[Bug] tt.assert

[longRookie]

When I execute python test_vec_add.py, I notice an error with the following message:

triton_shared/include/triton-shared/Conversion/TritonArithToLinalg/ConversionPatterns.hpp:843: virtual llvm::LogicalResult {anonymous}::AssertConverter::matchAndRewrite(mlir::triton::AssertOp, mlir::OpConversionPattern<mlir::triton::AssertOp>::OpAdaptor, mlir::ConversionPatternRewriter&) const: Assertion `condVal && isa<mlir::IntegerType>(condVal.getType()) && "Only asserts on scalars are currently supported"' failed.

I observed that my TTIR contains tt.assertfor tensors.

    %13 = arith.cmpi sle, %12, %cst_0 : tensor<1024xi64> loc(#loc5)
    %14 = arith.cmpi sge, %12, %cst : tensor<1024xi64> loc(#loc5)
    %15 = arith.andi %13, %14 : tensor<1024xi1> loc(#loc5)
    tt.assert %15, "int32 overflow detected for operation add" : tensor<1024xi1> loc(#loc5)
    %16 = arith.addi %8, %7 : tensor<1024xi32> loc(#loc5)
    %17 = tt.splat %arg3 : i32 -> tensor<1024xi32> loc(#loc6)

When I commented out the assertin the TTIR and then executed the command line again, I was able to obtain ttshared.mlir. How can I avoid this issue from occurring?

 assert(condVal && isa<mlir::IntegerType>(condVal.getType()) &&
           "Only asserts on scalars are currently supported");