Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[ppu] Fix arm64 trampoline #12242

Merged
merged 3 commits into from
Jun 20, 2022
Merged

[ppu] Fix arm64 trampoline #12242

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
195ee4a
[ppu] fix a macOS arm64 regression
sguo35 Jun 17, 2022
f8574d0
[3rdparty] Disable warnings for soundtouch dependency
sguo35 Jun 18, 2022
e10bb81
[ppu] Fix function trampoline for arm64
sguo35 Jun 18, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions 3rdparty/SoundTouch/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,8 @@ target_compile_definitions(soundtouch PUBLIC
SOUNDTOUCH_FLOAT_SAMPLES;
)

target_compile_options(soundtouch PUBLIC "-w")

if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86|X86|amd64|AMD64|em64t|EM64T)")
target_compile_definitions(soundtouch PUBLIC
SOUNDTOUCH_ALLOW_SSE
Expand Down
4 changes: 4 additions & 0 deletions Utilities/JIT.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -871,6 +871,10 @@ jit_compiler::jit_compiler(const std::unordered_map<std::string, u64>& _link, co
std::string result;

auto null_mod = std::make_unique<llvm::Module> ("null_", *m_context);
#if defined(__APPLE__) && defined(ARCH_ARM64)
// Force override triple on Apple arm64 or we'll get linking errors.
null_mod->setTargetTriple(llvm::Triple::normalize(utils::c_llvm_default_triple));
#endif

if (_link.empty())
{
Expand Down
98 changes: 57 additions & 41 deletions rpcs3/Emu/Cell/PPUFunction.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1889,6 +1889,60 @@ extern std::string ppu_get_variable_name(const std::string& _module, u32 vnid)
return fmt::format("0x%08X", vnid);
}

#if defined(ARCH_X64)

auto gen_ghc_cpp_trampoline(ppu_intrp_func_t fn_target)
{
return [fn_target] (native_asm& c, auto& args)
{
using namespace asmjit;

// Take second ghc arg
c.mov(args[0], x86::rbp);
c.mov(args[2].r32(), x86::dword_ptr(args[0], ::offset32(&ppu_thread::cia)));
c.add(args[2], x86::qword_ptr(reinterpret_cast<u64>(&vm::g_base_addr)));
c.jmp(fn_target);
};
}

#elif defined(ARCH_ARM64)

auto gen_ghc_cpp_trampoline(ppu_intrp_func_t fn_target)
{
return [fn_target] (native_asm& c, auto& args)
{
using namespace asmjit;

// Take second ghc arg
c.mov(args[0], a64::x20);

Label cia_offset = c.newLabel();
c.ldr(a64::x11, arm::Mem(cia_offset));
c.ldr(a64::x26, arm::Mem(args[0], a64::x11));

Label base_addr = c.newLabel();
c.ldr(a64::x22, arm::Mem(base_addr));
c.ldr(a64::x22, arm::Mem(a64::x22));

c.add(args[2], a64::x22, a64::x26);

Label jmp_target = c.newLabel();
c.ldr(a64::x22, arm::Mem(jmp_target));
c.br(a64::x22);

c.bind(base_addr);
c.embedUInt64(reinterpret_cast<u64>(&vm::g_base_addr));
c.bind(cia_offset);
c.embedUInt64(static_cast<u64>(::offset32(&ppu_thread::cia)));
c.bind(jmp_target);
c.embedUInt64(reinterpret_cast<u64>(fn_target));
};
}

#else
#error "Not implemented!"
#endif

std::vector<ppu_intrp_func_t>& ppu_function_manager::access(bool ghc)
{
static std::vector<ppu_intrp_func_t> list
Expand All @@ -1907,33 +1961,11 @@ std::vector<ppu_intrp_func_t>& ppu_function_manager::access(bool ghc)
},
};

#if defined(ARCH_X64)
static std::vector<ppu_intrp_func_t> list_ghc
{
build_function_asm<ppu_intrp_func_t>("ppu_unregistered", [](native_asm& c, auto& args)
{
using namespace asmjit;

// Take second ghc arg
c.mov(args[0], x86::rbp);
c.mov(args[2].r32(), x86::dword_ptr(args[0], ::offset32(&ppu_thread::cia)));
c.add(args[2], x86::qword_ptr(reinterpret_cast<u64>(&vm::g_base_addr)));
c.jmp(list[0]);
}),
build_function_asm<ppu_intrp_func_t>("ppu_return", [](native_asm& c, auto& args)
{
using namespace asmjit;

// Take second ghc arg
c.mov(args[0], x86::rbp);
c.mov(args[2].r32(), x86::dword_ptr(args[0], ::offset32(&ppu_thread::cia)));
c.add(args[2], x86::qword_ptr(reinterpret_cast<u64>(&vm::g_base_addr)));
c.jmp(list[1]);
}),
build_function_asm<ppu_intrp_func_t>("ppu_unregistered", gen_ghc_cpp_trampoline(list[0])),
build_function_asm<ppu_intrp_func_t>("ppu_return", gen_ghc_cpp_trampoline(list[1])),
};
#elif defined(ARCH_ARM64)
static std::vector<ppu_intrp_func_t> list_ghc(list);
#endif

return ghc ? list_ghc : list;
}
Expand All @@ -1945,23 +1977,7 @@ u32 ppu_function_manager::add_function(ppu_intrp_func_t function)

list.push_back(function);

// Generate trampoline
#if defined(ARCH_X64)
list2.push_back(build_function_asm<ppu_intrp_func_t>("", [&](native_asm& c, auto& args)
{
using namespace asmjit;

// Take second ghc arg
c.mov(args[0], x86::rbp);
c.mov(args[2].r32(), x86::dword_ptr(args[0], ::offset32(&ppu_thread::cia)));
c.add(args[2], x86::qword_ptr(reinterpret_cast<u64>(&vm::g_base_addr)));
c.jmp(function);
}));
#elif defined(ARCH_ARM64)
list2.push_back(function);
#else
#error "Not implemented"
#endif
list2.push_back(build_function_asm<ppu_intrp_func_t>("", gen_ghc_cpp_trampoline(function)));

return ::size32(list) - 1;
}
6 changes: 6 additions & 0 deletions rpcs3/Emu/Cell/PPUThread.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1330,6 +1330,9 @@ void ppu_thread::cpu_task()
}
case ppu_cmd::lle_call:
{
#ifdef __APPLE__
pthread_jit_write_protect_np(true);
#endif
const vm::ptr<u32> opd(arg < 32 ? vm::cast(gpr[arg]) : vm::cast(arg));
cmd_pop(), fast_call(opd[0], opd[1]);
break;
Expand All @@ -1341,6 +1344,9 @@ void ppu_thread::cpu_task()
}
case ppu_cmd::opd_call:
{
#ifdef __APPLE__
pthread_jit_write_protect_np(true);
#endif
const ppu_func_opd_t opd = cmd_get(1).as<ppu_func_opd_t>();
cmd_pop(1), fast_call(opd.addr, opd.rtoc);
break;
Expand Down