value_type,
 
         endianness endian,
 
         std::size_t alignment>
 
inline void write(void *memory, value_type value) {
 
  write<value_type, alignment>(memory, value, endian);
 
}
 
 
 
template <typename value_type>
 
using make_unsigned_t = std::make_unsigned_t<value_type>;
 
 
 
/// Read a value of a particular endianness from memory, for a location
 
/// that starts at the given bit offset within the first byte.
 
template <typename value_type, endianness endian, std::size_t alignment>
 
inline value_type readAtBitAlignment(const void *memory, uint64_t startBit) {
 
  assert(startBit < 8);
 
  if (startBit == 0)
 
    return read<value_type, endian, alignment>(memory);
 
  else {
 
    // Read two values and compose the result from them.
 
    value_type val[2];
 
    memcpy(&val[0],
 
           LLVM_ASSUME_ALIGNED(
 
               memory, (detail::PickAlignment<value_type, alignment>::value)),
 
           sizeof(value_type) * 2);
 
    val[0] = byte_swap<value_type, endian>(val[0]);
 
    val[1] = byte_swap<value_type, endian>(val[1]);
 
 
 
    // Shift bits from the lower value into place.
 
    make_unsigned_t<value_type> lowerVal = val[0] >> startBit;
 
    // Mask off upper bits after right shift in case of signed type.
 
    make_unsigned_t<value_type> numBitsFirstVal =
 
        (sizeof(value_type) * 8) - startBit;
 
    lowerVal &= ((make_unsigned_t<value_type>)1 << numBitsFirstVal) - 1;
 
 
 
    // Get the bits from the upper value.
 
    make_unsigned_t<value_type> upperVal =
 
        val[1] & (((make_unsigned_t<value_type>)1 << startBit) - 1);
 
    // Shift them in to place.
 
    upperVal <<= numBitsFirstVal;
 
 
 
    return lowerVal | upperVal;
 
  }
 
}
 
 
 
/// Write a value to memory with a particular endianness, for a location
 
/// that starts at the given bit offset within the first byte.
 
template <typename value_type, endianness endian, std::size_t alignment>
 
inline void writeAtBitAlignment(void *memory, value_type value,
 
                                uint64_t startBit) {
 
  assert(startBit < 8);
 
  if (startBit == 0)
 
    write<value_type, endian, alignment>(memory, value);
 
  else {
 
    // Read two values and shift the result into them.
 
    value_type val[2];
 
    memcpy(&val[0],
 
           LLVM_ASSUME_ALIGNED(
 
               memory, (detail::PickAlignment<value_type, alignment>::value)),
 
           sizeof(value_type) * 2);
 
    val[0] = byte_swap<value_type, endian>(val[0]);
 
    val[1] = byte_swap<value_type, endian>(val[1]);
 
 
 
    // Mask off any existing bits in the upper part of the lower value that
 
    // we want to replace.
 
    val[0] &= ((make_unsigned_t<value_type>)1 << startBit) - 1;
 
    make_unsigned_t<value_type> numBitsFirstVal =
 
        (sizeof(value_type) * 8) - startBit;
 
    make_unsigned_t<value_type> lowerVal = value;
 
    if (startBit > 0) {
 
      // Mask off the upper bits in the new value that are not going to go into
 
      // the lower value. This avoids a left shift of a negative value, which
 
      // is undefined behavior.
 
      lowerVal &= (((make_unsigned_t<value_type>)1 << numBitsFirstVal) - 1);
 
      // Now shift the new bits into place
 
      lowerVal <<= startBit;
 
    }
 
    val[0] |= lowerVal;
 
 
 
    // Mask off any existing bits in the lower part of the upper value that
 
    // we want to replace.
 
    val[1] &= ~(((make_unsigned_t<value_type>)1 << startBit) - 1);
 
    // Next shift the bits that go into the upper value into position.
 
    make_unsigned_t<value_type> upperVal = value >> numBitsFirstVal;
 
    // Mask off upper bits after right shift in case of signed type.
 
    upperVal &= ((make_unsigned_t<value_type>)1 << startBit) - 1;
 
    val[1] |= upperVal;
 
 
 
    // Finally, rewrite values.
 
    val[0] = byte_swap<value_type, endian>(val[0]);
 
    val[1] = byte_swap<value_type, endian>(val[1]);
 
    memcpy(LLVM_ASSUME_ALIGNED(
 
               memory, (detail::PickAlignment<value_type, alignment>::value)),
 
           &val[0], sizeof(value_type) * 2);
 
  }
 
}
 
 
 
} // end namespace endian
 
 
 
namespace detail {
 
 
 
template <typename ValueType, endianness Endian, std::size_t Alignment,
 
          std::size_t ALIGN = PickAlignment<ValueType, Alignment>::value>
 
struct packed_endian_specific_integral {
 
  using value_type = ValueType;
 
  static constexpr endianness endian = Endian;
 
  static constexpr std::size_t alignment = Alignment;
 
 
 
  packed_endian_specific_integral() = default;
 
 
 
  explicit packed_endian_specific_integral(value_type val) { *this = val; }
 
 
 
  operator value_type() const {
 
    return endian::read<value_type, endian, alignment>(
 
      (const void*)Value.buffer);
 
  }
 
 
 
  void operator=(value_type newValue) {
 
    endian::write<value_type, endian, alignment>(
 
      (void*)Value.buffer, newValue);
 
  }
 
 
 
  packed_endian_specific_integral &operator+=(value_type newValue) {
 
    *this = *this + newValue;
 
    return *this;
 
  }
 
 
 
  packed_endian_specific_integral &operator-=(value_type newValue) {
 
    *this = *this - newValue;
 
    return *this;
 
  }
 
 
 
  packed_endian_specific_integral &operator|=(value_type newValue) {
 
    *this = *this | newValue;
 
    return *this;
 
  }
 
 
 
  packed_endian_specific_integral &operator&=(value_type newValue) {
 
    *this = *this & newValue;
 
    return *this;
 
  }
 
 
 
private:
 
  struct {
 
    alignas(ALIGN) char buffer[sizeof(value_type)];
 
  } Value;
 
 
 
public:
 
  struct ref {
 
    explicit ref(void *Ptr) : Ptr(Ptr) {}
 
 
 
    operator value_type() const {
 
      return endian::read<value_type, endian, alignment>(Ptr);
 
    }
 
 
 
    void operator=(value_type NewValue) {
 
      endian::write<value_type, endian, alignment>(Ptr, NewValue);
 
    }
 
 
 
  private:
 
    void *Ptr;
 
  };
 
};
 
 
 
} // end namespace detail
 
 
 
using ulittle16_t =
 
    detail::packed_endian_specific_integral<uint16_t, little, unaligned>;
 
using ulittle32_t =
 
    detail::packed_endian_specific_integral<uint32_t, little, unaligned>;
 
using ulittle64_t =
 
    detail::packed_endian_specific_integral<uint64_t, little, unaligned>;
 
 
 
using little16_t =
 
    detail::packed_endian_specific_integral<int16_t, little, unaligned>;
 
using little32_t =
 
    detail::packed_endian_specific_integral<int32_t, little, unaligned>;
 
using little64_t =
 
    detail::packed_endian_specific_integral<int64_t, little, unaligned>;
 
 
 
using aligned_ulittle16_t =
 
    detail::packed_endian_specific_integral<uint16_t, little, aligned>;
 
using aligned_ulittle32_t =
 
    detail::packed_endian_specific_integral<uint32_t, little, aligned>;
 
using aligned_ulittle64_t =
 
    detail::packed_endian_specific_integral<uint64_t, little, aligned>;
 
 
 
using aligned_little16_t =
 
    detail::packed_endian_specific_integral<int16_t, little, aligned>;
 
using aligned_little32_t =
 
    detail::packed_endian_specific_integral<int32_t, little, aligned>;
 
using aligned_little64_t =
 
    detail::packed_endian_specific_integral<int64_t, little, aligned>;
 
 
 
using ubig16_t =
 
    detail::packed_endian_specific_integral<uint16_t, big, unaligned>;
 
using ubig32_t =
 
    detail::packed_endian_specific_integral<uint32_t, big, unaligned>;
 
using ubig64_t =
 
    detail::packed_endian_specific_integral<uint64_t, big, unaligned>;
 
 
 
using big16_t =
 
    detail::packed_endian_specific_integral<int16_t, big, unaligned>;
 
using big32_t =
 
    detail::packed_endian_specific_integral<int32_t, big, unaligned>;
 
using big64_t =
 
    detail::packed_endian_specific_integral<int64_t, big, unaligned>;
 
 
 
using aligned_ubig16_t =
 
    detail::packed_endian_specific_integral<uint16_t, big, aligned>;
 
using aligned_ubig32_t =
 
    detail::packed_endian_specific_integral<uint32_t, big, aligned>;
 
using aligned_ubig64_t =
 
    detail::packed_endian_specific_integral<uint64_t, big, aligned>;
 
 
 
using aligned_big16_t =
 
    detail::packed_endian_specific_integral<int16_t, big, aligned>;
 
using aligned_big32_t =
 
    detail::packed_endian_specific_integral<int32_t, big, aligned>;
 
using aligned_big64_t =
 
    detail::packed_endian_specific_integral<int64_t, big, aligned>;
 
 
 
using unaligned_uint16_t =
 
    detail::packed_endian_specific_integral<uint16_t, native, unaligned>;
 
using unaligned_uint32_t =
 
    detail::packed_endian_specific_integral<uint32_t, native, unaligned>;
 
using unaligned_uint64_t =
 
    detail::packed_endian_specific_integral<uint64_t, native, unaligned>;
 
 
 
using unaligned_int16_t =
 
    detail::packed_endian_specific_integral<int16_t, native, unaligned>;
 
using unaligned_int32_t =
 
    detail::packed_endian_specific_integral<int32_t, native, unaligned>;
 
using unaligned_int64_t =
 
    detail::packed_endian_specific_integral<int64_t, native, unaligned>;
 
 
 
template <typename T>
 
using little_t = detail::packed_endian_specific_integral<T, little, unaligned>;
 
template <typename T>
 
using big_t = detail::packed_endian_specific_integral<T, big, unaligned>;
 
 
 
template <typename T>
 
using aligned_little_t =
 
    detail::packed_endian_specific_integral<T, little, aligned>;
 
template <typename T>
 
using aligned_big_t = detail::packed_endian_specific_integral<T, big, aligned>;
 
 
 
namespace endian {
 
 
 
template <typename T> inline T read(const void *P, endianness E) {
 
  return read<T, unaligned>(P, E);
 
}
 
 
 
template <typename T, endianness E> inline T read(const void *P) {
 
  return *(const detail::packed_endian_specific_integral<T, E, unaligned> *)P;
 
}
 
 
 
inline uint16_t read16(const void *P, endianness E) {
 
  return read<uint16_t>(P, E);
 
}
 
inline uint32_t read32(const void *P, endianness E) {
 
  return read<uint32_t>(P, E);
 
}
 
inline uint64_t read64(const void *P, endianness E) {
 
  return read<uint64_t>(P, E);
 
}
 
 
 
template <endianness E> inline uint16_t read16(const void *P) {
 
  return read<uint16_t, E>(P);
 
}
 
template <endianness E> inline uint32_t read32(const void *P) {
 
  return read<uint32_t, E>(P);
 
}
 
template <endianness E> inline uint64_t read64(const void *P) {
 
  return read<uint64_t, E>(P);
 
}
 
 
 
inline uint16_t read16le(const void *P) { return read16<little>(P); }
 
inline uint32_t read32le(const void *P) { return read32<little>(P); }
 
inline uint64_t read64le(const void *P) { return read64<little>(P); }
 
inline uint16_t read16be(const void *P) { return read16<big>(P); }
 
inline uint32_t read32be(const void *P) { return read32<big>(P); }
 
inline uint64_t read64be(const void *P) { return read64<big>(P); }
 
 
 
template <typename T> inline void write(void *P, T V, endianness E) {
 
  write<T, unaligned>(P, V, E);
 
}
 
 
 
template <typename T, endianness E> inline void write(void *P, T V) {
 
  *(detail::packed_endian_specific_integral<T, E, unaligned> *)P = V;
 
}
 
 
 
inline void write16(void *P, uint16_t V, endianness E) {
 
  write<uint16_t>(P, V, E);
 
}
 
inline void write32(void *P, uint32_t V, endianness E) {
 
  write<uint32_t>(P, V, E);
 
}
 
inline void write64(void *P, uint64_t V, endianness E) {
 
  write<uint64_t>(P, V, E);
 
}
 
 
 
template <endianness E> inline void write16(void *P, uint16_t V) {
 
  write<uint16_t, E>(P, V);
 
}
 
template <endianness E> inline void write32(void *P, uint32_t V) {
 
  write<uint32_t, E>(P, V);
 
}
 
template <endianness E> inline void write64(void *P, uint64_t V) {
 
  write<uint64_t, E>(P, V);
 
}
 
 
 
inline void write16le(void *P, uint16_t V) { write16<little>(P, V); }
 
inline void write32le(void *P, uint32_t V) { write32<little>(P, V); }
 
inline void write64le(void *P, uint64_t V) { write64<little>(P, V); }
 
inline void write16be(void *P, uint16_t V) { write16<big>(P, V); }
 
inline void write32be(void *P, uint32_t V) { write32<big>(P, V); }
 
inline void write64be(void *P, uint64_t V) { write64<big>(P, V); }
 
 
 
} // end namespace endian
 
 
 
} // end namespace support
 
} // end namespace llvm
 
 
 
#endif // LLVM_SUPPORT_ENDIAN_H