ppc: Unify Memory Operation 3

Clean up 32 bit Load/Store

Change-Id: I5bab0d33830039d3c4a501eba6e7cf95f4b9559e
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2933597Reviewed-by: Milad Fa <mfarazma@redhat.com>
Commit-Queue: Junliang Yan <junyan@redhat.com>
Cr-Commit-Position: refs/heads/master@{#74927}

src/builtins/ppc/builtins-ppc.cc

@@ -871,7 +871,8 @@ static void TailCallOptimizedCodeSlot(MacroAssembler* masm,
       r0);
   __ LoadS32(
       scratch,
-      FieldMemOperand(scratch, CodeDataContainer::kKindSpecificFlagsOffset));
+      FieldMemOperand(scratch, CodeDataContainer::kKindSpecificFlagsOffset),
+      r0);
   __ TestBit(scratch, Code::kMarkedForDeoptimizationBit, r0);
   __ bne(&heal_optimized_code_slot, cr0);
 
@@ -1170,7 +1171,8 @@ void Builtins::Generate_InterpreterEntryTrampoline(MacroAssembler* masm) {
   __ LoadS32(r8,
              FieldMemOperand(
                  kInterpreterBytecodeArrayRegister,
-                 BytecodeArray::kIncomingNewTargetOrGeneratorRegisterOffset));
+                 BytecodeArray::kIncomingNewTargetOrGeneratorRegisterOffset),
+             r0);
   __ cmpi(r8, Operand::Zero());
   __ beq(&no_incoming_new_target_or_generator_register);
   __ ShiftLeftImm(r8, r8, Operand(kSystemPointerSizeLog2));

src/codegen/ppc/macro-assembler-ppc.cc

@@ -2766,56 +2766,21 @@ void TurboAssembler::StoreU64WithUpdate(Register src, const MemOperand& mem,
 
 void TurboAssembler::LoadS32(Register dst, const MemOperand& mem,
                              Register scratch) {
-  int offset = mem.offset();
-
-  if (!is_int16(offset)) {
-    CHECK(scratch != no_reg);
-    mov(scratch, Operand(offset));
-    lwax(dst, MemOperand(mem.ra(), scratch));
-  } else {
-    int misaligned = (offset & 3);
-    if (misaligned) {
-      // adjust base to conform to offset alignment requirements
-      // Todo: enhance to use scratch if dst is unsuitable
-      CHECK(dst != r0);
-      addi(dst, mem.ra(), Operand((offset & 3) - 4));
-      lwa(dst, MemOperand(dst, (offset & ~3) + 4));
-    } else {
-      lwa(dst, mem);
-    }
-  }
+  GenerateMemoryOperationWithAlign(dst, mem, lwa, lwax);
 }
 
 // Variable length depending on whether offset fits into immediate field
 // MemOperand currently only supports d-form
 void TurboAssembler::LoadU32(Register dst, const MemOperand& mem,
                              Register scratch) {
-  Register base = mem.ra();
-  int offset = mem.offset();
-
-  if (!is_int16(offset)) {
-    CHECK(scratch != no_reg);
-    mov(scratch, Operand(offset));
-    lwzx(dst, MemOperand(base, scratch));
-  } else {
-    // lwz can handle offset misalign
-    lwz(dst, mem);
-  }
+  GenerateMemoryOperation(dst, mem, lwz, lwzx);
 }
 
 // Variable length depending on whether offset fits into immediate field
 // MemOperand current only supports d-form
 void TurboAssembler::StoreU32(Register src, const MemOperand& mem,
                               Register scratch) {
-  Register base = mem.ra();
-  int offset = mem.offset();
-
-  if (!is_int16(offset)) {
-    LoadIntLiteral(scratch, offset);
-    stwx(src, MemOperand(base, scratch));
-  } else {
-    stw(src, mem);
-  }
+  GenerateMemoryOperation(src, mem, stw, stwx);
 }
 
 void TurboAssembler::LoadS16(Register dst, const MemOperand& mem,
@@ -3253,7 +3218,7 @@ void TurboAssembler::LoadCodeObjectEntry(Register destination,
     // Check whether the Code object is an off-heap trampoline. If so, call its
     // (off-heap) entry point directly without going through the (on-heap)
     // trampoline.  Otherwise, just call the Code object as always.
-    LoadS32(scratch, FieldMemOperand(code_object, Code::kFlagsOffset));
+    LoadS32(scratch, FieldMemOperand(code_object, Code::kFlagsOffset), r0);
     mov(r0, Operand(Code::IsOffHeapTrampoline::kMask));
     and_(r0, scratch, r0, SetRC);
     bne(&if_code_is_off_heap, cr0);
@@ -3266,7 +3231,8 @@ void TurboAssembler::LoadCodeObjectEntry(Register destination,
     // An off-heap trampoline, the entry point is loaded from the builtin entry
     // table.
     bind(&if_code_is_off_heap);
-    LoadS32(scratch, FieldMemOperand(code_object, Code::kBuiltinIndexOffset));
+    LoadS32(scratch, FieldMemOperand(code_object, Code::kBuiltinIndexOffset),
+            r0);
     ShiftLeftImm(destination, scratch, Operand(kSystemPointerSizeLog2));
     add(destination, destination, kRootRegister);
     LoadU64(destination,

src/compiler/backend/ppc/code-generator-ppc.cc

@@ -829,7 +829,8 @@ void CodeGenerator::BailoutIfDeoptimized() {
   __ LoadTaggedPointerField(
       r11, MemOperand(kJavaScriptCallCodeStartRegister, offset), r0);
   __ LoadS32(r11,
-             FieldMemOperand(r11, CodeDataContainer::kKindSpecificFlagsOffset));
+             FieldMemOperand(r11, CodeDataContainer::kKindSpecificFlagsOffset),
+             r0);
   __ TestBit(r11, Code::kMarkedForDeoptimizationBit);
   __ Jump(BUILTIN_CODE(isolate(), CompileLazyDeoptimizedCode),
           RelocInfo::CODE_TARGET, ne, cr0);