[coroutines] Adding builtins for coroutine intrinsics and backendutil support.

Summary:
With this commit simple coroutines can be created in plain C using coroutine builtins.

Reviewers: rnk, EricWF, rsmith

Subscribers: modocache, mgorny, mehdi_amini, beanz, cfe-commits

Differential Revision: https://reviews.llvm.org/D24373

llvm-svn: 283155

clang/docs/LanguageExtensions.rst

@@ -1865,6 +1865,82 @@ The types ``T`` currently supported are:
 Note that the compiler does not guarantee that non-temporal loads or stores
 will be used.
 
+C++ Coroutines support builtins
+--------------------------------
+
+.. warning::
+  This is a work in progress. Compatibility across Clang/LLVM releases is not 
+  guaranteed.
+
+Clang provides experimental builtins to support C++ Coroutines as defined by
+http://wg21.link/P0057. The following four are intended to be used by the
+standard library to implement `std::experimental::coroutine_handle` type.
+
+**Syntax**:
+
+.. code-block:: c
+
+  void  __builtin_coro_resume(void *addr);
+  void  __builtin_coro_destroy(void *addr);
+  bool  __builtin_coro_done(void *addr);
+  void *__builtin_coro_promise(void *addr, int alignment, bool from_promise)
+
+**Example of use**:
+
+.. code-block:: c++
+
+  template <> struct coroutine_handle<void> {
+    void resume() const { __builtin_coro_resume(ptr); }
+    void destroy() const { __builtin_coro_destroy(ptr); }
+    bool done() const { return __builtin_coro_done(ptr); }
+    // ...
+  protected:
+    void *ptr;
+  };
+
+  template <typename Promise> struct coroutine_handle : coroutine_handle<> {
+    // ...
+    Promise &promise() const {
+      return *reinterpret_cast<Promise *>(
+        __builtin_coro_promise(ptr, alignof(Promise), /*from-promise=*/false));
+    }
+    static coroutine_handle from_promise(Promise &promise) {
+      coroutine_handle p;
+      p.ptr = __builtin_coro_promise(&promise, alignof(Promise),
+                                                      /*from-promise=*/true);
+      return p;
+    }
+  };
+
+
+Other coroutine builtins are either for internal clang use or for use during
+development of the coroutine feature. See `Coroutines in LLVM
+<http://llvm.org/docs/Coroutines.html#intrinsics>`_ for
+more information on their semantics. Note that builtins matching the intrinsics
+that take token as the first parameter (llvm.coro.begin, llvm.coro.alloc, 
+llvm.coro.free and llvm.coro.suspend) omit the token parameter and fill it to
+an appropriate value during the emission.
+
+**Syntax**:
+
+.. code-block:: c
+
+  size_t __builtin_coro_size()
+  void  *__builtin_coro_frame()
+  void  *__builtin_coro_free(void *coro_frame)
+
+  void  *__builtin_coro_id(int align, void *promise, void *fnaddr, void *parts)
+  bool   __builtin_coro_alloc()
+  void  *__builtin_coro_begin(void *memory)
+  void   __builtin_coro_end(void *coro_frame, bool unwind)
+  char   __builtin_coro_suspend(bool final)
+  bool   __builtin_coro_param(void *original, void *copy)
+
+Note that there is no builtin matching the `llvm.coro.save` intrinsic. LLVM
+automatically will insert one if the first argument to `llvm.coro.suspend` is
+token `none`. If a user calls `__builin_suspend`, clang will insert `token none`
+as the first argument to the intrinsic.
+
 Non-standard C++11 Attributes
 =============================
 

clang/include/clang/Basic/Builtins.def

@@ -1330,6 +1330,22 @@ BUILTIN(__builtin___get_unsafe_stack_ptr, "v*", "Fn")
 BUILTIN(__builtin_nontemporal_store, "v.", "t")
 BUILTIN(__builtin_nontemporal_load, "v.", "t")
 
+// Coroutine intrinsics.
+BUILTIN(__builtin_coro_resume, "vv*", "")
+BUILTIN(__builtin_coro_destroy, "vv*", "")
+BUILTIN(__builtin_coro_done, "bv*", "n")
+BUILTIN(__builtin_coro_promise, "v*v*IiIb", "n")
+
+BUILTIN(__builtin_coro_size, "z", "n")
+BUILTIN(__builtin_coro_frame, "v*", "n")
+BUILTIN(__builtin_coro_free, "v*v*", "n")
+
+BUILTIN(__builtin_coro_id, "v*Iiv*v*v*", "n")
+BUILTIN(__builtin_coro_alloc, "b", "n")
+BUILTIN(__builtin_coro_begin, "v*v*", "n")
+BUILTIN(__builtin_coro_end, "vv*Ib", "n")
+BUILTIN(__builtin_coro_suspend, "cIb", "n")
+BUILTIN(__builtin_coro_param, "bv*v*", "n")
 // OpenCL v2.0 s6.13.16, s9.17.3.5 - Pipe functions.
 // We need the generic prototype, since the packet type could be anything.
 LANGBUILTIN(read_pipe, "i.", "tn", OCLC20_LANG)

clang/lib/CodeGen/BackendUtil.cpp

@@ -41,6 +41,7 @@
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/Target/TargetSubtargetInfo.h"
+#include "llvm/Transforms/Coroutines.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/IPO/AlwaysInliner.h"
 #include "llvm/Transforms/IPO/PassManagerBuilder.h"
@@ -401,6 +402,9 @@ void EmitAssemblyHelper::CreatePasses(legacy::PassManager &MPM,
                            addDataFlowSanitizerPass);
   }
 
+  if (LangOpts.CoroutinesTS)
+    addCoroutinePassesToExtensionPoints(PMBuilder);
+
   if (LangOpts.Sanitize.hasOneOf(SanitizerKind::Efficiency)) {
     PMBuilder.addExtension(PassManagerBuilder::EP_OptimizerLast,
                            addEfficiencySanitizerPass);

clang/lib/CodeGen/CGBuiltin.cpp

@@ -2135,6 +2135,39 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     break;
   }
 
+  case Builtin::BI__builtin_coro_size: {
+    auto & Context = getContext();
+    auto SizeTy = Context.getSizeType();
+    auto T = Builder.getIntNTy(Context.getTypeSize(SizeTy));
+    Value *F = CGM.getIntrinsic(Intrinsic::coro_size, T);
+    return RValue::get(Builder.CreateCall(F));
+  }
+
+  case Builtin::BI__builtin_coro_id:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_id);
+  case Builtin::BI__builtin_coro_promise:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_promise);
+  case Builtin::BI__builtin_coro_resume:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_resume);
+  case Builtin::BI__builtin_coro_frame:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_frame);
+  case Builtin::BI__builtin_coro_free:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_free);
+  case Builtin::BI__builtin_coro_destroy:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_destroy);
+  case Builtin::BI__builtin_coro_done:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_done);
+  case Builtin::BI__builtin_coro_alloc:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_alloc);
+  case Builtin::BI__builtin_coro_begin:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_begin);
+  case Builtin::BI__builtin_coro_end:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_end);
+  case Builtin::BI__builtin_coro_suspend:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_suspend);
+  case Builtin::BI__builtin_coro_param:
+    return EmitCoroutineIntrinsic(E, Intrinsic::coro_param);
+
   // OpenCL v2.0 s6.13.16.2, Built-in pipe read and write functions
   case Builtin::BIread_pipe:
   case Builtin::BIwrite_pipe: {

clang/lib/CodeGen/CGCoroutine.cpp

@@ -0,0 +1,100 @@
+//===----- CGCoroutine.cpp - Emit LLVM Code for C++ coroutines ------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This contains code dealing with C++ code generation of coroutines.
+//
+//===----------------------------------------------------------------------===//
+
+#include "CodeGenFunction.h"
+
+using namespace clang;
+using namespace CodeGen;
+
+namespace clang {
+namespace CodeGen {
+
+struct CGCoroData {
+  // Stores the llvm.coro.id emitted in the function so that we can supply it
+  // as the first argument to coro.begin, coro.alloc and coro.free intrinsics.
+  // Note: llvm.coro.id returns a token that cannot be directly expressed in a
+  // builtin.
+  llvm::CallInst *CoroId = nullptr;
+  // If coro.id came from the builtin, remember the expression to give better
+  // diagnostic. If CoroIdExpr is nullptr, the coro.id was created by
+  // EmitCoroutineBody.
+  CallExpr const *CoroIdExpr = nullptr;
+};
+}
+}
+
+clang::CodeGen::CodeGenFunction::CGCoroInfo::CGCoroInfo() {}
+CodeGenFunction::CGCoroInfo::~CGCoroInfo() {}
+
+static bool createCoroData(CodeGenFunction &CGF,
+                           CodeGenFunction::CGCoroInfo &CurCoro,
+                           llvm::CallInst *CoroId, CallExpr const *CoroIdExpr) {
+  if (CurCoro.Data) {
+    if (CurCoro.Data->CoroIdExpr)
+      CGF.CGM.Error(CoroIdExpr->getLocStart(),
+                    "only one __builtin_coro_id can be used in a function");
+    else if (CoroIdExpr)
+      CGF.CGM.Error(CoroIdExpr->getLocStart(),
+                    "__builtin_coro_id shall not be used in a C++ coroutine");
+    else
+      llvm_unreachable("EmitCoroutineBodyStatement called twice?");
+
+    return false;
+  }
+
+  CurCoro.Data = std::unique_ptr<CGCoroData>(new CGCoroData);
+  CurCoro.Data->CoroId = CoroId;
+  CurCoro.Data->CoroIdExpr = CoroIdExpr;
+  return true;
+}
+
+// Emit coroutine intrinsic and patch up arguments of the token type.
+RValue CodeGenFunction::EmitCoroutineIntrinsic(const CallExpr *E,
+                                               unsigned int IID) {
+  SmallVector<llvm::Value *, 8> Args;
+  switch (IID) {
+  default:
+    break;
+  // The following three intrinsics take a token parameter referring to a token
+  // returned by earlier call to @llvm.coro.id. Since we cannot represent it in
+  // builtins, we patch it up here.
+  case llvm::Intrinsic::coro_alloc:
+  case llvm::Intrinsic::coro_begin:
+  case llvm::Intrinsic::coro_free: {
+    if (CurCoro.Data && CurCoro.Data->CoroId) {
+      Args.push_back(CurCoro.Data->CoroId);
+      break;
+    }
+    CGM.Error(E->getLocStart(), "this builtin expect that __builtin_coro_id has"
+                                " been used earlier in this function");
+    // Fallthrough to the next case to add TokenNone as the first argument.
+  }
+  // @llvm.coro.suspend takes a token parameter. Add token 'none' as the first
+  // argument.
+  case llvm::Intrinsic::coro_suspend:
+    Args.push_back(llvm::ConstantTokenNone::get(getLLVMContext()));
+    break;
+  }
+  for (auto &Arg : E->arguments())
+    Args.push_back(EmitScalarExpr(Arg));
+
+  llvm::Value *F = CGM.getIntrinsic(IID);
+  llvm::CallInst *Call = Builder.CreateCall(F, Args);
+
+  // If we see @llvm.coro.id remember it in the CoroData. We will update
+  // coro.alloc, coro.begin and coro.free intrinsics to refer to it.
+  if (IID == llvm::Intrinsic::coro_id) {
+    createCoroData(*this, CurCoro, Call, E);
+  }
+  return RValue::get(Call);
+}

clang/lib/CodeGen/CMakeLists.txt

@@ -3,6 +3,7 @@ set(LLVM_LINK_COMPONENTS
   BitReader
   BitWriter
   Core
+  Coroutines
   Coverage
   IPO
   IRReader
@@ -42,6 +43,7 @@ add_clang_library(clangCodeGen
   CGCall.cpp
   CGClass.cpp
   CGCleanup.cpp
+  CGCoroutine.cpp
   CGDebugInfo.cpp
   CGDecl.cpp
   CGDeclCXX.cpp

clang/lib/CodeGen/CodeGenFunction.h

@@ -88,6 +88,7 @@ class BlockFieldFlags;
 class RegionCodeGenTy;
 class TargetCodeGenInfo;
 struct OMPTaskDataTy;
+struct CGCoroData;
 
 /// The kind of evaluation to perform on values of a particular
 /// type.  Basically, is the code in CGExprScalar, CGExprComplex, or
@@ -155,6 +156,16 @@ public:
   QualType FnRetTy;
   llvm::Function *CurFn;
 
+  // Holds coroutine data if the current function is a coroutine. We use a
+  // wrapper to manage its lifetime, so that we don't have to define CGCoroData
+  // in this header.
+  struct CGCoroInfo {
+    std::unique_ptr<CGCoroData> Data;
+    CGCoroInfo();
+    ~CGCoroInfo();
+  };
+  CGCoroInfo CurCoro;
+
   /// CurGD - The GlobalDecl for the current function being compiled.
   GlobalDecl CurGD;
 
@@ -2290,6 +2301,8 @@ public:
   void EmitObjCAtSynchronizedStmt(const ObjCAtSynchronizedStmt &S);
   void EmitObjCAutoreleasePoolStmt(const ObjCAutoreleasePoolStmt &S);
 
+  RValue EmitCoroutineIntrinsic(const CallExpr *E, unsigned int IID);
+
   void EnterCXXTryStmt(const CXXTryStmt &S, bool IsFnTryBlock = false);
   void ExitCXXTryStmt(const CXXTryStmt &S, bool IsFnTryBlock = false);
 

clang/test/CodeGenCoroutines/coro-builtins-err.c

@@ -0,0 +1,10 @@
+// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc18.0.0 -fcoroutines-ts -emit-llvm %s -o - -verify
+
+void f() {
+  __builtin_coro_alloc(); // expected-error {{this builtin expect that __builtin_coro_id}}
+  __builtin_coro_begin(0); // expected-error {{this builtin expect that __builtin_coro_id}}
+  __builtin_coro_free(0); // expected-error {{this builtin expect that __builtin_coro_id}}
+
+  __builtin_coro_id(32, 0, 0, 0);
+  __builtin_coro_id(32, 0, 0, 0); // expected-error {{only one __builtin_coro_id}}
+}

clang/test/CodeGenCoroutines/coro-builtins.c

@@ -0,0 +1,56 @@
+// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc18.0.0 -fcoroutines-ts -emit-llvm %s -o - -disable-llvm-passes | FileCheck %s
+
+void *myAlloc(long long);
+
+// CHECK-LABEL: f( 
+void f(int n) {
+  // CHECK: %n.addr = alloca i32
+  // CHECK: %n_copy = alloca i32
+  // CHECK: %promise = alloca i32
+  int n_copy;
+  int promise;
+
+  // CHECK: %[[PROM_ADDR:.+]] = bitcast i32* %promise to i8*
+  // CHECK-NEXT: %[[COROID:.+]] = call token @llvm.coro.id(i32 32, i8* %[[PROM_ADDR]], i8* null, i8* null)
+  __builtin_coro_id(32, &promise, 0, 0);
+
+  // CHECK-NEXT: call i1 @llvm.coro.alloc(token %[[COROID]])
+  __builtin_coro_alloc();
+
+  // CHECK-NEXT: %[[SIZE:.+]] = call i64 @llvm.coro.size.i64()
+  // CHECK-NEXT: %[[MEM:.+]] = call i8* @myAlloc(i64 %[[SIZE]])
+  // CHECK-NEXT: %[[BEG:.+]] = call i8* @llvm.coro.begin(token %[[COROID]], i8* %[[MEM]])
+  __builtin_coro_begin(myAlloc(__builtin_coro_size()));
+
+  // CHECK-NEXT: %[[FRAME1:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call void @llvm.coro.resume(i8* %[[FRAME1]])
+  __builtin_coro_resume(__builtin_coro_frame());
+
+  // CHECK-NEXT: %[[FRAME2:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call void @llvm.coro.destroy(i8* %[[FRAME2]])
+  __builtin_coro_destroy(__builtin_coro_frame());
+
+  // CHECK-NEXT: %[[FRAME3:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call i1 @llvm.coro.done(i8* %[[FRAME3]])
+  __builtin_coro_done(__builtin_coro_frame());
+
+  // CHECK-NEXT: %[[FRAME4:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call i8* @llvm.coro.promise(i8* %[[FRAME4]], i32 48, i1 false)
+  __builtin_coro_promise(__builtin_coro_frame(), 48, 0);
+
+  // CHECK-NEXT: %[[FRAME5:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call i8* @llvm.coro.free(token %[[COROID]], i8* %[[FRAME5]])
+  __builtin_coro_free(__builtin_coro_frame());
+
+  // CHECK-NEXT: %[[FRAME6:.+]] = call i8* @llvm.coro.frame() 
+  // CHECK-NEXT: call void @llvm.coro.end(i8* %[[FRAME6]], i1 false)
+  __builtin_coro_end(__builtin_coro_frame(), 0);
+
+  // CHECK-NEXT: call i8 @llvm.coro.suspend(token none, i1 true)
+  __builtin_coro_suspend(1);
+
+  // CHECK-NEXT: %[[N_ADDR:.+]] = bitcast i32* %n.addr to i8*
+  // CHECK-NEXT: %[[N_COPY_ADDR:.+]] = bitcast i32* %n_copy to i8*
+  // CHECK-NEXT: call i1 @llvm.coro.param(i8* %[[N_ADDR]], i8* %[[N_COPY_ADDR]])
+  __builtin_coro_param(&n, &n_copy);
+}