Stride validation fails for empty tensor input

[parthchadha]

For the below IR at runtime we get the following error:

summary = 'InvalidArgument: InvalidArgument: Input argument 0 validation failed against corresponding function signature arg 0. Reason: InvalidArgument: Runtime stride mismatch. Expected [-9223372036854775808, 1] but received [0, 0]'

module @ins_t_outs_t251_t252_t253_t254_t255_20 {
  func.func @main(%arg0: tensor<3x0xf32> {tensorrt.shape_profile = #tensorrt.shape_profile<min = [3, 0], opt = [3, 0], max = [3, 0]>}) -> (tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>) {
    %c = stablehlo.constant dense<0> : tensor<1xi32>
    %c_0 = stablehlo.constant dense<3> : tensor<i32>
    %c_1 = stablehlo.constant dense<1> : tensor<1xi32>
    %c_2 = stablehlo.constant dense<3> : tensor<1xi32>
    %c_3 = stablehlo.constant dense<0> : tensor<i32>
    %c_4 = stablehlo.constant dense<0> : tensor<1xi32>
    %0 = stablehlo.concatenate %c_2, %c_4, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %c_5 = stablehlo.constant dense<2> : tensor<1xi32>
    %1 = stablehlo.real_dynamic_slice %0, %c_1, %c_5, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %c_6 = stablehlo.constant dense<5> : tensor<1xi32>
    %2 = stablehlo.divide %1, %c_6 : (tensor<?xi32>, tensor<1xi32>) -> tensor<1xi32>
    %3 = stablehlo.multiply %2, %c : tensor<1xi32>
    %4 = stablehlo.concatenate %c, %3, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %5 = stablehlo.real_dynamic_slice %0, %c, %c_1, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %6 = stablehlo.multiply %2, %c_1 : tensor<1xi32>
    %7 = stablehlo.concatenate %5, %6, dim = 0 : (tensor<?xi32>, tensor<1xi32>) -> tensor<2xi32>
    %8 = stablehlo.concatenate %c_1, %c_1, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %9 = stablehlo.real_dynamic_slice %arg0, %4, %7, %8 : (tensor<3x0xf32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<?x?xf32>
    %10 = stablehlo.multiply %2, %c_1 : tensor<1xi32>
    %11 = stablehlo.concatenate %c, %10, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %12 = stablehlo.real_dynamic_slice %0, %c, %c_1, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %13 = stablehlo.multiply %2, %c_5 : tensor<1xi32>
    %14 = stablehlo.concatenate %12, %13, dim = 0 : (tensor<?xi32>, tensor<1xi32>) -> tensor<2xi32>
    %15 = stablehlo.concatenate %c_1, %c_1, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %16 = stablehlo.real_dynamic_slice %arg0, %11, %14, %15 : (tensor<3x0xf32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<?x?xf32>
    %17 = stablehlo.multiply %2, %c_5 : tensor<1xi32>
    %18 = stablehlo.concatenate %c, %17, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %19 = stablehlo.real_dynamic_slice %0, %c, %c_1, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %c_7 = stablehlo.constant dense<3> : tensor<1xi32>
    %20 = stablehlo.multiply %2, %c_7 : tensor<1xi32>
    %21 = stablehlo.concatenate %19, %20, dim = 0 : (tensor<?xi32>, tensor<1xi32>) -> tensor<2xi32>
    %22 = stablehlo.concatenate %c_1, %c_1, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %23 = stablehlo.real_dynamic_slice %arg0, %18, %21, %22 : (tensor<3x0xf32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<?x?xf32>
    %24 = stablehlo.multiply %2, %c_7 : tensor<1xi32>
    %25 = stablehlo.concatenate %c, %24, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %26 = stablehlo.real_dynamic_slice %0, %c, %c_1, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %c_8 = stablehlo.constant dense<4> : tensor<1xi32>
    %27 = stablehlo.multiply %2, %c_8 : tensor<1xi32>
    %28 = stablehlo.concatenate %26, %27, dim = 0 : (tensor<?xi32>, tensor<1xi32>) -> tensor<2xi32>
    %29 = stablehlo.concatenate %c_1, %c_1, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %30 = stablehlo.real_dynamic_slice %arg0, %25, %28, %29 : (tensor<3x0xf32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<?x?xf32>
    %31 = stablehlo.multiply %2, %c_8 : tensor<1xi32>
    %32 = stablehlo.concatenate %c, %31, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %33 = stablehlo.real_dynamic_slice %0, %c, %c_1, %c_1 : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<?xi32>
    %34 = stablehlo.multiply %2, %c_6 : tensor<1xi32>
    %35 = stablehlo.concatenate %33, %34, dim = 0 : (tensor<?xi32>, tensor<1xi32>) -> tensor<2xi32>
    %36 = stablehlo.concatenate %c_1, %c_1, dim = 0 : (tensor<1xi32>, tensor<1xi32>) -> tensor<2xi32>
    %37 = stablehlo.real_dynamic_slice %arg0, %32, %35, %36 : (tensor<3x0xf32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<?x?xf32>
    return %9, %16, %23, %30, %37 : tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>
  }
}

[akhilg-nv]

Fixed last week

[pranavm-nvidia]

This seems to be failing again with the TRT dialect. Re-opening so we can investigate.