[mlir][quant] output spec verification check for quant.region

[Lewuathe]

Summary

We can have validation test for quant.region having incompatible output spec.

[stellaraccident]

mlir/test/Dialect/Quant/quant_region.mlir

73

Hi - thank you for the patch.

Just so you know, we are having a discussion about possibly removing this op: https://discourse.llvm.org/t/rfc-removing-the-ops-from-the-quant-dialect/3643

To our knowledge it has no live uses, so if you are using it, please speak up on that thread.

[Lewuathe]

@stellaraccident Thank you for letting me know. For now, we are not using the quant dialect for real usage. Just looking at the dialect as an option to achieve and ease the quantization logic for our deployment. So basically, I'm on the same page on the direction they are discussing.

mlir/test/Dialect/Quant/quant_region.mlir

func.func @unmatched_primitive(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>, %arg2: tensor<4xf32>) -> (tensor<4xf32>) {
   }) {input_specs = [!quant.uniform<i8:f32, 1.0>, !quant.uniform<i8:f32, 2.0>, i32],
       output_specs = [!quant.uniform<i8:f32, 4.0>], logical_kernel = "xyz"}
     : (tensor<4xf32>, tensor<4xf32>, tensor<4xf32>) -> (tensor<4xf32>)
   return %0 : tensor<4xf32>
 }
 
 // -----
 
+func.func @unmatched_quantize(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>, %arg2: tensor<4xf32>) -> (tensor<4xf32>) {
+  // @expected-error @+1 {{'quant.region' op has incompatible specification !quant.uniform<i32:f16, 4.000000e+00> and output type 'tensor<4xf32>'}}
+  %0 = "quant.region"(%arg0, %arg1, %arg2) ({
+    ^bb0(%10: tensor<4xf32>, %11: tensor<4xf32>, %12: tensor<4xf32>):
+      %13 = "foo"(%10, %11) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
+      %14 = "bar"(%13, %12) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
+      "quant.return"(%14) : (tensor<4xf32>) -> ()
+  }) {input_specs = [!quant.uniform<i8:f32, 1.0>, !quant.uniform<i8:f32, 2.0>, !quant.uniform<i8:f32, 3.0>],
+      output_specs = [!quant.uniform<i32:f16, 4.0>], logical_kernel = "xyz"}
+    : (tensor<4xf32>, tensor<4xf32>, tensor<4xf32>) -> (tensor<4xf32>)
+  return %0 : tensor<4xf32>
+}
+
+// -----
+
+func.func @unmatched_primitive(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>, %arg2: tensor<4xf32>) -> (tensor<4xf32>) {
+  // @expected-error @+1 {{'quant.region' op has incompatible specification i32 and output type 'tensor<4xf32>'}}
+  %0 = "quant.region"(%arg0, %arg1, %arg2) ({
+    ^bb0(%10: tensor<4xf32>, %11: tensor<4xf32>, %12: tensor<4xf32>):
+      %13 = "foo"(%10, %11) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
+      %14 = "bar"(%13, %12) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
+      "quant.return"(%14) : (tensor<4xf32>) -> ()
+  }) {input_specs = [!quant.uniform<i8:f32, 1.0>, !quant.uniform<i8:f32, 2.0>, !quant.uniform<i32:f32, 2.0>],
+      output_specs = [i32], logical_kernel = "xyz"}
+    : (tensor<4xf32>, tensor<4xf32>, tensor<4xf32>) -> (tensor<4xf32>)
+  return %0 : tensor<4xf32>
+}
+
+// -----
+
 func.func @unmatched_number(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>, %arg2: tensor<4xf32>) -> (tensor<4xf32>) {
   // @expected-error @+1 {{'quant.region' op has unmatched operands/results number and spec attributes number}}
   %0 = "quant.region"(%arg0, %arg1, %arg2) ({
     ^bb0(%10: tensor<4xf32>, %11: tensor<4xf32>, %12: tensor<4xf32>):
       %13 = "foo"(%10, %11) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
       %14 = "bar"(%13, %12) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
       "quant.return"(%14) : (tensor<4xf32>) -> ()
   }) {input_specs = [!quant.uniform<i8:f32, 1.0>, !quant.uniform<i8:f32, 2.0>],