rpcsx/rpcs3/util/asm.hpp
Nekotekina b42fae0989 PPU: refactor shift and splat instructions
Fix utils::rol32/64 functions.
Fix immediate clamping in splat instructions.
Other fixes.
2022-01-19 23:01:31 +03:00

424 lines
8.5 KiB
C++

#pragma once
#include "util/types.hpp"
#include "util/tsc.hpp"
#include <functional>
extern bool g_use_rtm;
extern u64 g_rtm_tx_limit1;
#ifdef _M_X64
extern "C"
{
u32 _xbegin();
void _xend();
void _mm_pause();
void _mm_prefetch(const char*, int);
void _m_prefetchw(const volatile void*);
uchar _rotl8(uchar, uchar);
ushort _rotl16(ushort, uchar);
uint _rotl(uint, int);
u64 _rotl64(u64, int);
u64 __popcnt64(u64);
s64 __mulh(s64, s64);
u64 __umulh(u64, u64);
s64 _div128(s64, s64, s64, s64*);
u64 _udiv128(u64, u64, u64, u64*);
void __debugbreak();
}
#endif
namespace utils
{
// Transaction helper (result = pair of success and op result, or just bool)
template <typename F, typename R = std::invoke_result_t<F>>
inline auto tx_start(F op)
{
#if defined(ARCH_X64)
uint status = -1;
for (auto stamp0 = get_tsc(), stamp1 = stamp0; g_use_rtm && stamp1 - stamp0 <= g_rtm_tx_limit1; stamp1 = get_tsc())
{
#ifndef _MSC_VER
__asm__ goto ("xbegin %l[retry];" ::: "memory" : retry);
#else
status = _xbegin();
if (status != umax) [[unlikely]]
{
goto retry;
}
#endif
if constexpr (std::is_void_v<R>)
{
std::invoke(op);
#ifndef _MSC_VER
__asm__ volatile ("xend;" ::: "memory");
#else
_xend();
#endif
return true;
}
else
{
auto result = std::invoke(op);
#ifndef _MSC_VER
__asm__ volatile ("xend;" ::: "memory");
#else
_xend();
#endif
return std::make_pair(true, std::move(result));
}
retry:
#ifndef _MSC_VER
__asm__ volatile ("movl %%eax, %0;" : "=r" (status) :: "memory");
#endif
if (!status) [[unlikely]]
{
break;
}
}
#else
static_cast<void>(op);
#endif
if constexpr (std::is_void_v<R>)
{
return false;
}
else
{
return std::make_pair(false, R());
}
};
// Try to prefetch to Level 2 cache since it's not split to data/code on most processors
template <typename T>
constexpr void prefetch_exec(T func)
{
if (std::is_constant_evaluated())
{
return;
}
const u64 value = reinterpret_cast<u64>(func);
const void* ptr = reinterpret_cast<const void*>(value);
#ifdef _M_X64
return _mm_prefetch(static_cast<const char*>(ptr), 2);
#else
return __builtin_prefetch(ptr, 0, 2);
#endif
}
// Try to prefetch to Level 1 cache
constexpr void prefetch_read(const void* ptr)
{
if (std::is_constant_evaluated())
{
return;
}
#ifdef _M_X64
return _mm_prefetch(static_cast<const char*>(ptr), 3);
#else
return __builtin_prefetch(ptr, 0, 3);
#endif
}
constexpr void prefetch_write(void* ptr)
{
if (std::is_constant_evaluated())
{
return;
}
#ifdef _M_X64
return _m_prefetchw(ptr);
#else
return __builtin_prefetch(ptr, 1, 0);
#endif
}
constexpr u8 rol8(u8 x, u8 n)
{
if (std::is_constant_evaluated())
{
return (x << (n & 7)) | (x >> ((-n & 7)));
}
#ifdef _MSC_VER
return _rotl8(x, n);
#elif defined(__clang__)
return __builtin_rotateleft8(x, n);
#elif defined(ARCH_X64)
return __builtin_ia32_rolqi(x, n);
#else
return (x << (n & 7)) | (x >> ((-n & 7)));
#endif
}
constexpr u16 rol16(u16 x, u16 n)
{
if (std::is_constant_evaluated())
{
return (x << (n & 15)) | (x >> ((-n & 15)));
}
#ifdef _MSC_VER
return _rotl16(x, static_cast<uchar>(n));
#elif defined(__clang__)
return __builtin_rotateleft16(x, n);
#elif defined(ARCH_X64)
return __builtin_ia32_rolhi(x, n);
#else
return (x << (n & 15)) | (x >> ((-n & 15)));
#endif
}
constexpr u32 rol32(u32 x, u32 n)
{
if (std::is_constant_evaluated())
{
return (x << (n & 31)) | (x >> (((0 - n) & 31)));
}
#ifdef _MSC_VER
return _rotl(x, n);
#elif defined(__clang__)
return __builtin_rotateleft32(x, n);
#else
return (x << (n & 31)) | (x >> (((0 - n) & 31)));
#endif
}
constexpr u64 rol64(u64 x, u64 n)
{
if (std::is_constant_evaluated())
{
return (x << (n & 63)) | (x >> (((0 - n) & 63)));
}
#ifdef _MSC_VER
return _rotl64(x, static_cast<int>(n));
#elif defined(__clang__)
return __builtin_rotateleft64(x, n);
#else
return (x << (n & 63)) | (x >> (((0 - n) & 63)));
#endif
}
constexpr u32 popcnt64(u64 v)
{
#if !defined(_MSC_VER) || defined(__SSE4_2__)
if (std::is_constant_evaluated())
#endif
{
v = (v & 0xaaaaaaaaaaaaaaaa) / 2 + (v & 0x5555555555555555);
v = (v & 0xcccccccccccccccc) / 4 + (v & 0x3333333333333333);
v = (v & 0xf0f0f0f0f0f0f0f0) / 16 + (v & 0x0f0f0f0f0f0f0f0f);
v = (v & 0xff00ff00ff00ff00) / 256 + (v & 0x00ff00ff00ff00ff);
v = ((v & 0xffff0000ffff0000) >> 16) + (v & 0x0000ffff0000ffff);
return static_cast<u32>((v >> 32) + v);
}
#if !defined(_MSC_VER) || defined(__SSE4_2__)
#ifdef _MSC_VER
return static_cast<u32>(__popcnt64(v));
#else
return __builtin_popcountll(v);
#endif
#endif
}
constexpr u32 popcnt128(const u128& v)
{
#ifdef _MSC_VER
return popcnt64(v.lo) + popcnt64(v.hi);
#else
return popcnt64(v) + popcnt64(v >> 64);
#endif
}
constexpr u64 umulh64(u64 x, u64 y)
{
#ifdef _MSC_VER
if (std::is_constant_evaluated())
#endif
{
return static_cast<u64>((u128{x} * u128{y}) >> 64);
}
#ifdef _MSC_VER
return __umulh(x, y);
#endif
}
inline s64 mulh64(s64 x, s64 y)
{
#ifdef _MSC_VER
return __mulh(x, y);
#else
return (s128{x} * s128{y}) >> 64;
#endif
}
inline s64 div128(s64 high, s64 low, s64 divisor, s64* remainder = nullptr)
{
#ifdef _MSC_VER
s64 rem = 0;
s64 r = _div128(high, low, divisor, &rem);
if (remainder)
{
*remainder = rem;
}
#else
const s128 x = (u128{static_cast<u64>(high)} << 64) | u64(low);
const s128 r = x / divisor;
if (remainder)
{
*remainder = x % divisor;
}
#endif
return r;
}
inline u64 udiv128(u64 high, u64 low, u64 divisor, u64* remainder = nullptr)
{
#ifdef _MSC_VER
u64 rem = 0;
u64 r = _udiv128(high, low, divisor, &rem);
if (remainder)
{
*remainder = rem;
}
#else
const u128 x = (u128{high} << 64) | low;
const u128 r = x / divisor;
if (remainder)
{
*remainder = x % divisor;
}
#endif
return r;
}
constexpr u32 ctz128(u128 arg)
{
#ifdef _MSC_VER
if (!arg.lo)
return std::countr_zero(arg.hi) + 64u;
else
return std::countr_zero(arg.lo);
#else
if (u64 lo = static_cast<u64>(arg))
return std::countr_zero<u64>(lo);
else
return std::countr_zero<u64>(arg >> 64) + 64;
#endif
}
constexpr u32 clz128(u128 arg)
{
#ifdef _MSC_VER
if (arg.hi)
return std::countl_zero(arg.hi);
else
return std::countl_zero(arg.lo) + 64;
#else
if (u64 hi = static_cast<u64>(arg >> 64))
return std::countl_zero<u64>(hi);
else
return std::countl_zero<u64>(arg) + 64;
#endif
}
inline void pause()
{
#if defined(ARCH_ARM64)
__asm__ volatile("yield");
#elif defined(_M_X64)
_mm_pause();
#elif defined(ARCH_X64)
__builtin_ia32_pause();
#else
#error "Missing utils::pause() implementation"
#endif
}
// Synchronization helper (cache-friendly busy waiting)
inline void busy_wait(usz cycles = 3000)
{
const u64 start = get_tsc();
do pause();
while (get_tsc() - start < cycles);
}
// Align to power of 2
template <typename T, typename = std::enable_if_t<std::is_unsigned_v<T>>>
constexpr T align(T value, std::type_identity_t<T> align)
{
return static_cast<T>((value + (align - 1)) & (T{0} - align));
}
// General purpose aligned division, the result is rounded up not truncated
template <typename T, typename = std::enable_if_t<std::is_unsigned_v<T>>>
constexpr T aligned_div(T value, std::type_identity_t<T> align)
{
return static_cast<T>(value / align + T{!!(value % align)});
}
// General purpose aligned division, the result is rounded to nearest
template <typename T, typename = std::enable_if_t<std::is_integral_v<T>>>
constexpr T rounded_div(T value, std::type_identity_t<T> align)
{
if constexpr (std::is_unsigned_v<T>)
{
return static_cast<T>(value / align + T{(value % align) > (align / 2)});
}
return static_cast<T>(value / align + (value > 0 ? T{(value % align) > (align / 2)} : 0 - T{(value % align) < (align / 2)}));
}
// Hack. Pointer cast util to workaround UB. Use with extreme care.
template <typename T, typename U>
[[nodiscard]] T* bless(U* ptr)
{
#ifdef _MSC_VER
return (T*)ptr;
#elif defined(ARCH_X64)
T* result;
__asm__("movq %1, %0;" : "=r" (result) : "r" (ptr) : "memory");
return result;
#elif defined(ARCH_ARM64)
T* result;
__asm__("mov %0, %1" : "=r" (result) : "r" (ptr) : "memory");
return result;
#endif
}
inline void trap()
{
#ifdef _M_X64
__debugbreak();
#elif defined(ARCH_X64)
__asm__ volatile("int3");
#elif defined(ARCH_ARM64)
__asm__ volatile("brk 0x42");
#else
#error "Missing utils::trap() implementation"
#endif
}
} // namespace utils
using utils::busy_wait;