1
0
Fork 0
forked from suyu/suyu
suyu/src/common/swap.h
Lioncash 0d8ef2d3b9 common/swap: Improve codegen of the default swap fallbacks
Uses arithmetic that can be identified more trivially by compilers for
optimizations. e.g. Rather than shifting the halves of the value and
then swapping and combining them, we can swap them in place.

e.g. for the original swap32 code on x86-64, clang 8.0 would generate:

    mov     ecx, edi
    rol     cx, 8
    shl     ecx, 16
    shr     edi, 16
    rol     di, 8
    movzx   eax, di
    or      eax, ecx
    ret

while GCC 8.3 would generate the ideal:

    mov     eax, edi
    bswap   eax
    ret

now both generate the same optimal output.

MSVC used to generate the following with the old code:

    mov     eax, ecx
    rol     cx, 8
    shr     eax, 16
    rol     ax, 8
    movzx   ecx, cx
    movzx   eax, ax
    shl     ecx, 16
    or      eax, ecx
    ret     0

Now MSVC also generates a similar, but equally optimal result as clang/GCC:

    bswap   ecx
    mov     eax, ecx
    ret     0

====

In the swap64 case, for the original code, clang 8.0 would generate:

    mov     eax, edi
    bswap   eax
    shl     rax, 32
    shr     rdi, 32
    bswap   edi
    or      rax, rdi
    ret

(almost there, but still missing the mark)

while, again, GCC 8.3 would generate the more ideal:

    mov     rax, rdi
    bswap   rax
    ret

now clang also generates the optimal sequence for this fallback as well.

This is a case where MSVC unfortunately falls short, despite the new
code, this one still generates a doozy of an output.

    mov     r8, rcx
    mov     r9, rcx
    mov     rax, 71776119061217280
    mov     rdx, r8
    and     r9, rax
    and     edx, 65280
    mov     rax, rcx
    shr     rax, 16
    or      r9, rax
    mov     rax, rcx
    shr     r9, 16
    mov     rcx, 280375465082880
    and     rax, rcx
    mov     rcx, 1095216660480
    or      r9, rax
    mov     rax, r8
    and     rax, rcx
    shr     r9, 16
    or      r9, rax
    mov     rcx, r8
    mov     rax, r8
    shr     r9, 8
    shl     rax, 16
    and     ecx, 16711680
    or      rdx, rax
    mov     eax, -16777216
    and     rax, r8
    shl     rdx, 16
    or      rdx, rcx
    shl     rdx, 16
    or      rax, rdx
    shl     rax, 8
    or      rax, r9
    ret     0

which is pretty unfortunate.
2019-04-12 00:07:39 -04:00

718 lines
18 KiB
C++

// Copyright (c) 2012- PPSSPP Project / Dolphin Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0 or later versions.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
#pragma once
#include <type_traits>
#if defined(_MSC_VER)
#include <cstdlib>
#endif
#include <cstring>
#include "common/common_types.h"
// GCC
#ifdef __GNUC__
#if __BYTE_ORDER__ && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) && !defined(COMMON_LITTLE_ENDIAN)
#define COMMON_LITTLE_ENDIAN 1
#elif __BYTE_ORDER__ && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) && !defined(COMMON_BIG_ENDIAN)
#define COMMON_BIG_ENDIAN 1
#endif
// LLVM/clang
#elif defined(__clang__)
#if __LITTLE_ENDIAN__ && !defined(COMMON_LITTLE_ENDIAN)
#define COMMON_LITTLE_ENDIAN 1
#elif __BIG_ENDIAN__ && !defined(COMMON_BIG_ENDIAN)
#define COMMON_BIG_ENDIAN 1
#endif
// MSVC
#elif defined(_MSC_VER) && !defined(COMMON_BIG_ENDIAN) && !defined(COMMON_LITTLE_ENDIAN)
#define COMMON_LITTLE_ENDIAN 1
#endif
// Worst case, default to little endian.
#if !COMMON_BIG_ENDIAN && !COMMON_LITTLE_ENDIAN
#define COMMON_LITTLE_ENDIAN 1
#endif
namespace Common {
#ifdef _MSC_VER
[[nodiscard]] inline u16 swap16(u16 data) noexcept {
return _byteswap_ushort(data);
}
[[nodiscard]] inline u32 swap32(u32 data) noexcept {
return _byteswap_ulong(data);
}
[[nodiscard]] inline u64 swap64(u64 data) noexcept {
return _byteswap_uint64(data);
}
#elif defined(__clang__) || defined(__GNUC__)
#if defined(__Bitrig__) || defined(__OpenBSD__)
// redefine swap16, swap32, swap64 as inline functions
#undef swap16
#undef swap32
#undef swap64
#endif
[[nodiscard]] inline u16 swap16(u16 data) noexcept {
return __builtin_bswap16(data);
}
[[nodiscard]] inline u32 swap32(u32 data) noexcept {
return __builtin_bswap32(data);
}
[[nodiscard]] inline u64 swap64(u64 data) noexcept {
return __builtin_bswap64(data);
}
#else
// Generic implementation.
[[nodiscard]] inline u16 swap16(u16 data) noexcept {
return (data >> 8) | (data << 8);
}
[[nodiscard]] inline u32 swap32(u32 data) noexcept {
return ((data & 0xFF000000U) >> 24) | ((data & 0x00FF0000U) >> 8) |
((data & 0x0000FF00U) << 8) | ((data & 0x000000FFU) << 24);
}
[[nodiscard]] inline u64 swap64(u64 data) noexcept {
return ((data & 0xFF00000000000000ULL) >> 56) | ((data & 0x00FF000000000000ULL) >> 40) |
((data & 0x0000FF0000000000ULL) >> 24) | ((data & 0x000000FF00000000ULL) >> 8) |
((data & 0x00000000FF000000ULL) << 8) | ((data & 0x0000000000FF0000ULL) << 24) |
((data & 0x000000000000FF00ULL) << 40) | ((data & 0x00000000000000FFULL) << 56);
}
#endif
[[nodiscard]] inline float swapf(float f) noexcept {
static_assert(sizeof(u32) == sizeof(float), "float must be the same size as uint32_t.");
u32 value;
std::memcpy(&value, &f, sizeof(u32));
value = swap32(value);
std::memcpy(&f, &value, sizeof(u32));
return f;
}
[[nodiscard]] inline double swapd(double f) noexcept {
static_assert(sizeof(u64) == sizeof(double), "double must be the same size as uint64_t.");
u64 value;
std::memcpy(&value, &f, sizeof(u64));
value = swap64(value);
std::memcpy(&f, &value, sizeof(u64));
return f;
}
} // Namespace Common
template <typename T, typename F>
struct swap_struct_t {
using swapped_t = swap_struct_t;
protected:
T value;
static T swap(T v) {
return F::swap(v);
}
public:
T swap() const {
return swap(value);
}
swap_struct_t() = default;
swap_struct_t(const T& v) : value(swap(v)) {}
template <typename S>
swapped_t& operator=(const S& source) {
value = swap(static_cast<T>(source));
return *this;
}
operator s8() const {
return static_cast<s8>(swap());
}
operator u8() const {
return static_cast<u8>(swap());
}
operator s16() const {
return static_cast<s16>(swap());
}
operator u16() const {
return static_cast<u16>(swap());
}
operator s32() const {
return static_cast<s32>(swap());
}
operator u32() const {
return static_cast<u32>(swap());
}
operator s64() const {
return static_cast<s64>(swap());
}
operator u64() const {
return static_cast<u64>(swap());
}
operator float() const {
return static_cast<float>(swap());
}
operator double() const {
return static_cast<double>(swap());
}
// +v
swapped_t operator+() const {
return +swap();
}
// -v
swapped_t operator-() const {
return -swap();
}
// v / 5
swapped_t operator/(const swapped_t& i) const {
return swap() / i.swap();
}
template <typename S>
swapped_t operator/(const S& i) const {
return swap() / i;
}
// v * 5
swapped_t operator*(const swapped_t& i) const {
return swap() * i.swap();
}
template <typename S>
swapped_t operator*(const S& i) const {
return swap() * i;
}
// v + 5
swapped_t operator+(const swapped_t& i) const {
return swap() + i.swap();
}
template <typename S>
swapped_t operator+(const S& i) const {
return swap() + static_cast<T>(i);
}
// v - 5
swapped_t operator-(const swapped_t& i) const {
return swap() - i.swap();
}
template <typename S>
swapped_t operator-(const S& i) const {
return swap() - static_cast<T>(i);
}
// v += 5
swapped_t& operator+=(const swapped_t& i) {
value = swap(swap() + i.swap());
return *this;
}
template <typename S>
swapped_t& operator+=(const S& i) {
value = swap(swap() + static_cast<T>(i));
return *this;
}
// v -= 5
swapped_t& operator-=(const swapped_t& i) {
value = swap(swap() - i.swap());
return *this;
}
template <typename S>
swapped_t& operator-=(const S& i) {
value = swap(swap() - static_cast<T>(i));
return *this;
}
// ++v
swapped_t& operator++() {
value = swap(swap() + 1);
return *this;
}
// --v
swapped_t& operator--() {
value = swap(swap() - 1);
return *this;
}
// v++
swapped_t operator++(int) {
swapped_t old = *this;
value = swap(swap() + 1);
return old;
}
// v--
swapped_t operator--(int) {
swapped_t old = *this;
value = swap(swap() - 1);
return old;
}
// Comparaison
// v == i
bool operator==(const swapped_t& i) const {
return swap() == i.swap();
}
template <typename S>
bool operator==(const S& i) const {
return swap() == i;
}
// v != i
bool operator!=(const swapped_t& i) const {
return swap() != i.swap();
}
template <typename S>
bool operator!=(const S& i) const {
return swap() != i;
}
// v > i
bool operator>(const swapped_t& i) const {
return swap() > i.swap();
}
template <typename S>
bool operator>(const S& i) const {
return swap() > i;
}
// v < i
bool operator<(const swapped_t& i) const {
return swap() < i.swap();
}
template <typename S>
bool operator<(const S& i) const {
return swap() < i;
}
// v >= i
bool operator>=(const swapped_t& i) const {
return swap() >= i.swap();
}
template <typename S>
bool operator>=(const S& i) const {
return swap() >= i;
}
// v <= i
bool operator<=(const swapped_t& i) const {
return swap() <= i.swap();
}
template <typename S>
bool operator<=(const S& i) const {
return swap() <= i;
}
// logical
swapped_t operator!() const {
return !swap();
}
// bitmath
swapped_t operator~() const {
return ~swap();
}
swapped_t operator&(const swapped_t& b) const {
return swap() & b.swap();
}
template <typename S>
swapped_t operator&(const S& b) const {
return swap() & b;
}
swapped_t& operator&=(const swapped_t& b) {
value = swap(swap() & b.swap());
return *this;
}
template <typename S>
swapped_t& operator&=(const S b) {
value = swap(swap() & b);
return *this;
}
swapped_t operator|(const swapped_t& b) const {
return swap() | b.swap();
}
template <typename S>
swapped_t operator|(const S& b) const {
return swap() | b;
}
swapped_t& operator|=(const swapped_t& b) {
value = swap(swap() | b.swap());
return *this;
}
template <typename S>
swapped_t& operator|=(const S& b) {
value = swap(swap() | b);
return *this;
}
swapped_t operator^(const swapped_t& b) const {
return swap() ^ b.swap();
}
template <typename S>
swapped_t operator^(const S& b) const {
return swap() ^ b;
}
swapped_t& operator^=(const swapped_t& b) {
value = swap(swap() ^ b.swap());
return *this;
}
template <typename S>
swapped_t& operator^=(const S& b) {
value = swap(swap() ^ b);
return *this;
}
template <typename S>
swapped_t operator<<(const S& b) const {
return swap() << b;
}
template <typename S>
swapped_t& operator<<=(const S& b) const {
value = swap(swap() << b);
return *this;
}
template <typename S>
swapped_t operator>>(const S& b) const {
return swap() >> b;
}
template <typename S>
swapped_t& operator>>=(const S& b) const {
value = swap(swap() >> b);
return *this;
}
// Member
/** todo **/
// Arithmetics
template <typename S, typename T2, typename F2>
friend S operator+(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend S operator-(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend S operator/(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend S operator*(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend S operator%(const S& p, const swapped_t v);
// Arithmetics + assignements
template <typename S, typename T2, typename F2>
friend S operator+=(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend S operator-=(const S& p, const swapped_t v);
// Bitmath
template <typename S, typename T2, typename F2>
friend S operator&(const S& p, const swapped_t v);
// Comparison
template <typename S, typename T2, typename F2>
friend bool operator<(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend bool operator>(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend bool operator<=(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend bool operator>=(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend bool operator!=(const S& p, const swapped_t v);
template <typename S, typename T2, typename F2>
friend bool operator==(const S& p, const swapped_t v);
};
// Arithmetics
template <typename S, typename T, typename F>
S operator+(const S& i, const swap_struct_t<T, F> v) {
return i + v.swap();
}
template <typename S, typename T, typename F>
S operator-(const S& i, const swap_struct_t<T, F> v) {
return i - v.swap();
}
template <typename S, typename T, typename F>
S operator/(const S& i, const swap_struct_t<T, F> v) {
return i / v.swap();
}
template <typename S, typename T, typename F>
S operator*(const S& i, const swap_struct_t<T, F> v) {
return i * v.swap();
}
template <typename S, typename T, typename F>
S operator%(const S& i, const swap_struct_t<T, F> v) {
return i % v.swap();
}
// Arithmetics + assignements
template <typename S, typename T, typename F>
S& operator+=(S& i, const swap_struct_t<T, F> v) {
i += v.swap();
return i;
}
template <typename S, typename T, typename F>
S& operator-=(S& i, const swap_struct_t<T, F> v) {
i -= v.swap();
return i;
}
// Logical
template <typename S, typename T, typename F>
S operator&(const S& i, const swap_struct_t<T, F> v) {
return i & v.swap();
}
template <typename S, typename T, typename F>
S operator&(const swap_struct_t<T, F> v, const S& i) {
return static_cast<S>(v.swap() & i);
}
// Comparaison
template <typename S, typename T, typename F>
bool operator<(const S& p, const swap_struct_t<T, F> v) {
return p < v.swap();
}
template <typename S, typename T, typename F>
bool operator>(const S& p, const swap_struct_t<T, F> v) {
return p > v.swap();
}
template <typename S, typename T, typename F>
bool operator<=(const S& p, const swap_struct_t<T, F> v) {
return p <= v.swap();
}
template <typename S, typename T, typename F>
bool operator>=(const S& p, const swap_struct_t<T, F> v) {
return p >= v.swap();
}
template <typename S, typename T, typename F>
bool operator!=(const S& p, const swap_struct_t<T, F> v) {
return p != v.swap();
}
template <typename S, typename T, typename F>
bool operator==(const S& p, const swap_struct_t<T, F> v) {
return p == v.swap();
}
template <typename T>
struct swap_64_t {
static T swap(T x) {
return static_cast<T>(Common::swap64(x));
}
};
template <typename T>
struct swap_32_t {
static T swap(T x) {
return static_cast<T>(Common::swap32(x));
}
};
template <typename T>
struct swap_16_t {
static T swap(T x) {
return static_cast<T>(Common::swap16(x));
}
};
template <typename T>
struct swap_float_t {
static T swap(T x) {
return static_cast<T>(Common::swapf(x));
}
};
template <typename T>
struct swap_double_t {
static T swap(T x) {
return static_cast<T>(Common::swapd(x));
}
};
template <typename T>
struct swap_enum_t {
static_assert(std::is_enum_v<T>);
using base = std::underlying_type_t<T>;
public:
swap_enum_t() = default;
swap_enum_t(const T& v) : value(swap(v)) {}
swap_enum_t& operator=(const T& v) {
value = swap(v);
return *this;
}
operator T() const {
return swap(value);
}
explicit operator base() const {
return static_cast<base>(swap(value));
}
protected:
T value{};
// clang-format off
using swap_t = std::conditional_t<
std::is_same_v<base, u16>, swap_16_t<u16>, std::conditional_t<
std::is_same_v<base, s16>, swap_16_t<s16>, std::conditional_t<
std::is_same_v<base, u32>, swap_32_t<u32>, std::conditional_t<
std::is_same_v<base, s32>, swap_32_t<s32>, std::conditional_t<
std::is_same_v<base, u64>, swap_64_t<u64>, std::conditional_t<
std::is_same_v<base, s64>, swap_64_t<s64>, void>>>>>>;
// clang-format on
static T swap(T x) {
return static_cast<T>(swap_t::swap(static_cast<base>(x)));
}
};
struct SwapTag {}; // Use the different endianness from the system
struct KeepTag {}; // Use the same endianness as the system
template <typename T, typename Tag>
struct AddEndian;
// KeepTag specializations
template <typename T>
struct AddEndian<T, KeepTag> {
using type = T;
};
// SwapTag specializations
template <>
struct AddEndian<u8, SwapTag> {
using type = u8;
};
template <>
struct AddEndian<u16, SwapTag> {
using type = swap_struct_t<u16, swap_16_t<u16>>;
};
template <>
struct AddEndian<u32, SwapTag> {
using type = swap_struct_t<u32, swap_32_t<u32>>;
};
template <>
struct AddEndian<u64, SwapTag> {
using type = swap_struct_t<u64, swap_64_t<u64>>;
};
template <>
struct AddEndian<s8, SwapTag> {
using type = s8;
};
template <>
struct AddEndian<s16, SwapTag> {
using type = swap_struct_t<s16, swap_16_t<s16>>;
};
template <>
struct AddEndian<s32, SwapTag> {
using type = swap_struct_t<s32, swap_32_t<s32>>;
};
template <>
struct AddEndian<s64, SwapTag> {
using type = swap_struct_t<s64, swap_64_t<s64>>;
};
template <>
struct AddEndian<float, SwapTag> {
using type = swap_struct_t<float, swap_float_t<float>>;
};
template <>
struct AddEndian<double, SwapTag> {
using type = swap_struct_t<double, swap_double_t<double>>;
};
template <typename T>
struct AddEndian<T, SwapTag> {
static_assert(std::is_enum_v<T>);
using type = swap_enum_t<T>;
};
// Alias LETag/BETag as KeepTag/SwapTag depending on the system
#if COMMON_LITTLE_ENDIAN
using LETag = KeepTag;
using BETag = SwapTag;
#else
using BETag = KeepTag;
using LETag = SwapTag;
#endif
// Aliases for LE types
using u16_le = AddEndian<u16, LETag>::type;
using u32_le = AddEndian<u32, LETag>::type;
using u64_le = AddEndian<u64, LETag>::type;
using s16_le = AddEndian<s16, LETag>::type;
using s32_le = AddEndian<s32, LETag>::type;
using s64_le = AddEndian<s64, LETag>::type;
template <typename T>
using enum_le = std::enable_if_t<std::is_enum_v<T>, typename AddEndian<T, LETag>::type>;
using float_le = AddEndian<float, LETag>::type;
using double_le = AddEndian<double, LETag>::type;
// Aliases for BE types
using u16_be = AddEndian<u16, BETag>::type;
using u32_be = AddEndian<u32, BETag>::type;
using u64_be = AddEndian<u64, BETag>::type;
using s16_be = AddEndian<s16, BETag>::type;
using s32_be = AddEndian<s32, BETag>::type;
using s64_be = AddEndian<s64, BETag>::type;
template <typename T>
using enum_be = std::enable_if_t<std::is_enum_v<T>, typename AddEndian<T, BETag>::type>;
using float_be = AddEndian<float, BETag>::type;
using double_be = AddEndian<double, BETag>::type;