Fun/Yae
1
0
Fork 0
mirror of https://github.com/HolographicHat/Yae.git synced 2026-03-16 09:03:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

use sse

Browse Source
This commit is contained in:
REL
2024-10-15 07:32:14 -04:00
parent f73dbdc4fe
commit 618a9189ad
2 changed files with 26 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
lib/src/globals.h
View File

@@ -23,8 +23,8 @@ namespace Globals
class Offsets
{
public:
//PROPERTY2(uintptr_t, BitConverter_ToUInt16, 0, 0);
PROPERTY2(uintptr_t, BitConverter_ToUInt16, 0x0F826CF0, 0x0F825F10); // use non-zero to override dynamic search
PROPERTY2(uintptr_t, BitConverter_ToUInt16, 0, 0);
//PROPERTY2(uintptr_t, BitConverter_ToUInt16, 0x0F826CF0, 0x0F825F10); // use non-zero to override dynamic search
};
inline Offsets Offset;

49
lib/src/il2cpp-init.cpp
View File

@@ -8,6 +8,7 @@
#include <unordered_map>
#include <future>
#include <mutex>
#include <immintrin.h>
#include "globals.h"
#include "Zydis.h"
@@ -222,33 +223,36 @@ namespace
const auto sectionEnd = sectionAddress + sectionSize;
int32_t count = 0;
const __m128i callOpcode = _mm_set1_epi8(0xE8);
const size_t simdEnd = sectionSize / 16 * 16;
ZydisDecoder decoder{};
ZydisDecoderInit(&decoder, ZYDIS_MACHINE_MODE_LONG_64, ZYDIS_STACK_WIDTH_64);
for (size_t i = 0; i < simdEnd; i += 16) {
// load 16 bytes from the current address
const __m128i chunk = _mm_loadu_si128((__m128i*)(sectionAddress + i));
ZydisDecodedInstruction instruction{};
ZydisDecoderContext context{};
// compare the loaded chunk with 0xE8 in all 16 bytes
const __m128i result = _mm_cmpeq_epi8(chunk, callOpcode);
auto rip = (uint8_t*)sectionAddress;
while (rip < (uint8_t*)sectionEnd)
{
auto status = ZydisDecoderDecodeInstruction(&decoder, &context, rip, ZYDIS_MAX_INSTRUCTION_LENGTH, &instruction);
if (!ZYAN_SUCCESS(status))
{
rip += 1;
continue;
}
// move the comparison results into a mask
int mask = _mm_movemask_epi8(result);
if (instruction.mnemonic == ZYDIS_MNEMONIC_CALL)
{
const auto offset = *(int32_t*)(rip + 1);
const auto destination = rip + 5 + offset;
if (destination == target) {
while (mask != 0) {
DWORD first_match_idx = 0;
_BitScanForward(&first_match_idx, mask); // index of the first set bit (match)
// index of the instruction
const size_t instruction_index = i + first_match_idx;
const int32_t delta = *(int32_t*)(sectionAddress + instruction_index + 1);
const uintptr_t dest = sectionAddress + instruction_index + 5 + delta;
if (dest == (uintptr_t)target) {
count++;
}
}
rip += instruction.length;
// clear the bit we just processed and continue with the next match
mask &= ~(1 << first_match_idx);
}
}
return count;
@@ -277,11 +281,6 @@ namespace
return address;
}
/// <summary>
/// can be very slow to resolve on low-end machine,
/// consider updating static offset after it is resolved in development environment
/// </summary>
/// <returns></returns>
uintptr_t Resolve_BitConverter_ToUInt16()
{
size_t sectionSize;