Merge remote-tracking branch 'origin/main' into vplan-predication

Author: fhahn

.github/new-prs-labeler.yml

@@ -702,6 +702,7 @@ mlgo:
   - llvm/unittests/CodeGen/ML*
   - llvm/test/CodeGen/MLRegAlloc/**
   - llvm/utils/mlgo-utils/**
+  - llvm/docs/MLGO.rst
 
 tools:llvm-exegesis:
   - llvm/tools/llvm-exegesis/**

bolt/include/bolt/Profile/Heatmap.h

@@ -52,6 +52,9 @@ class Heatmap {
       : BucketSize(BucketSize), MinAddress(MinAddress), MaxAddress(MaxAddress),
         TextSections(TextSections) {}
 
+  uint64_t HotStart{0};
+  uint64_t HotEnd{0};
+
   inline bool ignoreAddress(uint64_t Address) const {
     return (Address > MaxAddress) || (Address < MinAddress);
   }

bolt/lib/Core/DIEBuilder.cpp

@@ -437,10 +437,10 @@ getUnitForOffset(DIEBuilder &Builder, DWARFContext &DWCtx,
       // This is a work around for XCode clang. There is a build error when we
       // pass DWCtx.compile_units() to llvm::upper_bound
       std::call_once(InitVectorFlag, initCUVector);
-      auto CUIter = std::upper_bound(CUOffsets.begin(), CUOffsets.end(), Offset,
-                                     [](uint64_t LHS, const DWARFUnit *RHS) {
-                                       return LHS < RHS->getNextUnitOffset();
-                                     });
+      auto CUIter = llvm::upper_bound(CUOffsets, Offset,
+                                      [](uint64_t LHS, const DWARFUnit *RHS) {
+                                        return LHS < RHS->getNextUnitOffset();
+                                      });
       CU = CUIter != CUOffsets.end() ? (*CUIter) : nullptr;
     }
     return CU;

bolt/lib/Profile/DataAggregator.cpp

@@ -1316,6 +1316,14 @@ std::error_code DataAggregator::printLBRHeatMap() {
   }
   Heatmap HM(opts::HeatmapBlock, opts::HeatmapMinAddress,
              opts::HeatmapMaxAddress, getTextSections(BC));
+  auto getSymbolValue = [&](const MCSymbol *Symbol) -> uint64_t {
+    if (Symbol)
+      if (ErrorOr<uint64_t> SymValue = BC->getSymbolValue(*Symbol))
+        return SymValue.get();
+    return 0;
+  };
+  HM.HotStart = getSymbolValue(BC->getHotTextStartSymbol());
+  HM.HotEnd = getSymbolValue(BC->getHotTextEndSymbol());
 
   if (!NumTotalSamples) {
     if (opts::BasicAggregation) {

bolt/lib/Profile/Heatmap.cpp

@@ -8,6 +8,7 @@
 
 #include "bolt/Profile/Heatmap.h"
 #include "bolt/Utils/CommandLineOpts.h"
+#include "llvm/ADT/AddressRanges.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/Support/Debug.h"
@@ -313,6 +314,9 @@ void Heatmap::printSectionHotness(raw_ostream &OS) const {
     UnmappedHotness += Frequency;
   };
 
+  AddressRange HotTextRange(HotStart, HotEnd);
+  StringRef HotTextName = "[hot text]";
+
   for (const std::pair<const uint64_t, uint64_t> &KV : Map) {
     NumTotalCounts += KV.second;
     // We map an address bucket to the first section (lowest address)
@@ -328,15 +332,24 @@ void Heatmap::printSectionHotness(raw_ostream &OS) const {
     }
     SectionHotness[TextSections[TextSectionIndex].Name] += KV.second;
     ++BucketUtilization[TextSections[TextSectionIndex].Name];
+    if (HotTextRange.contains(Address)) {
+      SectionHotness[HotTextName] += KV.second;
+      ++BucketUtilization[HotTextName];
+    }
   }
 
+  std::vector<SectionNameAndRange> Sections(TextSections);
+  // Append synthetic hot text section to TextSections
+  if (!HotTextRange.empty())
+    Sections.emplace_back(SectionNameAndRange{HotTextName, HotStart, HotEnd});
+
   assert(NumTotalCounts > 0 &&
          "total number of heatmap buckets should be greater than 0");
 
   OS << "Section Name, Begin Address, End Address, Percentage Hotness, "
      << "Utilization Pct, Partition Score\n";
   const uint64_t MappedCounts = NumTotalCounts - UnmappedHotness;
-  for (const auto [Name, Begin, End] : TextSections) {
+  for (const auto [Name, Begin, End] : Sections) {
     const float Hotness = 1. * SectionHotness[Name] / NumTotalCounts;
     const float MappedHotness =
         MappedCounts ? 1. * SectionHotness[Name] / MappedCounts : 0;

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -968,8 +968,9 @@ void RewriteInstance::discoverFileObjects() {
       continue;
     }
 
-    // Ignore input hot markers
-    if (SymName == "__hot_start" || SymName == "__hot_end")
+    // Ignore input hot markers unless in heatmap mode
+    if ((SymName == "__hot_start" || SymName == "__hot_end") &&
+        !opts::HeatmapMode)
       continue;
 
     FileSymRefs.emplace(SymbolAddress, Symbol);

bolt/test/X86/callcont-fallthru.s

@@ -6,7 +6,7 @@
 # RUN: %clangxx %cxxflags %s %t.so -o %t -Wl,-q -nostdlib
 # RUN: link_fdata %s %t %t.pat PREAGGT1
 # RUN: link_fdata %s %t %t.pat2 PREAGGT2
-# RUN: link_fdata %s %t %t.patplt PREAGGPLT
+# RUN-DISABLED: link_fdata %s %t %t.patplt PREAGGPLT
 
 # RUN: llvm-strip --strip-unneeded %t -o %t.strip
 # RUN: llvm-objcopy --remove-section=.eh_frame %t.strip %t.noeh
@@ -26,8 +26,8 @@
 
 ## Check pre-aggregated traces don't report zero-sized PLT fall-through as
 ## invalid trace
-# RUN: llvm-bolt %t.strip --pa -p %t.patplt -o %t.out | FileCheck %s \
-# RUN:   --check-prefix=CHECK-PLT
+# RUN-DISABLED: llvm-bolt %t.strip --pa -p %t.patplt -o %t.out | FileCheck %s \
+# RUN-DISABLED:   --check-prefix=CHECK-PLT
 # CHECK-PLT: traces mismatching disassembled function contents: 0
 
   .globl foo

bolt/test/X86/heatmap-preagg.test

@@ -13,6 +13,7 @@ RUN:   --reorder-functions=cdsort --enable-bat --dyno-stats --skip-funcs=main
 RUN: llvm-bolt-heatmap %t.out -o %t2 --pa -p %p/Inputs/blarge_new_bat.preagg.txt \
 RUN:   2>&1 | FileCheck --check-prefix CHECK-HEATMAP-BAT %s
 RUN: FileCheck %s --check-prefix CHECK-SEC-HOT-BAT --input-file %t2-section-hotness.csv
+RUN: llvm-nm -n %t.out | FileCheck %s --check-prefix=CHECK-HOT-SYMS
 
 CHECK-HEATMAP: PERF2BOLT: read 81 aggregated LBR entries
 CHECK-HEATMAP: HEATMAP: invalid traces: 1
@@ -33,3 +34,6 @@ CHECK-SEC-HOT-BAT-NEXT: .bolt.org.text, 0x4010b0, 0x401c25, 38.3385, 51.0638, 0.
 CHECK-SEC-HOT-BAT-NEXT: .fini, 0x401c28, 0x401c35, 0.0000, 0.0000, 0.0000
 CHECK-SEC-HOT-BAT-NEXT: .text, 0x800000, 0x8002cc, 38.7595, 91.6667, 0.3553
 CHECK-SEC-HOT-BAT-NEXT: .text.cold, 0x800300, 0x800415, 0.0000, 0.0000, 0.0000
+CHECK-SEC-HOT-BAT-NEXT: [hot text], 0x800000, 0x8002cc, 38.7595, 91.6667, 0.3553
+CHECK-HOT-SYMS: 800000 W __hot_start
+CHECK-HOT-SYMS: 8002cc W __hot_end