Fix the bug when SampleProfileWriter writes out number of callsites.

[danielcdh]

Summary

As we support multiple callsites for the same location, we need to traverse all locations to get the number of callsites.

[davidxl]

test/tools/llvm-profdata/Inputs/sample-profile.proftext

13

What does this actually test?

[danielcdh]

test/tools/llvm-profdata/Inputs/sample-profile.proftext

13

It tests if the profile reader/writer can handle 2 functions inlined at the same location. Without the patch, the test will fail.

[davidxl]

lgtm

lib/ProfileData/SampleProfWriter.cpp

for (const auto &J : Sample.getCallTargets()) {
       uint64_t CalleeSamples = J.second;
       if (std::error_code EC = writeNameIdx(Callee))
         return EC;
       encodeULEB128(CalleeSamples, OS);
     }
   }
 
   // Recursively emit all the callsite samples.
-  encodeULEB128(S.getCallsiteSamples().size(), OS);
+  uint64_t NumCallsites = 0;
+  for (const auto &J : S.getCallsiteSamples())
+    NumCallsites += J.second.size();
+  encodeULEB128(NumCallsites, OS);
   for (const auto &J : S.getCallsiteSamples())
     for (const auto &FS : J.second) {
       LineLocation Loc = J.first;
       const FunctionSamples &CalleeSamples = FS.second;
       encodeULEB128(Loc.LineOffset, OS);
       encodeULEB128(Loc.Discriminator, OS);
       if (std::error_code EC = writeBody(CalleeSamples))
         return EC;

test/tools/llvm-profdata/Inputs/sample-profile.proftext

 _Z3bari:20301:1437
  1: 1437
 _Z3fooi:7711:610
  1: 610
 main:184019:0
  4: 534
  4.2: 534
  5: 1075
  5.1: 1075
  6: 2080
  7: 534
  9: 2064 _Z3bari:1471 _Z3fooi:631
+ 10: inline1:1000
+  1: 1000
+ 10: inline2:2000
+  1: 2000