[debuginfo][coro] Emit debug info labels for coroutine resume points

[vogelsgesang]

RFC on discourse: https://discourse.llvm.org/t/rfc-debug-info-for-coroutine-suspension-locations-take-2/86606

With this commit, we add DILabel debug infos to the resume points of a coroutine. Those labels can be used by debugging scripts to figure out the exact line and column at which a coroutine was suspended by looking up current __coro_index value inside the coroutines frame, and then searching for the corresponding label inside the coroutine's resume function.

The DWARF information generated for such a label looks like:

0x00000f71:     DW_TAG_label
                  DW_AT_name    ("__coro_resume_1")
                  DW_AT_decl_file       ("generator-example.cpp")
                  DW_AT_decl_line       (5)
                  DW_AT_decl_column     (3)
                  DW_AT_artificial      (true)
                  DW_AT_LLVM_coro_suspend_idx   (0x01)
                  DW_AT_low_pc  (0x00000000000019be)

The labels can be mapped to their corresponding __coro_idx values either via their naming convention __coro_resume_<N> or using the new DW_AT_LLVM_coro_suspend_idx attribute. In gdb, those line numebrs can be looked up using info line -function my_coroutine -label __coro_resume_1. LLDB unfortunately does not understand DW_TAG_label debug information, yet.

Given this is an artificial compiler-generated label, I did apply the DW_AT_artificial tag to it. The DWARFv5 standard only allows that tag on type and variable definitions, but this is a natural extension and was also blessed in the RFC on discourse.

Also, this commit adds DW_AT_decl_column to labels, not only for coroutines but also for normal C and C++ labels. While not strictly necessary, I am doing so now because it would be harder to do so later without breaking the binary LLVM-IR format

Drive-by fixes: While reading the existing test cases to understand how to write my own test case, I did a couple of small typo fixes and comment improvements

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[llvmbot]

@llvm/pr-subscribers-mlir
@llvm/pr-subscribers-mlir-llvm
@llvm/pr-subscribers-llvm-binary-utilities
@llvm/pr-subscribers-clang-codegen
@llvm/pr-subscribers-llvm-ir
@llvm/pr-subscribers-clang
@llvm/pr-subscribers-debuginfo
@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-coroutines

Author: Adrian Vogelsgesang (vogelsgesang)

Changes

With this commit, we add DILabel debug infos (lowering to DW_TAG_label in DWARF) to the resume points of a coroutine. Those labels use the naming convention __coro_resume_<N> where <N> is the suspension point id.

That way, debugging scripts can figure out the exact line / column at which a coroutine was suspended by looking up current __coro_index value inside the coroutines frame, and then searching for the corresponding __coro_resume_<N> inside the coroutine's resume function.

While this is an artificial compiler-generated label, I did not apply the DW_AT_artificial tag to it. The DWARFv5 standard only allows that tag on type and variable definitions, but not on labels.

In gdb, those line numebers can then be looked up using the command info line -function my_coroutine -label __coro_resume_1.

LLDB unfortunately does not parse DW_TAG_label debug information, yet. As such, this debug information is currently not useful in LLDB.

Corresponding RFC in https://discourse.llvm.org/t/rfc-debug-info-for-coroutine-suspension-locations-take-2/86606

---

Full diff: https://github.com/llvm/llvm-project/pull/141937.diff

2 Files Affected:

• (modified) llvm/lib/Transforms/Coroutines/CoroFrame.cpp (+2-3)
• (modified) llvm/lib/Transforms/Coroutines/CoroSplit.cpp (+22-1)

diff --git a/llvm/lib/Transforms/Coroutines/CoroFrame.cpp b/llvm/lib/Transforms/Coroutines/CoroFrame.cpp
index 2f5f1089067bf..7fe19ca0dd02a 100644
--- a/llvm/lib/Transforms/Coroutines/CoroFrame.cpp
+++ b/llvm/lib/Transforms/Coroutines/CoroFrame.cpp
@@ -689,9 +689,8 @@ static DIType *solveDIType(DIBuilder &Builder, Type *Ty,
 static void buildFrameDebugInfo(Function &F, coro::Shape &Shape,
                                 FrameDataInfo &FrameData) {
   DISubprogram *DIS = F.getSubprogram();
-  // If there is no DISubprogram for F, it implies the Function are not compiled
-  // with debug info. So we also don't need to generate debug info for the frame
-  // neither.
+  // If there is no DISubprogram for F, it implies the function is compiled without
+  // debug info. So we also don't generate debug info for the frame, either.
   if (!DIS || !DIS->getUnit() ||
       !dwarf::isCPlusPlus(
           (dwarf::SourceLanguage)DIS->getUnit()->getSourceLanguage()) ||
diff --git a/llvm/lib/Transforms/Coroutines/CoroSplit.cpp b/llvm/lib/Transforms/Coroutines/CoroSplit.cpp
index f9a6c70fedc2d..d84886cd1ca4e 100644
--- a/llvm/lib/Transforms/Coroutines/CoroSplit.cpp
+++ b/llvm/lib/Transforms/Coroutines/CoroSplit.cpp
@@ -42,6 +42,7 @@
 #include "llvm/IR/CFG.h"
 #include "llvm/IR/CallingConv.h"
 #include "llvm/IR/Constants.h"
+#include "llvm/IR/DIBuilder.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/DebugInfo.h"
 #include "llvm/IR/DerivedTypes.h"
@@ -302,7 +303,7 @@ static void replaceFallthroughCoroEnd(AnyCoroEndInst *End,
 }
 
 // Mark a coroutine as done, which implies that the coroutine is finished and
-// never get resumed.
+// never gets resumed.
 //
 // In resume-switched ABI, the done state is represented by storing zero in
 // ResumeFnAddr.
@@ -1492,6 +1493,14 @@ struct SwitchCoroutineSplitter {
   static void createResumeEntryBlock(Function &F, coro::Shape &Shape) {
     LLVMContext &C = F.getContext();
 
+    DIBuilder DBuilder(*F.getParent(), /*AllowUnresolved*/ false);
+    DISubprogram *DIS = F.getSubprogram();
+    // If there is no DISubprogram for F, it implies the function is compiled without
+    // debug info. So we also don't generate debug info for the suspension points, either.
+    bool AddDebugLabels = (DIS && DIS->getUnit() &&
+        (DIS->getUnit()->getEmissionKind() == DICompileUnit::DebugEmissionKind::FullDebug ||
+         DIS->getUnit()->getEmissionKind() == DICompileUnit::DebugEmissionKind::LineTablesOnly));
+
     // resume.entry:
     //  %index.addr = getelementptr inbounds %f.Frame, %f.Frame* %FramePtr, i32
     //  0, i32 2 % index = load i32, i32* %index.addr switch i32 %index, label
@@ -1514,6 +1523,7 @@ struct SwitchCoroutineSplitter {
         Builder.CreateSwitch(Index, UnreachBB, Shape.CoroSuspends.size());
     Shape.SwitchLowering.ResumeSwitch = Switch;
 
+    // Split all coro.suspend calls
     size_t SuspendIndex = 0;
     for (auto *AnyS : Shape.CoroSuspends) {
       auto *S = cast<CoroSuspendInst>(AnyS);
@@ -1552,6 +1562,7 @@ struct SwitchCoroutineSplitter {
       //     br label %resume.0.landing
       //
       //  resume.0: ; <--- jump from the switch in the resume.entry
+      //     XXX: label
       //     %0 = tail call i8 @llvm.coro.suspend(token none, i1 false)
       //     br label %resume.0.landing
       //
@@ -1574,11 +1585,21 @@ struct SwitchCoroutineSplitter {
       PN->addIncoming(Builder.getInt8(-1), SuspendBB);
       PN->addIncoming(S, ResumeBB);
 
+      if (AddDebugLabels) {
+         if (DebugLoc SuspendLoc = S->getDebugLoc()) {
+            std::string labelName = ("__coro_resume_" + Twine(SuspendIndex)).str();
+            DILocation& DILoc = *SuspendLoc.get();
+            DILabel *ResumeLabel = DBuilder.createLabel(DIS, labelName, DILoc.getFile(), SuspendLoc.getLine());
+            DBuilder.insertLabel(ResumeLabel, &DILoc, ResumeBB->begin());
+         }
+      }
+
       ++SuspendIndex;
     }
 
     Builder.SetInsertPoint(UnreachBB);
     Builder.CreateUnreachable();
+    DBuilder.finalize();
 
     Shape.SwitchLowering.ResumeEntryBlock = NewEntry;
   }

[vogelsgesang]

CC @iains @ArsenArsen, since from the gcc commit log it seems you are working on coroutine support in gcc. Would be great if clang/LLVM and gcc could agree on a common approach to emit debug info for coroutine suspension points 🙂

See https://discourse.llvm.org/t/rfc-debug-info-for-coroutine-suspension-locations-take-2/86606 for the actual discussion

[ChuanqiXu9]

I love this really.

[vogelsgesang]

@ChuanqiXu9 @dwblaikie please let me know if you feel comfortable reviewing this change or if we also need input from others 🙂

[ChuanqiXu9]

@ChuanqiXu9 @dwblaikie please let me know if you feel comfortable reviewing this change or if we also need input from others 🙂

I like the work but my knowledge to dwarf is limited. But I'd like to approve this if no one gives feedbacks. I do think it is better to move on.

[vogelsgesang]

Could you review clang and the coroutine-specific pieces? I think the DWARF and DILabel aspects @dwblaikie might be able to give feedback?

[ChuanqiXu9]

I don't feel there are coroutine specific things, but if you want, I can accept this.

[vogelsgesang]

I don't feel there are coroutine specific things

What I would have had in mind: The integration in the CoroSplit pass and the test case changes

I can accept this

Thanks! I will still wait for a 2nd review from someone with debuginfo expertise. We still have 3 weeks before the branch cut for clang-21, so we are not in a hurry :)