[PseudoProbe] Extend to skip instrumenting probe into the dests of invoke

llvm/include/llvm/Analysis/EHUtils.h

@@ -79,7 +79,6 @@ static void computeEHOnlyBlocks(FunctionT &F, DenseSet<BlockT *> &EHBlocks) {
     }
   }
 
-  EHBlocks.clear();
   for (auto Entry : Statuses) {
     if (Entry.second == EH)
       EHBlocks.insert(Entry.first);

llvm/include/llvm/Transforms/IPO/SampleProfileProbe.h

@@ -81,8 +81,17 @@ class SampleProfileProber {
   uint64_t getFunctionHash() const { return FunctionHash; }
   uint32_t getBlockId(const BasicBlock *BB) const;
   uint32_t getCallsiteId(const Instruction *Call) const;
-  void computeCFGHash();
-  void computeProbeIdForBlocks();
+  void findUnreachableBlocks(DenseSet<BasicBlock *> &BlocksToIgnore);
+  void findInvokeNormalDests(DenseSet<BasicBlock *> &InvokeNormalDests);
+  void computeBlocksToIgnore(DenseSet<BasicBlock *> &BlocksToIgnore,
+                             DenseSet<BasicBlock *> &BlocksAndCallsToIgnore);
+  void computeProbeIdForCallsites(
+      const DenseSet<BasicBlock *> &BlocksAndCallsToIgnore);
+  const Instruction *
+  getOriginalTerminator(const BasicBlock *Head,
+                        const DenseSet<BasicBlock *> &BlocksToIgnore);
+  void computeCFGHash(const DenseSet<BasicBlock *> &BlocksToIgnore);
+  void computeProbeIdForBlocks(const DenseSet<BasicBlock *> &BlocksToIgnore);
   void computeProbeIdForCallsites();
 
   Function *F;

llvm/lib/Transforms/IPO/SampleProfileProbe.cpp

@@ -173,21 +173,114 @@ SampleProfileProber::SampleProfileProber(Function &Func,
   BlockProbeIds.clear();
   CallProbeIds.clear();
   LastProbeId = (uint32_t)PseudoProbeReservedId::Last;
-  computeProbeIdForBlocks();
-  computeProbeIdForCallsites();
-  computeCFGHash();
+
+  DenseSet<BasicBlock *> BlocksToIgnore;
+  DenseSet<BasicBlock *> BlocksAndCallsToIgnore;
+  computeBlocksToIgnore(BlocksToIgnore, BlocksAndCallsToIgnore);
+
+  computeProbeIdForBlocks(BlocksToIgnore);
+  computeProbeIdForCallsites(BlocksAndCallsToIgnore);
+  computeCFGHash(BlocksToIgnore);
+}
+
+// Two purposes to compute the blocks to ignore:
+// 1. Reduce the IR size.
+// 2. Make the instrumentation(checksum) stable. e.g. the frondend may
+// generate unstable IR while optimizing nounwind attribute, some versions are
+// optimized with the call-to-invoke conversion, while other versions do not.
+// This discrepancy in probe ID could cause profile mismatching issues.
+// Note that those ignored blocks are either cold blocks or new split blocks
+// whose original blocks are instrumented, so it shouldn't degrade the profile
+// quality.
+void SampleProfileProber::computeBlocksToIgnore(
+    DenseSet<BasicBlock *> &BlocksToIgnore,
+    DenseSet<BasicBlock *> &BlocksAndCallsToIgnore) {
+  // Ignore the cold EH and unreachable blocks and calls.
+  computeEHOnlyBlocks(*F, BlocksAndCallsToIgnore);
+  findUnreachableBlocks(BlocksAndCallsToIgnore);
+
+  BlocksToIgnore.insert(BlocksAndCallsToIgnore.begin(),
+                        BlocksAndCallsToIgnore.end());
+
+  // Handle the call-to-invoke conversion case: make sure that the probe id and
+  // callsite id are consistent before and after the block split. For block
+  // probe, we only keep the head block probe id and ignore the block ids of the
+  // normal dests. For callsite probe, it's different to block probe, there is
+  // no additional callsite in the normal dests, so we don't ignore the
+  // callsites.
+  findInvokeNormalDests(BlocksToIgnore);
+}
+
+// Unreachable blocks and calls are always cold, ignore them.
+void SampleProfileProber::findUnreachableBlocks(
+    DenseSet<BasicBlock *> &BlocksToIgnore) {
+  for (auto &BB : *F) {
+    if (&BB != &F->getEntryBlock() && pred_size(&BB) == 0)
+      BlocksToIgnore.insert(&BB);
+  }
+}
+
+// In call-to-invoke conversion, basic block can be split into multiple blocks,
+// only instrument probe in the head block, ignore the normal dests.
+void SampleProfileProber::findInvokeNormalDests(
+    DenseSet<BasicBlock *> &InvokeNormalDests) {
+  for (auto &BB : *F) {
+    auto *TI = BB.getTerminator();
+    if (auto *II = dyn_cast<InvokeInst>(TI)) {
+      auto *ND = II->getNormalDest();
+      InvokeNormalDests.insert(ND);
+
+      // The normal dest and the try/catch block are connected by an
+      // unconditional branch.
+      while (pred_size(ND) == 1) {
+        auto *Pred = *pred_begin(ND);
+        if (succ_size(Pred) == 1) {
+          InvokeNormalDests.insert(Pred);
+          ND = Pred;
+        } else
+          break;
+      }
+    }
+  }
+}
+
+// The call-to-invoke conversion splits the original block into a list of block,
+// we need to compute the hash using the original block's successors to keep the
+// CFG Hash consistent. For a given head block, we keep searching the
+// succesor(normal dest or unconditional branch dest) to find the tail block,
+// the tail block's successors are the original block's successors.
+const Instruction *SampleProfileProber::getOriginalTerminator(
+    const BasicBlock *Head, const DenseSet<BasicBlock *> &BlocksToIgnore) {
+  auto *TI = Head->getTerminator();
+  if (auto *II = dyn_cast<InvokeInst>(TI)) {
+    return getOriginalTerminator(II->getNormalDest(), BlocksToIgnore);
+  } else if (succ_size(Head) == 1 &&
+             BlocksToIgnore.contains(*succ_begin(Head))) {
+    // Go to the unconditional branch dest.
+    return getOriginalTerminator(*succ_begin(Head), BlocksToIgnore);
+  }
+  return TI;
 }
 
 // Compute Hash value for the CFG: the lower 32 bits are CRC32 of the index
 // value of each BB in the CFG. The higher 32 bits record the number of edges
 // preceded by the number of indirect calls.
 // This is derived from FuncPGOInstrumentation<Edge, BBInfo>::computeCFGHash().
-void SampleProfileProber::computeCFGHash() {
+void SampleProfileProber::computeCFGHash(
+    const DenseSet<BasicBlock *> &BlocksToIgnore) {
   std::vector<uint8_t> Indexes;
   JamCRC JC;
   for (auto &BB : *F) {
-    for (BasicBlock *Succ : successors(&BB)) {
+    if (BlocksToIgnore.contains(&BB))
+      continue;
+
+    auto *TI = getOriginalTerminator(&BB, BlocksToIgnore);
+    for (unsigned I = 0, E = TI->getNumSuccessors(); I != E; ++I) {
+      auto *Succ = TI->getSuccessor(I);
       auto Index = getBlockId(Succ);
+      // Ingore ignored-block(zero ID) to avoid unstable checksum.
+      if (Index == 0)
+        continue;
       for (int J = 0; J < 4; J++)
         Indexes.push_back((uint8_t)(Index >> (J * 8)));
     }
@@ -207,23 +300,23 @@ void SampleProfileProber::computeCFGHash() {
                     << ", Hash = " << FunctionHash << "\n");
 }
 
-void SampleProfileProber::computeProbeIdForBlocks() {
-  DenseSet<BasicBlock *> KnownColdBlocks;
-  computeEHOnlyBlocks(*F, KnownColdBlocks);
-  // Insert pseudo probe to non-cold blocks only. This will reduce IR size as
-  // well as the binary size while retaining the profile quality.
+void SampleProfileProber::computeProbeIdForBlocks(
+    const DenseSet<BasicBlock *> &BlocksToIgnore) {
   for (auto &BB : *F) {
-    ++LastProbeId;
-    if (!KnownColdBlocks.contains(&BB))
-      BlockProbeIds[&BB] = LastProbeId;
+    if (BlocksToIgnore.contains(&BB))
+      continue;
+    BlockProbeIds[&BB] = ++LastProbeId;
   }
 }
 
-void SampleProfileProber::computeProbeIdForCallsites() {
+void SampleProfileProber::computeProbeIdForCallsites(
+    const DenseSet<BasicBlock *> &BlocksAndCallsToIgnore) {
   LLVMContext &Ctx = F->getContext();
   Module *M = F->getParent();
 
   for (auto &BB : *F) {
+    if (BlocksAndCallsToIgnore.contains(&BB))
+      continue;
     for (auto &I : BB) {
       if (!isa<CallBase>(I))
         continue;

llvm/test/ThinLTO/X86/pseudo-probe-desc-import.ll

@@ -12,8 +12,8 @@
 ; RUN: llvm-lto -thinlto-action=import %t3.bc -thinlto-index=%t3.index.bc -o /dev/null 2>&1 | FileCheck %s  --check-prefix=WARN
 
 
-; CHECK-NOT: {i64 6699318081062747564, i64 4294967295, !"foo"
-; CHECK: !{i64 -2624081020897602054, i64 281479271677951, !"main"
+; CHECK-NOT: {i64 6699318081062747564, i64 [[#]], !"foo"
+; CHECK: !{i64 -2624081020897602054, i64 [[#]], !"main"
 
 ; WARN: warning: Pseudo-probe ignored: source module '{{.*}}' is compiled with -fpseudo-probe-for-profiling while destination module '{{.*}}' is not
 

llvm/test/Transforms/SampleProfile/pseudo-probe-eh.ll

@@ -18,7 +18,7 @@ entry:
           to label %ret unwind label %lpad
 
 ret:
-; CHECK: call void @llvm.pseudoprobe
+; CHECK-NOT: call void @llvm.pseudoprobe
   ret void
 
 lpad:                                             ; preds = %entry

llvm/test/Transforms/SampleProfile/pseudo-probe-invoke.ll

@@ -0,0 +1,155 @@
+; REQUIRES: x86_64-linux
+; RUN: opt < %s -passes=pseudo-probe -S -o - | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+$__clang_call_terminate = comdat any
+
+@x = dso_local global i32 0, align 4, !dbg !0
+
+; Function Attrs: mustprogress noinline nounwind uwtable
+define dso_local void @_Z3barv() #0 personality ptr @__gxx_personality_v0 !dbg !14 {
+entry:
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 1
+  %0 = load volatile i32, ptr @x, align 4, !dbg !17, !tbaa !19
+  %tobool = icmp ne i32 %0, 0, !dbg !17
+  br i1 %tobool, label %if.then, label %if.else, !dbg !23
+
+if.then:                                          ; preds = %entry
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 2
+  invoke void @_Z3foov()
+          to label %invoke.cont unwind label %terminate.lpad, !dbg !24
+
+invoke.cont:                                      ; preds = %if.then
+; CHECK-NOT: call void @llvm.pseudoprobe(i64 -1069303473483922844,
+  invoke void @_Z3bazv()
+          to label %invoke.cont1 unwind label %terminate.lpad, !dbg !26
+
+invoke.cont1:                                     ; preds = %invoke.cont
+; CHECK-NOT: call void @llvm.pseudoprobe(i64 -1069303473483922844,
+  br label %if.end, !dbg !27
+
+if.else:                                          ; preds = %entry
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 3
+  invoke void @_Z3foov()
+          to label %invoke.cont2 unwind label %terminate.lpad, !dbg !28
+
+invoke.cont2:                                     ; preds = %if.else
+; CHECK-NOT: call void @llvm.pseudoprobe(i64 -1069303473483922844,
+  br label %if.end
+
+if.end:                                           ; preds = %invoke.cont2, %invoke.cont1
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 4
+  invoke void @_Z3foov()
+          to label %invoke.cont3 unwind label %terminate.lpad, !dbg !29
+
+invoke.cont3:                                     ; preds = %if.end
+; CHECK-NOT: call void @llvm.pseudoprobe(i64 -1069303473483922844,
+  %1 = load volatile i32, ptr @x, align 4, !dbg !30, !tbaa !19
+  %tobool4 = icmp ne i32 %1, 0, !dbg !30
+  br i1 %tobool4, label %if.then5, label %if.end6, !dbg !32
+
+if.then5:                                         ; preds = %invoke.cont3
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 5
+  %2 = load volatile i32, ptr @x, align 4, !dbg !33, !tbaa !19
+  %inc = add nsw i32 %2, 1, !dbg !33
+  store volatile i32 %inc, ptr @x, align 4, !dbg !33, !tbaa !19
+  br label %if.end6, !dbg !35
+
+if.end6:                                          ; preds = %if.then5, %invoke.cont3
+; CHECK: call void @llvm.pseudoprobe(i64 -1069303473483922844, i64 6
+  ret void, !dbg !36
+
+terminate.lpad:                                   ; preds = %if.end, %if.else, %invoke.cont, %if.then
+; CHECK-NOT: call void @llvm.pseudoprobe(i64 -1069303473483922844,
+  %3 = landingpad { ptr, i32 }
+          catch ptr null, !dbg !24
+  %4 = extractvalue { ptr, i32 } %3, 0, !dbg !24
+  call void @__clang_call_terminate(ptr %4) #3, !dbg !24
+  unreachable, !dbg !24
+}
+
+; Function Attrs: mustprogress noinline nounwind uwtable
+define dso_local void @_Z3foov() #0 !dbg !37 {
+entry:
+  ret void, !dbg !38
+}
+
+declare i32 @__gxx_personality_v0(...)
+
+; Function Attrs: noinline noreturn nounwind uwtable
+define linkonce_odr hidden void @__clang_call_terminate(ptr noundef %0) #1 comdat {
+  %2 = call ptr @__cxa_begin_catch(ptr %0) #4
+  call void @_ZSt9terminatev() #3
+  unreachable
+}
+
+declare ptr @__cxa_begin_catch(ptr)
+
+declare void @_ZSt9terminatev()
+
+; Function Attrs: mustprogress noinline nounwind uwtable
+define dso_local void @_Z3bazv() #0 !dbg !39 {
+entry:
+  ret void, !dbg !40
+}
+
+; CHECK: ![[#]] = !{i64 -3270123626113159616, i64 4294967295, !"_Z3bazv"}
+
+attributes #0 = { mustprogress noinline nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #1 = { noinline noreturn nounwind uwtable "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #2 = { mustprogress noinline norecurse nounwind uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #3 = { noreturn nounwind }
+attributes #4 = { nounwind }
+
+!llvm.dbg.cu = !{!2}
+!llvm.module.flags = !{!7, !8, !9, !10, !11, !12}
+!llvm.ident = !{!13}
+
+!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
+!1 = distinct !DIGlobalVariable(name: "x", scope: !2, file: !3, line: 1, type: !5, isLocal: false, isDefinition: true)
+!2 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !3, producer: "clang version 19.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: None)
+!3 = !DIFile(filename: "test.cpp", directory: "/home", checksumkind: CSK_MD5, checksum: "a4c7b0392f3fd9c8ebb85065159dbb02")
+!4 = !{!0}
+!5 = !DIDerivedType(tag: DW_TAG_volatile_type, baseType: !6)
+!6 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!7 = !{i32 7, !"Dwarf Version", i32 5}
+!8 = !{i32 2, !"Debug Info Version", i32 3}
+!9 = !{i32 1, !"wchar_size", i32 4}
+!10 = !{i32 8, !"PIC Level", i32 2}
+!11 = !{i32 7, !"PIE Level", i32 2}
+!12 = !{i32 7, !"uwtable", i32 2}
+!13 = !{!"clang version 19.0.0"}
+!14 = distinct !DISubprogram(name: "bar", linkageName: "_Z3barv", scope: !3, file: !3, line: 4, type: !15, scopeLine: 4, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
+!15 = !DISubroutineType(types: !16)
+!16 = !{null}
+!17 = !DILocation(line: 5, column: 6, scope: !18)
+!18 = distinct !DILexicalBlock(scope: !14, file: !3, line: 5, column: 6)
+!19 = !{!20, !20, i64 0}
+!20 = !{!"int", !21, i64 0}
+!21 = !{!"omnipotent char", !22, i64 0}
+!22 = !{!"Simple C++ TBAA"}
+!23 = !DILocation(line: 5, column: 6, scope: !14)
+!24 = !DILocation(line: 6, column: 5, scope: !25)
+!25 = distinct !DILexicalBlock(scope: !18, file: !3, line: 5, column: 9)
+!26 = !DILocation(line: 7, column: 5, scope: !25)
+!27 = !DILocation(line: 8, column: 3, scope: !25)
+!28 = !DILocation(line: 9, column: 5, scope: !18)
+!29 = !DILocation(line: 11, column: 3, scope: !14)
+!30 = !DILocation(line: 12, column: 6, scope: !31)
+!31 = distinct !DILexicalBlock(scope: !14, file: !3, line: 12, column: 6)
+!32 = !DILocation(line: 12, column: 6, scope: !14)
+!33 = !DILocation(line: 13, column: 5, scope: !34)
+!34 = distinct !DILexicalBlock(scope: !31, file: !3, line: 12, column: 9)
+!35 = !DILocation(line: 14, column: 5, scope: !34)
+!36 = !DILocation(line: 17, column: 1, scope: !14)
+!37 = distinct !DISubprogram(name: "foo", linkageName: "_Z3foov", scope: !3, file: !3, line: 19, type: !15, scopeLine: 19, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
+!38 = !DILocation(line: 19, column: 13, scope: !37)
+!39 = distinct !DISubprogram(name: "baz", linkageName: "_Z3bazv", scope: !3, file: !3, line: 18, type: !15, scopeLine: 18, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
+!40 = !DILocation(line: 18, column: 13, scope: !39)
+!41 = distinct !DISubprogram(name: "main", scope: !3, file: !3, line: 22, type: !42, scopeLine: 22, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
+!42 = !DISubroutineType(types: !43)
+!43 = !{!6}
+!44 = !DILocation(line: 23, column: 3, scope: !41)
+!45 = !DILocation(line: 24, column: 1, scope: !41)