[llvm-branch-commits] [RISCV] Attach VLMAX range attribute for vsetvli/vsetvlimax in InstCombine (PR #218652)

[Ramkumar Ramachandra via llvm-branch-commits]

================
@@ -3752,6 +3752,58 @@ bool RISCVTTIImpl::shouldCopyAttributeWhenOutliningFrom(
 
 std::optional<Instruction *>
 RISCVTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
+  // Attach a range return attribute describing the result of vsetvli/vsetvlimax
+  // so generic value analyses can reason about it. The verifier guarantees an
+  // XLen result and constant VSEW/VLMUL encoding a valid vtype, so no defensive
+  // validation is needed here.
+  if (II.getIntrinsicID() == Intrinsic::riscv_vsetvli ||
+      II.getIntrinsicID() == Intrinsic::riscv_vsetvlimax) {
+    bool HasAVL = II.getIntrinsicID() == Intrinsic::riscv_vsetvli;
+    unsigned Offset = HasAVL ? 1 : 0;
+    unsigned Width = II.getType()->getScalarSizeInBits();
+    ConstantRange VLenRange(APInt(Width, ST->getRealMinVLen()),
+                            APInt(Width, ST->getRealMaxVLen()) + 1);
+
+    uint64_t VSEW = cast<ConstantInt>(II.getArgOperand(Offset))->getZExtValue();
+    auto VLMUL = static_cast<RISCVVType::VLMUL>(
+        cast<ConstantInt>(II.getArgOperand(Offset + 1))->getZExtValue());
----------------
artagnon wrote:

Will the zext value overflow 64 bits?

https://github.com/llvm/llvm-project/pull/218652