Merge branch 'main' into user_and_settings

This commit is contained in:
georgemoralis
2025-12-09 14:43:15 +02:00
committed by GitHub
21 changed files with 453 additions and 106 deletions

View File

@@ -190,6 +190,7 @@ static ConfigEntry<bool> pipelineCacheArchive(false);
// Debug
static ConfigEntry<bool> isFpsColor(true);
static ConfigEntry<bool> showFpsCounter(false);
// GUI
static std::vector<GameInstallDir> settings_install_dirs = {};
@@ -376,6 +377,14 @@ bool fpsColor() {
return isFpsColor.get();
}
bool getShowFpsCounter() {
return showFpsCounter.get();
}
void setShowFpsCounter(bool enable, bool is_game_specific) {
showFpsCounter.set(enable, is_game_specific);
}
u32 vblankFreq() {
if (vblankFrequency.get() < 60) {
vblankFrequency = 60;
@@ -683,8 +692,8 @@ void load(const std::filesystem::path& path, bool is_game_specific) {
string current_version = {};
if (data.contains("Debug")) {
const toml::value& debug = data.at("Debug");
isFpsColor.setFromToml(debug, "FPSColor", is_game_specific);
showFpsCounter.setFromToml(debug, "showFpsCounter", is_game_specific);
current_version = toml::find_or<std::string>(debug, "ConfigVersion", current_version);
}
@@ -870,6 +879,7 @@ void save(const std::filesystem::path& path, bool is_game_specific) {
data["GPU"]["internalScreenHeight"] = internalScreenHeight.base_value;
data["GPU"]["patchShaders"] = shouldPatchShaders.base_value;
data["Debug"]["FPSColor"] = isFpsColor.base_value;
data["Debug"]["showFpsCounter"] = showFpsCounter.base_value;
}
// Sorting of TOML sections
@@ -952,6 +962,7 @@ void setDefaultValues(bool is_game_specific) {
// Debug
isFpsColor.base_value = true;
showFpsCounter.base_value = false;
}
}

View File

@@ -93,6 +93,8 @@ void setSpecialPadClass(int type);
int getSpecialPadClass();
bool patchShaders(); // no set
bool fpsColor(); // no set
bool getShowFpsCounter();
void setShowFpsCounter(bool enable, bool is_game_specific = false);
bool getIsMotionControlsEnabled();
void setIsMotionControlsEnabled(bool use, bool is_game_specific = false);
std::string getDefaultControllerID();

View File

@@ -5,6 +5,7 @@
#include <memory>
#include <mutex>
#include <set>
#include <vector>
#include <Zydis/Zydis.h>
#include <xbyak/xbyak.h>
#include <xbyak/xbyak_util.h>
@@ -122,6 +123,30 @@ static void GenerateTcbAccess(void* /* address */, const ZydisDecodedOperand* op
#endif
}
static bool FilterStackCheck(const ZydisDecodedOperand* operands) {
const auto& dst_op = operands[0];
const auto& src_op = operands[1];
// Some compilers emit stack checks by starting a function with
// 'mov (64-bit register), fs:[0x28]', then checking with `xor (64-bit register), fs:[0x28]`
return src_op.type == ZYDIS_OPERAND_TYPE_MEMORY && src_op.mem.segment == ZYDIS_REGISTER_FS &&
src_op.mem.base == ZYDIS_REGISTER_NONE && src_op.mem.index == ZYDIS_REGISTER_NONE &&
src_op.mem.disp.value == 0x28 && dst_op.reg.value >= ZYDIS_REGISTER_RAX &&
dst_op.reg.value <= ZYDIS_REGISTER_R15;
}
static void GenerateStackCheck(void* /* address */, const ZydisDecodedOperand* operands,
Xbyak::CodeGenerator& c) {
const auto dst = ZydisToXbyakRegisterOperand(operands[0]);
c.xor_(dst, 0);
}
static void GenerateStackCanary(void* /* address */, const ZydisDecodedOperand* operands,
Xbyak::CodeGenerator& c) {
const auto dst = ZydisToXbyakRegisterOperand(operands[0]);
c.mov(dst, 0);
}
static bool FilterNoSSE4a(const ZydisDecodedOperand*) {
Cpu cpu;
return !cpu.has(Cpu::tSSE4a);
@@ -440,18 +465,26 @@ struct PatchInfo {
bool trampoline;
};
static const std::unordered_map<ZydisMnemonic, PatchInfo> Patches = {
static const std::unordered_map<ZydisMnemonic, std::vector<PatchInfo>> Patches = {
// SSE4a
{ZYDIS_MNEMONIC_EXTRQ, {FilterNoSSE4a, GenerateEXTRQ, true}},
{ZYDIS_MNEMONIC_INSERTQ, {FilterNoSSE4a, GenerateINSERTQ, true}},
{ZYDIS_MNEMONIC_MOVNTSS, {FilterNoSSE4a, ReplaceMOVNTSS, false}},
{ZYDIS_MNEMONIC_MOVNTSD, {FilterNoSSE4a, ReplaceMOVNTSD, false}},
{ZYDIS_MNEMONIC_EXTRQ, {{FilterNoSSE4a, GenerateEXTRQ, true}}},
{ZYDIS_MNEMONIC_INSERTQ, {{FilterNoSSE4a, GenerateINSERTQ, true}}},
{ZYDIS_MNEMONIC_MOVNTSS, {{FilterNoSSE4a, ReplaceMOVNTSS, false}}},
{ZYDIS_MNEMONIC_MOVNTSD, {{FilterNoSSE4a, ReplaceMOVNTSD, false}}},
#if !defined(__APPLE__)
// FS segment patches
// These first two patches are for accesses to the stack canary, fs:[0x28]
{ZYDIS_MNEMONIC_XOR, {{FilterStackCheck, GenerateStackCheck, false}}},
{ZYDIS_MNEMONIC_MOV,
{{FilterStackCheck, GenerateStackCanary, false},
#if defined(_WIN32)
// Windows needs a trampoline.
{ZYDIS_MNEMONIC_MOV, {FilterTcbAccess, GenerateTcbAccess, true}},
#elif !defined(__APPLE__)
{ZYDIS_MNEMONIC_MOV, {FilterTcbAccess, GenerateTcbAccess, false}},
// Windows needs a trampoline for Tcb accesses.
{FilterTcbAccess, GenerateTcbAccess, true}
#else
{FilterTcbAccess, GenerateTcbAccess, false}
#endif
}},
#endif
};
@@ -503,51 +536,53 @@ static std::pair<bool, u64> TryPatch(u8* code, PatchModule* module) {
}
if (Patches.contains(instruction.mnemonic)) {
const auto& patch_info = Patches.at(instruction.mnemonic);
bool needs_trampoline = patch_info.trampoline;
if (patch_info.filter(operands)) {
auto& patch_gen = module->patch_gen;
const auto& patches = Patches.at(instruction.mnemonic);
for (const auto& patch_info : patches) {
bool needs_trampoline = patch_info.trampoline;
if (patch_info.filter(operands)) {
auto& patch_gen = module->patch_gen;
if (needs_trampoline && instruction.length < 5) {
// Trampoline is needed but instruction is too short to patch.
// Return false and length to signal to AOT compilation that this instruction
// should be skipped and handled at runtime.
return std::make_pair(false, instruction.length);
}
if (needs_trampoline && instruction.length < 5) {
// Trampoline is needed but instruction is too short to patch.
// Return false and length to signal to AOT compilation that this instruction
// should be skipped and handled at runtime.
return std::make_pair(false, instruction.length);
}
// Reset state and move to current code position.
patch_gen.reset();
patch_gen.setSize(code - patch_gen.getCode());
// Reset state and move to current code position.
patch_gen.reset();
patch_gen.setSize(code - patch_gen.getCode());
if (needs_trampoline) {
auto& trampoline_gen = module->trampoline_gen;
const auto trampoline_ptr = trampoline_gen.getCurr();
if (needs_trampoline) {
auto& trampoline_gen = module->trampoline_gen;
const auto trampoline_ptr = trampoline_gen.getCurr();
patch_info.generator(code, operands, trampoline_gen);
patch_info.generator(code, operands, trampoline_gen);
// Return to the following instruction at the end of the trampoline.
trampoline_gen.jmp(code + instruction.length);
// Return to the following instruction at the end of the trampoline.
trampoline_gen.jmp(code + instruction.length);
// Replace instruction with near jump to the trampoline.
patch_gen.jmp(trampoline_ptr, Xbyak::CodeGenerator::LabelType::T_NEAR);
} else {
patch_info.generator(code, operands, patch_gen);
}
// Replace instruction with near jump to the trampoline.
patch_gen.jmp(trampoline_ptr, Xbyak::CodeGenerator::LabelType::T_NEAR);
} else {
patch_info.generator(code, operands, patch_gen);
}
const auto patch_size = patch_gen.getCurr() - code;
if (patch_size > 0) {
ASSERT_MSG(instruction.length >= patch_size,
"Instruction {} with length {} is too short to replace at: {}",
ZydisMnemonicGetString(instruction.mnemonic), instruction.length,
fmt::ptr(code));
const auto patch_size = patch_gen.getCurr() - code;
if (patch_size > 0) {
ASSERT_MSG(instruction.length >= patch_size,
"Instruction {} with length {} is too short to replace at: {}",
ZydisMnemonicGetString(instruction.mnemonic), instruction.length,
fmt::ptr(code));
// Fill remaining space with nops.
patch_gen.nop(instruction.length - patch_size);
// Fill remaining space with nops.
patch_gen.nop(instruction.length - patch_size);
module->patched.insert(code);
LOG_DEBUG(Core, "Patched instruction '{}' at: {}",
ZydisMnemonicGetString(instruction.mnemonic), fmt::ptr(code));
return std::make_pair(true, instruction.length);
module->patched.insert(code);
LOG_DEBUG(Core, "Patched instruction '{}' at: {}",
ZydisMnemonicGetString(instruction.mnemonic), fmt::ptr(code));
return std::make_pair(true, instruction.length);
}
}
}
}

View File

@@ -311,6 +311,7 @@ static void LoadSettings(const char* line) {
void L::SetupSettings() {
frame_graph.is_open = true;
show_simple_fps = Config::getShowFpsCounter();
using SettingLoader = void (*)(const char*);
@@ -475,6 +476,11 @@ void ToggleSimpleFps() {
visibility_toggled = true;
}
void SetSimpleFps(bool enabled) {
show_simple_fps = enabled;
visibility_toggled = true;
}
void ToggleQuitWindow() {
show_quit_window = !show_quit_window;
}

View File

@@ -30,6 +30,7 @@ private:
namespace Overlay {
void ToggleSimpleFps();
void SetSimpleFps(bool enabled);
void ToggleQuitWindow();
} // namespace Overlay

View File

@@ -6,8 +6,8 @@
#include "common/elf_info.h"
#include "common/logging/log.h"
#include "core/libraries/fiber/fiber_error.h"
#include "core/libraries/kernel/threads/pthread.h"
#include "core/libraries/libs.h"
#include "core/tls.h"
namespace Libraries::Fiber {
@@ -20,7 +20,7 @@ static constexpr u64 kFiberStackSizeCheck = 0xdeadbeefdeadbeef;
static std::atomic<u32> context_size_check = false;
OrbisFiberContext* GetFiberContext() {
return Libraries::Kernel::g_curthread->tcb->tcb_fiber;
return Core::GetTcbBase()->tcb_fiber;
}
extern "C" s32 PS4_SYSV_ABI _sceFiberSetJmp(OrbisFiberContext* ctx) asm("_sceFiberSetJmp");
@@ -269,7 +269,7 @@ s32 PS4_SYSV_ABI sceFiberRunImpl(OrbisFiber* fiber, void* addr_context, u64 size
return ORBIS_FIBER_ERROR_INVALID;
}
Core::Tcb* tcb = Libraries::Kernel::g_curthread->tcb;
Core::Tcb* tcb = Core::GetTcbBase();
if (tcb->tcb_fiber) {
return ORBIS_FIBER_ERROR_PERMISSION;
}

View File

@@ -5,24 +5,35 @@
#include "common/alignment.h"
#include "common/assert.h"
#include "common/elf_info.h"
#include "common/logging/log.h"
#include "common/scope_exit.h"
#include "common/singleton.h"
#include "core/libraries/kernel/kernel.h"
#include "core/libraries/kernel/memory.h"
#include "core/libraries/kernel/orbis_error.h"
#include "core/libraries/kernel/process.h"
#include "core/libraries/libs.h"
#include "core/linker.h"
#include "core/memory.h"
namespace Libraries::Kernel {
static s32 g_sdk_version = -1;
static bool g_alias_dmem = false;
u64 PS4_SYSV_ABI sceKernelGetDirectMemorySize() {
LOG_TRACE(Kernel_Vmm, "called");
const auto* memory = Core::Memory::Instance();
return memory->GetTotalDirectSize();
}
s32 PS4_SYSV_ABI sceKernelEnableDmemAliasing() {
LOG_DEBUG(Kernel_Vmm, "called");
g_alias_dmem = true;
return ORBIS_OK;
}
s32 PS4_SYSV_ABI sceKernelAllocateDirectMemory(s64 searchStart, s64 searchEnd, u64 len,
u64 alignment, s32 memoryType, s64* physAddrOut) {
if (searchStart < 0 || searchEnd < 0) {
@@ -197,8 +208,14 @@ s32 PS4_SYSV_ABI sceKernelMapNamedDirectMemory(void** addr, u64 len, s32 prot, s
const VAddr in_addr = reinterpret_cast<VAddr>(*addr);
auto* memory = Core::Memory::Instance();
const auto ret = memory->MapMemory(addr, in_addr, len, mem_prot, map_flags,
Core::VMAType::Direct, name, false, phys_addr, alignment);
bool should_check = false;
if (g_sdk_version >= Common::ElfInfo::FW_25 && False(map_flags & Core::MemoryMapFlags::Stack)) {
// Under these conditions, this would normally redirect to sceKernelMapDirectMemory2.
should_check = !g_alias_dmem;
}
const auto ret =
memory->MapMemory(addr, in_addr, len, mem_prot, map_flags, Core::VMAType::Direct, name,
should_check, phys_addr, alignment);
LOG_INFO(Kernel_Vmm, "out_addr = {}", fmt::ptr(*addr));
return ret;
@@ -244,8 +261,9 @@ s32 PS4_SYSV_ABI sceKernelMapDirectMemory2(void** addr, u64 len, s32 type, s32 p
const VAddr in_addr = reinterpret_cast<VAddr>(*addr);
auto* memory = Core::Memory::Instance();
const auto ret = memory->MapMemory(addr, in_addr, len, mem_prot, map_flags,
Core::VMAType::Direct, "anon", true, phys_addr, alignment);
const auto ret =
memory->MapMemory(addr, in_addr, len, mem_prot, map_flags, Core::VMAType::Direct, "anon",
!g_alias_dmem, phys_addr, alignment);
if (ret == 0) {
// If the map call succeeds, set the direct memory type using the output address.
@@ -668,10 +686,21 @@ void* PS4_SYSV_ABI posix_mmap(void* addr, u64 len, s32 prot, s32 flags, s32 fd,
}
s32 result = ORBIS_OK;
if (fd == -1) {
if (True(mem_flags & Core::MemoryMapFlags::Anon)) {
// Maps flexible memory
result = memory->MapMemory(&addr_out, aligned_addr, aligned_size, mem_prot, mem_flags,
Core::VMAType::Flexible, "anon", false);
} else if (True(mem_flags & Core::MemoryMapFlags::Stack)) {
// Maps stack memory
result = memory->MapMemory(&addr_out, aligned_addr, aligned_size, mem_prot, mem_flags,
Core::VMAType::Stack, "anon", false);
} else if (True(mem_flags & Core::MemoryMapFlags::Void)) {
// Reserves memory
result =
memory->MapMemory(&addr_out, aligned_addr, aligned_size, Core::MemoryProt::NoAccess,
mem_flags, Core::VMAType::Reserved, "anon", false);
} else {
// Default to file mapping
result = memory->MapFile(&addr_out, aligned_addr, aligned_size, mem_prot, mem_flags, fd,
phys_addr);
}
@@ -769,6 +798,12 @@ s32 PS4_SYSV_ABI sceKernelGetPrtAperture(s32 id, VAddr* address, u64* size) {
}
void RegisterMemory(Core::Loader::SymbolsResolver* sym) {
ASSERT_MSG(sceKernelGetCompiledSdkVersion(&g_sdk_version) == ORBIS_OK,
"Failed to get compiled SDK verision.");
LIB_FUNCTION("usHTMoFoBTM", "libkernel_dmem_aliasing2", 1, "libkernel",
sceKernelEnableDmemAliasing);
LIB_FUNCTION("usHTMoFoBTM", "libkernel", 1, "libkernel", sceKernelEnableDmemAliasing);
LIB_FUNCTION("rTXw65xmLIA", "libkernel", 1, "libkernel", sceKernelAllocateDirectMemory);
LIB_FUNCTION("B+vc2AO2Zrc", "libkernel", 1, "libkernel", sceKernelAllocateMainDirectMemory);
LIB_FUNCTION("C0f7TJcbfac", "libkernel", 1, "libkernel", sceKernelAvailableDirectMemorySize);

View File

@@ -173,6 +173,8 @@ void RegisterException(Core::Loader::SymbolsResolver* sym) {
LIB_FUNCTION("OMDRKKAZ8I4", "libkernel", 1, "libkernel", sceKernelDebugRaiseException);
LIB_FUNCTION("zE-wXIZjLoM", "libkernel", 1, "libkernel",
sceKernelDebugRaiseExceptionOnReleaseMode);
LIB_FUNCTION("WkwEd3N7w0Y", "libkernel", 1, "libkernel", sceKernelInstallExceptionHandler);
LIB_FUNCTION("Qhv5ARAoOEc", "libkernel", 1, "libkernel", sceKernelRemoveExceptionHandler);
}
} // namespace Libraries::Kernel

View File

@@ -663,6 +663,10 @@ void RegisterThread(Core::Loader::SymbolsResolver* sym) {
LIB_FUNCTION("Z4QosVuAsA0", "libkernel", 1, "libkernel", posix_pthread_once);
LIB_FUNCTION("EotR8a3ASf4", "libkernel", 1, "libkernel", posix_pthread_self);
LIB_FUNCTION("OxhIB8LB-PQ", "libkernel", 1, "libkernel", posix_pthread_create);
LIB_FUNCTION("lZzFeSxPl08", "libkernel", 1, "libkernel", posix_pthread_setcancelstate);
LIB_FUNCTION("CBNtXOoef-E", "libkernel", 1, "libkernel", posix_sched_get_priority_max);
LIB_FUNCTION("m0iS6jNsXds", "libkernel", 1, "libkernel", posix_sched_get_priority_min);
LIB_FUNCTION("Xs9hdiD7sAA", "libkernel", 1, "libkernel", posix_pthread_setschedparam);
LIB_FUNCTION("+U1R4WtXvoc", "libkernel", 1, "libkernel", posix_pthread_detach);
LIB_FUNCTION("7Xl257M4VNI", "libkernel", 1, "libkernel", posix_pthread_equal);
LIB_FUNCTION("h9CcP3J0oVM", "libkernel", 1, "libkernel", posix_pthread_join);

View File

@@ -712,8 +712,61 @@ int PS4_SYSV_ABI sceHttpUriCopy() {
return ORBIS_OK;
}
int PS4_SYSV_ABI sceHttpUriEscape() {
LOG_ERROR(Lib_Http, "(STUBBED) called");
int PS4_SYSV_ABI sceHttpUriEscape(char* out, u64* require, u64 prepare, const char* in) {
LOG_TRACE(Lib_Http, "called");
if (!in) {
LOG_ERROR(Lib_Http, "Invalid input string");
return ORBIS_HTTP_ERROR_INVALID_VALUE;
}
auto IsUnreserved = [](unsigned char c) -> bool {
return (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') ||
c == '-' || c == '_' || c == '.' || c == '~';
};
u64 needed = 0;
const char* src = in;
while (*src) {
unsigned char c = static_cast<unsigned char>(*src);
if (IsUnreserved(c)) {
needed++;
} else {
needed += 3; // %XX format
}
src++;
}
needed++; // null terminator
if (require) {
*require = needed;
}
if (!out) {
return ORBIS_OK;
}
if (prepare < needed) {
LOG_ERROR(Lib_Http, "Buffer too small: need {} but only {} available", needed, prepare);
return ORBIS_HTTP_ERROR_OUT_OF_MEMORY;
}
static const char hex_chars[] = "0123456789ABCDEF";
src = in;
char* dst = out;
while (*src) {
unsigned char c = static_cast<unsigned char>(*src);
if (IsUnreserved(c)) {
*dst++ = *src;
} else {
*dst++ = '%';
*dst++ = hex_chars[(c >> 4) & 0x0F];
*dst++ = hex_chars[c & 0x0F];
}
src++;
}
*dst = '\0';
return ORBIS_OK;
}
@@ -1072,12 +1125,163 @@ int PS4_SYSV_ABI sceHttpUriParse(OrbisHttpUriElement* out, const char* srcUri, v
}
int PS4_SYSV_ABI sceHttpUriSweepPath(char* dst, const char* src, u64 srcSize) {
LOG_ERROR(Lib_Http, "(STUBBED) called");
LOG_TRACE(Lib_Http, "called");
if (!dst || !src) {
LOG_ERROR(Lib_Http, "Invalid parameters");
return ORBIS_HTTP_ERROR_INVALID_VALUE;
}
if (srcSize == 0) {
dst[0] = '\0';
return ORBIS_OK;
}
u64 len = 0;
while (len < srcSize && src[len] != '\0') {
len++;
}
for (u64 i = 0; i < len; i++) {
dst[i] = src[i];
}
dst[len] = '\0';
char* read = dst;
char* write = dst;
while (*read) {
if (read[0] == '.' && read[1] == '.' && read[2] == '/') {
read += 3;
continue;
}
if (read[0] == '.' && read[1] == '/') {
read += 2;
continue;
}
if (read[0] == '/' && read[1] == '.' && read[2] == '/') {
read += 2;
continue;
}
if (read[0] == '/' && read[1] == '.' && read[2] == '\0') {
if (write == dst) {
*write++ = '/';
}
break;
}
bool is_dotdot_mid = (read[0] == '/' && read[1] == '.' && read[2] == '.' && read[3] == '/');
bool is_dotdot_end =
(read[0] == '/' && read[1] == '.' && read[2] == '.' && read[3] == '\0');
if (is_dotdot_mid || is_dotdot_end) {
if (write > dst) {
if (*(write - 1) == '/') {
write--;
}
while (write > dst && *(write - 1) != '/') {
write--;
}
if (is_dotdot_mid && write > dst) {
write--;
}
}
if (is_dotdot_mid) {
read += 3;
} else {
break;
}
continue;
}
if ((read[0] == '.' && read[1] == '\0') ||
(read[0] == '.' && read[1] == '.' && read[2] == '\0')) {
break;
}
if (read[0] == '/') {
*write++ = *read++;
}
while (*read && *read != '/') {
*write++ = *read++;
}
}
*write = '\0';
return ORBIS_OK;
}
int PS4_SYSV_ABI sceHttpUriUnescape(char* out, u64* require, u64 prepare, const char* in) {
LOG_ERROR(Lib_Http, "(STUBBED) called");
LOG_TRACE(Lib_Http, "called");
if (!in) {
LOG_ERROR(Lib_Http, "Invalid input string");
return ORBIS_HTTP_ERROR_INVALID_VALUE;
}
// Locale-independent hex digit check
auto IsHex = [](char c) -> bool {
return (c >= '0' && c <= '9') || (c >= 'A' && c <= 'F') || (c >= 'a' && c <= 'f');
};
// Convert hex char to int value
auto HexToInt = [](char c) -> int {
if (c >= '0' && c <= '9')
return c - '0';
if (c >= 'A' && c <= 'F')
return c - 'A' + 10;
if (c >= 'a' && c <= 'f')
return c - 'a' + 10;
return 0;
};
// Check for valid percent-encoded sequence (%XX)
auto IsValidPercentSequence = [&](const char* s) -> bool {
return s[0] == '%' && s[1] != '\0' && s[2] != '\0' && IsHex(s[1]) && IsHex(s[2]);
};
u64 needed = 0;
const char* src = in;
while (*src) {
if (IsValidPercentSequence(src)) {
src += 3;
} else {
src++;
}
needed++;
}
needed++; // null terminator
if (require) {
*require = needed;
}
if (!out) {
return ORBIS_OK;
}
if (prepare < needed) {
LOG_ERROR(Lib_Http, "Buffer too small: need {} but only {} available", needed, prepare);
return ORBIS_HTTP_ERROR_OUT_OF_MEMORY;
}
src = in;
char* dst = out;
while (*src) {
if (IsValidPercentSequence(src)) {
*dst++ = static_cast<char>((HexToInt(src[1]) << 4) | HexToInt(src[2]));
src += 3;
} else {
*dst++ = *src++;
}
}
*dst = '\0';
return ORBIS_OK;
}

View File

@@ -148,7 +148,7 @@ int PS4_SYSV_ABI sceHttpUnsetEpoll();
int PS4_SYSV_ABI sceHttpUriBuild(char* out, u64* require, u64 prepare,
const OrbisHttpUriElement* srcElement, u32 option);
int PS4_SYSV_ABI sceHttpUriCopy();
int PS4_SYSV_ABI sceHttpUriEscape();
int PS4_SYSV_ABI sceHttpUriEscape(char* out, u64* require, u64 prepare, const char* in);
int PS4_SYSV_ABI sceHttpUriMerge(char* mergedUrl, char* url, char* relativeUri, u64* require,
u64 prepare, u32 option);
int PS4_SYSV_ABI sceHttpUriParse(OrbisHttpUriElement* out, const char* srcUri, void* pool,

View File

@@ -369,7 +369,7 @@ bool Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
void* Linker::TlsGetAddr(u64 module_index, u64 offset) {
std::scoped_lock lk{mutex};
DtvEntry* dtv_table = Libraries::Kernel::g_curthread->tcb->tcb_dtv;
DtvEntry* dtv_table = GetTcbBase()->tcb_dtv;
if (dtv_table[0].counter != dtv_generation_counter) {
// Generation counter changed, a dynamic module was either loaded or unloaded.
const u32 old_num_dtvs = dtv_table[1].counter;
@@ -382,7 +382,7 @@ void* Linker::TlsGetAddr(u64 module_index, u64 offset) {
delete[] dtv_table;
// Update TCB pointer.
Libraries::Kernel::g_curthread->tcb->tcb_dtv = new_dtv_table;
GetTcbBase()->tcb_dtv = new_dtv_table;
dtv_table = new_dtv_table;
}

View File

@@ -45,7 +45,10 @@ enum class MemoryMapFlags : u32 {
Private = 2,
Fixed = 0x10,
NoOverwrite = 0x80,
Void = 0x100,
Stack = 0x400,
NoSync = 0x800,
Anon = 0x1000,
NoCore = 0x20000,
NoCoalesce = 0x400000,
};

View File

@@ -46,6 +46,10 @@ void SetTcbBase(void* image_address) {
ASSERT(result != 0);
}
Tcb* GetTcbBase() {
return reinterpret_cast<Tcb*>(TlsGetValue(GetTcbKey()));
}
#elif defined(__APPLE__) && defined(ARCH_X86_64)
// Apple x86_64
@@ -145,6 +149,12 @@ void SetTcbBase(void* image_address) {
"Failed to store thread LDT page pointer: {}", errno);
}
Tcb* GetTcbBase() {
Tcb* tcb;
asm volatile("mov %%fs:0x0, %0" : "=r"(tcb));
return tcb;
}
#elif defined(ARCH_X86_64)
// Other POSIX x86_64
@@ -154,6 +164,10 @@ void SetTcbBase(void* image_address) {
ASSERT_MSG(ret == 0, "Failed to set GS base: errno {}", errno);
}
Tcb* GetTcbBase() {
return Libraries::Kernel::g_curthread->tcb;
}
#else
// POSIX non-x86_64
@@ -176,6 +190,10 @@ void SetTcbBase(void* image_address) {
ASSERT(pthread_setspecific(GetTcbKey(), image_address) == 0);
}
Tcb* GetTcbBase() {
return static_cast<Tcb*>(pthread_getspecific(GetTcbKey()));
}
#endif
thread_local std::once_flag init_tls_flag;

View File

@@ -39,6 +39,9 @@ u32 GetTcbKey();
/// Sets the data pointer to the TCB block.
void SetTcbBase(void* image_address);
/// Retrieves Tcb structure for the calling thread.
Tcb* GetTcbBase();
/// Makes sure TLS is initialized for the thread before entering guest.
void EnsureThreadInitialized();

View File

@@ -71,6 +71,7 @@ int main(int argc, char* argv[]) {
"values, ignores the config file(s) entirely.\n"
" --config-global Run the emulator with the base config file "
"only, ignores game specific configs.\n"
" --show-fps Enable FPS counter display at startup\n"
" -h, --help Display this help message\n";
exit(0);
}},
@@ -178,13 +179,15 @@ int main(int argc, char* argv[]) {
game_folder = folder;
}},
{"--wait-for-debugger", [&](int& i) { waitForDebugger = true; }},
{"--wait-for-pid", [&](int& i) {
{"--wait-for-pid",
[&](int& i) {
if (++i >= argc) {
std::cerr << "Error: Missing argument for --wait-for-pid\n";
exit(1);
}
waitPid = std::stoi(argv[i]);
}}};
}},
{"--show-fps", [&](int& i) { Config::setShowFpsCounter(true); }}};
if (argc == 1) {
if (!SDL_ShowSimpleMessageBox(

View File

@@ -1050,7 +1050,14 @@ void Translator::V_CMP_U32(ConditionOp op, bool is_signed, bool set_exec, const
}
void Translator::V_CMP_U64(ConditionOp op, bool is_signed, bool set_exec, const GcnInst& inst) {
ASSERT(inst.src[1].field == OperandField::ConstZero);
const bool is_zero = inst.src[1].field == OperandField::ConstZero;
const bool is_neg_one = inst.src[1].field == OperandField::SignedConstIntNeg;
ASSERT(is_zero || is_neg_one);
if (is_neg_one) {
ASSERT_MSG(-s32(inst.src[1].code) + SignedConstIntNegMin - 1 == -1,
"SignedConstIntNeg must be -1");
}
const IR::U1 src0 = [&] {
switch (inst.src[0].field) {
case OperandField::ScalarGPR:
@@ -1064,10 +1071,11 @@ void Translator::V_CMP_U64(ConditionOp op, bool is_signed, bool set_exec, const
const IR::U1 result = [&] {
switch (op) {
case ConditionOp::EQ:
return ir.LogicalNot(src0);
return is_zero ? ir.LogicalNot(src0) : src0;
case ConditionOp::LG: // NE
return src0;
return is_zero ? src0 : ir.LogicalNot(src0);
case ConditionOp::GT:
ASSERT(is_zero);
return ir.GroupAny(ir.GetThreadBitScalarReg(IR::ScalarReg(inst.src[0].code)));
default:
UNREACHABLE_MSG("Unsupported V_CMP_U64 condition operation: {}", u32(op));

View File

@@ -3,6 +3,7 @@
#include "common/assert.h"
#include "common/debug.h"
#include "common/thread.h"
#include "imgui/renderer/texture_manager.h"
#include "video_core/renderer_vulkan/vk_instance.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
@@ -17,6 +18,8 @@ Scheduler::Scheduler(const Instance& instance)
profiler_scope = reinterpret_cast<tracy::VkCtxScope*>(std::malloc(sizeof(tracy::VkCtxScope)));
#endif
AllocateWorkerCommandBuffers();
priority_pending_ops_thread =
std::jthread(std::bind_front(&Scheduler::PriorityPendingOpsThread, this));
}
Scheduler::~Scheduler() {
@@ -167,6 +170,32 @@ void Scheduler::SubmitExecution(SubmitInfo& info) {
PopPendingOperations();
}
void Scheduler::PriorityPendingOpsThread(std::stop_token stoken) {
Common::SetCurrentThreadName("shadPS4:GpuSchedPriorityPendingOpsRunner");
while (!stoken.stop_requested()) {
PendingOp op;
{
std::unique_lock lk(priority_pending_ops_mutex);
priority_pending_ops_cv.wait(lk, stoken,
[this] { return !priority_pending_ops.empty(); });
if (stoken.stop_requested()) {
break;
}
op = std::move(priority_pending_ops.front());
priority_pending_ops.pop();
}
master_semaphore.Wait(op.gpu_tick);
if (stoken.stop_requested()) {
break;
}
op.callback();
}
}
void DynamicState::Commit(const Instance& instance, const vk::CommandBuffer& cmdbuf) {
if (dirty_state.viewports) {
dirty_state.viewports = false;

View File

@@ -5,6 +5,7 @@
#include <condition_variable>
#include <mutex>
#include <thread>
#include <queue>
#include "common/unique_function.h"
@@ -401,10 +402,21 @@ public:
}
/// Defers an operation until the gpu has reached the current cpu tick.
/// Will be run when submitting or calling PopPendingOperations.
void DeferOperation(Common::UniqueFunction<void>&& func) {
pending_ops.emplace(std::move(func), CurrentTick());
}
/// Defers an operation until the gpu has reached the current cpu tick.
/// Runs as soon as possible in another thread.
void DeferPriorityOperation(Common::UniqueFunction<void>&& func) {
{
std::unique_lock lk(priority_pending_ops_mutex);
priority_pending_ops.emplace(std::move(func), CurrentTick());
}
priority_pending_ops_cv.notify_one();
}
static std::mutex submit_mutex;
private:
@@ -412,6 +424,8 @@ private:
void SubmitExecution(SubmitInfo& info);
void PriorityPendingOpsThread(std::stop_token stoken);
private:
const Instance& instance;
MasterSemaphore master_semaphore;
@@ -424,6 +438,10 @@ private:
u64 gpu_tick;
};
std::queue<PendingOp> pending_ops;
std::queue<PendingOp> priority_pending_ops;
std::mutex priority_pending_ops_mutex;
std::condition_variable_any priority_pending_ops_cv;
std::jthread priority_pending_ops_thread;
RenderState render_state;
bool is_rendering = false;
tracy::VkCtxScope* profiler_scope{};

View File

@@ -52,9 +52,6 @@ TextureCache::TextureCache(const Vulkan::Instance& instance_, Vulkan::Scheduler&
std::max<u64>(std::min(device_local_memory - min_vacancy_critical, min_spacing_critical),
DEFAULT_CRITICAL_GC_MEMORY));
trigger_gc_memory = static_cast<u64>((device_local_memory - mem_threshold) / 2);
downloaded_images_thread =
std::jthread([&](const std::stop_token& token) { DownloadedImagesThread(token); });
}
TextureCache::~TextureCache() = default;
@@ -125,33 +122,11 @@ void TextureCache::DownloadImageMemory(ImageId image_id) {
cmdbuf.copyImageToBuffer(image.GetImage(), vk::ImageLayout::eTransferSrcOptimal,
download_buffer.Handle(), image_download);
{
std::unique_lock lock(downloaded_images_mutex);
downloaded_images_queue.emplace(scheduler.CurrentTick(), image.info.guest_address, download,
download_size);
downloaded_images_cv.notify_one();
}
}
void TextureCache::DownloadedImagesThread(const std::stop_token& token) {
auto* memory = Core::Memory::Instance();
while (!token.stop_requested()) {
DownloadedImage image;
{
std::unique_lock lock{downloaded_images_mutex};
downloaded_images_cv.wait(lock, token,
[this] { return !downloaded_images_queue.empty(); });
if (token.stop_requested()) {
break;
}
image = downloaded_images_queue.front();
downloaded_images_queue.pop();
}
scheduler.GetMasterSemaphore()->Wait(image.tick);
memory->TryWriteBacking(std::bit_cast<u8*>(image.device_addr), image.download,
image.download_size);
}
scheduler.DeferPriorityOperation(
[this, device_addr = image.info.guest_address, download, download_size] {
Core::Memory::Instance()->TryWriteBacking(std::bit_cast<u8*>(device_addr), download,
download_size);
});
}
void TextureCache::MarkAsMaybeDirty(ImageId image_id, Image& image) {

View File

@@ -314,16 +314,6 @@ private:
Common::LeastRecentlyUsedCache<ImageId, u64> lru_cache;
PageTable page_table;
std::mutex mutex;
struct DownloadedImage {
u64 tick;
VAddr device_addr;
void* download;
size_t download_size;
};
std::queue<DownloadedImage> downloaded_images_queue;
std::mutex downloaded_images_mutex;
std::condition_variable_any downloaded_images_cv;
std::jthread downloaded_images_thread;
struct MetaDataInfo {
enum class Type {
CMask,