alignment info for cost estimation in SLP vectorization

[chentong]

Summary

When the cost of load/store is estimated in SLP vectorization, constant "1" was always given to the cost model. This caused over estimate the cost of sequential version in some architectures that care about the alignment (such as PPC). The fix is to add alignment info.

[mzolotukhin]

lib/Transforms/Vectorize/SLPVectorizer.cpp

1620–1624

Please run clang-format on the patch.

1621–1624

What about alignment here? Should we still use 1?

[chentong]

1. yes, the alignment for vector type is changed
2. clang-format is used

[chentong]

updates based on the comments:
yes, the alignment for vector type is changed
clang-format is used

[hfinkel]

Please add a test case. It should go in test/Transforms/SLPVectorizer/PowerPC (this directory does not currently exist, but please create it -- you'll need to add a lit.local.cfg file to the directory as well (just like test/Transforms/SLPVectorizer/X86/lit.local.cfg, but editing the file to change 'X86' to 'PowerPC').

lib/Transforms/Vectorize/SLPVectorizer.cpp

case Instruction::GetElementPtr: {
           TTI->getArithmeticInstrCost(Instruction::Add, ScalarTy, Op1VK, Op2VK);
       int VecCost =
           TTI->getArithmeticInstrCost(Instruction::Add, VecTy, Op1VK, Op2VK);
 
       return VecCost - ScalarCost;
     }
     case Instruction::Load: {
       // Cost of wide load - cost of scalar loads.
+      LoadInst *LoadI = cast<LoadInst>(VL[0]);
       int ScalarLdCost = VecTy->getNumElements() *
-      TTI->getMemoryOpCost(Instruction::Load, ScalarTy, 1, 0);
-      int VecLdCost = TTI->getMemoryOpCost(Instruction::Load, VecTy, 1, 0);
+                         TTI->getMemoryOpCost(Instruction::Load, ScalarTy,
+                                              LoadI->getAlignment(), 0);
+      int VecLdCost = TTI->getMemoryOpCost(Instruction::Load, VecTy,
+                                           LoadI->getAlignment(), 0);
       return VecLdCost - ScalarLdCost;
     }
     case Instruction::Store: {
       // We know that we can merge the stores. Calculate the cost.
+      StoreInst *StoreI = cast<StoreInst>(VL[0]);
       int ScalarStCost = VecTy->getNumElements() *
-      TTI->getMemoryOpCost(Instruction::Store, ScalarTy, 1, 0);
-      int VecStCost = TTI->getMemoryOpCost(Instruction::Store, VecTy, 1, 0);
+                         TTI->getMemoryOpCost(Instruction::Store, ScalarTy,
+                                              StoreI->getAlignment(), 0);
+      int VecStCost = TTI->getMemoryOpCost(Instruction::Store, VecTy,
+                                           StoreI->getAlignment(), 0);
       return VecStCost - ScalarStCost;
     }
     case Instruction::Call: {
       CallInst *CI = cast<CallInst>(VL0);
       Intrinsic::ID ID = getIntrinsicIDForCall(CI, TLI);
 
       // Calculate the cost of the scalar and vector calls.
       SmallVector<Type*, 4> ScalarTys, VecTys;