[llvm-profdata] Emit Error when Invalid MemOpSize Section is Created by llvm-profdata

[ormris]

Summary

Under certain (currently unknown) conditions, llvm-profdata is outputting
profiles that have two consecutive entries in the MemOPSize section for the
value 0. This causes the PGOMemOPSizeOpt pass to output an invalid switch
instruction with two cases for 0. As mentioned, we’re not quite sure what’s
causing this to happen, but this patch prevents llvm-profdata from outputting a
profile that has this problem and gives an error with a request for a
reproducible.

[vsk]

The error handling additions look good to me, but I don't have the background to lgtm InstrProfWriter::validateRecord. cc @xur.

[xur]

I'm not sure llvm-profdata is good place to fix. I would rather to fix this in the memop size transformation -- to detect this in the heuristic.
At the same, I'm interested to know what is the cause of this. Should we do something in profile runtime to prevent this.

[ormris]

I would rather to fix this in the memop size transformation -- to detect this in the heuristic.

I think it would be valuable to guard against this pattern within the transform, but I'm concerned that putting the error in the pass would make it harder to debug this issue. If we detect this in llvm-profdata, the user is more likely to have/provide the profraw files causing this behavior.

Should we do something in profile runtime to prevent this.

Do you think llvm-profdata could be causing this issue? I was thinking that profdata might be merging two profiles incorrectly. Either way, having the profraw files would be valuable for debugging...

[ormris]

ping

[ormris]

ping ping

[ormris]

ping ping ping

[ormris]

ping

[xur]

Sorry for the delay.

I replied to the comments on Dec 17 via email. But it did not come through. I don't know the reason. I put my comments here.

I think it would be valuable to guard against this pattern within the transform, but I'm concerned that putting the error in the pass would make it harder to debug this issue. If we detect this in llvm-profdata, the user is more likely to have/provide the profraw files causing this behavior.

I don't think putting this into transformation will make it more difficult to debug. There are ready many legality heuristics there.
To me, the merge operation should be a straightforward translation from raw format to index format. We should not put filters there.
A warning is OK but not a filter of profile records.

Should we do something in profile runtime to prevent this.

Do you think llvm-profdata could be causing this issue? I was thinking that profdata might be merging two profiles incorrectly. Either way, having the profraw files would be valuable for debugging...

There are chances of llvm-prodate bugs but more chances in the compiler-rt runtime.
Providing the raw profile and the index profile will be very helpful for the debuging.

[davidxl]

I agree with Rong here -- the semantic related error checking belongs to the consumer/client, not in the tool.

In profile use pass, the error message can be more meaningful. On the other hand, finding the root cause of it will be more valuable.

[ormris]

OK. That makes sense. I'll go ahead and update the patch.

[ormris]

• Add in-pass profile data check
• Change error to warning in llvm-profdata
• Update tests

[xur]

I think the patch is going the right direction. Some minor comments:
(1) If we have two entries with the same value (other than 0), will we have the same issue? I think we will have same invalid switch statement. So can we extend this patch to handle this? That should be a small change.
(2) can we change "invalid merge" to "invalid profile"? This does not look like a merge issue -- at least from the test case where the problem is from the raw profile. This points to profile runtime issue. I think invalid profile is more appropriate here.

[ormris]

• Update in-pass check to detect all duplicates
• Remove duplicate from MemOp pass test case
• Change warning to "Invalid profile"

[ormris]

ping

[davidxl]

llvm/lib/Transforms/Instrumentation/PGOMemOPSizeOpt.cpp

342

This initial value mean leads to false warning. Use -1.

[ormris]

• Change initial value to -1 to avoid false warnings.

[ormris]

ping

[xur]

The transformation part looks good to me.
Func validateRecord() still checks entries with 0 size only.
But it's OK, as the transformation will catch bad profiles.
I'll approve this patch. Sorry for taking so long.

[ormris]

Thanks for the review @xur and @davidxl

llvm/include/llvm/ProfileData/InstrProf.h

enum class instrprof_error {
   bad_magic,
   bad_header,
   unsupported_version,
   unsupported_hash_type,
   too_large,
   truncated,
   malformed,
   unknown_function,
+  invalid_prof,
   hash_mismatch,
   count_mismatch,
   counter_overflow,
   value_site_count_mismatch,
   compress_failed,
   uncompress_failed,
   empty_raw_profile,
   zlib_unavailable

llvm/include/llvm/ProfileData/InstrProfWriter.h

void addRecord(NamedInstrProfRecord &&I, function_ref<void(Error)> Warn) {
     addRecord(std::move(I), 1, Warn);
   }
 
   /// Merge existing function counts from the given writer.
   void mergeRecordsFromWriter(InstrProfWriter &&IPW,
                               function_ref<void(Error)> Warn);
 
   /// Write the profile to \c OS
-  void write(raw_fd_ostream &OS);
+  Error write(raw_fd_ostream &OS);
 
   /// Write the profile in text format to \c OS
   Error writeText(raw_fd_ostream &OS);
 
+  Error validateRecord(const InstrProfRecord &Func);
+
   /// Write \c Record in text format to \c OS
   static void writeRecordInText(StringRef Name, uint64_t Hash,
                                 const InstrProfRecord &Counters,
                                 InstrProfSymtab &Symtab, raw_fd_ostream &OS);
 
   /// Write the profile, returning the raw data. For testing.
   std::unique_ptr<MemoryBuffer> writeBuffer();
 
public:
   void overlapRecord(NamedInstrProfRecord &&Other, OverlapStats &Overlap,
                      OverlapStats &FuncLevelOverlap,
                      const OverlapFuncFilters &FuncFilter);
 
 private:
   void addRecord(StringRef Name, uint64_t Hash, InstrProfRecord &&I,
                  uint64_t Weight, function_ref<void(Error)> Warn);
   bool shouldEncodeData(const ProfilingData &PD);
-  void writeImpl(ProfOStream &OS);
+
+  Error writeImpl(ProfOStream &OS);
 };
 
 } // end namespace llvm
 
 #endif // LLVM_PROFILEDATA_INSTRPROFWRITER_H

llvm/lib/ProfileData/InstrProf.cpp

 
 #include "llvm/ProfileData/InstrProf.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/Triple.h"
+#include "llvm/Config/config.h"
 #include "llvm/IR/Constant.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalValue.h"
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/MDBuilder.h"
static std::string getInstrProfErrString(instrprof_error Err) {
   case instrprof_error::unsupported_hash_type:
     return "Unsupported instrumentation profile hash type";
   case instrprof_error::too_large:
     return "Too much profile data";
   case instrprof_error::truncated:
     return "Truncated profile data";
   case instrprof_error::malformed:
     return "Malformed instrumentation profile data";
+  case instrprof_error::invalid_prof:
+    return "Invalid profile created. Please file a bug "
+           "at: " BUG_REPORT_URL
+           " and include the profraw files that caused this error.";
   case instrprof_error::unknown_function:
     return "No profile data available for function";
   case instrprof_error::hash_mismatch:
     return "Function control flow change detected (hash mismatch)";
   case instrprof_error::count_mismatch:
     return "Function basic block count change detected (counter mismatch)";
   case instrprof_error::counter_overflow:
     return "Counter overflow";

llvm/lib/ProfileData/InstrProfWriter.cpp

static void setSummary(IndexedInstrProf::Summary *TheSummary,
   TheSummary->set(Summary::MaxInternalBlockCount, PS.getMaxInternalCount());
   TheSummary->set(Summary::TotalBlockCount, PS.getTotalCount());
   TheSummary->set(Summary::TotalNumBlocks, PS.getNumCounts());
   TheSummary->set(Summary::TotalNumFunctions, PS.getNumFunctions());
   for (unsigned I = 0; I < Res.size(); I++)
     TheSummary->setEntry(I, Res[I]);
 }
 
-void InstrProfWriter::writeImpl(ProfOStream &OS) {
+Error InstrProfWriter::writeImpl(ProfOStream &OS) {
   using namespace IndexedInstrProf;
 
   OnDiskChainedHashTableGenerator<InstrProfRecordWriterTrait> Generator;
 
   InstrProfSummaryBuilder ISB(ProfileSummaryBuilder::DefaultCutoffs);
   InfoObj->SummaryBuilder = &ISB;
   InstrProfSummaryBuilder CSISB(ProfileSummaryBuilder::DefaultCutoffs);
   InfoObj->CSSummaryBuilder = &CSISB;
PatchItem PatchItems[] = {
       {HashTableStartFieldOffset, &HashTableStart, 1},
       // Patch the summary data.
       {SummaryOffset, reinterpret_cast<uint64_t *>(TheSummary.get()),
        (int)(SummarySize / sizeof(uint64_t))},
       {CSSummaryOffset, reinterpret_cast<uint64_t *>(TheCSSummary.get()),
        (int)CSSummarySize}};
 
   OS.patch(PatchItems, sizeof(PatchItems) / sizeof(*PatchItems));
+
+  for (const auto &I : FunctionData)
+    for (const auto &F : I.getValue())
+      if (Error E = validateRecord(F.second))
+        return E;
+
+  return Error::success();
 }
 
-void InstrProfWriter::write(raw_fd_ostream &OS) {
+Error InstrProfWriter::write(raw_fd_ostream &OS) {
   // Write the hash table.
   ProfOStream POS(OS);
-  writeImpl(POS);
+  return writeImpl(POS);
 }
 
 std::unique_ptr<MemoryBuffer> InstrProfWriter::writeBuffer() {
   std::string Data;
   raw_string_ostream OS(Data);
   ProfOStream POS(OS);
   // Write the hash table.
-  writeImpl(POS);
+  if (Error E = writeImpl(POS))
+    return nullptr;
   // Return this in an aligned memory buffer.
   return MemoryBuffer::getMemBufferCopy(Data);
 }
 
 static const char *ValueProfKindStr[] = {
 #define VALUE_PROF_KIND(Enumerator, Value, Descr) #Enumerator,
 #include "llvm/ProfileData/InstrProfData.inc"
 };
 
+Error InstrProfWriter::validateRecord(const InstrProfRecord &Func) {
+  for (uint32_t VK = 0; VK <= IPVK_Last; VK++) {
+    uint32_t NS = Func.getNumValueSites(VK);
+    if (!NS)
+      continue;
+    for (uint32_t S = 0; S < NS; S++) {
+      uint32_t ND = Func.getNumValueDataForSite(VK, S);
+      std::unique_ptr<InstrProfValueData[]> VD = Func.getValueForSite(VK, S);
+      bool WasZero = false;
+      for (uint32_t I = 0; I < ND; I++)
+        if ((VK != IPVK_IndirectCallTarget) && (VD[I].Value == 0)) {
+          if (WasZero)
+            return make_error<InstrProfError>(instrprof_error::invalid_prof);
+          WasZero = true;
+        }
+    }
+  }
+
+  return Error::success();
+}
+
 void InstrProfWriter::writeRecordInText(StringRef Name, uint64_t Hash,
                                         const InstrProfRecord &Func,
                                         InstrProfSymtab &Symtab,
                                         raw_fd_ostream &OS) {
   OS << Name << "\n";
   OS << "# Func Hash:\n" << Hash << "\n";
   OS << "# Num Counters:\n" << Func.Counts.size() << "\n";
   OS << "# Counter Values:\n";
Error InstrProfWriter::writeText(raw_fd_ostream &OS) {
   });
 
   for (const auto &record : OrderedFuncData) {
     const StringRef &Name = record.first;
     const FuncPair &Func = record.second;
     writeRecordInText(Name, Func.first, Func.second, Symtab, OS);
   }
 
+  for (const auto &record : OrderedFuncData) {
+    const FuncPair &Func = record.second;
+    if (Error E = validateRecord(Func.second))
+      return E;
+  }
+
   return Error::success();
 }

llvm/lib/Transforms/Instrumentation/PGOMemOPSizeOpt.cpp

 #include "llvm/ProfileData/InstrProf.h"
 #define INSTR_PROF_VALUE_PROF_MEMOP_API
 #include "llvm/ProfileData/InstrProfData.inc"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
+#include "llvm/Support/WithColor.h"
 #include "llvm/Transforms/Instrumentation.h"
 #include "llvm/Transforms/Instrumentation/PGOInstrumentation.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include <cassert>
 #include <cstdint>
 #include <vector>
 
 using namespace llvm;
bool MemOPSizeOpt::perform(MemOp MO) {
 
   // Keeping track of the count of the default case:
   uint64_t RemainCount = TotalCount;
   uint64_t SavedRemainCount = SavedTotalCount;
   SmallVector<uint64_t, 16> SizeIds;
   SmallVector<uint64_t, 16> CaseCounts;
   uint64_t MaxCount = 0;
   unsigned Version = 0;
+  int64_t LastV = -1;
   // Default case is in the front -- save the slot here.
   CaseCounts.push_back(0);
   for (auto &VD : VDs) {
     int64_t V = VD.Value;
     uint64_t C = VD.Count;
     if (MemOPScaleCount)
       C = getScaledCount(C, ActualCount, SavedTotalCount);
 
     if (!InstrProfIsSingleValRange(V) || V > MemOpMaxOptSize)
       continue;
 
     // ValueCounts are sorted on the count. Break at the first un-profitable
     // value.
     if (!isProfitable(C, RemainCount))
       break;
 
+    if (V == LastV) {
+      LLVM_DEBUG(dbgs() << "Invalid Profile Data in Function " << Func.getName()
+                        << ": Two consecutive, identical values in MemOp value"
+                           "counts.\n");
+      return false;
+    }
+
+    LastV = V;
+
     SizeIds.push_back(V);
     CaseCounts.push_back(C);
     if (C > MaxCount)
       MaxCount = C;
 
     assert(RemainCount >= C);
     RemainCount -= C;
     assert(SavedRemainCount >= VD.Count);

llvm/test/Transforms/PGOProfile/Inputs/consecutive-zeros.proftext

+# IR level Instrumentation Flag
+:ir
+foo
+# Func Hash:
+687116424982578944
+# Num Counters:
+3
+# Counter Values:
+523
+20
+1
+# Num Value Kinds:
+1
+# ValueKind = IPVK_MemOPSize:
+1
+# NumValueSites:
+3
+9
+0:99
+0:88
+3:77
+9:72
+4:66
+5:55
+6:44
+7:33
+8:22
+9
+7:33
+2:88
+9:72
+4:66
+1:99
+5:55
+6:44
+3:77
+8:22
+9
+7:33
+2:88
+9:72
+4:66
+1:99
+5:55
+6:44
+3:77
+8:22

llvm/test/Transforms/PGOProfile/consecutive-zeros.ll

+; RUN: llvm-profdata merge %S/Inputs/consecutive-zeros.proftext -o %t.profdata
+; RUN: opt < %s -debug -pgo-instr-use -pgo-memop-opt -pgo-memop-count-threshold=0 -pgo-memop-percent-threshold=0 -pgo-test-profile-file=%t.profdata -S 2>&1 | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @foo(i8* %dst, i8* %src, i32* %a, i32 %n) {
+; CHECK: Invalid Profile
+entry:
+  br label %for.cond
+
+for.cond:
+  %i.0 = phi i32 [ 0, %entry ], [ %inc5, %for.inc4 ]
+  %cmp = icmp slt i32 %i.0, %n
+  br i1 %cmp, label %for.body, label %for.end6
+
+for.body:
+  br label %for.cond1
+
+for.cond1:
+  %j.0 = phi i32 [ 0, %for.body ], [ %inc, %for.inc ]
+  %idx.ext = sext i32 %i.0 to i64
+  %add.ptr = getelementptr inbounds i32, i32* %a, i64 %idx.ext
+  %0 = load i32, i32* %add.ptr, align 4
+  %cmp2 = icmp slt i32 %j.0, %0
+  br i1 %cmp2, label %for.body3, label %for.end
+
+for.body3:
+  %add = add nsw i32 %i.0, 1
+  %conv = sext i32 %add to i64
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %conv, i1 false)
+  %memcmp = call i32 @memcmp(i8* %dst, i8* %src, i64 %conv)
+  %bcmp = call i32 @bcmp(i8* %dst, i8* %src, i64 %conv)
+  br label %for.inc
+
+for.inc:
+  %inc = add nsw i32 %j.0, 1
+  br label %for.cond1
+
+for.end:
+  br label %for.inc4
+
+for.inc4:
+  %inc5 = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end6:
+  ret void
+}
+
+declare void @llvm.lifetime.start(i64, i8* nocapture)
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i1)
+
+declare i32 @memcmp(i8*, i8*, i64)
+declare i32 @bcmp(i8*, i8*, i64)
+
+declare void @llvm.lifetime.end(i64, i8* nocapture)

llvm/test/Transforms/PGOProfile/memop_size_opt.ll

 for.body3:
   %add = add nsw i32 %i.0, 1
   %conv = sext i32 %add to i64
   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %conv, i1 false), !prof !30
   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src2, i64 %conv, i1 false), !prof !31
   br label %for.inc
 
 ; MEMOP_OPT:  switch i64 %conv, label %[[DEFAULT_LABEL:.*]] [
-; MEMOP_OPT:    i64 1, label %[[CASE_1_LABEL:.*]]
+; MEMOP_OPT:    i64 0, label %[[CASE_1_LABEL:.*]]
 ; MEMOP_OPT:  ], !prof [[SWITCH_BW:![0-9]+]] 
 ; MEMOP_OPT: [[CASE_1_LABEL]]:
-; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 1, i1 false)
+; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 0, i1 false)
 ; MEMOP_OPT:   br label %[[MERGE_LABEL:.*]]
 ; MEMOP_OPT: [[DEFAULT_LABEL]]:
 ; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %conv, i1 false), !prof [[NEWVP:![0-9]+]]
 ; MEMOP_OPT:   br label %[[MERGE_LABEL]]
 ; MEMOP_OPT: [[MERGE_LABEL]]:
 ; MEMOP_OPT:  switch i64 %conv, label %[[DEFAULT_LABEL2:.*]] [
-; MEMOP_OPT:    i64 1, label %[[CASE_1_LABEL2:.*]]
+; MEMOP_OPT:    i64 0, label %[[CASE_1_LABEL2:.*]]
 ; MEMOP_OPT:  ], !prof [[SWITCH_BW:![0-9]+]] 
 ; MEMOP_OPT: [[CASE_1_LABEL2]]:
-; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src2, i64 1, i1 false)
+; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src2, i64 0, i1 false)
 ; MEMOP_OPT:   br label %[[MERGE_LABEL2:.*]]
 ; MEMOP_OPT: [[DEFAULT_LABEL2]]:
 ; MEMOP_OPT:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src2, i64 %conv, i1 false), !prof [[NEWVP]]
 ; MEMOP_OPT:   br label %[[MERGE_LABEL2]]
 ; MEMOP_OPT: [[MERGE_LABEL2]]:
 ; MEMOP_OPT:   br label %for.inc
 
 for.inc:
for.body3:
   %add = add nsw i32 %i.0, 1
   %conv = sext i32 %add to i64
   %memcmp = call i32 @memcmp(i8* %dst, i8* %src, i64 %conv), !prof !30
   %bcmp = call i32 @bcmp(i8* %dst2, i8* %src2, i64 %conv), !prof !31
   call void @consume(i32 %memcmp, i32 %bcmp)
   br label %for.inc
 
 ; MEMOP_OPT:  switch i64 %conv, label %[[DEFAULT_LABEL:.*]] [
-; MEMOP_OPT:    i64 1, label %[[CASE_1_LABEL:.*]]
+; MEMOP_OPT:    i64 0, label %[[CASE_1_LABEL:.*]]
 ; MEMOP_OPT:  ], !prof [[SWITCH_BW:![0-9]+]]
 ; MEMOP_OPT: [[CASE_1_LABEL]]:
-; MEMOP_OPT:   %[[RV:.*]] = call i32 @memcmp(i8* %dst, i8* %src, i64 1)
+; MEMOP_OPT:   %[[RV:.*]] = call i32 @memcmp(i8* %dst, i8* %src, i64 0)
 ; MEMOP_OPT:   br label %[[MERGE_LABEL:.*]]
 ; MEMOP_OPT: [[DEFAULT_LABEL]]:
 ; MEMOP_OPT:   %[[RVD:.*]] = call i32 @memcmp(i8* %dst, i8* %src, i64 %conv), !prof [[NEWVP:![0-9]+]]
 ; MEMOP_OPT:   br label %[[MERGE_LABEL]]
 ; MEMOP_OPT: [[MERGE_LABEL]]:
 ; MEMOP_OPT:  %[[PHI:.*]] = phi i32 [ %[[RVD]], %[[DEFAULT_LABEL]] ], [ %[[RV]], %[[CASE_1_LABEL]] ]
 ; MEMOP_OPT:  switch i64 %conv, label %[[DEFAULT_LABEL2:.*]] [
-; MEMOP_OPT:    i64 1, label %[[CASE_1_LABEL2:.*]]
+; MEMOP_OPT:    i64 0, label %[[CASE_1_LABEL2:.*]]
 ; MEMOP_OPT:  ], !prof [[SWITCH_BW:![0-9]+]]
 ; MEMOP_OPT: [[CASE_1_LABEL2]]:
-; MEMOP_OPT:   %[[RV2:.*]] = call i32 @bcmp(i8* %dst2, i8* %src2, i64 1)
+; MEMOP_OPT:   %[[RV2:.*]] = call i32 @bcmp(i8* %dst2, i8* %src2, i64 0)
 ; MEMOP_OPT:   br label %[[MERGE_LABEL2:.*]]
 ; MEMOP_OPT: [[DEFAULT_LABEL2]]:
 ; MEMOP_OPT:   %[[RVD2:.*]] = call i32 @bcmp(i8* %dst2, i8* %src2, i64 %conv), !prof [[NEWVP]]
 ; MEMOP_OPT:   br label %[[MERGE_LABEL2]]
 ; MEMOP_OPT: [[MERGE_LABEL2]]:
 ; MEMOP_OPT:   %[[PHI2:.*]] = phi i32 [ %[[RVD2]], %[[DEFAULT_LABEL2]] ], [ %[[RV2]], %[[CASE_1_LABEL2]] ]
 ; MEMOP_OPT:   call void @consume(i32 %[[PHI]], i32 %[[PHI2]])
 ; MEMOP_OPT:   br label %for.inc

 !22 = !{i32 990000, i64 20, i32 2}
 !23 = !{i32 999000, i64 1, i32 5}
 !24 = !{i32 999900, i64 1, i32 5}
 !25 = !{i32 999990, i64 1, i32 5}
 !26 = !{i32 999999, i64 1, i32 5}
 !27 = !{!"function_entry_count", i64 1}
 !28 = !{!"branch_weights", i32 20, i32 1}
 !29 = !{!"branch_weights", i32 556, i32 20}
-!30 = !{!"VP", i32 1, i64 556, i64 1, i64 99, i64 2, i64 88, i64 3, i64 77, i64 9, i64 72, i64 4, i64 66, i64 5, i64 55, i64 6, i64 44, i64 7, i64 33, i64 8, i64 22}
-!31 = !{!"VP", i32 1, i64 556, i64 1, i64 99, i64 2, i64 88, i64 3, i64 77, i64 9, i64 72, i64 4, i64 66, i64 5, i64 55, i64 6, i64 44, i64 7, i64 33, i64 8, i64 22}
+!30 = !{!"VP", i32 1, i64 556, i64 0, i64 99, i64 2, i64 88, i64 3, i64 77, i64 9, i64 72, i64 4, i64 66, i64 5, i64 55, i64 6, i64 44, i64 7, i64 33, i64 8, i64 22}
+!31 = !{!"VP", i32 1, i64 556, i64 0, i64 99, i64 2, i64 88, i64 3, i64 77, i64 9, i64 72, i64 4, i64 66, i64 5, i64 55, i64 6, i64 44, i64 7, i64 33, i64 8, i64 22}
 
 declare void @llvm.lifetime.start(i64, i8* nocapture)
 
 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i1)
 
 declare i32 @memcmp(i8*, i8*, i64)
 declare i32 @bcmp(i8*, i8*, i64)
 

llvm/test/tools/llvm-profdata/invalid-profile-gen-zeros.proftext

+# RUN: llvm-profdata merge --text -j 4 %s %s %s %s -o %t 2>&1 | FileCheck %s
+# RUN: llvm-profdata merge --binary -j 4 %s %s %s %s -o %t 2>&1 | FileCheck %s
+# IR level Instrumentation Flag
+# CHECK: Invalid profile
+:ir
+foo
+# Func Hash:
[phone]
+# Num Counters:
+3
+# Counter Values:
+20
+556
+1
+# Num Value Kinds:
+1
+# ValueKind = IPVK_MemOPSize:
+1
+# NumValueSites:
+1
+9
+0:99
+0:88
+3:77
+9:72
+4:66
+5:55
+6:44
+7:33
+8:22

llvm/tools/llvm-profdata/llvm-profdata.cpp

static void warn(Twine Message, std::string Whence = "",
   WithColor::warning();
   if (!Whence.empty())
     errs() << Whence << ": ";
   errs() << Message << "\n";
   if (!Hint.empty())
     WithColor::note() << Hint << "\n";
 }
 
+static void warn(Error E, StringRef Whence = "") {
+  if (E.isA<InstrProfError>()) {
+    handleAllErrors(std::move(E), [&](const InstrProfError &IPE) {
+      warn(IPE.message(), std::string(Whence), std::string(""));
+    });
+  }
+}
+
 static void exitWithError(Twine Message, std::string Whence = "",
                           std::string Hint = "") {
   WithColor::error();
   if (!Whence.empty())
     errs() << Whence << ": ";
   errs() << Message << "\n";
   if (!Hint.empty())
     WithColor::note() << Hint << "\n";
static void writeInstrProfile(StringRef OutputFilename,
   raw_fd_ostream Output(OutputFilename.data(), EC,
                         OutputFormat == PF_Text ? sys::fs::OF_Text
                                                 : sys::fs::OF_None);
   if (EC)
     exitWithErrorCode(EC, OutputFilename);
 
   if (OutputFormat == PF_Text) {
     if (Error E = Writer.writeText(Output))
-      exitWithError(std::move(E));
+      warn(std::move(E));
   } else {
-    Writer.write(Output);
+    if (Error E = Writer.write(Output))
+      warn(std::move(E));
   }
 }
 
 static void mergeInstrProfile(const WeightedFileVector &Inputs,
                               SymbolRemapper *Remapper,
                               StringRef OutputFilename,
                               ProfileFormat OutputFormat, bool OutputSparse,
                               unsigned NumThreads, FailureMode FailMode) {