Add --iree-flow-demote-i64-to-i32 and --iree-flow-demote-f64-to-f32

[powderluv]

Add --iree-flow-demote-i64-to-i32 (default false) and --iree-flow-demote-f64-to-f32 to compile MiniLM via Torch-mlir and run on IREE.

[powderluv]

See upstream issue iree-org/iree#9087 (comment)

[pashu123]

<stdin>:295:11: error: failed to materialize conversion for result #0 of operation 'hal.interface.binding.subspan' that remained live after conversion
    %55 = tensor.expand_shape %54 [[0], [1, 2, 3]] : tensor<1x?xf32> into tensor<1x1x1x?xf32>
          ^
<stdin>:23:3: note: called from
  func.func @forward(%arg0: tensor<1x128xi64>) -> tensor<1x2xf32> {
  ^
<stdin>:315:11: note: see existing live user here: %19 = memref.subview %13[0, 0, 0, %18] [1, 1, 1, %5] [1, 1, 1, 1] : memref<1x1x1x?xf32> to memref<?xf32, affine_map<(d0)[s0] -> (d0 + s0)>> loc(callsite("<stdin>":315:11 at "<stdin>":23:3))
    %75 = tensor.extract_slice %55[0, 0, 0, %64] [1, 1, 1, %74] [1, 1, 1, 1] : tensor<1x1x1x?xf32> to tensor<1x1x1x?xf32>
          ^
<stdin>:325:11: error: failed to run translation of source executable to target executable for backend #hal.executable.target<"vulkan", "vulkan-spirv-fb", {spv.target_env = #spv.target_env<#spv.vce<v1.3, [Shader, GroupNonUniform, GroupNonUniformVote, GroupNonUniformArithmetic, GroupNonUniformBallot, GroupNonUniformShuffle, GroupNonUniformShuffleRelative], [SPV_KHR_storage_buffer_storage_class]>, SwiftShader:CPU, {cooperative_matrix_properties_nv = [], max_compute_shared_memory_size = 16384 : i32, max_compute_workgroup_invocations = 128 : i32, max_compute_workgroup_size = dense<[128, 128, 64]> : vector<3xi32>, subgroup_size = 4 : i32}>}>
    %78 = linalg.generic {indexing_maps = [#map3, #map5, #map4], iterator_types = ["parallel", "parallel", "parallel", "parallel"]} ins(%77, %cst_6 : tensor<1x1x1x?xf32>, tensor<f64>) outs(%76 : tensor<1x1x1x?xf32>) {
          ^
<stdin>:23:3: note: called from
  func.func @forward(%arg0: tensor<1x128xi64>) -> tensor<1x2xf32> {
  ^
<stdin>:325:11: error: failed to serialize executables
    %78 = linalg.generic {indexing_maps = [#map3, #map5, #map4], iterator_types = ["parallel", "parallel", "parallel", "parallel"]} ins(%77, %cst_6 : tensor<1x1x1x?xf32>, tensor<f64>) outs(%76 : tensor<1x1x1x?xf32>) {
          ^
<stdin>:23:3: note: called from
  func.func @forward(%arg0: tensor<1x128xi64>) -> tensor<1x2xf32> {
  ^
compilation from source to vm failed


Invoked with:
 iree-compile /home/prashant/SHARK/shark.venv/lib/python3.9/site-packages/iree/compiler/tools/../_mlir_libs/iree-compile - --iree-input-type=none --iree-vm-bytecode-module-output-format=flatbuffer-binary --iree-hal-target-backends=vulkan --iree-mlir-to-vm-bytecode-module --iree-llvm-embedded-linker-path=/home/prashant/SHARK/shark.venv/lib/python3.9/site-packages/iree/compiler/tools/../_mlir_libs/iree-lld --mlir-print-debuginfo --mlir-print-op-on-diagnostic=false --iree-llvm-target-cpu-features=host --iree-flow-demote-i64-to-i32 --iree-flow-demote-f64-to-f32