[mlir][tosa] Improve invalid operator data types error message #140756
+68
−7
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The error message on invalid operator data types in the validation pass was not very clear. This commit improves the error message as follows: Current: ``` 'tosa.add' op illegal: operand/result data types not supported ``` Improved: ``` 'tosa.add' op illegal: operation operand/result data types did not align with any profile or extension, got (i1,i1,i1), did you mean (i32,i32,i32)? Otherwise, please refer to the 'supported data types' for 'tosa.add' in the specification. ``` Change-Id: I844bdc27f16100b1434c85fa9b79f756ae519c8c
Member
|
@llvm/pr-subscribers-mlir-tosa @llvm/pr-subscribers-mlir Author: Luke Hutton (lhutton1) ChangesThe error message on invalid operator data types in the validation pass was not very clear. This commit improves the error message as follows: Current: Improved: Full diff: https://github.com/llvm/llvm-project/pull/140756.diff 4 Files Affected:
diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaProfileCompliance.h b/mlir/include/mlir/Dialect/Tosa/IR/TosaProfileCompliance.h
index 716e55706c625..8f5c72bc5f7a9 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaProfileCompliance.h
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaProfileCompliance.h
@@ -164,6 +164,8 @@ class TosaProfileCompliance {
SmallVector<StringRef>
stringifyProfile(const SmallVector<ArrayRef<T>> &profileSet);
+ static llvm::SmallString<7> stringifyTypeInfo(const TypeInfo &typeInfo);
+
private:
template <typename T>
FailureOr<SmallVector<T>> getOperatorDefinition(Operation *op,
diff --git a/mlir/lib/Dialect/Tosa/Transforms/TosaProfileCompliance.cpp b/mlir/lib/Dialect/Tosa/Transforms/TosaProfileCompliance.cpp
index 66ea00b23b9d4..1a896c1464e1c 100644
--- a/mlir/lib/Dialect/Tosa/Transforms/TosaProfileCompliance.cpp
+++ b/mlir/lib/Dialect/Tosa/Transforms/TosaProfileCompliance.cpp
@@ -485,9 +485,52 @@ LogicalResult TosaProfileCompliance::checkInvalid(Operation *op) {
CheckCondition condition = CheckCondition::invalid;
const auto maybeProfDef = getOperatorDefinition<Profile>(op, condition);
const auto maybeExtDef = getOperatorDefinition<Extension>(op, condition);
+
if (!failed(maybeProfDef) && !failed(maybeExtDef) &&
- !maybeProfDef.value().size() && !maybeExtDef.value().size())
+ !maybeProfDef.value().size() && !maybeExtDef.value().size()) {
+ std::string message;
+ llvm::raw_string_ostream os(message);
+ os << "illegal: operation operand/result data types did not align with any "
+ "profile or extension, got (";
+
+ ProfileInfoDepot depot(op);
+ SmallVector<TypeInfo> current = depot.getInfo();
+ for (const auto &typeInfo : llvm::drop_end(current))
+ os << stringifyTypeInfo(typeInfo) << ",";
+ os << stringifyTypeInfo(current.back()) << ")";
+
+ // avoid polluting the error message output by outputting only
+ // the best match
+ const std::string opName = op->getName().getStringRef().str();
+ int maxMatches = -1;
+ SmallVector<TypeInfo> bestTypeInfo;
+ const auto searchBestMatch = [&](auto map) {
+ for (const auto &complianceInfos : map[opName]) {
+ for (const auto &typeInfos : complianceInfos.operandTypeInfoSet) {
+ const int matches = llvm::count_if(
+ llvm::zip_equal(current, typeInfos), [&](const auto zipType) {
+ return isSameTypeInfo(std::get<0>(zipType),
+ std::get<1>(zipType));
+ });
+ if (matches > maxMatches) {
+ maxMatches = matches;
+ bestTypeInfo = typeInfos;
+ }
+ }
+ }
+ };
+ searchBestMatch(getProfileComplianceMap<Profile>());
+ searchBestMatch(getProfileComplianceMap<Extension>());
+
+ os << ", did you mean (";
+ for (const auto &typeInfo : llvm::drop_end(bestTypeInfo))
+ os << stringifyTypeInfo(typeInfo) << ",";
+ os << stringifyTypeInfo(bestTypeInfo.back()) << ")? ";
+ os << "Otherwise, please refer to the 'supported data types' for '"
+ << opName << "' in the specification.";
+ op->emitOpError(message);
return failure();
+ }
return success();
}
@@ -562,3 +605,21 @@ SmallVector<StringRef> TosaProfileCompliance::stringifyProfile(
return debugStrings;
}
+
+llvm::SmallString<7>
+TosaProfileCompliance::stringifyTypeInfo(const TypeInfo &typeInfo) {
+ if (typeInfo.typeID == mlir::IntegerType::getTypeID()) {
+ return {"i" + llvm::utostr(typeInfo.bitWidth)};
+ } else if (typeInfo.typeID == mlir::Float16Type::getTypeID()) {
+ return {"f16"};
+ } else if (typeInfo.typeID == mlir::Float32Type::getTypeID()) {
+ return {"f32"};
+ } else if (typeInfo.typeID == mlir::BFloat16Type::getTypeID()) {
+ return {"bf16"};
+ } else if (typeInfo.typeID == mlir::Float8E4M3FNType::getTypeID()) {
+ return {"fp8e4m3"};
+ } else if (typeInfo.typeID == mlir::Float8E5M2Type::getTypeID()) {
+ return {"fp8e5m2"};
+ }
+ llvm_unreachable("unknown type");
+}
diff --git a/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp b/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
index feedc5057bea0..e6d8e7834bf2c 100644
--- a/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
+++ b/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
@@ -1248,10 +1248,8 @@ void TosaValidation::runOnOperation() {
return signalPassFailure();
if (!allowInvalidOpDatatypeCombinations &&
- failed(profileComp.checkInvalid(op))) {
- op->emitOpError("illegal: operand/result data types not supported");
+ failed(profileComp.checkInvalid(op)))
return signalPassFailure();
- }
// Some uses of TOSA rely on the constant operands of particular
// operations.
diff --git a/mlir/test/Dialect/Tosa/invalid.mlir b/mlir/test/Dialect/Tosa/invalid.mlir
index 732c980f3ab92..7b589fa839b44 100644
--- a/mlir/test/Dialect/Tosa/invalid.mlir
+++ b/mlir/test/Dialect/Tosa/invalid.mlir
@@ -35,7 +35,7 @@ func.func @test_conv2d(%arg0: tensor<*xf32>, %arg1: tensor<16x3x3x4xi8>, %arg2:
func.func @test_conv2d(%arg0: tensor<1x29x29x4xi8>, %arg1: tensor<*xi8>, %arg2: tensor<16xi8>) -> tensor<1x27x27x16xi8> {
%zp = "tosa.const"() {values = dense<0> : tensor<1xi8>} : () -> tensor<1xi8>
- // expected-error@+1 {{'tosa.conv2d' op illegal: operand/result data types not supported}}
+ // expected-error@+1 {{'tosa.conv2d' op illegal: operation operand/result data types did not align with any profile or extension, got (i8,i8,i8,i8,i8,i32,i8), did you mean (i8,i8,i32,i8,i8,i32,i32)?}}
%0 = tosa.conv2d %arg0, %arg1, %arg2, %zp, %zp {acc_type = i32, dilation = array<i64: 1, 1>, pad = array<i64: 0, 0, 0, 0>, stride = array<i64: 1, 1>}
: (tensor<1x29x29x4xi8>, tensor<*xi8>, tensor<16xi8>, tensor<1xi8>, tensor<1xi8>) -> tensor<1x27x27x16xi8>
return %0 : tensor<1x27x27x16xi8>
@@ -1888,7 +1888,7 @@ func.func @test_scalar_tile(%arg0: tensor<f32>) -> tensor<*xf32> {
// CHECK-LABEL: test_add_i1
func.func @test_add_i1(%arg0: tensor<13x21x1xi1>, %arg1: tensor<13x21x3xi1>) -> tensor<13x21x3xi1> {
- // expected-error@+1 {{'tosa.add' op illegal: operand/result data types not supported}}
+ // expected-error@+1 {{'tosa.add' op illegal: operation operand/result data types did not align with any profile or extension, got (i1,i1,i1), did you mean (i32,i32,i32)? Otherwise, please refer to the 'supported data types' for 'tosa.add' in the specification.}}
%0 = tosa.add %arg0, %arg1 : (tensor<13x21x1xi1>, tensor<13x21x3xi1>) -> tensor<13x21x3xi1>
return %0 : tensor<13x21x3xi1>
}
@@ -1897,7 +1897,7 @@ func.func @test_add_i1(%arg0: tensor<13x21x1xi1>, %arg1: tensor<13x21x3xi1>) ->
// CHECK-LABEL: test_mul_out_i16
func.func @test_mul_out_i16(%arg0: tensor<13x21x3xi8>, %arg1: tensor<13x1x3xi8>, %shift: tensor<1xi8>) -> tensor<13x21x3xi16> {
- // expected-error@+1 {{'tosa.mul' op illegal: operand/result data types not supported}}
+ // expected-error@+1 {{'tosa.mul' op illegal: operation operand/result data types did not align with any profile or extension, got (i8,i8,i16), did you mean (i8,i8,i32)?}}
%0 = tosa.mul %arg0, %arg1, %shift : (tensor<13x21x3xi8>, tensor<13x1x3xi8>, tensor<1xi8>) -> tensor<13x21x3xi16>
return %0 : tensor<13x21x3xi16>
}
|
Contributor
|
Very useful, thanks |
Jerry-Ge
approved these changes
May 22, 2025
Member
Jerry-Ge
left a comment
Jerry-Ge
left a comment
There was a problem hiding this comment.
LGTM, thanks for this improvement!
leandron
approved these changes
May 22, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The error message on invalid operator data types in the validation pass was not very clear. This commit improves the error message as follows:
Current:
Improved: