Added some shifts

src/binary_translation/ARMFuncs.cpp

@@ -0,0 +1,168 @@
+#include "ARMFuncs.h"
+#include "InstructionBlock.h"
+
+ARMFuncs::ShiftTN ARMFuncs::DecodeImmShift(InstructionBlock* instruction, u32 type, u32 imm5)
+{
+    auto ir_builder = instruction->IrBuilder();
+    switch (type)
+    {
+    case 0: return{ SRType::LSL, ir_builder->getInt32(imm5) };
+    case 1: return{ SRType::LSR, ir_builder->getInt32(imm5 ? imm5 : 32) };
+    case 2: return{ SRType::ASR, ir_builder->getInt32(imm5 ? imm5 : 32) };
+    case 3:
+        if (imm5)
+            return{ SRType::ROR, ir_builder->getInt32(imm5) };
+        else
+            return{ SRType::RRX, ir_builder->getInt32(1) };
+    default: assert(false, "Invalid shift type");
+    }
+}
+
+ARMFuncs::SRType ARMFuncs::DecodeRegShift(u32 type)
+{
+    switch (type)
+    {
+    case 0: return SRType::LSL;
+    case 1: return SRType::LSR;
+    case 2: return SRType::ASR;
+    case 3: return SRType::ROR;
+    default: assert(false, "Invalid shift type");
+    }
+}
+
+llvm::Value* ARMFuncs::Shift(InstructionBlock* instruction, llvm::Value* value, SRType type, llvm::Value* amount, llvm::Value* carry_in)
+{
+    return Shift_C(instruction, value, type, amount, carry_in).result;
+}
+
+ARMFuncs::ResultCarry ARMFuncs::Shift_C(InstructionBlock* instruction, llvm::Value* value, SRType type, llvm::Value* amount, llvm::Value* carry_in)
+{
+    auto ir_builder = instruction->IrBuilder();
+
+    // amount_zero_basic_block will not recieve any code, it used only for the phi
+    auto amount_zero_basic_block = instruction->CreateBasicBlock("ShiftCAmount0");
+    auto amount_not_zero_basic_block = instruction->CreateBasicBlock("ShiftCAmountNot0");
+    auto phi_basic_block = instruction->CreateBasicBlock("ShiftCPhi");
+
+    ir_builder->CreateCondBr(ir_builder->CreateICmpEQ(amount, ir_builder->getInt32(0)), amount_zero_basic_block, amount_not_zero_basic_block);
+
+    ir_builder->SetInsertPoint(amount_zero_basic_block);
+    ir_builder->CreateBr(phi_basic_block);
+
+    ir_builder->SetInsertPoint(amount_not_zero_basic_block);
+    ResultCarry result_amount_not_zero = {};
+    switch (type)
+    {
+    case SRType::LSL: result_amount_not_zero = LSL_C(instruction, value, amount); break;
+    case SRType::LSR: result_amount_not_zero = LSR_C(instruction, value, amount); break;
+    case SRType::ASR: result_amount_not_zero = ASR_C(instruction, value, amount); break;
+    case SRType::ROR: result_amount_not_zero = ROR_C(instruction, value, amount); break;
+    case SRType::RRX: result_amount_not_zero = RRX_C(instruction, value, carry_in); break;
+    default: assert(false, "Invalid shift type");
+    }
+    auto pred = ir_builder->GetInsertBlock(); // The basic block might have changed and needs to be current for the phi
+    ir_builder->CreateBr(phi_basic_block);
+
+    ir_builder->SetInsertPoint(phi_basic_block);
+    auto result_phi = ir_builder->CreatePHI(ir_builder->getInt32Ty(), 2);
+    auto carry_phi = ir_builder->CreatePHI(ir_builder->getInt1Ty(), 2);
+
+    result_phi->addIncoming(value, amount_zero_basic_block);
+    result_phi->addIncoming(result_amount_not_zero.result, pred);
+    carry_phi->addIncoming(carry_in, amount_zero_basic_block);
+    carry_phi->addIncoming(result_amount_not_zero.carry, pred);
+
+    return{ result_phi, carry_phi };
+}
+
+// Generates code for LSL, LSR that checks for 0 shift
+llvm::Value* ShiftZeroCheck(
+    InstructionBlock *instruction, llvm::Value* x, llvm::Value* shift,
+    std::function<ARMFuncs::ResultCarry(InstructionBlock *, llvm::Value*, llvm::Value*)> non_zero_function)
+{
+    auto ir_builder = instruction->IrBuilder();
+
+    // amount_zero_basic_block will not recieve any code, it used only for the phi
+    auto amount_zero_basic_block = instruction->CreateBasicBlock("ShiftZeroCheckAmount0");
+    auto amount_not_zero_basic_block = instruction->CreateBasicBlock("ShiftZeroCheckAmountNot0");
+    auto phi_basic_block = instruction->CreateBasicBlock("ShiftZeroCheckPhi");
+
+    ir_builder->CreateCondBr(ir_builder->CreateICmpEQ(shift, ir_builder->getInt32(0)), amount_zero_basic_block, amount_not_zero_basic_block);
+
+    ir_builder->SetInsertPoint(amount_zero_basic_block);
+    ir_builder->CreateBr(phi_basic_block);
+
+    ir_builder->SetInsertPoint(amount_not_zero_basic_block);
+    auto result_amount_not_zero = non_zero_function(instruction, x, shift);
+    auto pred = ir_builder->GetInsertBlock(); // The basic block might have changed and needs to be current for the phi
+    ir_builder->CreateBr(phi_basic_block);
+
+    ir_builder->SetInsertPoint(phi_basic_block);
+    auto phi = ir_builder->CreatePHI(ir_builder->getInt32Ty(), 2);
+
+    phi->addIncoming(x, amount_zero_basic_block);
+    phi->addIncoming(result_amount_not_zero.result, pred);
+
+    return phi;
+}
+
+ARMFuncs::ResultCarry ARMFuncs::LSL_C(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    auto ir_builder = instruction->IrBuilder();
+    auto N = ir_builder->getInt32(32);
+
+    auto result = ir_builder->CreateShl(x, shift);
+    auto carry = ir_builder->CreateTrunc(ir_builder->CreateLShr(x, ir_builder->CreateSub(N, shift, "", true, true)), ir_builder->getInt1Ty());
+    return{ result, carry };
+}
+
+llvm::Value* ARMFuncs::LSL(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    return ShiftZeroCheck(instruction, x, shift, &ARMFuncs::LSL_C);
+}
+
+ARMFuncs::ResultCarry ARMFuncs::LSR_C(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    auto ir_builder = instruction->IrBuilder();
+    auto one = ir_builder->getInt32(1);
+
+    auto result = ir_builder->CreateLShr(x, shift);
+    auto carry = ir_builder->CreateTrunc(ir_builder->CreateLShr(x, ir_builder->CreateSub(shift, one, "", true, true)), ir_builder->getInt1Ty());
+    return{ result, carry };
+}
+
+llvm::Value* ARMFuncs::LSR(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    return ShiftZeroCheck(instruction, x, shift, &ARMFuncs::LSR_C);
+}
+
+ARMFuncs::ResultCarry ARMFuncs::ASR_C(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    auto ir_builder = instruction->IrBuilder();
+    auto one = ir_builder->getInt32(1);
+
+    auto result = ir_builder->CreateAShr(x, shift);
+    auto carry = ir_builder->CreateTrunc(ir_builder->CreateLShr(x, ir_builder->CreateSub(shift, one, "", true, true)), ir_builder->getInt1Ty());
+    return{ result, carry };
+}
+
+ARMFuncs::ResultCarry ARMFuncs::ROR_C(InstructionBlock* instruction, llvm::Value* x, llvm::Value* shift)
+{
+    auto ir_builder = instruction->IrBuilder();
+    auto N = ir_builder->getInt32(32);
+    auto m = ir_builder->CreateURem(shift, N);
+
+    auto result = ir_builder->CreateOr(LSR(instruction, x, m), LSL(instruction, x, ir_builder->CreateSub(N, m)));
+    auto carry = ir_builder->CreateTrunc(ir_builder->CreateLShr(result, ir_builder->getInt32(31)), ir_builder->getInt1Ty());
+    return{ result, carry };
+}
+
+ARMFuncs::ResultCarry ARMFuncs::RRX_C(InstructionBlock* instruction, llvm::Value* x, llvm::Value* carry_in)
+{
+    auto ir_builder = instruction->IrBuilder();
+
+    auto result = ir_builder->CreateLShr(x, 1);
+    result = ir_builder->CreateOr(result, ir_builder->CreateShl(ir_builder->CreateZExt(carry_in, ir_builder->getInt32Ty()), 31));
+    auto carry = ir_builder->CreateTrunc(x, ir_builder->getInt1Ty());
+    return{ result, carry };
+}

src/binary_translation/ARMFuncs.h

@@ -0,0 +1,42 @@
+#include <common/common_types.h>
+
+/*
+ * Functions from the manual,
+ * A8.4.3 Pseudocode details of instruction-specified shifts and rotates
+ * A2.2.1 Integer arithmetic 
+ */
+
+class InstructionBlock;
+
+namespace llvm
+{
+    class Value;
+}
+
+class ARMFuncs
+{
+public:
+    enum class SRType { LSL, LSR, ASR, RRX, ROR };
+    struct ShiftTN
+    {
+        SRType type;
+        llvm::Value *amount;
+    };
+    struct ResultCarry
+    {
+        llvm::Value *result, *carry;
+    };
+
+    static ShiftTN DecodeImmShift(InstructionBlock *instruction, u32 type, u32 imm5);
+    static SRType DecodeRegShift(u32 type);
+
+    static llvm::Value *Shift(InstructionBlock *instruction, llvm::Value *value, SRType type, llvm::Value *amount, llvm::Value *carry_in);
+    static ResultCarry Shift_C(InstructionBlock *instruction, llvm::Value *value, SRType type, llvm::Value *amount, llvm::Value *carry_in);
+    static ResultCarry LSL_C(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static llvm::Value *LSL(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static ResultCarry LSR_C(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static llvm::Value *LSR(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static ResultCarry ASR_C(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static ResultCarry ROR_C(InstructionBlock *instruction, llvm::Value *x, llvm::Value *shift);
+    static ResultCarry RRX_C(InstructionBlock *instruction, llvm::Value *x, llvm::Value *carry_in);
+};

src/binary_translation/CMakeLists.txt

@@ -6,6 +6,7 @@ set(SRCS
 			InstructionBlock.cpp
 			MachineState.cpp
 			TBAA.cpp
+			ARMFuncs.cpp
 			
 			Instructions/Instruction.cpp
 			Instructions/DataProcessing.cpp
@@ -19,6 +20,7 @@ set(HEADERS
 			MachineState.h
 			TBAA.h
 			BinarySearch.h
+			ARMFuncs.h
 			
 			Instructions/Types.h
 			Instructions/Instruction.h

src/binary_translation/InstructionBlock.h

@@ -1,6 +1,8 @@
 #include <memory>
 #include <string>
 #include <common/common_types.h>
+#include <llvm/IR/IRBuilder.h>
+#include "ModuleGen.h"
 
 namespace llvm
 {
@@ -8,7 +10,6 @@ namespace llvm
     class BasicBlock;
 }
 
-class ModuleGen;
 class Instruction;
 
 enum class Register;
@@ -51,6 +52,7 @@ public:
 
     u32 Address();
     ModuleGen *Module() { return module; }
+    llvm::IRBuilder<> *IrBuilder() { return module->IrBuilder(); }
 
     llvm::BasicBlock *GetEntryBasicBlock() { return entry_basic_block; }
 private:

src/binary_translation/Instructions/DataProcessing.cpp

@@ -2,6 +2,7 @@
 #include "Disassembler.h"
 #include "InstructionBlock.h"
 #include "ModuleGen.h"
+#include "ARMFuncs.h"
 
 static RegisterInstruction<DataProcessing> register_instruction;
 
@@ -9,12 +10,11 @@ bool DataProcessing::Decode()
 {
     // Mov and shifts must have zeroes at some operands of different data processing instructions
     if (ReadFields({ CondDef(), FieldDef<3>(0), FieldDef<4>((u32)ShortOpType::MoveAndShifts), FieldDef<1>(&s), FieldDef<4>(0),
-                     FieldDef<4>(&rd), FieldDef<5>(&imm5), FieldDef<3>(0), FieldDef<4>(&rm) }))
+                     FieldDef<4>(&rd), FieldDef<5>(&imm5), FieldDef<2>(&op2), FieldDef<1>(0), FieldDef<4>(&rm) }))
     {
         form = Form::Register;
-        if (imm5 != 0) return false; // Shifts
+        if (rm == Register::PC) return false;
-        if (s != 0) return false; // Set flags
+        if (rd == Register::PC && s) return false; // SEE SUBS PC, LR and related instructions;
-        if (rm == Register::PC) return false; // Jump
         return true;
     }
     if (ReadFields({ CondDef(), FieldDef<3>(1), FieldDef<4>(&short_op), FieldDef<1>(&s), FieldDef<4>(&rn),
@@ -29,10 +29,51 @@ bool DataProcessing::Decode()
 
 void DataProcessing::GenerateInstructionCode(InstructionBlock* instruction_block)
 {
-    // Currently supports only mov reg, reg
+    auto ir_builder = instruction_block->IrBuilder();
 
-    auto value = instruction_block->Read(rm);
+    auto original_carry = instruction_block->Read(Register::C);
-    instruction_block->Write(rd, value);
+    ARMFuncs::ResultCarry result = { instruction_block->Read(rm), original_carry };
+
+    switch (op2)
+    {
+    case Op2Type::MoveAndLSL:
+        if (imm5 != 0)
+        {
+            result = ARMFuncs::Shift_C(instruction_block, result.result, ARMFuncs::SRType::LSL,
+                ARMFuncs::DecodeImmShift(instruction_block, 0, imm5).amount, result.carry);
+        }
+        break;
+    case Op2Type::LSR:
+        result = ARMFuncs::Shift_C(instruction_block, result.result, ARMFuncs::SRType::LSR,
+            ARMFuncs::DecodeImmShift(instruction_block, 1, imm5).amount, result.carry);
+        break;
+    case Op2Type::ASR:
+        result = ARMFuncs::Shift_C(instruction_block, result.result, ARMFuncs::SRType::ASR,
+            ARMFuncs::DecodeImmShift(instruction_block, 2, imm5).amount, result.carry);
+        break;
+    case Op2Type::RRXAndROR:
+        if (imm5 == 0)
+        {
+            result = ARMFuncs::Shift_C(instruction_block, result.result, ARMFuncs::SRType::RRX,
+                ir_builder->getInt32(1), result.carry);
+        }
+        else
+        {
+            result = ARMFuncs::Shift_C(instruction_block, result.result, ARMFuncs::SRType::ROR,
+                ARMFuncs::DecodeImmShift(instruction_block, 3, imm5).amount, result.carry);
+        }
+        break;
+    }
+
+    instruction_block->Write(rd, result.result);
+
+    if (s)
+    {
+        instruction_block->Write(Register::N, ir_builder->CreateTrunc(ir_builder->CreateLShr(result.result, 31), ir_builder->getInt1Ty()));
+        instruction_block->Write(Register::Z, ir_builder->CreateICmpEQ(result.result, ir_builder->getInt32(0)));
+        if (result.carry != original_carry)
+            instruction_block->Write(Register::C, result.carry);
+    }
 
     if (rd == Register::PC)
     {

src/binary_translation/Instructions/DataProcessing.h

@@ -18,6 +18,10 @@ public:
         // Compare, Test, Misc
         BitwiseOr = 12, MoveAndShifts, BitwiseBitClear, BitwiseNot
     };
+    enum class Op2Type
+    {
+        MoveAndLSL, LSR, ASR, RRXAndROR
+    };
     enum class Form
     {
         Register, RegisterShiftedRegister, Immediate
@@ -35,4 +39,5 @@ private:
     Register rm;
     u32 imm12;
     u32 imm5;
+    Op2Type op2;
 };

src/binary_translation/MachineState.cpp

@@ -5,6 +5,7 @@
 #include <llvm/IR/Constants.h>
 #include <llvm/IR/LLVMContext.h>
 #include <llvm/IR/GlobalVariable.h>
+#include "TBAA.h"
 
 using namespace llvm;
 

src/binary_translation/MachineState.h

@@ -6,6 +6,7 @@ class ModuleGen;
 namespace llvm
 {
     class Value;
+    class Instruction;
     class GlobalVariable;
 }
 

src/binary_translation/ModuleGen.cpp

@@ -9,6 +9,7 @@
 #include <llvm/IR/GlobalVariable.h>
 #include <stack>
 #include "MachineState.h"
+#include "TBAA.h"
 
 using namespace llvm;
 

src/binary_translation/ModuleGen.h

@@ -1,12 +1,13 @@
+#pragma once
 #include <llvm/IR/IRBuilder.h>
 #include <unordered_map>
 #include <common/common_types.h>
-#include "TBAA.h"
 
 enum class Register;
 
 class InstructionBlock;
 class MachineState;
+class TBAA;
 
 namespace llvm
 {

src/binary_translation/TBAA.h

@@ -1,3 +1,4 @@
+#pragma once
 #include "Instructions/Types.h"
 
 namespace llvm