[Coroutines] Add remarks in CoroSplit and CoroElide passes

[weiwang]

Summary

Add remarks to show frame size and alignment.

[weiwang]

update test.

[ChuanqiXu]

LGTM. Thanks.

llvm/lib/Transforms/Coroutines/CoroElide.cpp

 //===- CoroElide.cpp - Coroutine Frame Allocation Elision Pass ------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Coroutines/CoroElide.h"
 #include "CoroInternal.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/AliasAnalysis.h"
 #include "llvm/Analysis/InstructionSimplify.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/IR/Dominators.h"
 #include "llvm/IR/InstIterator.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/FileSystem.h"
 #include <optional>
 
 using namespace llvm;
 
struct Lowerer : coro::LowererBase {
   SmallPtrSet<const SwitchInst *, 4> CoroSuspendSwitches;
 
   Lowerer(Module &M) : LowererBase(M) {}
 
   void elideHeapAllocations(Function *F, uint64_t FrameSize, Align FrameAlign,
                             AAResults &AA);
   bool shouldElide(Function *F, DominatorTree &DT) const;
   void collectPostSplitCoroIds(Function *F);
-  bool processCoroId(CoroIdInst *, AAResults &AA, DominatorTree &DT);
+  bool processCoroId(CoroIdInst *, AAResults &AA, DominatorTree &DT,
+                     OptimizationRemarkEmitter &ORE);
   bool hasEscapePath(const CoroBeginInst *,
                      const SmallPtrSetImpl<BasicBlock *> &) const;
 };
 } // end anonymous namespace
 
 // Go through the list of coro.subfn.addr intrinsics and replace them with the
 // provided constant.
 static void replaceWithConstant(Constant *Value,
if (auto *CSI = dyn_cast<CoroSuspendInst>(&I))
         SwitchInst *SWI = cast<SwitchInst>(CSI->use_begin()->getUser());
         if (SWI->getNumCases() == 2)
           CoroSuspendSwitches.insert(SWI);
       }
   }
 }
 
 bool Lowerer::processCoroId(CoroIdInst *CoroId, AAResults &AA,
-                            DominatorTree &DT) {
+                            DominatorTree &DT, OptimizationRemarkEmitter &ORE) {
   CoroBegins.clear();
   CoroAllocs.clear();
   ResumeAddr.clear();
   DestroyAddr.clear();
 
   // Collect all coro.begin and coro.allocs associated with this coro.id.
   for (User *U : CoroId->users()) {
     if (auto *CB = dyn_cast<CoroBeginInst>(U))
bool Lowerer::processCoroId(CoroIdInst *CoroId, AAResults &AA,
   assert(Resumers && "PostSplit coro.id Info argument must refer to an array"
                      "of coroutine subfunctions");
   auto *ResumeAddrConstant =
       Resumers->getAggregateElement(CoroSubFnInst::ResumeIndex);
 
   replaceWithConstant(ResumeAddrConstant, ResumeAddr);
 
   bool ShouldElide = shouldElide(CoroId->getFunction(), DT);
+  if (!ShouldElide)
+    ORE.emit([&]() {
+      if (auto FrameSizeAndAlign =
+              getFrameLayout(cast<Function>(ResumeAddrConstant)))
+        return OptimizationRemarkMissed(DEBUG_TYPE, "CoroElide", CoroId)
+               << "'" << ore::NV("callee", CoroId->getCoroutine()->getName())
+               << "' not elided in '"
+               << ore::NV("caller", CoroId->getFunction()->getName())
+               << "' (frame_size="
+               << ore::NV("frame_size", FrameSizeAndAlign->first) << ", align="
+               << ore::NV("align", FrameSizeAndAlign->second.value()) << ")";
+      else
+        return OptimizationRemarkMissed(DEBUG_TYPE, "CoroElide", CoroId)
+               << "'" << ore::NV("callee", CoroId->getCoroutine()->getName())
+               << "' not elided in '"
+               << ore::NV("caller", CoroId->getFunction()->getName())
+               << "' (frame_size=unknown, align=unknown)";
+    });
 
   auto *DestroyAddrConstant = Resumers->getAggregateElement(
       ShouldElide ? CoroSubFnInst::CleanupIndex : CoroSubFnInst::DestroyIndex);
 
   for (auto &It : DestroyAddr)
     replaceWithConstant(DestroyAddrConstant, It.second);
 
   if (ShouldElide) {
     if (auto FrameSizeAndAlign =
             getFrameLayout(cast<Function>(ResumeAddrConstant))) {
       elideHeapAllocations(CoroId->getFunction(), FrameSizeAndAlign->first,
                            FrameSizeAndAlign->second, AA);
       coro::replaceCoroFree(CoroId, /*Elide=*/true);
       NumOfCoroElided++;
 #ifndef NDEBUG
       if (!CoroElideInfoOutputFilename.empty())
         *getOrCreateLogFile()
             << "Elide " << CoroId->getCoroutine()->getName() << " in "
             << CoroId->getFunction()->getName() << "\n";
 #endif
+      ORE.emit([&]() {
+        return OptimizationRemark(DEBUG_TYPE, "CoroElide", CoroId)
+               << "'" << ore::NV("callee", CoroId->getCoroutine()->getName())
+               << "' elided in '"
+               << ore::NV("caller", CoroId->getFunction()->getName())
+               << "' (frame_size="
+               << ore::NV("frame_size", FrameSizeAndAlign->first) << ", align="
+               << ore::NV("align", FrameSizeAndAlign->second.value()) << ")";
+      });
+    } else {
+      ORE.emit([&]() {
+        return OptimizationRemarkMissed(DEBUG_TYPE, "CoroElide", CoroId)
+               << "'" << ore::NV("callee", CoroId->getCoroutine()->getName())
+               << "' not elided in '"
+               << ore::NV("caller", CoroId->getFunction()->getName())
+               << "' (frame_size=unknown, align=unknown)";
+      });
     }
   }
 
   return true;
 }
 
 static bool declaresCoroElideIntrinsics(Module &M) {
   return coro::declaresIntrinsics(M, {"llvm.coro.id", "llvm.coro.id.async"});
 }
 
 PreservedAnalyses CoroElidePass::run(Function &F, FunctionAnalysisManager &AM) {
   auto &M = *F.getParent();
   if (!declaresCoroElideIntrinsics(M))
     return PreservedAnalyses::all();
 
   Lowerer L(M);
   L.CoroIds.clear();
   L.collectPostSplitCoroIds(&F);
   // If we did not find any coro.id, there is nothing to do.
   if (L.CoroIds.empty())
     return PreservedAnalyses::all();
 
   AAResults &AA = AM.getResult<AAManager>(F);
   DominatorTree &DT = AM.getResult<DominatorTreeAnalysis>(F);
+  auto &ORE = AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
 
   bool Changed = false;
   for (auto *CII : L.CoroIds)
-    Changed |= L.processCoroId(CII, AA, DT);
+    Changed |= L.processCoroId(CII, AA, DT, ORE);
 
   return Changed ? PreservedAnalyses::none() : PreservedAnalyses::all();
 }

llvm/lib/Transforms/Coroutines/CoroSplit.cpp

 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/Analysis/CFG.h"
 #include "llvm/Analysis/CallGraph.h"
 #include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/LazyCallGraph.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/BinaryFormat/Dwarf.h"
 #include "llvm/IR/Argument.h"
 #include "llvm/IR/Attributes.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/CFG.h"
 #include "llvm/IR/CallingConv.h"
 #include "llvm/IR/Constants.h"
PreservedAnalyses CoroSplitPass::run(LazyCallGraph::SCC &C,
   // Split all the coroutines.
   for (LazyCallGraph::Node *N : Coroutines) {
     Function &F = N->getFunction();
     LLVM_DEBUG(dbgs() << "CoroSplit: Processing coroutine '" << F.getName()
                       << "\n");
     F.setSplittedCoroutine();
 
     SmallVector<Function *, 4> Clones;
+    auto &ORE = FAM.getResult<OptimizationRemarkEmitterAnalysis>(F);
     const coro::Shape Shape =
         splitCoroutine(F, Clones, FAM.getResult<TargetIRAnalysis>(F),
                        OptimizeFrame, MaterializableCallback);
     updateCallGraphAfterCoroutineSplit(*N, Shape, Clones, C, CG, AM, UR, FAM);
 
+    ORE.emit([&]() {
+      return OptimizationRemark(DEBUG_TYPE, "CoroSplit", &F)
+             << "Split '" << ore::NV("function", F.getName())
+             << "' (frame_size=" << ore::NV("frame_size", Shape.FrameSize)
+             << ", align=" << ore::NV("align", Shape.FrameAlign.value()) << ")";
+    });
+
     if (!Shape.CoroSuspends.empty()) {
       // Run the CGSCC pipeline on the original and newly split functions.
       UR.CWorklist.insert(&C);
       for (Function *Clone : Clones)
         UR.CWorklist.insert(CG.lookupSCC(CG.get(*Clone)));
     }
   }
 
   if (!PrepareFns.empty()) {
     for (auto *PrepareFn : PrepareFns) {
       replaceAllPrepares(PrepareFn, CG, C);
     }
   }
 
   return PreservedAnalyses::none();
 }

llvm/test/Transforms/Coroutines/remarks.ll

+; Tests coroutine remarks.
+; RUN: opt < %s --disable-output -S -passes='default<O1>' \
+; RUN: --pass-remarks="coro-split|coro-elide" \
+; RUN: --pass-remarks-missed="coro-split|coro-elide" \
+; RUN: --pass-remarks-with-hotness 2>&1 | FileCheck %s
+
+; CHECK: Split 'foo' (frame_size=24, align=8) (hotness: 400)
+; CHECK-NEXT: 'foo' not elided in 'bar' (frame_size=24, align=8) (hotness: 100)
+; CHECK-NEXT: 'foo' not elided in 'bar' (frame_size=24, align=8) (hotness: 100)
+; CHECK-NEXT: 'foo' elided in 'baz' (frame_size=24, align=8) (hotness: 200)
+
+define ptr @foo() presplitcoroutine !prof !0 {
+entry:
+  %id = call token @llvm.coro.id(i32 0, ptr null, ptr null, ptr null)
+  %need.dyn.alloc = call i1 @llvm.coro.alloc(token %id)
+  br i1 %need.dyn.alloc, label %dyn.alloc, label %coro.begin
+dyn.alloc:
+  %size = call i32 @llvm.coro.size.i32()
+  %alloc = call ptr @malloc(i32 %size)
+  br label %coro.begin
+coro.begin:
+  %phi = phi ptr [ null, %entry ], [ %alloc, %dyn.alloc ]
+  %hdl = call ptr @llvm.coro.begin(token %id, ptr %phi)
+  call void @print(i32 0)
+  %0 = call i8 @llvm.coro.suspend(token none, i1 false)
+  switch i8 %0, label %suspend [i8 0, label %resume
+                                i8 1, label %cleanup]
+resume:
+  call void @print(i32 1)
+  br label %cleanup
+
+cleanup:
+  %mem = call ptr @llvm.coro.free(token %id, ptr %hdl)
+  call void @free(ptr %mem)
+  br label %suspend
+suspend:
+  call i1 @llvm.coro.end(ptr %hdl, i1 0)
+  ret ptr %hdl
+}
+
+define i32 @bar() !prof !1 {
+entry:
+  %hdl = call ptr @foo()
+  call void @llvm.coro.resume(ptr %hdl)
+  ret i32 0
+}
+
+define i32 @baz() !prof !2 {
+entry:
+  %hdl = call ptr @foo()
+  call void @llvm.coro.destroy(ptr %hdl)
+  ret i32 0
+}
+
+declare ptr @llvm.coro.free(token, ptr)
+declare i32 @llvm.coro.size.i32()
+declare i8  @llvm.coro.suspend(token, i1)
+declare void @llvm.coro.resume(ptr)
+declare void @llvm.coro.destroy(ptr)
+
+declare token @llvm.coro.id(i32, ptr, ptr, ptr)
+declare i1 @llvm.coro.alloc(token)
+declare ptr @llvm.coro.begin(token, ptr)
+declare i1 @llvm.coro.end(ptr, i1)
+
+declare noalias ptr @malloc(i32)
+declare void @print(i32)
+declare void @free(ptr)
+
+!0 = !{!"function_entry_count", i64 400}
+!1 = !{!"function_entry_count", i64 100}
+!2 = !{!"function_entry_count", i64 200}