[mlir][EmitC] Add `verbatim` op #79584

simon-camp · 2024-01-26T11:54:58Z

The verbatim operation produces no results and the value is emitted as is followed by a line break ('\n' character) during translation.

Note: Use with caution. This operation can have arbitrary effects on the semantics of the emitted code. Use semantically more meaningful operations whenever possible. Additionally this op is NOT intended to be used to inject large snippets of code.

This operation can be used in situations where a more suitable operation is not yet implemented in the dialect or where preprocessor directives interfere with the structure of the code.

Co-authored-by: Marius Brehler [email protected]

This operation can be used in situations where a more suitable operation is not yet implemented in the dialect or where preprocessor directives interfere with the structure of the code.

llvmbot · 2024-01-26T11:55:28Z

@llvm/pr-subscribers-mlir-emitc

@llvm/pr-subscribers-mlir

Author: Simon Camphausen (simon-camp)

Changes

The verbatim operation produces no results and the value is emitted as is durung translation. If trailing_semicolon is present an additional semicolon is emitted.

Note: Use with caution. This operation can have arbitrary effects on the semantics of the emitted code. Use semantically more meaningful operations whenever possible. Additionally this op is NOT intended to be used to inject large snippets of code.

This operation can be used in situations where a more suitable operation is not yet implemented in the dialect or where preprocessor directives interfere with the structure of the code.

Full diff: https://github.com/llvm/llvm-project/pull/79584.diff

6 Files Affected:

(modified) mlir/include/mlir/Dialect/EmitC/IR/EmitC.td (+39)
(modified) mlir/lib/Dialect/EmitC/IR/EmitC.cpp (+11)
(modified) mlir/lib/Target/Cpp/TranslateToCpp.cpp (+17-6)
(modified) mlir/test/Dialect/EmitC/invalid_ops.mlir (+8)
(modified) mlir/test/Dialect/EmitC/ops.mlir (+11)
(added) mlir/test/Target/Cpp/verbatim.mlir (+21)

diff --git a/mlir/include/mlir/Dialect/EmitC/IR/EmitC.td b/mlir/include/mlir/Dialect/EmitC/IR/EmitC.td
index b8f8f1e2d818d5..4d3d6c20e9b2c3 100644
--- a/mlir/include/mlir/Dialect/EmitC/IR/EmitC.td
+++ b/mlir/include/mlir/Dialect/EmitC/IR/EmitC.td
@@ -544,6 +544,45 @@ def EmitC_VariableOp : EmitC_Op<"variable", []> {
   let hasVerifier = 1;
 }
 
+def EmitC_VerbatimOp : EmitC_Op<"verbatim"> {
+  let summary = "Verbatim operation";
+  let description = [{
+    The `verbatim` operation produces no results and the value is emitted as is
+    durung translation. If `trailing_semicolon` is present an additional 
+    semicolon is emitted.
+
+    Note: Use with caution. This operation can have arbitrary effects on the
+    semantics of the emitted code. Use semantically more meaningful operations
+    whenever possible. Additionally this op is *NOT* intended to be used to
+    inject large snippets of code.
+
+    This operation can be used in situations where a more suitable operation is
+    not yet implemented in the dialect or where preprocessor directives
+    interfere with the structure of the code. One example of this is to declare
+    the linkage of external symbols to make the generated code usable in both C
+    and C++ contexts:
+
+    ```c++
+    #ifdef __cplusplus
+    extern "C" {
+    #endif
+
+    ...
+    
+    #ifdef __cplusplus
+    }
+    #endif
+    ```
+  }];
+
+  let arguments = (ins
+    StrAttr:$value,
+    UnitAttr:$trailing_semicolon
+  );
+  let assemblyFormat = "$value (`trailing_semicolon` $trailing_semicolon^)? attr-dict";
+  let hasVerifier = 1;
+}
+
 def EmitC_AssignOp : EmitC_Op<"assign", []> {
   let summary = "Assign operation";
   let description = [{
diff --git a/mlir/lib/Dialect/EmitC/IR/EmitC.cpp b/mlir/lib/Dialect/EmitC/IR/EmitC.cpp
index 5f502f1f7a1714..921cce58294315 100644
--- a/mlir/lib/Dialect/EmitC/IR/EmitC.cpp
+++ b/mlir/lib/Dialect/EmitC/IR/EmitC.cpp
@@ -584,6 +584,17 @@ LogicalResult emitc::VariableOp::verify() {
   return verifyInitializationAttribute(getOperation(), getValueAttr());
 }
 
+//===----------------------------------------------------------------------===//
+// VerbatimOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult emitc::VerbatimOp::verify() {
+  if (getValue().back() == ';')
+    return emitOpError() << "';' not allowed as the last character, use the "
+                            "`trailing_semicolon` attribute instead";
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // YieldOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Target/Cpp/TranslateToCpp.cpp b/mlir/lib/Target/Cpp/TranslateToCpp.cpp
index c32cb03caf9db6..657f6ae092aef8 100644
--- a/mlir/lib/Target/Cpp/TranslateToCpp.cpp
+++ b/mlir/lib/Target/Cpp/TranslateToCpp.cpp
@@ -429,6 +429,17 @@ static LogicalResult printOperation(CppEmitter &emitter, emitc::CmpOp cmpOp) {
   return printBinaryOperation(emitter, operation, binaryOperator);
 }
 
+static LogicalResult printOperation(CppEmitter &emitter,
+                                    emitc::VerbatimOp verbatimOp) {
+  raw_ostream &os = emitter.ostream();
+
+  os << verbatimOp.getValue();
+
+  if (verbatimOp.getTrailingSemicolon())
+    os << ";";
+  return success();
+}
+
 static LogicalResult printOperation(CppEmitter &emitter,
                                     cf::BranchOp branchOp) {
   raw_ostream &os = emitter.ostream();
@@ -814,11 +825,10 @@ static LogicalResult printOperation(CppEmitter &emitter,
     for (Operation &op : block.getOperations()) {
       // When generating code for an emitc.if or cf.cond_br op no semicolon
       // needs to be printed after the closing brace.
-      // When generating code for an emitc.for op, printing a trailing semicolon
-      // is handled within the printOperation function.
-      bool trailingSemicolon =
-          !isa<cf::CondBranchOp, emitc::ForOp, emitc::IfOp, emitc::LiteralOp>(
-              op);
+      // When generating code for an emitc.for and emitc.verbatim op, printing a
+      // trailing semicolon is handled within the printOperation function.
+      bool trailingSemicolon = !isa<cf::CondBranchOp, emitc::ForOp, emitc::IfOp,
+                                    emitc::LiteralOp, emitc::VerbatimOp>(op);
 
       if (failed(emitter.emitOperation(
               op, /*trailingSemicolon=*/trailingSemicolon)))
@@ -1144,7 +1154,8 @@ LogicalResult CppEmitter::emitOperation(Operation &op, bool trailingSemicolon) {
                 emitc::CallOpaqueOp, emitc::CastOp, emitc::CmpOp,
                 emitc::ConstantOp, emitc::DivOp, emitc::ExpressionOp,
                 emitc::ForOp, emitc::IfOp, emitc::IncludeOp, emitc::MulOp,
-                emitc::RemOp, emitc::SubOp, emitc::VariableOp>(
+                emitc::RemOp, emitc::SubOp, emitc::VariableOp,
+                emitc::VerbatimOp>(
               [&](auto op) { return printOperation(*this, op); })
           // Func ops.
           .Case<func::CallOp, func::ConstantOp, func::FuncOp, func::ReturnOp>(
diff --git a/mlir/test/Dialect/EmitC/invalid_ops.mlir b/mlir/test/Dialect/EmitC/invalid_ops.mlir
index 46eccb1c24eea2..2d98c505ac2435 100644
--- a/mlir/test/Dialect/EmitC/invalid_ops.mlir
+++ b/mlir/test/Dialect/EmitC/invalid_ops.mlir
@@ -289,3 +289,11 @@ func.func @test_expression_multiple_results(%arg0: i32) -> i32 {
   }
   return %r : i32
 }
+
+// -----
+
+func.func @verbatim_trailing_semicolon() {
+    // expected-error @+1 {{'emitc.verbatim' op ';' not allowed as the last character, use the `trailing_semicolon` attribute instead}}
+    emitc.verbatim "typedef int32_t i32;"
+    return
+}
\ No newline at end of file
diff --git a/mlir/test/Dialect/EmitC/ops.mlir b/mlir/test/Dialect/EmitC/ops.mlir
index 45ce2bcb99092c..a50883b38d14a2 100644
--- a/mlir/test/Dialect/EmitC/ops.mlir
+++ b/mlir/test/Dialect/EmitC/ops.mlir
@@ -166,3 +166,14 @@ func.func @test_for_not_index_induction(%arg0 : i16, %arg1 : i16, %arg2 : i16) {
   }
   return
 }
+
+emitc.verbatim "#ifdef __cplusplus"
+emitc.verbatim "extern \"C\" {"
+emitc.verbatim "#endif  // __cplusplus"
+
+emitc.verbatim "#ifdef __cplusplus"
+emitc.verbatim "}  // extern \"C\""
+emitc.verbatim "#endif  // __cplusplus"
+
+emitc.verbatim "typedef int32_t i32" {trailing_semicolon = unit}
+emitc.verbatim "typedef float f32" trailing_semicolon
diff --git a/mlir/test/Target/Cpp/verbatim.mlir b/mlir/test/Target/Cpp/verbatim.mlir
new file mode 100644
index 00000000000000..cf17486050f981
--- /dev/null
+++ b/mlir/test/Target/Cpp/verbatim.mlir
@@ -0,0 +1,21 @@
+// RUN: mlir-translate -mlir-to-cpp %s | FileCheck %s
+// RUN: mlir-translate -mlir-to-cpp -declare-variables-at-top %s | FileCheck %s
+
+
+emitc.verbatim "#ifdef __cplusplus"
+// CHECK: #ifdef __cplusplus
+emitc.verbatim "extern \"C\" {"
+// CHECK-NEXT: extern "C" {
+emitc.verbatim "#endif  // __cplusplus"
+// CHECK-NEXT: #endif  // __cplusplus
+emitc.verbatim "#ifdef __cplusplus"
+// CHECK-NEXT: #ifdef __cplusplus
+emitc.verbatim "}  // extern \"C\""
+// CHECK-NEXT: }  // extern "C"
+emitc.verbatim "#endif  // __cplusplus"
+// CHECK-NEXT: #endif  // __cplusplus
+
+emitc.verbatim "typedef int32_t i32" {trailing_semicolon = unit}
+// CHECK-NEXT: typedef int32_t i32;
+emitc.verbatim "typedef float f32" trailing_semicolon
+// CHECK-NEXT: typedef float f32;

marbre

Thanks @simon-camp :) As we've already discussed this offline for quite a while, I don't have much to say. From a first quick review, there are only minor things to change.

mlir/test/Dialect/EmitC/invalid_ops.mlir

marbre · 2024-01-26T12:47:26Z

mlir/test/Dialect/EmitC/ops.mlir

+emitc.verbatim "typedef int32_t i32" {trailing_semicolon = unit}
+emitc.verbatim "typedef float f32" trailing_semicolon


Do you want to add one of those or even both to the example in the description?

AlexDenisov · 2024-01-30T19:57:04Z

This is very useful!

Out of curiosity, what makes the trailing_semicolon necessary? Isn't it easier to include ; into the string itself, and the overcome the validation by adding a space at the end of the string?

jpienaar · 2024-01-31T04:27:59Z

This is very useful!

Out of curiosity, what makes the trailing_semicolon necessary? Isn't it easier to include ; into the string itself, and the overcome the validation by adding a space at the end of the string?

+1, could the attribute be avoided?

mlir/include/mlir/Dialect/EmitC/IR/EmitC.td

simon-camp · 2024-01-31T08:25:18Z

This is very useful!

Out of curiosity, what makes the trailing_semicolon necessary? Isn't it easier to include ; into the string itself, and the overcome the validation by adding a space at the end of the string?

Good suggestion, thank you.

I'm not sure I follow the validation part. The semicolon is now part of the attribute, so there is nothing to verify. Nonetheless this made me think of the implicit \n generated during emission, so this is now mentioned in the op description.

Co-authored-by: Marius Brehler <[email protected]>

AlexDenisov · 2024-01-31T09:42:34Z

I'm not sure I follow the validation part. The semicolon is now part of the attribute, so there is nothing to verify.

I was referring to these lines (which are now gone).

The uses of the attribute were removed in code review of #79584, but it's definition was inadvertently kept.

The `verbatim` operation produces no results and the value is emitted as is followed by a line break ('\n' character) during translation. Note: Use with caution. This operation can have arbitrary effects on the semantics of the emitted code. Use semantically more meaningful operations whenever possible. Additionally this op is *NOT* intended to be used to inject large snippets of code. This operation can be used in situations where a more suitable operation is not yet implemented in the dialect or where preprocessor directives interfere with the structure of the code. Co-authored-by: Marius Brehler <[email protected]>

The uses of the attribute were removed in code review of llvm#79584, but it's definition was inadvertently kept.

[mlir][EmitC] Add verbatim op

ab8cc1b

This operation can be used in situations where a more suitable operation is not yet implemented in the dialect or where preprocessor directives interfere with the structure of the code.

simon-camp requested review from jpienaar, joker-eph and marbre January 26, 2024 11:54

llvmbot added mlir mlir:emitc labels Jan 26, 2024

marbre requested changes Jan 26, 2024

View reviewed changes

Remove trailing_semicolon attribute

73cd31b

simon-camp force-pushed the emitc.verbatim branch from 6131bfe to 73cd31b Compare January 31, 2024 08:15

marbre reviewed Jan 31, 2024

View reviewed changes

mlir/include/mlir/Dialect/EmitC/IR/EmitC.td Outdated Show resolved Hide resolved

Update mlir/include/mlir/Dialect/EmitC/IR/EmitC.td

0844ddb

Co-authored-by: Marius Brehler <[email protected]>

marbre approved these changes Jan 31, 2024

View reviewed changes

simon-camp merged commit e624648 into llvm:main Jan 31, 2024

simon-camp deleted the emitc.verbatim branch January 31, 2024 10:56

simon-camp mentioned this pull request Jan 31, 2024

[mlir][EmitC] Remove unused attribute from verbatim op #80142

Merged

simon-camp pushed a commit that referenced this pull request Jan 31, 2024

[mlir][EmitC] Remove unused attribute from verbatim op (#80142)

121a0ef

The uses of the attribute were removed in code review of #79584, but it's definition was inadvertently kept.

mgehre-amd pushed a commit to Xilinx/llvm-project that referenced this pull request Mar 11, 2024

[mlir][EmitC] Remove unused attribute from verbatim op (llvm#80142)

763ebf9

The uses of the attribute were removed in code review of llvm#79584, but it's definition was inadvertently kept.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[mlir][EmitC] Add `verbatim` op #79584

[mlir][EmitC] Add `verbatim` op #79584

Uh oh!

simon-camp commented Jan 26, 2024 •

edited

Loading

Uh oh!

llvmbot commented Jan 26, 2024 •

edited

Loading

Uh oh!

marbre left a comment

Uh oh!

Uh oh!

marbre Jan 26, 2024

Uh oh!

AlexDenisov commented Jan 30, 2024

Uh oh!

jpienaar commented Jan 31, 2024

Uh oh!

Uh oh!

simon-camp commented Jan 31, 2024

Uh oh!

AlexDenisov commented Jan 31, 2024

Uh oh!

Uh oh!

		emitc.verbatim "typedef int32_t i32" {trailing_semicolon = unit}
		emitc.verbatim "typedef float f32" trailing_semicolon

[mlir][EmitC] Add verbatim op #79584

[mlir][EmitC] Add verbatim op #79584

Uh oh!

Conversation

simon-camp commented Jan 26, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Jan 26, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

marbre left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

marbre Jan 26, 2024

Choose a reason for hiding this comment

Uh oh!

AlexDenisov commented Jan 30, 2024

Uh oh!

jpienaar commented Jan 31, 2024

Uh oh!

Uh oh!

simon-camp commented Jan 31, 2024

Uh oh!

AlexDenisov commented Jan 31, 2024

Uh oh!

Uh oh!

[mlir][EmitC] Add `verbatim` op #79584

[mlir][EmitC] Add `verbatim` op #79584

simon-camp commented Jan 26, 2024 •

edited

Loading

llvmbot commented Jan 26, 2024 •

edited

Loading