[Mlir-commits] [mlir] [mlir][vector]add foldConstantOp fold function and apply it to extractOp and insertOp. (PR #124399)

[lonely eagle]

================
@@ -1977,6 +1977,46 @@ static Value foldScalarExtractFromFromElements(ExtractOp extractOp) {
   return fromElementsOp.getElements()[flatIndex];
 }
 
+// If the dynamic operands of `extractOp` or `insertOp` is result of
+// `constantOp`, then fold it.
+template <typename T>
+static void foldConstantOp(T op, SmallVectorImpl<Value> &operands) {
+  auto staticPosition = op.getStaticPosition().vec();
+  OperandRange dynamicPosition = op.getDynamicPosition();
+
+  // If the dynamic operands is empty, it is returned directly.
+  if (!dynamicPosition.size())
+    return;
+  unsigned index = 0;
+
+  // `opChange` is a flog. If it is true, it means to update `op` in place.
+  bool opChange = false;
+  for (unsigned i = 0, e = staticPosition.size(); i < e; ++i) {
+    if (!ShapedType::isDynamic(staticPosition[i]))
+      continue;
+    Value position = dynamicPosition[index++];
+
+    // If it is a block parameter, proceed to the next iteration.
+    if (!position.getDefiningOp()) {
+      operands.push_back(position);
+      continue;
+    }
+
+    if (auto constantOp =
+            mlir::dyn_cast<arith::ConstantIndexOp>(position.getDefiningOp())) {
+      opChange = true;
+      staticPosition[i] = constantOp.value();
+      continue;
+    }
----------------
linuxlonelyeagle wrote:

I didn't test this further, I think the key point is that the operand returned by `adaptor` in the `insertOp` or `extractOp` case in the dynamic position is an array of `ArrayRef<Attribute>`,If it's `arith::constant`, it will return the correct value,so I'm making a more stringent judgment here that it's an `arith::constant`.

Talking about my other idea, at first I wanted to support fold `llvm.constant`.But later on I disabused of this idea, because I cited that `llvm.constant` and `vector.insert` are incompatible, `llvm.constant` returns `i64`, while `vector.insert` needs the `SSA Value of index`.They need `buildin.unrealized_conversion_cast` for adaptation.

https://github.com/llvm/llvm-project/pull/124399