[CIR] Upstream CIR await op

Author: Andres-Salamanca

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -790,8 +790,8 @@ def CIR_ConditionOp : CIR_Op<"condition", [
 //===----------------------------------------------------------------------===//
 
 defvar CIR_YieldableScopes = [
-  "ArrayCtor", "ArrayDtor", "CaseOp", "DoWhileOp", "ForOp", "GlobalOp", "IfOp",
-  "ScopeOp", "SwitchOp", "TernaryOp", "WhileOp", "TryOp"
+  "ArrayCtor", "ArrayDtor", "AwaitOp", "CaseOp", "DoWhileOp", "ForOp",
+  "GlobalOp", "IfOp", "ScopeOp", "SwitchOp", "TernaryOp", "WhileOp", "TryOp"
 ];
 
 def CIR_YieldOp : CIR_Op<"yield", [
@@ -2707,6 +2707,102 @@ def CIR_CallOp : CIR_CallOpBase<"call", [NoRegionArguments]> {
   ];
 }
 
+//===----------------------------------------------------------------------===//
+// AwaitOp
+//===----------------------------------------------------------------------===//
+
+def CIR_AwaitKind : CIR_I32EnumAttr<"AwaitKind", "await kind", [
+  I32EnumAttrCase<"Init", 0, "init">,
+  I32EnumAttrCase<"User", 1, "user">,
+  I32EnumAttrCase<"Yield", 2, "yield">,
+  I32EnumAttrCase<"Final", 3, "final">
+]>;
+
+def CIR_AwaitOp : CIR_Op<"await",[
+  DeclareOpInterfaceMethods<RegionBranchOpInterface>,
+  RecursivelySpeculatable, NoRegionArguments
+]> {
+  let summary = "Wraps C++ co_await implicit logic";
+  let description = [{
+    The under the hood effect of using C++ `co_await expr` roughly
+    translates to:
+
+    ```c++
+    // co_await expr;
+
+    auto &&x = CommonExpr();
+    if (!x.await_ready()) {
+       ...
+       x.await_suspend(...);
+       ...
+    }
+    x.await_resume();
+    ```
+
+    `cir.await` represents this logic by using 3 regions:
+      - ready: covers veto power from x.await_ready()
+      - suspend: wraps actual x.await_suspend() logic
+      - resume: handles x.await_resume()
+
+    Breaking this up in regions allow individual scrutiny of conditions
+    which might lead to folding some of them out. Lowerings coming out
+    of CIR, e.g. LLVM, should use the `suspend` region to track more
+    lower level codegen (e.g. intrinsic emission for coro.save/coro.suspend).
+
+    There are also 4 flavors of `cir.await` available:
+    - `init`: compiler generated initial suspend via implicit `co_await`.
+    - `user`: also known as normal, representing user written co_await's.
+    - `yield`: user written `co_yield` expressions.
+    - `final`: compiler generated final suspend via implicit `co_await`.
+
+    From the C++ snippet we get:
+
+    ```mlir
+      cir.scope {
+        ... // auto &&x = CommonExpr();
+        cir.await(user, ready : {
+          ... // x.await_ready()
+        }, suspend : {
+          ... // x.await_suspend()
+        }, resume : {
+          ... // x.await_resume()
+        })
+      }
+    ```
+
+    Note that resulution of the common expression is assumed to happen
+    as part of the enclosing await scope.
+  }];
+
+  let arguments = (ins CIR_AwaitKind:$kind);
+  let regions = (region SizedRegion<1>:$ready,
+                        SizedRegion<1>:$suspend,
+                        SizedRegion<1>:$resume);
+  let assemblyFormat = [{
+    `(` $kind `,`
+    `ready` `:` $ready `,`
+    `suspend` `:` $suspend `,`
+    `resume` `:` $resume `,`
+    `)`
+    attr-dict
+  }];
+
+  let skipDefaultBuilders = 1;
+  let builders = [
+    OpBuilder<(ins
+      "cir::AwaitKind":$kind,
+      CArg<"BuilderCallbackRef",
+           "nullptr">:$readyBuilder,
+      CArg<"BuilderCallbackRef",
+           "nullptr">:$suspendBuilder,
+      CArg<"BuilderCallbackRef",
+           "nullptr">:$resumeBuilder
+      )>
+  ];
+
+  let hasVerifier = 1;
+}
+
 //===----------------------------------------------------------------------===//
 // CopyOp
 //===----------------------------------------------------------------------===//

clang/lib/CIR/CodeGen/CIRGenCoroutine.cpp

@@ -22,6 +22,10 @@ using namespace clang;
 using namespace clang::CIRGen;
 
 struct clang::CIRGen::CGCoroData {
+  // What is the current await expression kind and how many
+  // await/yield expressions were encountered so far.
+  // These are used to generate pretty labels for await expressions in LLVM IR.
+  cir::AwaitKind currentAwaitKind = cir::AwaitKind::Init;
   // Stores the __builtin_coro_id emitted in the function so that we can supply
   // it as the first argument to other builtins.
   cir::CallOp coroId = nullptr;
@@ -249,7 +253,114 @@ CIRGenFunction::emitCoroutineBody(const CoroutineBodyStmt &s) {
       emitAnyExprToMem(s.getReturnValue(), returnValue,
                        s.getReturnValue()->getType().getQualifiers(),
                        /*isInit*/ true);
+
+    assert(!cir::MissingFeatures::ehCleanupScope());
+    // FIXME(cir): EHStack.pushCleanup<CallCoroEnd>(EHCleanup);
+    curCoro.data->currentAwaitKind = cir::AwaitKind::Init;
+    if (emitStmt(s.getInitSuspendStmt(), /*useCurrentScope=*/true).failed())
+      return mlir::failure();
     assert(!cir::MissingFeatures::emitBodyAndFallthrough());
   }
   return mlir::success();
 }
+// Given a suspend expression which roughly looks like:
+//
+//   auto && x = CommonExpr();
+//   if (!x.await_ready()) {
+//      x.await_suspend(...); (*)
+//   }
+//   x.await_resume();
+//
+// where the result of the entire expression is the result of x.await_resume()
+//
+//   (*) If x.await_suspend return type is bool, it allows to veto a suspend:
+//      if (x.await_suspend(...))
+//        llvm_coro_suspend();
+//
+// This is more higher level than LLVM codegen, for that one see llvm's
+// docs/Coroutines.rst for more details.
+namespace {
+struct LValueOrRValue {
+  LValue lv;
+  RValue rv;
+};
+} // namespace
+
+static LValueOrRValue
+emitSuspendExpression(CIRGenFunction &cgf, CGCoroData &coro,
+                      CoroutineSuspendExpr const &s, cir::AwaitKind kind,
+                      AggValueSlot aggSlot, bool ignoreResult,
+                      mlir::Block *scopeParentBlock,
+                      mlir::Value &tmpResumeRValAddr, bool forLValue) {
+  mlir::LogicalResult awaitBuild = mlir::success();
+  LValueOrRValue awaitRes;
+
+  CIRGenFunction::OpaqueValueMapping binder =
+      CIRGenFunction::OpaqueValueMapping(cgf, s.getOpaqueValue());
+  CIRGenBuilderTy &builder = cgf.getBuilder();
+  [[maybe_unused]] cir::AwaitOp awaitOp = cir::AwaitOp::create(
+      builder, cgf.getLoc(s.getSourceRange()), kind,
+      /*readyBuilder=*/
+      [&](mlir::OpBuilder &b, mlir::Location loc) {
+        builder.createCondition(
+            cgf.createDummyValue(loc, cgf.getContext().BoolTy));
+      },
+      /*suspendBuilder=*/
+      [&](mlir::OpBuilder &b, mlir::Location loc) {
+        cir::YieldOp::create(builder, loc);
+      },
+      /*resumeBuilder=*/
+      [&](mlir::OpBuilder &b, mlir::Location loc) {
+        cir::YieldOp::create(builder, loc);
+      });
+
+  assert(awaitBuild.succeeded() && "Should know how to codegen");
+  return awaitRes;
+}
+
+static RValue emitSuspendExpr(CIRGenFunction &cgf,
+                              const CoroutineSuspendExpr &e,
+                              cir::AwaitKind kind, AggValueSlot aggSlot,
+                              bool ignoreResult) {
+  RValue rval;
+  mlir::Location scopeLoc = cgf.getLoc(e.getSourceRange());
+
+  // Since we model suspend / resume as an inner region, we must store
+  // resume scalar results in a tmp alloca, and load it after we build the
+  // suspend expression. An alternative way to do this would be to make
+  // every region return a value when promise.return_value() is used, but
+  // it's a bit awkward given that resume is the only region that actually
+  // returns a value.
+  mlir::Block *currEntryBlock = cgf.curLexScope->getEntryBlock();
+  [[maybe_unused]] mlir::Value tmpResumeRValAddr;
+
+  // No need to explicitly wrap this into a scope since the AST already uses a
+  // ExprWithCleanups, which will wrap this into a cir.scope anyways.
+  rval = emitSuspendExpression(cgf, *cgf.curCoro.data, e, kind, aggSlot,
+                               ignoreResult, currEntryBlock, tmpResumeRValAddr,
+                               /*forLValue*/ false)
+             .rv;
+
+  if (ignoreResult || rval.isIgnored())
+    return rval;
+
+  if (rval.isScalar()) {
+    rval = RValue::get(cir::LoadOp::create(cgf.getBuilder(), scopeLoc,
+                                           rval.getValue().getType(),
+                                           tmpResumeRValAddr));
+  } else if (rval.isAggregate()) {
+    // This is probably already handled via AggSlot, remove this assertion
+    // once we have a testcase and prove all pieces work.
+    cgf.cgm.errorNYI("emitSuspendExpr Aggregate");
+  } else { // complex
+    cgf.cgm.errorNYI("emitSuspendExpr Complex");
+  }
+  return rval;
+}
+
+RValue CIRGenFunction::emitCoawaitExpr(const CoawaitExpr &e,
+                                       AggValueSlot aggSlot,
+                                       bool ignoreResult) {
+  return emitSuspendExpr(*this, e, curCoro.data->currentAwaitKind, aggSlot,
+                         ignoreResult);
+}

clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp

@@ -154,6 +154,10 @@ class ScalarExprEmitter : public StmtVisitor<ScalarExprEmitter, mlir::Value> {
     return cgf.emitLoadOfLValue(lv, e->getExprLoc()).getValue();
   }
 
+  mlir::Value VisitCoawaitExpr(CoawaitExpr *s) {
+    return cgf.emitCoawaitExpr(*s).getValue();
+  }
+
   mlir::Value emitLoadOfLValue(LValue lv, SourceLocation loc) {
     return cgf.emitLoadOfLValue(lv, loc).getValue();
   }

clang/lib/CIR/CodeGen/CIRGenFunction.h

@@ -1557,6 +1557,9 @@ class CIRGenFunction : public CIRGenTypeCache {
   void emitForwardingCallToLambda(const CXXMethodDecl *lambdaCallOperator,
                                   CallArgList &callArgs);
 
+  RValue emitCoawaitExpr(const CoawaitExpr &e,
+                         AggValueSlot aggSlot = AggValueSlot::ignored(),
+                         bool ignoreResult = false);
   /// Emit the computation of the specified expression of complex type,
   /// returning the result.
   mlir::Value emitComplexExpr(const Expr *e);

clang/lib/CIR/CodeGen/CIRGenValue.h

@@ -49,6 +49,7 @@ class RValue {
   bool isScalar() const { return flavor == Scalar; }
   bool isComplex() const { return flavor == Complex; }
   bool isAggregate() const { return flavor == Aggregate; }
+  bool isIgnored() const { return isScalar() && !getValue(); }
 
   bool isVolatileQualified() const { return isVolatile; }
 

clang/lib/CIR/Dialect/IR/CIRDialect.cpp

@@ -289,7 +289,10 @@ void cir::ConditionOp::getSuccessorRegions(
     regions.emplace_back(getOperation(), loopOp->getResults());
   }
 
-  assert(!cir::MissingFeatures::awaitOp());
+  // Parent is an await: condition may branch to resume or suspend regions.
+  auto await = cast<AwaitOp>(getOperation()->getParentOp());
+  regions.emplace_back(&await.getResume(), await.getResume().getArguments());
+  regions.emplace_back(&await.getSuspend(), await.getSuspend().getArguments());
 }
 
 MutableOperandRange
@@ -299,8 +302,7 @@ cir::ConditionOp::getMutableSuccessorOperands(RegionSuccessor point) {
 }
 
 LogicalResult cir::ConditionOp::verify() {
-  assert(!cir::MissingFeatures::awaitOp());
-  if (!isa<LoopOpInterface>(getOperation()->getParentOp()))
+  if (!isa<LoopOpInterface, AwaitOp>(getOperation()->getParentOp()))
     return emitOpError("condition must be within a conditional region");
   return success();
 }
@@ -1900,6 +1902,19 @@ void cir::FuncOp::print(OpAsmPrinter &p) {
 
 mlir::LogicalResult cir::FuncOp::verify() {
 
+  if (!isDeclaration() && getCoroutine()) {
+    bool foundAwait = false;
+    this->walk([&](Operation *op) {
+      if (auto await = dyn_cast<AwaitOp>(op)) {
+        foundAwait = true;
+        return;
+      }
+    });
+    if (!foundAwait)
+      return emitOpError()
+             << "coroutine body must use at least one cir.await op";
+  }
+
   llvm::SmallSet<llvm::StringRef, 16> labels;
   llvm::SmallSet<llvm::StringRef, 16> gotos;
 
@@ -2116,6 +2131,65 @@ OpFoldResult cir::UnaryOp::fold(FoldAdaptor adaptor) {
 
   return {};
 }
+//===----------------------------------------------------------------------===//
+// AwaitOp
+//===----------------------------------------------------------------------===//
+
+void cir::AwaitOp::build(OpBuilder &builder, OperationState &result,
+                         cir::AwaitKind kind, BuilderCallbackRef readyBuilder,
+                         BuilderCallbackRef suspendBuilder,
+                         BuilderCallbackRef resumeBuilder) {
+  result.addAttribute(getKindAttrName(result.name),
+                      cir::AwaitKindAttr::get(builder.getContext(), kind));
+  {
+    OpBuilder::InsertionGuard guard(builder);
+    Region *readyRegion = result.addRegion();
+    builder.createBlock(readyRegion);
+    readyBuilder(builder, result.location);
+  }
+
+  {
+    OpBuilder::InsertionGuard guard(builder);
+    Region *suspendRegion = result.addRegion();
+    builder.createBlock(suspendRegion);
+    suspendBuilder(builder, result.location);
+  }
+
+  {
+    OpBuilder::InsertionGuard guard(builder);
+    Region *resumeRegion = result.addRegion();
+    builder.createBlock(resumeRegion);
+    resumeBuilder(builder, result.location);
+  }
+}
+
+/// Given the region at `index`, or the parent operation if `index` is None,
+/// return the successor regions. These are the regions that may be selected
+/// during the flow of control. `operands` is a set of optional attributes
+/// that correspond to a constant value for each operand, or null if that
+/// operand is not a constant.
+void cir::AwaitOp::getSuccessorRegions(
+    mlir::RegionBranchPoint point, SmallVectorImpl<RegionSuccessor> &regions) {
+  // If any index all the underlying regions branch back to the parent
+  // operation.
+  if (!point.isParent()) {
+    regions.push_back(
+        RegionSuccessor(getOperation(), getOperation()->getResults()));
+    return;
+  }
+
+  // FIXME: we want to look at cond region for getting more accurate results
+  // if the other regions will get a chance to execute.
+  regions.push_back(RegionSuccessor(&this->getReady()));
+  regions.push_back(RegionSuccessor(&this->getSuspend()));
+  regions.push_back(RegionSuccessor(&this->getResume()));
+}
+
+LogicalResult cir::AwaitOp::verify() {
+  if (!isa<ConditionOp>(this->getReady().back().getTerminator()))
+    return emitOpError("ready region must end with cir.condition");
+  return success();
+}
 
 //===----------------------------------------------------------------------===//
 // CopyOp Definitions

clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp

@@ -3778,6 +3778,12 @@ mlir::LogicalResult CIRToLLVMVAArgOpLowering::matchAndRewrite(
   return mlir::success();
 }
 
+mlir::LogicalResult CIRToLLVMAwaitOpLowering::matchAndRewrite(
+    cir::AwaitOp op, OpAdaptor adaptor,
+    mlir::ConversionPatternRewriter &rewriter) const {
+  return mlir::failure();
+}
+
 std::unique_ptr<mlir::Pass> createConvertCIRToLLVMPass() {
   return std::make_unique<ConvertCIRToLLVMPass>();
 }

clang/lib/CodeGen/CGValue.h

@@ -64,6 +64,7 @@ class RValue {
   bool isScalar() const { return Flavor == Scalar; }
   bool isComplex() const { return Flavor == Complex; }
   bool isAggregate() const { return Flavor == Aggregate; }
+  bool isIgnored() const { return isScalar() && !getScalarVal(); }
 
   bool isVolatileQualified() const { return IsVolatile; }