[BOLT] Control aggregation mode output profile file format

[Amir]

Summary

In perf2bolt and -aggregate-only BOLT mode, the output profile file is written
in fdata format by default. Provide a knob -profile-format=[fdata,yaml] to
control the format.
Note that -w option still dumps in YAML format.

[Amir]

typo

[Amir]

typos

[Amir]

Fix build issue, extend X86/pre-aggregated-perf.test

[Amir]

Use sort for fdata output

[maksfb]

Can you s/ProfileOutput/ProfileFormat/g and s/profile-output/profile-format/g?

bolt/test/X86/pre-aggregated-perf.test

17

nit: use double dash for long options.

[Amir]

Address comments

[Amir]

In D133995#3796613, @maksfb wrote:

Can you s/ProfileOutput/ProfileFormat/g and s/profile-output/profile-format/g?

Done

[maksfb]

bolt/test/X86/pre-aggregated-perf.test

22

It's super confusing, but -pa is a long option.

[Amir]

Use --pa

[Amir]

bolt/test/X86/pre-aggregated-perf.test

22

Sorry, I didn't notice that while copying. Fixed.

[maksfb]

LGTM

bolt/include/bolt/Utils/CommandLineOpts.h

 extern llvm::cl::opt<bool> HotData;
 extern llvm::cl::opt<bool> HotFunctionsAtEnd;
 extern llvm::cl::opt<bool> HotText;
 extern llvm::cl::opt<bool> Instrument;
 extern llvm::cl::opt<std::string> OutputFilename;
 extern llvm::cl::opt<std::string> PerfData;
 extern llvm::cl::opt<bool> PrintCacheMetrics;
 extern llvm::cl::opt<bool> PrintSections;
+
+// The format to use with -o in aggregation mode (perf2bolt)
+enum ProfileFormatKind { PF_Fdata, PF_YAML };
+
+extern llvm::cl::opt<ProfileFormatKind> ProfileFormat;
 extern llvm::cl::opt<bool> SplitEH;
 extern llvm::cl::opt<bool> StrictMode;
 extern llvm::cl::opt<bool> TimeOpts;
 extern llvm::cl::opt<bool> UseOldText;
 extern llvm::cl::opt<bool> UpdateDebugSections;
 
 // The default verbosity level (0) is pretty terse, level 1 is fairly
 // verbose and usually prints some informational message for every

bolt/lib/Profile/DataAggregator.cpp

 static cl::opt<unsigned long long>
 MaxSamples("max-samples",
   cl::init(-1ULL),
   cl::desc("maximum number of samples to read from LBR profile"),
   cl::Optional,
   cl::Hidden,
   cl::cat(AggregatorCategory));
 
+extern cl::opt<opts::ProfileFormatKind> ProfileFormat;
+
 cl::opt<bool> ReadPreAggregated(
     "pa", cl::desc("skip perf and read data from a pre-aggregated file format"),
     cl::cat(AggregatorCategory));
 
 static cl::opt<bool>
 TimeAggregator("time-aggr",
   cl::desc("time BOLT aggregator"),
   cl::init(false),

 Error DataAggregator::readProfile(BinaryContext &BC) {
   processProfile(BC);
 
   for (auto &BFI : BC.getBinaryFunctions()) {
     BinaryFunction &Function = BFI.second;
     convertBranchData(Function);
   }
 
-  if (opts::AggregateOnly) {
+  if (opts::AggregateOnly &&
+      opts::ProfileFormat == opts::ProfileFormatKind::PF_Fdata) {
     if (std::error_code EC = writeAggregatedFile(opts::OutputFilename))
       report_error("cannot create output data file", EC);
   }
 
   return Error::success();
 }
 
 bool DataAggregator::mayHaveProfileData(const BinaryFunction &Function) {

bolt/lib/Rewrite/RewriteInstance.cpp

void RewriteInstance::processProfileData() {
 
   if (Error E = ProfileReader->readProfile(*BC.get()))
     report_error("cannot read profile", std::move(E));
 
   if (!opts::SaveProfile.empty()) {
     YAMLProfileWriter PW(opts::SaveProfile);
     PW.writeProfile(*this);
   }
+  if (opts::AggregateOnly &&
+      opts::ProfileFormat == opts::ProfileFormatKind::PF_YAML) {
+    YAMLProfileWriter PW(opts::OutputFilename);
+    PW.writeProfile(*this);
+  }
 
   // Release memory used by profile reader.
   ProfileReader.reset();
 
   if (opts::AggregateOnly)
     exit(0);
 }
 

bolt/lib/Utils/CommandLineOpts.cpp

cl::opt<bool> PrintCacheMetrics(
     "print-cache-metrics",
     cl::desc("calculate and print various metrics for instruction cache"),
     cl::cat(BoltOptCategory));
 
 cl::opt<bool> PrintSections("print-sections",
                             cl::desc("print all registered sections"),
                             cl::Hidden, cl::cat(BoltCategory));
 
+cl::opt<ProfileFormatKind> ProfileFormat(
+    "profile-format",
+    cl::desc(
+        "format to dump profile output in aggregation mode, default is fdata"),
+    cl::init(PF_Fdata),
+    cl::values(clEnumValN(PF_Fdata, "fdata", "offset-based plaintext format"),
+               clEnumValN(PF_YAML, "yaml", "dense YAML reprensentation")),
+    cl::ZeroOrMore, cl::Hidden, cl::cat(BoltCategory));
+
 cl::opt<bool> SplitEH("split-eh", cl::desc("split C++ exception handling code"),
                       cl::Hidden, cl::cat(BoltOptCategory));
 
 cl::opt<bool>
     StrictMode("strict",
                cl::desc("trust the input to be from a well-formed source"),
 
                cl::cat(BoltCategory));

bolt/test/X86/pre-aggregated-perf.test

 # This script checks that perf2bolt is reading pre-aggregated perf information
 # correctly for a simple example. The perf.data of this example was generated
 # with the following command:
 #
 #  $ perf record -j any,u -e branch -o perf.data -- ./blarge
 #
 #  blarge is the binary for "basicmath large inputs" taken from Mibench.
 
 # Currently failing in MacOS / generating different hash for usqrt
 REQUIRES: system-linux
 
 RUN: yaml2obj %p/Inputs/blarge.yaml &> %t.exe
-RUN: perf2bolt %t.exe -o %t -pa -p %p/Inputs/pre-aggregated.txt -w %t.new
+RUN: perf2bolt %t.exe -o %t --pa -p %p/Inputs/pre-aggregated.txt -w %t.new
 RUN: cat %t | sort | FileCheck %s -check-prefix=PERF2BOLT
 RUN: cat %t.new | FileCheck %s -check-prefix=NEWFORMAT
 
+# Test --profile-format option with perf2bolt
+RUN: perf2bolt %t.exe -o %t.fdata --pa -p %p/Inputs/pre-aggregated.txt \
+RUN:   --profile-format=fdata
+RUN: cat %t.fdata | sort | FileCheck %s -check-prefix=PERF2BOLT
+
+RUN: perf2bolt %t.exe -o %t.yaml --pa -p %p/Inputs/pre-aggregated.txt \
+RUN:   --profile-format=yaml
+RUN: cat %t.yaml | FileCheck %s -check-prefix=NEWFORMAT
+
+# Test --profile-format option with llvm-bolt --aggregate-only
+RUN: llvm-bolt %t.exe -o %t.bolt.fdata --pa -p %p/Inputs/pre-aggregated.txt \
+RUN:   --aggregate-only --profile-format=fdata
+RUN: cat %t.bolt.fdata | sort | FileCheck %s -check-prefix=PERF2BOLT
+
+RUN: llvm-bolt %t.exe -o %t.bolt.yaml --pa -p %p/Inputs/pre-aggregated.txt \
+RUN:   --aggregate-only --profile-format=yaml
+RUN: cat %t.bolt.yaml | FileCheck %s -check-prefix=NEWFORMAT
+
 PERF2BOLT: 0 [unknown] 7f36d18d60c0 1 main 53c 0 2
 PERF2BOLT: 1 main 451 1 SolveCubic 0 0 2
 PERF2BOLT: 1 main 490 0 [unknown] 4005f0 0 1
 PERF2BOLT: 1 main 537 0 [unknown] 400610 0 1
 PERF2BOLT: 1 usqrt 30 1 usqrt 32 0 22
 PERF2BOLT: 1 usqrt 30 1 usqrt 39 4 33
 PERF2BOLT: 1 usqrt 35 1 usqrt 39 0 22
 PERF2BOLT: 1 usqrt 3d 1 usqrt 10 0 58