emit_arm64: Remove is_a64

src/dynarmic/backend/arm64/a32_address_space.cpp

@@ -303,21 +303,30 @@ EmittedBlockInfo A32AddressSpace::Emit(IR::Block block) {
     mem.unprotect();
 
     const EmitConfig emit_conf{
-        .tpidr_el0{},
+        .optimizations = conf.unsafe_optimizations ? conf.optimizations : conf.optimizations & all_safe_optimizations,
-        .tpidrro_el0{},
+
-        .cntfreq_el0{},
-        .dczid_el0{},
-        .ctr_el0{},
-        .is_a64 = false,
         .hook_isb = conf.hook_isb,
-        .enable_cycle_counting = conf.enable_cycle_counting,
+
+        .cntfreq_el0{},
+        .ctr_el0{},
+        .dczid_el0{},
+        .tpidrro_el0{},
+        .tpidr_el0{},
+
         .wall_clock_cntpct = conf.wall_clock_cntpct,
+        .enable_cycle_counting = conf.enable_cycle_counting,
+
         .always_little_endian = conf.always_little_endian,
+
         .descriptor_to_fpcr = [](const IR::LocationDescriptor& location) { return FP::FPCR{A32::LocationDescriptor{location}.FPSCR().Value()}; },
+        .emit_cond = EmitA32Cond,
+        .emit_condition_failed_terminal = EmitA32ConditionFailedTerminal,
+        .emit_terminal = EmitA32Terminal,
+
         .state_nzcv_offset = offsetof(A32JitState, cpsr_nzcv),
         .state_fpsr_offset = offsetof(A32JitState, fpsr),
+
         .coprocessors = conf.coprocessors,
-        .optimizations = conf.unsafe_optimizations ? conf.optimizations : conf.optimizations & all_safe_optimizations,
     };
     EmittedBlockInfo block_info = EmitArm64(code, std::move(block), emit_conf);
 

src/dynarmic/backend/arm64/a64_address_space.cpp

@@ -433,21 +433,30 @@ EmittedBlockInfo A64AddressSpace::Emit(IR::Block block) {
     mem.unprotect();
 
     const EmitConfig emit_conf{
+        .optimizations = conf.unsafe_optimizations ? conf.optimizations : conf.optimizations & all_safe_optimizations,
+
+        .hook_isb = conf.hook_isb,
+
         .tpidr_el0 = conf.tpidr_el0,
         .tpidrro_el0 = conf.tpidrro_el0,
         .cntfreq_el0 = conf.cntfrq_el0,
         .dczid_el0 = conf.dczid_el0,
         .ctr_el0 = conf.ctr_el0,
-        .is_a64 = true,
+
-        .hook_isb = conf.hook_isb,
-        .enable_cycle_counting = conf.enable_cycle_counting,
         .wall_clock_cntpct = conf.wall_clock_cntpct,
+        .enable_cycle_counting = conf.enable_cycle_counting,
+
         .always_little_endian = true,
+
         .descriptor_to_fpcr = [](const IR::LocationDescriptor& location) { return A64::LocationDescriptor{location}.FPCR(); },
+        .emit_cond = EmitA64Cond,
+        .emit_condition_failed_terminal = EmitA64ConditionFailedTerminal,
+        .emit_terminal = EmitA64Terminal,
+
         .state_nzcv_offset = offsetof(A64JitState, cpsr_nzcv),
         .state_fpsr_offset = offsetof(A64JitState, fpsr),
+
         .coprocessors{},
-        .optimizations = conf.unsafe_optimizations ? conf.optimizations : conf.optimizations & all_safe_optimizations,
     };
     EmittedBlockInfo block_info = EmitArm64(code, std::move(block), emit_conf);
 

src/dynarmic/backend/arm64/emit_arm64.cpp

@@ -8,7 +8,6 @@
 #include <fmt/ostream.h>
 #include <oaknut/oaknut.hpp>
 
-#include "dynarmic/backend/arm64/a32_jitstate.h"
 #include "dynarmic/backend/arm64/abi.h"
 #include "dynarmic/backend/arm64/emit_context.h"
 #include "dynarmic/backend/arm64/fpsr_manager.h"
@@ -191,15 +190,9 @@ EmittedBlockInfo EmitArm64(oaknut::CodeGenerator& code, IR::Block block, const E
         ASSERT(ctx.block.HasConditionFailedLocation());
         oaknut::Label pass;
 
-        if (conf.is_a64) {
+        pass = conf.emit_cond(code, ctx, ctx.block.GetCondition());
-            pass = EmitA64Cond(code, ctx, ctx.block.GetCondition());
+        EmitAddCycles(code, ctx, ctx.block.ConditionFailedCycleCount());
-            EmitAddCycles(code, ctx, ctx.block.ConditionFailedCycleCount());
+        conf.emit_condition_failed_terminal(code, ctx);
-            EmitA64ConditionFailedTerminal(code, ctx);
-        } else {
-            pass = EmitA32Cond(code, ctx, ctx.block.GetCondition());
-            EmitAddCycles(code, ctx, ctx.block.ConditionFailedCycleCount());
-            EmitA32ConditionFailedTerminal(code, ctx);
-        }
 
         code.l(pass);
     }
@@ -238,11 +231,7 @@ EmittedBlockInfo EmitArm64(oaknut::CodeGenerator& code, IR::Block block, const E
     reg_alloc.AssertNoMoreUses();
 
     EmitAddCycles(code, ctx, block.CycleCount());
-    if (conf.is_a64) {
+    conf.emit_terminal(code, ctx);
-        EmitA64Terminal(code, ctx);
-    } else {
-        EmitA32Terminal(code, ctx);
-    }
 
     ebi.size = code.ptr<CodePtr>() - ebi.entry_point;
     return ebi;

src/dynarmic/backend/arm64/emit_arm64.h

@@ -38,6 +38,8 @@ enum class Opcode;
 
 namespace Dynarmic::Backend::Arm64 {
 
+struct EmitContext;
+
 using CodePtr = std::byte*;
 
 enum class LinkTarget {
@@ -90,31 +92,39 @@ struct EmittedBlockInfo {
 };
 
 struct EmitConfig {
-    u64* tpidr_el0;
+    OptimizationFlag optimizations;
-    const u64* tpidrro_el0;
+    bool HasOptimization(OptimizationFlag f) const { return (f & optimizations) != no_optimizations; }
-    u64 cntfreq_el0;
+
-    u32 dczid_el0;
-    u32 ctr_el0;
-    bool is_a64;
     bool hook_isb;
-    bool enable_cycle_counting;
+
+    // System registers
+    u64 cntfreq_el0;
+    u32 ctr_el0;
+    u32 dczid_el0;
+    const u64* tpidrro_el0;
+    u64* tpidr_el0;
+
+    // Timing
     bool wall_clock_cntpct;
+    bool enable_cycle_counting;
+
+    // Endianness
     bool always_little_endian;
 
+    // Frontend specific callbacks
     FP::FPCR (*descriptor_to_fpcr)(const IR::LocationDescriptor& descriptor);
+    oaknut::Label (*emit_cond)(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Cond cond);
+    void (*emit_condition_failed_terminal)(oaknut::CodeGenerator& code, EmitContext& ctx);
+    void (*emit_terminal)(oaknut::CodeGenerator& code, EmitContext& ctx);
 
+    // State offsets
     size_t state_nzcv_offset;
     size_t state_fpsr_offset;
 
+    // A32 specific
     std::array<std::shared_ptr<A32::Coprocessor>, 16> coprocessors{};
-
-    OptimizationFlag optimizations;
-
-    bool HasOptimization(OptimizationFlag f) const { return (f & optimizations) != no_optimizations; }
 };
 
-struct EmitContext;
-
 EmittedBlockInfo EmitArm64(oaknut::CodeGenerator& code, IR::Block block, const EmitConfig& emit_conf);
 
 template<IR::Opcode op>

src/dynarmic/backend/arm64/emit_arm64_a64.cpp

@@ -19,8 +19,6 @@ namespace Dynarmic::Backend::Arm64 {
 
 using namespace oaknut::util;
 
-static constexpr int nzcv_c_flag_shift = 29;
-
 oaknut::Label EmitA64Cond(oaknut::CodeGenerator& code, EmitContext&, IR::Cond cond) {
     oaknut::Label pass;
     // TODO: Flags in host flags
@@ -145,7 +143,7 @@ void EmitIR<IR::Opcode::A64GetCFlag>(oaknut::CodeGenerator& code, EmitContext& c
     auto Wflag = ctx.reg_alloc.WriteW(inst);
     RegAlloc::Realize(Wflag);
     code.LDR(Wflag, Xstate, offsetof(A64JitState, cpsr_nzcv));
-    code.AND(Wflag, Wflag, 1 << nzcv_c_flag_shift);
+    code.AND(Wflag, Wflag, 1 << 29);
 }
 
 template<>

src/dynarmic/backend/arm64/emit_arm64_vector_floating_point.cpp

@@ -258,7 +258,7 @@ void EmitToFixed(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Inst* inst)
     });
 }
 
-template<typename JitState, typename Lambda>
+template<typename Lambda>
 static void EmitTwoOpFallbackWithoutRegAlloc(oaknut::CodeGenerator& code, EmitContext& ctx, oaknut::QReg Qresult, oaknut::QReg Qarg1, Lambda lambda, bool fpcr_controlled) {
     const auto fn = static_cast<mcl::equivalent_function_type<Lambda>*>(lambda);
 
@@ -274,7 +274,7 @@ static void EmitTwoOpFallbackWithoutRegAlloc(oaknut::CodeGenerator& code, EmitCo
     code.ADD(X0, Xscratch0, 0 * 16);
     code.ADD(X1, Xscratch0, 1 * 16);
     code.MOV(X2, fpcr);
-    code.ADD(X3, Xstate, offsetof(JitState, fpsr));
+    code.ADD(X3, Xstate, ctx.conf.state_fpsr_offset);
     code.STR(Qarg1, X1);
     code.BLR(Xscratch1);
 
@@ -289,7 +289,7 @@ static void EmitTwoOpFallbackWithoutRegAlloc(oaknut::CodeGenerator& code, EmitCo
     code.l(end);
 }
 
-template<size_t fpcr_controlled_arg_index = 1, typename JitState, typename Lambda>
+template<size_t fpcr_controlled_arg_index = 1, typename Lambda>
 static void EmitTwoOpFallback(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Inst* inst, Lambda lambda) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
     auto Qarg1 = ctx.reg_alloc.ReadQ(args[0]);
@@ -300,7 +300,7 @@ static void EmitTwoOpFallback(oaknut::CodeGenerator& code, EmitContext& ctx, IR:
     ctx.fpsr.Spill();
 
     const bool fpcr_controlled = args[fpcr_controlled_arg_index].GetImmediateU1();
-    EmitTwoOpFallbackWithoutRegAlloc<JitState>(code, ctx, Qresult, Qarg1, lambda, fpcr_controlled);
+    EmitTwoOpFallbackWithoutRegAlloc(code, ctx, Qresult, Qarg1, lambda, fpcr_controlled);
 }
 
 template<>
@@ -592,11 +592,7 @@ void EmitIR<IR::Opcode::FPVectorRoundInt16>(oaknut::CodeGenerator& code, EmitCon
         },
         mp::cartesian_product<rounding_list, exact_list>{});
 
-    if (ctx.conf.is_a64) {
+    EmitTwoOpFallback<3>(code, ctx, inst, lut.at(std::make_tuple(rounding, exact)));
-        EmitTwoOpFallback<3, A64JitState>(code, ctx, inst, lut.at(std::make_tuple(rounding, exact)));
-    } else {
-        EmitTwoOpFallback<3, A32JitState>(code, ctx, inst, lut.at(std::make_tuple(rounding, exact)));
-    }
 }
 
 template<>