[llvm] 96e8d0f - [AArch64] Refactor creation of a shuffle mask for TBL (NFC) (#92529)

[via llvm-commits]

Author: Momchil Velikov
Date: 2024-06-17T14:17:07+01:00
New Revision: 96e8d0fd944e2732a3cf2cc125494e7235b25fa9

URL: https://github.com/llvm/llvm-project/commit/96e8d0fd944e2732a3cf2cc125494e7235b25fa9
DIFF: https://github.com/llvm/llvm-project/commit/96e8d0fd944e2732a3cf2cc125494e7235b25fa9.diff

LOG: [AArch64] Refactor creation of a shuffle mask for TBL (NFC) (#92529)

... in preparation for https://github.com/llvm/llvm-project/pull/92528

Added: 
    

Modified: 
    llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 394b741f1c1d0..c5c3ef02115ec 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -15824,48 +15824,49 @@ bool AArch64TargetLowering::shouldSinkOperands(
   return false;
 }
 
-static bool createTblShuffleForZExt(ZExtInst *ZExt, FixedVectorType *DstTy,
-                                    bool IsLittleEndian) {
-  Value *Op = ZExt->getOperand(0);
-  auto *SrcTy = cast<FixedVectorType>(Op->getType());
-  auto SrcWidth = cast<IntegerType>(SrcTy->getElementType())->getBitWidth();
-  auto DstWidth = cast<IntegerType>(DstTy->getElementType())->getBitWidth();
+static bool createTblShuffleMask(unsigned SrcWidth, unsigned DstWidth,
+                                 unsigned NumElts, bool IsLittleEndian,
+                                 SmallVectorImpl<int> &Mask) {
   if (DstWidth % 8 != 0 || DstWidth <= 16 || DstWidth >= 64)
     return false;
 
   assert(DstWidth % SrcWidth == 0 &&
-         "TBL lowering is not supported for a ZExt instruction with this "
-         "source & destination element type.");
-  unsigned ZExtFactor = DstWidth / SrcWidth;
+         "TBL lowering is not supported for a conversion instruction with this "
+         "source and destination element type.");
+
+  unsigned Factor = DstWidth / SrcWidth;
+  unsigned MaskLen = NumElts * Factor;
+
+  Mask.clear();
+  Mask.resize(MaskLen, NumElts);
+
+  unsigned SrcIndex = 0;
+  for (unsigned I = IsLittleEndian ? 0 : Factor - 1; I < MaskLen; I += Factor)
+    Mask[I] = SrcIndex++;
+
+  return true;
+}
+
+static Value *createTblShuffleForZExt(IRBuilderBase &Builder, Value *Op,
+                                      FixedVectorType *ZExtTy,
+                                      FixedVectorType *DstTy,
+                                      bool IsLittleEndian) {
+  auto *SrcTy = cast<FixedVectorType>(Op->getType());
   unsigned NumElts = SrcTy->getNumElements();
-  IRBuilder<> Builder(ZExt);
+  auto SrcWidth = cast<IntegerType>(SrcTy->getElementType())->getBitWidth();
+  auto DstWidth = cast<IntegerType>(DstTy->getElementType())->getBitWidth();
+
   SmallVector<int> Mask;
-  // Create a mask that selects <0,...,Op[i]> for each lane of the destination
-  // vector to replace the original ZExt. This can later be lowered to a set of
-  // tbl instructions.
-  for (unsigned i = 0; i < NumElts * ZExtFactor; i++) {
-    if (IsLittleEndian) {
-      if (i % ZExtFactor == 0)
-        Mask.push_back(i / ZExtFactor);
-      else
-        Mask.push_back(NumElts);
-    } else {
-      if ((i + 1) % ZExtFactor == 0)
-        Mask.push_back((i - ZExtFactor + 1) / ZExtFactor);
-      else
-        Mask.push_back(NumElts);
-    }
-  }
+  if (!createTblShuffleMask(SrcWidth, DstWidth, NumElts, IsLittleEndian, Mask))
+    return nullptr;
 
   auto *FirstEltZero = Builder.CreateInsertElement(
       PoisonValue::get(SrcTy), Builder.getInt8(0), uint64_t(0));
   Value *Result = Builder.CreateShuffleVector(Op, FirstEltZero, Mask);
   Result = Builder.CreateBitCast(Result, DstTy);
-  if (DstTy != ZExt->getType())
-    Result = Builder.CreateZExt(Result, ZExt->getType());
-  ZExt->replaceAllUsesWith(Result);
-  ZExt->eraseFromParent();
-  return true;
+  if (DstTy != ZExtTy)
+    Result = Builder.CreateZExt(Result, ZExtTy);
+  return Result;
 }
 
 static void createTblForTrunc(TruncInst *TI, bool IsLittleEndian) {
@@ -16030,21 +16031,30 @@ bool AArch64TargetLowering::optimizeExtendOrTruncateConversion(
 
       DstTy = TruncDstType;
     }
-
-    return createTblShuffleForZExt(ZExt, DstTy, Subtarget->isLittleEndian());
+    IRBuilder<> Builder(ZExt);
+    Value *Result = createTblShuffleForZExt(
+        Builder, ZExt->getOperand(0), cast<FixedVectorType>(ZExt->getType()),
+        DstTy, Subtarget->isLittleEndian());
+    if (!Result)
+      return false;
+    ZExt->replaceAllUsesWith(Result);
+    ZExt->eraseFromParent();
+    return true;
   }
 
   auto *UIToFP = dyn_cast<UIToFPInst>(I);
   if (UIToFP && SrcTy->getElementType()->isIntegerTy(8) &&
       DstTy->getElementType()->isFloatTy()) {
     IRBuilder<> Builder(I);
-    auto *ZExt = cast<ZExtInst>(
-        Builder.CreateZExt(I->getOperand(0), VectorType::getInteger(DstTy)));
+    Value *ZExt = createTblShuffleForZExt(
+        Builder, I->getOperand(0), FixedVectorType::getInteger(DstTy),
+        FixedVectorType::getInteger(DstTy), Subtarget->isLittleEndian());
+    if (!ZExt)
+      return false;
     auto *UI = Builder.CreateUIToFP(ZExt, DstTy);
     I->replaceAllUsesWith(UI);
     I->eraseFromParent();
-    return createTblShuffleForZExt(ZExt, cast<FixedVectorType>(ZExt->getType()),
-                                   Subtarget->isLittleEndian());
+    return true;
   }
 
   // Convert 'fptoui <(8|16) x float> to <(8|16) x i8>' to a wide fptoui