[dynarmic] Coalesce non-exclusive Write/Read fallback functions (#4158)

the general idea is to have a common procedure form whom to call
this way theres less "jumping around" for values of different sizes
additionally this **should** allow for better codegen since
most of the u8,u16,u32,u64 can be held within a u64
theoretically this means that you could deifne callbacks in suck
a way that it's essentially as costly as a `mov r64, m64`
but that's not doable due to the fact we have to do translations...

Is this a good change?
Primarily aimed for x86 and ARM to benefit, but I suppose LooooooongArch64 can benefit too.

Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4158
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>

6 files changed, 74 insertions, 129 deletions

diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.cpp b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
index 8be0abc4fa..a59aa3cac3 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_32.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
@@ -22,21 +22,15 @@ DynarmicCallbacks32::DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess
     , m_check_memory_access{m_debugger_enabled || !Settings::values.cpuopt_ignore_memory_aborts.GetValue()}
 {}
 
-u8 DynarmicCallbacks32::MemoryRead8(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read8(vaddr);
-}
-u16 DynarmicCallbacks32::MemoryRead16(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read16(vaddr);
-}
-u32 DynarmicCallbacks32::MemoryRead32(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read32(vaddr);
-}
-u64 DynarmicCallbacks32::MemoryRead64(u32 vaddr) {
-    CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read64(vaddr);
+u64 DynarmicCallbacks32::MemoryRead(u32 vaddr, size_t size) {
+    CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Read);
+    switch (size) {
+    case sizeof(u64): return m_memory.Read64(vaddr);
+    case sizeof(u32): return m_memory.Read32(vaddr);
+    case sizeof(u16): return m_memory.Read16(vaddr);
+    case sizeof(u8): return m_memory.Read8(vaddr);
+    default: UNREACHABLE();
+    }
 }
 
 std::optional<u32> DynarmicCallbacks32::MemoryReadCode(u32 vaddr) {
@@ -50,27 +44,17 @@ std::optional<u32> DynarmicCallbacks32::MemoryReadCode(u32 vaddr) {
     return cached_code_page.inst[(vaddr & Core::Memory::YUZU_PAGEMASK) / sizeof(u32)];
 }
 
-void DynarmicCallbacks32::MemoryWrite8(u32 vaddr, u8 value) {
-    if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write8(vaddr, value);
-    }
-}
-void DynarmicCallbacks32::MemoryWrite16(u32 vaddr, u16 value) {
-    if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write16(vaddr, value);
-    }
-}
-void DynarmicCallbacks32::MemoryWrite32(u32 vaddr, u32 value) {
-    if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write32(vaddr, value);
-    }
-}
-void DynarmicCallbacks32::MemoryWrite64(u32 vaddr, u64 value) {
-    if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write64(vaddr, value);
+void DynarmicCallbacks32::MemoryWrite(Dynarmic::A32::VAddr vaddr, u64 value, size_t size) {
+    if (CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Write)) {
+        switch (size) {
+        case sizeof(u64): return m_memory.Write64(vaddr, value);
+        case sizeof(u32): return m_memory.Write32(vaddr, u32(value));
+        case sizeof(u16): return m_memory.Write16(vaddr, u16(value));
+        case sizeof(u8): return m_memory.Write8(vaddr, u8(value));
+        default: UNREACHABLE();
+        }
     }
 }
-
 bool DynarmicCallbacks32::MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) {
     return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
             m_memory.WriteExclusive8(vaddr, value, expected);
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.h b/src/core/arm/dynarmic/arm_dynarmic_32.h
index 623597152f..22655dff47 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_32.h
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.h
@@ -30,18 +30,12 @@ class System;
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
     explicit DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess* process);
-    u8 MemoryRead8(u32 vaddr) override;
-    u16 MemoryRead16(u32 vaddr) override;
-    u32 MemoryRead32(u32 vaddr) override;
-    u64 MemoryRead64(u32 vaddr) override;
+    u64 MemoryRead(u32 vaddr, size_t size) override;
     std::optional<u32> MemoryReadCode(u32 vaddr) override;
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    void MemoryWrite8(u32 vaddr, u8 value) override;
-    void MemoryWrite16(u32 vaddr, u16 value) override;
-    void MemoryWrite32(u32 vaddr, u32 value) override;
-    void MemoryWrite64(u32 vaddr, u64 value) override;
+    void MemoryWrite(Dynarmic::A32::VAddr vaddr, u64 value, size_t size) override;
     bool MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) override;
     bool MemoryWriteExclusive16(u32 vaddr, u16 value, u16 expected) override;
     bool MemoryWriteExclusive32(u32 vaddr, u32 value, u32 expected) override;
diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
index 9db40b9356..894857a86d 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_64.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
@@ -10,6 +10,7 @@
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/k_process.h"
+#include "dynarmic/interface/A64/config.h"
 
 namespace Core {
 
@@ -21,21 +22,15 @@ DynarmicCallbacks64::DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess
     , m_check_memory_access{m_debugger_enabled || !Settings::values.cpuopt_ignore_memory_aborts.GetValue()}
 {}
 
-u8 DynarmicCallbacks64::MemoryRead8(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read8(vaddr);
-}
-u16 DynarmicCallbacks64::MemoryRead16(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read16(vaddr);
-}
-u32 DynarmicCallbacks64::MemoryRead32(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read32(vaddr);
-}
-u64 DynarmicCallbacks64::MemoryRead64(u64 vaddr) {
-    CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
-    return m_memory.Read64(vaddr);
+u64 DynarmicCallbacks64::MemoryRead(u64 vaddr, size_t size) {
+    CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Read);
+    switch (size) {
+    case sizeof(u64): return m_memory.Read64(vaddr);
+    case sizeof(u32): return m_memory.Read32(vaddr);
+    case sizeof(u16): return m_memory.Read16(vaddr);
+    case sizeof(u8): return m_memory.Read8(vaddr);
+    default: UNREACHABLE();
+    }
 }
 Dynarmic::A64::Vector DynarmicCallbacks64::MemoryRead128(u64 vaddr) {
     CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Read);
@@ -53,24 +48,15 @@ std::optional<u32> DynarmicCallbacks64::MemoryReadCode(u64 vaddr) {
     return cached_code_page.inst[(vaddr & Core::Memory::YUZU_PAGEMASK) / sizeof(u32)];
 }
 
-void DynarmicCallbacks64::MemoryWrite8(u64 vaddr, u8 value) {
-    if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write8(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite16(u64 vaddr, u16 value) {
-    if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write16(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite32(u64 vaddr, u32 value) {
-    if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write32(vaddr, value);
-    }
-}
-void DynarmicCallbacks64::MemoryWrite64(u64 vaddr, u64 value) {
-    if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
-        m_memory.Write64(vaddr, value);
+void DynarmicCallbacks64::MemoryWrite(Dynarmic::A64::VAddr vaddr, u64 value, std::size_t size) {
+    if (CheckMemoryAccess(vaddr, size, Kernel::DebugWatchpointType::Write)) {
+        switch (size) {
+        case sizeof(u64): return m_memory.Write64(vaddr, u64(value));
+        case sizeof(u32): return m_memory.Write32(vaddr, u32(value));
+        case sizeof(u16): return m_memory.Write16(vaddr, u16(value));
+        case sizeof(u8): return m_memory.Write8(vaddr, u8(value));
+        default: UNREACHABLE();
+        }
     }
 }
 void DynarmicCallbacks64::MemoryWrite128(u64 vaddr, Dynarmic::A64::Vector value) {
diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.h b/src/core/arm/dynarmic/arm_dynarmic_64.h
index 4c8570913f..5ea5dde8a6 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_64.h
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.h
@@ -16,6 +16,7 @@
 #include "common/hash.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/dynarmic/dynarmic_exclusive_monitor.h"
+#include "dynarmic/interface/A64/config.h"
 
 namespace Core::Memory {
 class Memory;
@@ -36,19 +37,13 @@ class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
     explicit DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess* process);
 
-    u8 MemoryRead8(u64 vaddr) override;
-    u16 MemoryRead16(u64 vaddr) override;
-    u32 MemoryRead32(u64 vaddr) override;
-    u64 MemoryRead64(u64 vaddr) override;
+    u64 MemoryRead(u64 vaddr, size_t size) override;
     Dynarmic::A64::Vector MemoryRead128(u64 vaddr) override;
     std::optional<u32> MemoryReadCode(u64 vaddr) override;
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    void MemoryWrite8(u64 vaddr, u8 value) override;
-    void MemoryWrite16(u64 vaddr, u16 value) override;
-    void MemoryWrite32(u64 vaddr, u32 value) override;
-    void MemoryWrite64(u64 vaddr, u64 value) override;
+    void MemoryWrite(Dynarmic::A64::VAddr vaddr, u64 value, std::size_t size) override;
     void MemoryWrite128(u64 vaddr, Dynarmic::A64::Vector value) override;
     bool MemoryWriteExclusive8(u64 vaddr, std::uint8_t value, std::uint8_t expected) override;
     bool MemoryWriteExclusive16(u64 vaddr, std::uint16_t value, std::uint16_t expected) override;
diff --git a/src/core/hle/service/jit/jit_context.cpp b/src/core/hle/service/jit/jit_context.cpp
index 36bf5d6649..e1e8eeadbe 100644
--- a/src/core/hle/service/jit/jit_context.cpp
+++ b/src/core/hle/service/jit/jit_context.cpp
@@ -71,17 +71,14 @@ public:
     void InstructionSynchronizationBarrierRaised() override {
         last_code_addr = u64(-1); //reset back, force refetch
     }
-    u8 MemoryRead8(u64 vaddr) override {
-        return ReadMemory<u8>(vaddr);
-    }
-    u16 MemoryRead16(u64 vaddr) override {
-        return ReadMemory<u16>(vaddr);
-    }
-    u32 MemoryRead32(u64 vaddr) override {
-        return ReadMemory<u32>(vaddr);
-    }
-    u64 MemoryRead64(u64 vaddr) override {
-        return ReadMemory<u64>(vaddr);
+    u64 MemoryRead(u64 vaddr, size_t size) override {
+        switch (size) {
+        case sizeof(u64): return ReadMemory<u64>(vaddr);
+        case sizeof(u32): return ReadMemory<u32>(vaddr);
+        case sizeof(u16): return ReadMemory<u16>(vaddr);
+        case sizeof(u8): return ReadMemory<u8>(vaddr);
+        default: UNREACHABLE();
+        }
     }
     u128 MemoryRead128(u64 vaddr) override {
         return ReadMemory<u128>(vaddr);
@@ -89,22 +86,19 @@ public:
     std::string MemoryReadCString(u64 vaddr) {
         std::string result{};
         u8 next;
-        while ((next = MemoryRead8(vaddr++)) != 0)
+        while ((next = u8(MemoryRead(vaddr++, sizeof(u8)))) != 0)
             result += char(next);
         return result;
     }
 
-    void MemoryWrite8(u64 vaddr, u8 value) override {
-        WriteMemory<u8>(vaddr, value);
-    }
-    void MemoryWrite16(u64 vaddr, u16 value) override {
-        WriteMemory<u16>(vaddr, value);
-    }
-    void MemoryWrite32(u64 vaddr, u32 value) override {
-        WriteMemory<u32>(vaddr, value);
-    }
-    void MemoryWrite64(u64 vaddr, u64 value) override {
-        WriteMemory<u64>(vaddr, value);
+    void MemoryWrite(u64 vaddr, u64 value, size_t size) override {
+        switch (size) {
+        case sizeof(u64): WriteMemory<u64>(vaddr, u64(value)); break;
+        case sizeof(u32): WriteMemory<u32>(vaddr, u32(value)); break;
+        case sizeof(u16): WriteMemory<u16>(vaddr, u16(value)); break;
+        case sizeof(u8): WriteMemory<u8>(vaddr, u8(value)); break;
+        default: UNREACHABLE();
+        }
     }
     void MemoryWrite128(u64 vaddr, u128 value) override {
         WriteMemory<u128>(vaddr, value);
@@ -193,14 +187,14 @@ public:
         // The loaded NRO file has ELF relocations that must be processed before it can run.
         // Normally this would be processed by RTLD, but in HLE context, we don't have
         // the linker available, so we have to do it ourselves.
-        const VAddr mod_offset{callbacks->MemoryRead32(4)};
-        if (callbacks->MemoryRead32(mod_offset) != Common::MakeMagic('M', 'O', 'D', '0'))
+        const VAddr mod_offset{callbacks->MemoryRead(4, sizeof(u32))};
+        if (callbacks->MemoryRead(mod_offset, sizeof(u32)) != Common::MakeMagic('M', 'O', 'D', '0'))
             return false;
 
         // For more info about dynamic entries, see the ELF ABI specification:
         // https://refspecs.linuxbase.org/elf/gabi4+/ch5.dynamic.html
         // https://refspecs.linuxbase.org/elf/gabi4+/ch4.reloc.html
-        VAddr dynamic_offset{mod_offset + callbacks->MemoryRead32(mod_offset + 4)};
+        VAddr dynamic_offset{mod_offset + callbacks->MemoryRead(mod_offset + 4, sizeof(u32))};
         VAddr rela_dyn = 0, relr_dyn = 0;
         size_t num_rela = 0, num_relr = 0;
         while (true) {
@@ -222,8 +216,8 @@ public:
         for (size_t i = 0; i < num_rela; i++) {
             const auto rela{callbacks->ReadMemory<Elf64_Rela>(rela_dyn + i * sizeof(Elf64_Rela))};
             if (Elf64RelType(rela.r_info) == ElfAArch64Relative) {
-                const VAddr contents{callbacks->MemoryRead64(rela.r_offset)};
-                callbacks->MemoryWrite64(rela.r_offset, contents + rela.r_addend);
+                const VAddr contents{callbacks->MemoryRead(rela.r_offset, sizeof(u64))};
+                callbacks->MemoryWrite(rela.r_offset, contents + rela.r_addend, sizeof(u64));
             }
         }
 
@@ -231,7 +225,7 @@ public:
         for (size_t i = 0; i < num_relr; i++) {
             const auto relr = callbacks->ReadMemory<Elf64_Relr>(relr_dyn + i * sizeof(Elf64_Relr));
             const auto incr = [&](VAddr where) {
-                callbacks->MemoryWrite64(where, callbacks->MemoryRead64(where) + relocbase);
+                callbacks->MemoryWrite(where, callbacks->MemoryRead(where, sizeof(u64)) + relocbase, sizeof(u64));
             };
             if ((relr & 1) == 0) {
                 // where pointer
@@ -294,7 +288,7 @@ public:
         if (argument_stack.size() > 8) {
             const VAddr new_sp = Common::AlignDown(top_of_stack - (argument_stack.size() - 8) * sizeof(u64), STACK_ALIGN);
             for (size_t i = 8; i < argument_stack.size(); i++)
-                callbacks->MemoryWrite64(new_sp + (i - 8) * sizeof(u64), argument_stack[i]);
+                callbacks->MemoryWrite(new_sp + (i - 8) * sizeof(u64), argument_stack[i], sizeof(u64));
             jit->SetSP(new_sp);
         }
         // Reset the call state for the next invocation
@@ -385,17 +379,17 @@ void DynarmicCallbacks64::CallSVC(u32 swi) {
 
         if (dest < src) {
             for (size_t i = 0; i < n; i++)
-                MemoryWrite8(dest + i, MemoryRead8(src + i));
+                MemoryWrite(dest + i, u8(MemoryRead(src + i, sizeof(u8))), sizeof(u8));
         } else {
             for (size_t i = n; i > 0; i--)
-                MemoryWrite8(dest + i - 1, MemoryRead8(src + i - 1));
+                MemoryWrite(dest + i - 1, u8(MemoryRead(src + i - 1, sizeof(u8))), sizeof(u8));
         }
     } else if (pc == parent.helpers[size_t(HelperFn::Memset)]) {
         const VAddr dest{parent.jit->GetRegister(0)};
         const u64 c{parent.jit->GetRegister(1)};
         const size_t n{parent.jit->GetRegister(2)};
         for (size_t i = 0; i < n; i++)
-            MemoryWrite8(dest + i, u8(c));
+            MemoryWrite(dest + i, u8(c), sizeof(u8));
     } else if (pc == parent.helpers[size_t(HelperFn::Resolve)]) {
         // X0 contains a char* for a symbol to resolve
         const auto name{MemoryReadCString(parent.jit->GetRegister(0))};
@@ -422,7 +416,7 @@ void DynarmicCallbacks64::CallSVC(u32 swi) {
 }
 
 void DynarmicCallbacks64::ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) {
-    auto const inst = MemoryRead32(pc);
+    auto const inst = MemoryRead(pc, sizeof(u32));
     LOG_CRITICAL(Service_JIT, "{} PC @ {:08x}, data = {:08x}", exception, pc, inst);
     parent.jit->HaltExecution();
 }
diff --git a/src/core/tools/freezer.cpp b/src/core/tools/freezer.cpp
index 73a86b3c8c..606ac2da42 100644
--- a/src/core/tools/freezer.cpp
+++ b/src/core/tools/freezer.cpp
@@ -33,18 +33,10 @@ u64 MemoryReadWidth(Core::Memory::Memory& memory, u32 width, VAddr addr) {
 
 void MemoryWriteWidth(Core::Memory::Memory& memory, u32 width, VAddr addr, u64 value) {
     switch (width) {
-    case 1:
-        memory.Write8(addr, static_cast<u8>(value));
-        break;
-    case 2:
-        memory.Write16(addr, static_cast<u16>(value));
-        break;
-    case 4:
-        memory.Write32(addr, static_cast<u32>(value));
-        break;
-    case 8:
-        memory.Write64(addr, value);
-        break;
+    case sizeof(u64): return memory.Write64(addr, value);
+    case sizeof(u32): return memory.Write32(addr, u32(value));
+    case sizeof(u16): return memory.Write16(addr, u16(value));
+    case sizeof(u8): return memory.Write8(addr, u8(value));
     default:
         UNREACHABLE();
     }