[llvm] [IR] Initial introduction of memset_pattern (PR #97583)

[Alex Bradbury via llvm-commits]

================
@@ -456,6 +456,109 @@ static void createMemMoveLoop(Instruction *InsertBefore, Value *SrcAddr,
   ElseTerm->eraseFromParent();
 }
 
+static void createMemSetPatternLoop(Instruction *InsertBefore, Value *DstAddr,
+                                    Value *CopyLen, Value *SetValue,
+                                    Align DstAlign, bool IsVolatile) {
+  BasicBlock *OrigBB = InsertBefore->getParent();
+  Function *F = OrigBB->getParent();
+  const DataLayout &DL = F->getDataLayout();
+
+  if (DL.isBigEndian())
+    report_fatal_error("memset_pattern.inline expansion not currently "
+                       "implemented for big-endian targets",
+                       false);
+
+  // To start with, let's assume SetValue is an i128 and bail out if it's not.
+  if (!isPowerOf2_32(SetValue->getType()->getScalarSizeInBits()))
+    report_fatal_error("Pattern width for memset_pattern must be a power of 2",
+                       false);
+  unsigned PatternSize = SetValue->getType()->getScalarSizeInBits() / 8;
+
+  Type *TypeOfCopyLen = CopyLen->getType();
+
+  BasicBlock *NewBB = OrigBB->splitBasicBlock(InsertBefore, "split");
+  BasicBlock *LoopBB =
+      BasicBlock::Create(F->getContext(), "storeloop", F, NewBB);
+  BasicBlock *RemCheckBB =
----------------
asb wrote:

Yes, though my intent was to generate the most generic "dumb" expansion that doesn't take advantage of any of these opportunities to simplify, use #98311 as a pipecleaner for that and then incrementally add such improvements. Happy to do it otherwise if you prefer though.

https://github.com/llvm/llvm-project/pull/97583