[llvm] [SLP][AMDGPU] Vectorize operands of non-trivially-vectorizable intrinsic calls (PR #189784)

[Alexey Bataev via llvm-commits]

================
@@ -243,6 +243,39 @@ static const int MinScheduleRegionSize = 16;
 /// Maximum allowed number of operands in the PHI nodes.
 static const unsigned MaxPHINumOperands = 128;
 
+/// For instructions that are not trivially vectorizable, try to vectorize their
+/// operands.
+/// FIXME: Extend for all non-vectorized functions.
+static Value *getNonTriviallyVectorizableIntrinsicCallOperand(Value *V) {
+  auto *CI = dyn_cast<CallInst>(V);
+  if (!CI)
+    return nullptr;
+  Intrinsic::ID ID = CI->getIntrinsicID();
----------------
alexey-bataev wrote:

I would add check for skipping debug and assume-like intrinsics

https://github.com/llvm/llvm-project/pull/189784