improve nop handling and proper behaviour for LDM^

fixes dslinux

src/ARM.cpp

@@ -768,6 +768,8 @@ void ARMv4::ExecuteJIT()
             return;
         }
 
+        //printf("executing armv4 at %08x\n", instrAddr);
+
         ARMJIT::JitBlockEntry block = ARMJIT::LookUpBlock<1>(instrAddr);
         if (block)
             Cycles += block();

src/ARMJIT.cpp

@@ -273,6 +273,8 @@ bool IsIdleLoop(FetchedInstr* instrs, int instrsCount)
 
 typedef void (*InterpreterFunc)(ARM* cpu);
 
+void NOP(ARM* cpu) {}
+
 #define F(x) &ARMInterpreter::A_##x
 #define F_ALU(name, s) \
 	F(name##_REG_LSL_IMM##s), F(name##_REG_LSR_IMM##s), F(name##_REG_ASR_IMM##s), F(name##_REG_ROR_IMM##s), \
@@ -320,7 +322,8 @@ InterpreterFunc InterpretARM[ARMInstrInfo::ak_Count] =
 	F(LDM), F(STM),
 
 	F(B), F(BL), F(BLX_IMM), F(BX), F(BLX_REG),
-	F(UNK), F(MSR_IMM), F(MSR_REG), F(MRS), F(MCR), F(MRC), F(SVC)
+	F(UNK), F(MSR_IMM), F(MSR_REG), F(MRS), F(MCR), F(MRC), F(SVC),
+	NOP
 };
 #undef F_ALU
 #undef F_MEM_WB
@@ -387,8 +390,8 @@ void CompileBlock(ARM* cpu)
     u32 nextInstr[2] = {cpu->NextInstr[0], cpu->NextInstr[1]};
 	u32 nextInstrAddr[2] = {blockAddr, r15};
 
-	JIT_DEBUGPRINT("start block %x (%x) %p %p (region invalidates %dx)\n", 
-		blockAddr, pseudoPhysicalAddr, FastBlockAccess[pseudoPhysicalAddr / 2], 
+	JIT_DEBUGPRINT("start block %x %08x (%x) %p %p (region invalidates %dx)\n",
+		blockAddr, cpu->CPSR, pseudoPhysicalAddr, FastBlockAccess[pseudoPhysicalAddr / 2], 
 		cpu->Num == 0 ? LookUpBlock<0>(blockAddr) : LookUpBlock<1>(blockAddr),
 		CodeRanges[pseudoPhysicalAddr / 512].TimesInvalidated);
 
@@ -473,7 +476,9 @@ void CompileBlock(ARM* cpu)
 			else
 			{
                 u32 icode = ((instrs[i].Instr >> 4) & 0xF) | ((instrs[i].Instr >> 16) & 0xFF0);
-				assert(InterpretARM[instrs[i].Info.Kind] == ARMInterpreter::ARMInstrTable[icode] || instrs[i].Info.Kind == ARMInstrInfo::ak_MOV_REG_LSL_IMM);
+				assert(InterpretARM[instrs[i].Info.Kind] == ARMInterpreter::ARMInstrTable[icode]
+					|| instrs[i].Info.Kind == ARMInstrInfo::ak_MOV_REG_LSL_IMM
+					|| instrs[i].Info.Kind == ARMInstrInfo::ak_Nop);
 				if (cpu->CheckCondition(instrs[i].Cond()))
 					InterpretARM[instrs[i].Info.Kind](cpu);
 				else

src/ARMJIT_RegisterCache.h

@@ -152,7 +152,7 @@ public:
                 needValueLoaded = BitSet16(instr.Info.SrcRegs);
             for (int reg : needToBeLoaded)
                 LoadRegister(reg, needValueLoaded[reg]);
-        } 
+        }
         {
             BitSet16 loadedSet(LoadedRegs);
             BitSet16 loadRegs(instr.Info.NotStrictlyNeeded & futureNeeded & ~LoadedRegs);

src/ARMJIT_x64/ARMJIT_Branch.cpp

@@ -134,7 +134,7 @@ void Compiler::Comp_JumpTo(Gen::X64Reg addr, bool restoreCPSR)
 {
     IrregularCycles = true;
 
-    BitSet16 hiRegsLoaded(RegCache.DirtyRegs & 0xFF00);
+    BitSet16 hiRegsLoaded(RegCache.LoadedRegs & 0x7F00);
     bool previouslyDirty = CPSRDirty;
     SaveCPSR();
 
@@ -156,12 +156,12 @@ void Compiler::Comp_JumpTo(Gen::X64Reg addr, bool restoreCPSR)
     if (!restoreCPSR)
         XOR(32, R(ABI_PARAM3), R(ABI_PARAM3));
     else
-        MOV(32, R(ABI_PARAM3), Imm32(restoreCPSR));
+        MOV(32, R(ABI_PARAM3), Imm32(true)); // what a waste
     if (Num == 0)
         CALL((void*)&ARMv5::JumpTo);
     else
         CALL((void*)&ARMv4::JumpTo);
-    
+
     if (!Thumb && restoreCPSR && CurInstr.Cond() < 0xE)
     {
         for (int reg : hiRegsLoaded)

src/ARMJIT_x64/ARMJIT_Compiler.cpp

@@ -308,6 +308,7 @@ const Compiler::CompileFunc A_Comp[ARMInstrInfo::ak_Count] =
     F(A_Comp_BranchImm), F(A_Comp_BranchImm), F(A_Comp_BranchImm), F(A_Comp_BranchXchangeReg), F(A_Comp_BranchXchangeReg),
     // system stuff
     NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+    F(Nop)
 };
 
 const Compiler::CompileFunc T_Comp[ARMInstrInfo::tk_Count] = {

src/ARMJIT_x64/ARMJIT_Compiler.h

@@ -79,6 +79,8 @@ public:
         opInvertOp2 = 1 << 5,
     };
 
+    void Nop() {}
+
     void A_Comp_Arith();
     void A_Comp_MovOp();
     void A_Comp_CmpOp();

src/ARMJIT_x64/ARMJIT_LoadStore.cpp

@@ -531,7 +531,7 @@ s32 Compiler::Comp_MemAccessBlock(int rn, BitSet16 regs, bool store, bool preinc
         {
             if (regs[reg])
             {
-                if (usermode && reg >= 8 && reg < 15)
+                if (usermode && !regs[15] && reg >= 8 && reg < 15)
                 {
                     if (firstUserMode)
                     {
@@ -545,7 +545,8 @@ s32 Compiler::Comp_MemAccessBlock(int rn, BitSet16 regs, bool store, bool preinc
                     FixupBranch sucessfulWritten = J_CC(CC_NC);
                     if (RegCache.Mapping[reg] != INVALID_REG)
                         MOV(32, R(RegCache.Mapping[reg]), R(ABI_PARAM3));
-                    SaveReg(reg, ABI_PARAM3);
+                    else
+                        SaveReg(reg, ABI_PARAM3);
                     SetJumpTarget(sucessfulWritten);
                 }
                 else if (RegCache.Mapping[reg] == INVALID_REG)

src/ARM_InstrInfo.cpp

@@ -392,6 +392,8 @@ Info Decode(bool thumb, u32 num, u32 instr)
         u32 data = ARMInstrTable[((instr >> 4) & 0xF) | ((instr >> 16) & 0xFF0)];
         if (num == 0 && (instr & 0xFE000000) == 0xFA000000)
             data = A_BLX_IMM;
+        else if ((instr >> 28) == 0xF)
+            data = ak(ak_Nop);
 
         if (data & A_UnkOnARM7 && num != 0)
             data = A_UNK;

src/ARM_InstrInfo.h

@@ -139,6 +139,8 @@ enum
     ak_MRC,
     ak_SVC,
 
+    ak_Nop,
+
     ak_Count,
 
     tk_LSL_IMM = 0,