Use type alias for large tuples

[jpienaar]

Summary

Tuples types can occupy quite a lot of space, instead of printing out tuple
type everywhere, just use the type alias if larger (arbitrarily chose a bound
for now).

[jpienaar]

Fix style

[rriddle]

LGTM, though we could even drop it lower than 16 IMO.

[mehdi_amini]

Can you add a test for this please?

[jpienaar]

In D95707#2531751, @mehdi_amini wrote:

Can you add a test for this please?

Yes, will do. I'll have it just verify this happens but not at a specific size (e.g., create tuple with 40 elements) as the exact number may change. I'll submit one later today

mlir/lib/IR/BuiltinDialect.cpp

if (attr.isa<IntegerSetAttr>()) {
       return success();
     }
     if (attr.isa<LocationAttr>()) {
       os << "loc";
       return success();
     }
     return failure();
   }
+
+  LogicalResult getAlias(Type type, raw_ostream &os) const final {
+    if (auto tupleType = type.dyn_cast<TupleType>()) {
+      if (tupleType.size() > 16) {
+        os << "tuple";
+        return success();
+      }
+    }
+    return failure();
+  }
 };
 } // end anonymous namespace.
 
 void BuiltinDialect::initialize() {
   addTypes<ComplexType, BFloat16Type, Float16Type, Float32Type, Float64Type,
            Float80Type, Float128Type, FunctionType, IndexType, IntegerType,
            MemRefType, UnrankedMemRefType, NoneType, OpaqueType,
            RankedTensorType, TupleType, UnrankedTensorType, VectorType>();