[BOLT] Report adjusted program stats from perf2bolt in BAT mode (#91683)

Author: aaupov

bolt/include/bolt/Passes/BinaryPasses.h

@@ -16,6 +16,7 @@
 #include "bolt/Core/BinaryContext.h"
 #include "bolt/Core/BinaryFunction.h"
 #include "bolt/Core/DynoStats.h"
+#include "bolt/Profile/BoltAddressTranslation.h"
 #include "llvm/Support/CommandLine.h"
 #include <atomic>
 #include <set>
@@ -399,8 +400,11 @@ class PrintProfileStats : public BinaryFunctionPass {
 /// Prints a list of the top 100 functions sorted by a set of
 /// dyno stats categories.
 class PrintProgramStats : public BinaryFunctionPass {
+  BoltAddressTranslation *BAT = nullptr;
+
 public:
-  explicit PrintProgramStats() : BinaryFunctionPass(false) {}
+  explicit PrintProgramStats(BoltAddressTranslation *BAT = nullptr)
+      : BinaryFunctionPass(false), BAT(BAT) {}
 
   const char *getName() const override { return "print-stats"; }
   bool shouldPrint(const BinaryFunction &) const override { return false; }

bolt/lib/Passes/BinaryPasses.cpp

@@ -1390,9 +1390,19 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
     if (Function.isPLTFunction())
       continue;
 
+    // Adjustment for BAT mode: the profile for BOLT split fragments is combined
+    // so only count the hot fragment.
+    const uint64_t Address = Function.getAddress();
+    bool IsHotParentOfBOLTSplitFunction = !Function.getFragments().empty() &&
+                                          BAT && BAT->isBATFunction(Address) &&
+                                          !BAT->fetchParentAddress(Address);
+
     ++NumRegularFunctions;
 
-    if (!Function.isSimple()) {
+    // In BOLTed binaries split functions are non-simple (due to non-relocation
+    // mode), but the original function is known to be simple and we have a
+    // valid profile for it.
+    if (!Function.isSimple() && !IsHotParentOfBOLTSplitFunction) {
       if (Function.hasProfile())
         ++NumNonSimpleProfiledFunctions;
       continue;

bolt/lib/Profile/DataAggregator.cpp

@@ -613,7 +613,8 @@ Error DataAggregator::readProfile(BinaryContext &BC) {
         if (std::error_code EC = writeBATYAML(BC, opts::SaveProfile))
           report_error("cannot create output data file", EC);
     }
-    BC.logBOLTErrorsAndQuitOnFatal(PrintProgramStats().runOnFunctions(BC));
+    PrintProgramStats PPS(BAT);
+    BC.logBOLTErrorsAndQuitOnFatal(PPS.runOnFunctions(BC));
   }
 
   return Error::success();

bolt/test/X86/bolt-address-translation-yaml.test

@@ -31,7 +31,8 @@ RUN: perf2bolt %t.out --pa -p %p/Inputs/blarge_new_bat.preagg.txt -w %t.yaml -o
 RUN:   2>&1 | FileCheck --check-prefix READ-BAT-CHECK %s
 RUN: FileCheck --input-file %t.yaml --check-prefix YAML-BAT-CHECK %s
 # Check that YAML converted from fdata matches YAML created directly with BAT.
-RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o /dev/null
+RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o /dev/null \
+RUN:   2>&1 | FileCheck --check-prefix READ-BAT-FDATA-CHECK %s
 RUN: FileCheck --input-file %t.yaml-fdata --check-prefix YAML-BAT-CHECK %s
 
 # Test resulting YAML profile with the original binary (no-stale mode)
@@ -45,6 +46,8 @@ WRITE-BAT-CHECK: BOLT-INFO: BAT section size (bytes): 384
 READ-BAT-CHECK-NOT: BOLT-ERROR: unable to save profile in YAML format for input file processed by BOLT
 READ-BAT-CHECK: BOLT-INFO: Parsed 5 BAT entries
 READ-BAT-CHECK: PERF2BOLT: read 79 aggregated LBR entries
+READ-BAT-CHECK: BOLT-INFO: 5 out of 21 functions in the binary (23.8%) have non-empty execution profile
+READ-BAT-FDATA-CHECK: BOLT-INFO: 5 out of 16 functions in the binary (31.2%) have non-empty execution profile
 
 YAML-BAT-CHECK:      functions:
 # Function not covered by BAT - has insns in basic block