a64_emit_x64: Implement fastmem for A64 frontend for 8-64 bit reads/writes

src/dynarmic/backend/x64/a32_emit_x64.cpp

@@ -100,7 +100,7 @@ A32EmitX64::BlockDescriptor A32EmitX64::Emit(IR::Block& block) {
     code.EnableWriting();
     SCOPE_EXIT { code.DisableWriting(); };
 
-    static const std::vector<HostLoc> gpr_order = [this] {
+    const std::vector<HostLoc> gpr_order = [this] {
         std::vector<HostLoc> gprs{any_gpr};
         if (conf.page_table) {
             gprs.erase(std::find(gprs.begin(), gprs.end(), HostLoc::R14));
@@ -1033,10 +1033,12 @@ void EmitWriteMemoryMov(BlockOfCode& code, const Xbyak::RegExp& addr, const Xbya
 }  // anonymous namespace
 
 template<std::size_t bitsize, auto callback>
-void A32EmitX64::ReadMemory(A32EmitContext& ctx, IR::Inst* inst) {
+void A32EmitX64::EmitMemoryRead(A32EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
+    const auto fastmem_marker = ShouldFastmem(ctx, inst);
 
-    if (!conf.page_table) {
+    if (!conf.page_table && !fastmem_marker) {
+        // Neither fastmem nor page table: Use callbacks
         ctx.reg_alloc.HostCall(inst, {}, args[0]);
         Devirtualize<callback>(conf.callbacks).EmitCall(code);
         return;
@@ -1047,22 +1049,27 @@ void A32EmitX64::ReadMemory(A32EmitContext& ctx, IR::Inst* inst) {
 
     const auto wrapped_fn = read_fallbacks[std::make_tuple(bitsize, vaddr.getIdx(), value.getIdx())];
 
-    if (const auto marker = ShouldFastmem(ctx, inst)) {
+    if (fastmem_marker) {
+        // Use fastmem
+        const auto src_ptr = r13 + vaddr;
+
         const auto location = code.getCurr();
-        EmitReadMemoryMov<bitsize>(code, value, r13 + vaddr);
+        EmitReadMemoryMov<bitsize>(code, value, src_ptr);
 
         fastmem_patch_info.emplace(
             Common::BitCast<u64>(location),
             FastmemPatchInfo{
                 Common::BitCast<u64>(code.getCurr()),
                 Common::BitCast<u64>(wrapped_fn),
-                *marker,
+                *fastmem_marker,
             });
 
         ctx.reg_alloc.DefineValue(inst, value);
         return;
     }
 
+    // Use page table
+    ASSERT(conf.page_table);
     Xbyak::Label abort, end;
 
     const auto src_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
@@ -1079,10 +1086,12 @@ void A32EmitX64::ReadMemory(A32EmitContext& ctx, IR::Inst* inst) {
 }
 
 template<std::size_t bitsize, auto callback>
-void A32EmitX64::WriteMemory(A32EmitContext& ctx, IR::Inst* inst) {
+void A32EmitX64::EmitMemoryWrite(A32EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
+    const auto fastmem_marker = ShouldFastmem(ctx, inst);
 
-    if (!conf.page_table) {
+    if (!conf.page_table && !fastmem_marker) {
+        // Neither fastmem nor page table: Use callbacks
         ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
         Devirtualize<callback>(conf.callbacks).EmitCall(code);
         return;
@@ -1093,21 +1102,26 @@ void A32EmitX64::WriteMemory(A32EmitContext& ctx, IR::Inst* inst) {
 
     const auto wrapped_fn = write_fallbacks[std::make_tuple(bitsize, vaddr.getIdx(), value.getIdx())];
 
-    if (const auto marker = ShouldFastmem(ctx, inst)) {
+    if (fastmem_marker) {
+        // Use fastmem
+        const auto dest_ptr = r13 + vaddr;
+
         const auto location = code.getCurr();
-        EmitWriteMemoryMov<bitsize>(code, r13 + vaddr, value);
+        EmitWriteMemoryMov<bitsize>(code, dest_ptr, value);
 
         fastmem_patch_info.emplace(
             Common::BitCast<u64>(location),
             FastmemPatchInfo{
                 Common::BitCast<u64>(code.getCurr()),
                 Common::BitCast<u64>(wrapped_fn),
-                *marker,
+                *fastmem_marker,
             });
 
         return;
     }
 
+    // Use page table
+    ASSERT(conf.page_table);
     Xbyak::Label abort, end;
 
     const auto dest_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
@@ -1122,35 +1136,35 @@ void A32EmitX64::WriteMemory(A32EmitContext& ctx, IR::Inst* inst) {
 }
 
 void A32EmitX64::EmitA32ReadMemory8(A32EmitContext& ctx, IR::Inst* inst) {
-    ReadMemory<8, &A32::UserCallbacks::MemoryRead8>(ctx, inst);
+    EmitMemoryRead<8, &A32::UserCallbacks::MemoryRead8>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory16(A32EmitContext& ctx, IR::Inst* inst) {
-    ReadMemory<16, &A32::UserCallbacks::MemoryRead16>(ctx, inst);
+    EmitMemoryRead<16, &A32::UserCallbacks::MemoryRead16>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory32(A32EmitContext& ctx, IR::Inst* inst) {
-    ReadMemory<32, &A32::UserCallbacks::MemoryRead32>(ctx, inst);
+    EmitMemoryRead<32, &A32::UserCallbacks::MemoryRead32>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32ReadMemory64(A32EmitContext& ctx, IR::Inst* inst) {
-    ReadMemory<64, &A32::UserCallbacks::MemoryRead64>(ctx, inst);
+    EmitMemoryRead<64, &A32::UserCallbacks::MemoryRead64>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory8(A32EmitContext& ctx, IR::Inst* inst) {
-    WriteMemory<8, &A32::UserCallbacks::MemoryWrite8>(ctx, inst);
+    EmitMemoryWrite<8, &A32::UserCallbacks::MemoryWrite8>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory16(A32EmitContext& ctx, IR::Inst* inst) {
-    WriteMemory<16, &A32::UserCallbacks::MemoryWrite16>(ctx, inst);
+    EmitMemoryWrite<16, &A32::UserCallbacks::MemoryWrite16>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory32(A32EmitContext& ctx, IR::Inst* inst) {
-    WriteMemory<32, &A32::UserCallbacks::MemoryWrite32>(ctx, inst);
+    EmitMemoryWrite<32, &A32::UserCallbacks::MemoryWrite32>(ctx, inst);
 }
 
 void A32EmitX64::EmitA32WriteMemory64(A32EmitContext& ctx, IR::Inst* inst) {
-    WriteMemory<64, &A32::UserCallbacks::MemoryWrite64>(ctx, inst);
+    EmitMemoryWrite<64, &A32::UserCallbacks::MemoryWrite64>(ctx, inst);
 }
 
 template<size_t bitsize, auto callback>

src/dynarmic/backend/x64/a32_emit_x64.h

@@ -106,9 +106,9 @@ protected:
 
     // Memory access helpers
     template<std::size_t bitsize, auto callback>
-    void ReadMemory(A32EmitContext& ctx, IR::Inst* inst);
+    void EmitMemoryRead(A32EmitContext& ctx, IR::Inst* inst);
     template<std::size_t bitsize, auto callback>
-    void WriteMemory(A32EmitContext& ctx, IR::Inst* inst);
+    void EmitMemoryWrite(A32EmitContext& ctx, IR::Inst* inst);
     template<std::size_t bitsize, auto callback>
     void ExclusiveReadMemory(A32EmitContext& ctx, IR::Inst* inst);
     template<std::size_t bitsize, auto callback>

src/dynarmic/backend/x64/a64_emit_x64.cpp

@@ -23,6 +23,7 @@
 #include "dynarmic/common/bit_util.h"
 #include "dynarmic/common/common_types.h"
 #include "dynarmic/common/scope_exit.h"
+#include "dynarmic/common/x64_disassemble.h"
 #include "dynarmic/frontend/A64/location_descriptor.h"
 #include "dynarmic/frontend/A64/types.h"
 #include "dynarmic/interface/exclusive_monitor.h"
@@ -60,6 +61,10 @@ A64EmitX64::A64EmitX64(BlockOfCode& code, A64::UserConfig conf, A64::Jit* jit_in
     GenTerminalHandlers();
     code.PreludeComplete();
     ClearFastDispatchTable();
+
+    exception_handler.SetFastmemCallback([this](u64 rip_){
+        return FastmemCallback(rip_);
+    });
 }
 
 A64EmitX64::~A64EmitX64() = default;
@@ -68,11 +73,14 @@ A64EmitX64::BlockDescriptor A64EmitX64::Emit(IR::Block& block) {
     code.EnableWriting();
     SCOPE_EXIT { code.DisableWriting(); };
 
-    static const std::vector<HostLoc> gpr_order = [this] {
+    const std::vector<HostLoc> gpr_order = [this] {
         std::vector<HostLoc> gprs{any_gpr};
         if (conf.page_table) {
             gprs.erase(std::find(gprs.begin(), gprs.end(), HostLoc::R14));
         }
+        if (conf.fastmem_pointer) {
+            gprs.erase(std::find(gprs.begin(), gprs.end(), HostLoc::R13));
+        }
         return gprs;
     }();
 
@@ -737,6 +745,40 @@ void A64EmitX64::EmitA64ClearExclusive(A64EmitContext&, IR::Inst*) {
     code.mov(code.byte[r15 + offsetof(A64JitState, exclusive_state)], u8(0));
 }
 
+std::optional<A64EmitX64::DoNotFastmemMarker> A64EmitX64::ShouldFastmem(A64EmitContext& ctx, IR::Inst* inst) const {
+    if (!conf.fastmem_pointer || !exception_handler.SupportsFastmem()) {
+        return std::nullopt;
+    }
+
+    const auto marker = std::make_tuple(ctx.Location(), ctx.GetInstOffset(inst));
+    if (do_not_fastmem.count(marker) > 0) {
+        return std::nullopt;
+    }
+    return marker;
+}
+
+FakeCall A64EmitX64::FastmemCallback(u64 rip_) {
+    const auto iter = fastmem_patch_info.find(rip_);
+
+    if (iter == fastmem_patch_info.end()) {
+        fmt::print("dynarmic: Segfault happened within JITted code at rip = {:016x}\n", rip_);
+        fmt::print("Segfault wasn't at a fastmem patch location!\n");
+        fmt::print("Now dumping code.......\n\n");
+        Common::DumpDisassembledX64((void*)(rip_ & ~u64(0xFFF)), 0x1000);
+        ASSERT_FALSE("iter != fastmem_patch_info.end()");
+    }
+
+    if (conf.recompile_on_fastmem_failure) {
+        const auto marker = iter->second.marker;
+        do_not_fastmem.emplace(marker);
+        InvalidateBasicBlocks({std::get<0>(marker)});
+    }
+    FakeCall ret;
+    ret.call_rip = iter->second.callback;
+    ret.ret_rip = iter->second.resume_rip;
+    return ret;
+}
+
 namespace {
 
 constexpr size_t page_bits = 12;
@@ -838,6 +880,39 @@ Xbyak::RegExp EmitVAddrLookup(BlockOfCode& code, A64EmitContext& ctx, size_t bit
     return page + tmp;
 }
 
+Xbyak::RegExp EmitFastmemVAddr(BlockOfCode& code, A64EmitContext& ctx, Xbyak::Label& abort, Xbyak::Reg64 vaddr) {
+    const size_t unused_top_bits = 64 - ctx.conf.page_table_address_space_bits;
+
+    if (unused_top_bits == 0) {
+        return r13 + vaddr;
+    } else if (ctx.conf.silently_mirror_page_table) {
+        Xbyak::Reg64 tmp = ctx.reg_alloc.ScratchGpr();
+        if (unused_top_bits < 32) {
+            code.mov(tmp, vaddr);
+            code.shl(tmp, int(unused_top_bits));
+            code.shr(tmp, int(unused_top_bits));
+        } else if (unused_top_bits == 32) {
+            code.mov(tmp.cvt32(), vaddr.cvt32());
+        } else {
+            code.mov(tmp.cvt32(), vaddr.cvt32());
+            code.and_(tmp, u32((1 << ctx.conf.page_table_address_space_bits) - 1));
+        }
+        return r13 + tmp;
+    } else {
+        if (ctx.conf.page_table_address_space_bits < 32) {
+            code.test(vaddr, u32(-(1 << ctx.conf.page_table_address_space_bits)));
+            code.jnz(abort, code.T_NEAR);
+        } else {
+            // TODO: Consider having TEST as above but coalesce 64-bit constant in register allocator
+            Xbyak::Reg64 tmp = ctx.reg_alloc.ScratchGpr();
+            code.mov(tmp, vaddr);
+            code.shr(tmp, int(ctx.conf.page_table_address_space_bits));
+            code.jnz(abort, code.T_NEAR);
+        }
+        return r13 + vaddr;
+    }
+}
+
 template<std::size_t bitsize>
 void EmitReadMemoryMov(BlockOfCode& code, const Xbyak::Reg64& value, const Xbyak::RegExp& addr) {
     switch (bitsize) {
@@ -880,9 +955,17 @@ void EmitWriteMemoryMov(BlockOfCode& code, const Xbyak::RegExp& addr, const Xbya
 
 }  // namespace
 
-template<std::size_t bitsize>
+template<std::size_t bitsize, auto callback>
-void A64EmitX64::EmitDirectPageTableMemoryRead(A64EmitContext& ctx, IR::Inst* inst) {
+void A64EmitX64::EmitMemoryRead(A64EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
+    const auto fastmem_marker = ShouldFastmem(ctx, inst);
+
+    if (!conf.page_table && !fastmem_marker) {
+        // Neither fastmem nor page table: Use callbacks
+        ctx.reg_alloc.HostCall(inst, {}, args[0]);
+        Devirtualize<callback>(conf.callbacks).EmitCall(code);
+        return;
+    }
 
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(args[0]);
     const Xbyak::Reg64 value = ctx.reg_alloc.ScratchGpr();
@@ -891,8 +974,27 @@ void A64EmitX64::EmitDirectPageTableMemoryRead(A64EmitContext& ctx, IR::Inst* in
 
     Xbyak::Label abort, end;
 
-    const auto src_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
+    if (fastmem_marker) {
-    EmitReadMemoryMov<bitsize>(code, value, src_ptr);
+        // Use fastmem
+        const auto src_ptr = EmitFastmemVAddr(code, ctx, abort, vaddr);
+
+        const auto location = code.getCurr();
+        EmitReadMemoryMov<bitsize>(code, value, src_ptr);
+
+        fastmem_patch_info.emplace(
+            Common::BitCast<u64>(location),
+            FastmemPatchInfo{
+                Common::BitCast<u64>(code.getCurr()),
+                Common::BitCast<u64>(wrapped_fn),
+                *fastmem_marker,
+            }
+        );
+    } else {
+        // Use page table
+        ASSERT(conf.page_table);
+        const auto src_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
+        EmitReadMemoryMov<bitsize>(code, value, src_ptr);
+    }
     code.L(end);
 
     code.SwitchToFarCode();
@@ -904,9 +1006,17 @@ void A64EmitX64::EmitDirectPageTableMemoryRead(A64EmitContext& ctx, IR::Inst* in
     ctx.reg_alloc.DefineValue(inst, value);
 }
 
-template<std::size_t bitsize>
+template<std::size_t bitsize, auto callback>
-void A64EmitX64::EmitDirectPageTableMemoryWrite(A64EmitContext& ctx, IR::Inst* inst) {
+void A64EmitX64::EmitMemoryWrite(A64EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
+    const auto fastmem_marker = ShouldFastmem(ctx, inst);
+
+    if (!conf.page_table && !fastmem_marker) {
+        // Neither fastmem nor page table: Use callbacks
+        ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
+        Devirtualize<callback>(conf.callbacks).EmitCall(code);
+        return;
+    }
 
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(args[0]);
     const Xbyak::Reg64 value = ctx.reg_alloc.UseGpr(args[1]);
@@ -915,8 +1025,27 @@ void A64EmitX64::EmitDirectPageTableMemoryWrite(A64EmitContext& ctx, IR::Inst* i
 
     Xbyak::Label abort, end;
 
-    const auto dest_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
+    if (fastmem_marker) {
-    EmitWriteMemoryMov<bitsize>(code, dest_ptr, value);
+        // Use fastmem
+        const auto dest_ptr = EmitFastmemVAddr(code, ctx, abort, vaddr);
+
+        const auto location = code.getCurr();
+        EmitWriteMemoryMov<bitsize>(code, dest_ptr, value);
+
+        fastmem_patch_info.emplace(
+            Common::BitCast<u64>(location),
+            FastmemPatchInfo{
+                Common::BitCast<u64>(code.getCurr()),
+                Common::BitCast<u64>(wrapped_fn),
+                *fastmem_marker,
+            }
+        );
+    } else {
+        // Use page table
+        ASSERT(conf.page_table);
+        const auto dest_ptr = EmitVAddrLookup(code, ctx, bitsize, abort, vaddr);
+        EmitWriteMemoryMov<bitsize>(code, dest_ptr, value);
+    }
     code.L(end);
 
     code.SwitchToFarCode();
@@ -927,47 +1056,19 @@ void A64EmitX64::EmitDirectPageTableMemoryWrite(A64EmitContext& ctx, IR::Inst* i
 }
 
 void A64EmitX64::EmitA64ReadMemory8(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryRead<8, &A64::UserCallbacks::MemoryRead8>(ctx, inst);
-        EmitDirectPageTableMemoryRead<8>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(inst, {}, args[0]);
-    Devirtualize<&A64::UserCallbacks::MemoryRead8>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64ReadMemory16(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryRead<16, &A64::UserCallbacks::MemoryRead16>(ctx, inst);
-        EmitDirectPageTableMemoryRead<16>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(inst, {}, args[0]);
-    Devirtualize<&A64::UserCallbacks::MemoryRead16>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64ReadMemory32(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryRead<32, &A64::UserCallbacks::MemoryRead32>(ctx, inst);
-        EmitDirectPageTableMemoryRead<32>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(inst, {}, args[0]);
-    Devirtualize<&A64::UserCallbacks::MemoryRead32>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64ReadMemory64(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryRead<64, &A64::UserCallbacks::MemoryRead64>(ctx, inst);
-        EmitDirectPageTableMemoryRead<64>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(inst, {}, args[0]);
-    Devirtualize<&A64::UserCallbacks::MemoryRead64>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64ReadMemory128(A64EmitContext& ctx, IR::Inst* inst) {
@@ -999,47 +1100,19 @@ void A64EmitX64::EmitA64ReadMemory128(A64EmitContext& ctx, IR::Inst* inst) {
 }
 
 void A64EmitX64::EmitA64WriteMemory8(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryWrite<8, &A64::UserCallbacks::MemoryWrite8>(ctx, inst);
-        EmitDirectPageTableMemoryWrite<8>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
-    Devirtualize<&A64::UserCallbacks::MemoryWrite8>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64WriteMemory16(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryWrite<16, &A64::UserCallbacks::MemoryWrite16>(ctx, inst);
-        EmitDirectPageTableMemoryWrite<16>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
-    Devirtualize<&A64::UserCallbacks::MemoryWrite16>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64WriteMemory32(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryWrite<32, &A64::UserCallbacks::MemoryWrite32>(ctx, inst);
-        EmitDirectPageTableMemoryWrite<32>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
-    Devirtualize<&A64::UserCallbacks::MemoryWrite32>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64WriteMemory64(A64EmitContext& ctx, IR::Inst* inst) {
-    if (conf.page_table) {
+    EmitMemoryWrite<64, &A64::UserCallbacks::MemoryWrite64>(ctx, inst);
-        EmitDirectPageTableMemoryWrite<64>(ctx, inst);
-        return;
-    }
-
-    auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    ctx.reg_alloc.HostCall(nullptr, {}, args[0], args[1]);
-    Devirtualize<&A64::UserCallbacks::MemoryWrite64>(conf.callbacks).EmitCall(code);
 }
 
 void A64EmitX64::EmitA64WriteMemory128(A64EmitContext& ctx, IR::Inst* inst) {

src/dynarmic/backend/x64/a64_emit_x64.h

@@ -78,15 +78,6 @@ protected:
     FastDispatchEntry& (*fast_dispatch_table_lookup)(u64) = nullptr;
     void GenTerminalHandlers();
 
-    template<std::size_t bitsize>
-    void EmitDirectPageTableMemoryRead(A64EmitContext& ctx, IR::Inst* inst);
-    template<std::size_t bitsize>
-    void EmitDirectPageTableMemoryWrite(A64EmitContext& ctx, IR::Inst* inst);
-    template<std::size_t bitsize, auto callback>
-    void EmitExclusiveReadMemory(A64EmitContext& ctx, IR::Inst* inst);
-    template<std::size_t bitsize, auto callback>
-    void EmitExclusiveWriteMemory(A64EmitContext& ctx, IR::Inst* inst);
-
     // Microinstruction emitters
     void EmitPushRSB(EmitContext& ctx, IR::Inst* inst);
 #define OPCODE(...)
@@ -100,6 +91,28 @@ protected:
     // Helpers
     std::string LocationDescriptorToFriendlyName(const IR::LocationDescriptor&) const override;
 
+    // Fastmem information
+    using DoNotFastmemMarker = std::tuple<IR::LocationDescriptor, std::ptrdiff_t>;
+    struct FastmemPatchInfo {
+        u64 resume_rip;
+        u64 callback;
+        DoNotFastmemMarker marker;
+    };
+    tsl::robin_map<u64, FastmemPatchInfo> fastmem_patch_info;
+    std::set<DoNotFastmemMarker> do_not_fastmem;
+    std::optional<DoNotFastmemMarker> ShouldFastmem(A64EmitContext& ctx, IR::Inst* inst) const;
+    FakeCall FastmemCallback(u64 rip);
+
+    // Memory access helpers
+    template<std::size_t bitsize, auto callback>
+    void EmitMemoryRead(A64EmitContext& ctx, IR::Inst* inst);
+    template<std::size_t bitsize, auto callback>
+    void EmitMemoryWrite(A64EmitContext& ctx, IR::Inst* inst);
+    template<std::size_t bitsize, auto callback>
+    void EmitExclusiveReadMemory(A64EmitContext& ctx, IR::Inst* inst);
+    template<std::size_t bitsize, auto callback>
+    void EmitExclusiveWriteMemory(A64EmitContext& ctx, IR::Inst* inst);
+
     // Terminal instruction emitters
     void EmitTerminalImpl(IR::Term::Interpret terminal, IR::LocationDescriptor initial_location, bool is_single_step) override;
     void EmitTerminalImpl(IR::Term::ReturnToDispatch terminal, IR::LocationDescriptor initial_location, bool is_single_step) override;

src/dynarmic/backend/x64/a64_interface.cpp

@@ -38,6 +38,9 @@ static std::function<void(BlockOfCode&)> GenRCP(const A64::UserConfig& conf) {
         if (conf.page_table) {
             code.mov(code.r14, Common::BitCast<u64>(conf.page_table));
         }
+        if (conf.fastmem_pointer) {
+            code.mov(code.r13, Common::BitCast<u64>(conf.fastmem_pointer));
+        }
     };
 }
 

src/dynarmic/interface/A32/config.h

@@ -173,6 +173,8 @@ struct UserConfig {
     void* fastmem_pointer = nullptr;
     /// Determines if instructions that pagefault should cause recompilation of that block
     /// with fastmem disabled.
+    /// Recompiled code will use the page_table if this is available, otherwise memory
+    /// accesses will hit the memory callbacks.
     bool recompile_on_fastmem_failure = true;
 
     // Coprocessors

src/dynarmic/interface/A64/config.h

@@ -202,7 +202,7 @@ struct UserConfig {
     void** page_table = nullptr;
     /// Declares how many valid address bits are there in virtual addresses.
     /// Determines the size of page_table. Valid values are between 12 and 64 inclusive.
-    /// This is only used if page_table is not nullptr.
+    /// This is only used if page_table or fastmem_pointer is not nullptr.
     size_t page_table_address_space_bits = 36;
     /// Masks out the first N bits in host pointers from the page table.
     /// The intention behind this is to allow users of Dynarmic to pack attributes in the
@@ -213,7 +213,7 @@ struct UserConfig {
     /// page table. If true, Dynarmic will silently mirror page_table's address space. If
     /// false, accessing memory outside of page_table bounds will result in a call to the
     /// relevant memory callback.
-    /// This is only used if page_table is not nullptr.
+    /// This is only used if page_table or fastmem_pointer is not nullptr.
     bool silently_mirror_page_table = true;
     /// Determines if the pointer in the page_table shall be offseted locally or globally.
     /// 'false' will access page_table[addr >> bits][addr & mask]
@@ -232,6 +232,18 @@ struct UserConfig {
     /// page boundary.
     bool only_detect_misalignment_via_page_table_on_page_boundary = false;
 
+    /// Fastmem Pointer
+    /// This should point to the beginning of a 2^page_table_address_space_bits bytes
+    /// address space which is in arranged just like what you wish for emulated memory to
+    /// be. If the host page faults on an address, the JIT will fallback to calling the
+    /// MemoryRead*/MemoryWrite* callbacks.
+    void* fastmem_pointer = nullptr;
+    /// Determines if instructions that pagefault should cause recompilation of that block
+    /// with fastmem disabled.
+    /// Recompiled code will use the page_table if this is available, otherwise memory
+    /// accesses will hit the memory callbacks.
+    bool recompile_on_fastmem_failure = true;
+
     /// This option relates to translation. Generally when we run into an unpredictable
     /// instruction the ExceptionRaised callback is called. If this is true, we define
     /// definite behaviour for some unpredictable instructions.