[SHT_LLVM_BB_ADDR_MAP] Add a new PGOAnalysisMap feature to emit dynamic instruction count

llvm/include/llvm/Object/ELFTypes.h

@@ -831,8 +831,11 @@ struct BBAddrMap {
     bool BrProb : 1;
     bool MultiBBRange : 1;
     bool OmitBBEntries : 1;
+    bool DynamicInstCount : 1;
 
-    bool hasPGOAnalysis() const { return FuncEntryCount || BBFreq || BrProb; }
+    bool hasPGOAnalysis() const {
+      return FuncEntryCount || BBFreq || BrProb || DynamicInstCount;
+    }
 
     bool hasPGOAnalysisBBData() const { return BBFreq || BrProb; }
 
@@ -842,7 +845,8 @@ struct BBAddrMap {
              (static_cast<uint8_t>(BBFreq) << 1) |
              (static_cast<uint8_t>(BrProb) << 2) |
              (static_cast<uint8_t>(MultiBBRange) << 3) |
-             (static_cast<uint8_t>(OmitBBEntries) << 4);
+             (static_cast<uint8_t>(OmitBBEntries) << 4) |
+             (static_cast<uint8_t>(DynamicInstCount) << 5);
     }
 
     // Decodes from minimum bit width representation and validates no
@@ -851,7 +855,7 @@ struct BBAddrMap {
       Features Feat{
           static_cast<bool>(Val & (1 << 0)), static_cast<bool>(Val & (1 << 1)),
           static_cast<bool>(Val & (1 << 2)), static_cast<bool>(Val & (1 << 3)),
-          static_cast<bool>(Val & (1 << 4))};
+          static_cast<bool>(Val & (1 << 4)), static_cast<bool>(Val & (1 << 5))};
       if (Feat.encode() != Val)
         return createStringError(
             std::error_code(), "invalid encoding for BBAddrMap::Features: 0x%x",
@@ -861,9 +865,10 @@ struct BBAddrMap {
 
     bool operator==(const Features &Other) const {
       return std::tie(FuncEntryCount, BBFreq, BrProb, MultiBBRange,
-                      OmitBBEntries) ==
+                      OmitBBEntries, DynamicInstCount) ==
              std::tie(Other.FuncEntryCount, Other.BBFreq, Other.BrProb,
-                      Other.MultiBBRange, Other.OmitBBEntries);
+                      Other.MultiBBRange, Other.OmitBBEntries,
+                      Other.DynamicInstCount);
     }
   };
 
@@ -1016,14 +1021,16 @@ struct PGOAnalysisMap {
   };
 
   uint64_t FuncEntryCount;           // Prof count from IR function
+  uint64_t DynamicInstCount;         // Dynamic instruction count
   std::vector<PGOBBEntry> BBEntries; // Extended basic block entries
 
   // Flags to indicate if each PGO related info was enabled in this function
   BBAddrMap::Features FeatEnable;
 
   bool operator==(const PGOAnalysisMap &Other) const {
-    return std::tie(FuncEntryCount, BBEntries, FeatEnable) ==
-           std::tie(Other.FuncEntryCount, Other.BBEntries, Other.FeatEnable);
+    return std::tie(FuncEntryCount, DynamicInstCount, BBEntries, FeatEnable) ==
+           std::tie(Other.FuncEntryCount, Other.DynamicInstCount,
+                    Other.BBEntries, Other.FeatEnable);
   }
 };
 

llvm/include/llvm/ObjectYAML/ELFYAML.h

@@ -192,6 +192,7 @@ struct PGOAnalysisMapEntry {
     std::optional<std::vector<SuccessorEntry>> Successors;
   };
   std::optional<uint64_t> FuncEntryCount;
+  std::optional<uint64_t> DynamicInstCount;
   std::optional<std::vector<PGOBBEntry>> PGOBBEntries;
 };
 

llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp

@@ -143,6 +143,7 @@ using namespace llvm;
 enum class PGOMapFeaturesEnum {
   None,
   FuncEntryCount,
+  DynInstCount,
   BBFreq,
   BrProb,
   All,
@@ -153,6 +154,8 @@ static cl::bits<PGOMapFeaturesEnum> PgoAnalysisMapFeatures(
         clEnumValN(PGOMapFeaturesEnum::None, "none", "Disable all options"),
         clEnumValN(PGOMapFeaturesEnum::FuncEntryCount, "func-entry-count",
                    "Function Entry Count"),
+        clEnumValN(PGOMapFeaturesEnum::DynInstCount, "dyn-inst-count",
+                   "Dynamic instruction count"),
         clEnumValN(PGOMapFeaturesEnum::BBFreq, "bb-freq",
                    "Basic Block Frequency"),
         clEnumValN(PGOMapFeaturesEnum::BrProb, "br-prob", "Branch Probability"),
@@ -1412,6 +1415,9 @@ getBBAddrMapFeature(const MachineFunction &MF, int NumMBBSectionRanges) {
   bool FuncEntryCountEnabled =
       AllFeatures || (!NoFeatures && PgoAnalysisMapFeatures.isSet(
                                          PGOMapFeaturesEnum::FuncEntryCount));
+  bool DynInstCountEnabled =
+      AllFeatures || (!NoFeatures && PgoAnalysisMapFeatures.isSet(
+                                         PGOMapFeaturesEnum::DynInstCount));
   bool BBFreqEnabled =
       AllFeatures ||
       (!NoFeatures && PgoAnalysisMapFeatures.isSet(PGOMapFeaturesEnum::BBFreq));
@@ -1424,9 +1430,12 @@ getBBAddrMapFeature(const MachineFunction &MF, int NumMBBSectionRanges) {
         "BB entries info is required for BBFreq and BrProb "
         "features");
   }
-  return {FuncEntryCountEnabled, BBFreqEnabled, BrProbEnabled,
+  return {FuncEntryCountEnabled,
+          BBFreqEnabled,
+          BrProbEnabled,
           MF.hasBBSections() && NumMBBSectionRanges > 1,
-          static_cast<bool>(BBAddrMapSkipEmitBBEntries)};
+          static_cast<bool>(BBAddrMapSkipEmitBBEntries),
+          DynInstCountEnabled};
 }
 
 void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) {
@@ -1519,16 +1528,21 @@ void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) {
           MaybeEntryCount ? MaybeEntryCount->getCount() : 0);
     }
     const MachineBlockFrequencyInfo *MBFI =
-        Features.BBFreq
+        Features.BBFreq || Features.DynamicInstCount
             ? &getAnalysis<LazyMachineBlockFrequencyInfoPass>().getBFI()
             : nullptr;
     const MachineBranchProbabilityInfo *MBPI =
         Features.BrProb
             ? &getAnalysis<MachineBranchProbabilityInfoWrapperPass>().getMBPI()
             : nullptr;
 
-    if (Features.BBFreq || Features.BrProb) {
+    if (Features.BBFreq || Features.BrProb || Features.DynamicInstCount) {
+      uint64_t DynInstCount = 0;
       for (const MachineBasicBlock &MBB : MF) {
+        if (Features.DynamicInstCount) {
+          DynInstCount +=
+              MBFI->getBlockProfileCount(&MBB).value_or(0) * MBB.size();
+        }
         if (Features.BBFreq) {
           OutStreamer->AddComment("basic block frequency");
           OutStreamer->emitULEB128IntValue(
@@ -1547,6 +1561,10 @@ void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) {
           }
         }
       }
+      if (Features.DynamicInstCount) {
+        OutStreamer->AddComment("Dynamic instruction count");
+        OutStreamer->emitULEB128IntValue(DynInstCount);
+      }
     }
   }
 

llvm/lib/Object/ELF.cpp

@@ -887,6 +887,12 @@ decodeBBAddrMapImpl(const ELFFile<ELFT> &EF,
               ? readULEB128As<uint64_t>(Data, Cur, ULEBSizeErr)
               : 0;
 
+      // Dynamic instruction count
+      uint64_t DynInstCount =
+          FeatEnable.DynamicInstCount
+              ? readULEB128As<uint64_t>(Data, Cur, ULEBSizeErr)
+              : 0;
+
       std::vector<PGOAnalysisMap::PGOBBEntry> PGOBBEntries;
       for (uint32_t BlockIndex = 0;
            FeatEnable.hasPGOAnalysisBBData() && !MetadataDecodeErr &&
@@ -915,8 +921,8 @@ decodeBBAddrMapImpl(const ELFFile<ELFT> &EF,
       }
 
       if (PGOAnalyses)
-        PGOAnalyses->push_back(
-            {FuncEntryCount, std::move(PGOBBEntries), FeatEnable});
+        PGOAnalyses->push_back({FuncEntryCount, DynInstCount,
+                                std::move(PGOBBEntries), FeatEnable});
     }
   }
   // Either Cur is in the error state, or we have an error in ULEBSizeErr or

llvm/lib/ObjectYAML/ELFEmitter.cpp

@@ -1515,6 +1515,9 @@ void ELFState<ELFT>::writeSectionContent(
     if (PGOEntry.FuncEntryCount)
       SHeader.sh_size += CBA.writeULEB128(*PGOEntry.FuncEntryCount);
 
+    if (PGOEntry.DynamicInstCount)
+      SHeader.sh_size += CBA.writeULEB128(*PGOEntry.DynamicInstCount);
+
     if (!PGOEntry.PGOBBEntries)
       continue;
 

llvm/lib/ObjectYAML/ELFYAML.cpp

@@ -1872,6 +1872,7 @@ void MappingTraits<ELFYAML::PGOAnalysisMapEntry>::mapping(
     IO &IO, ELFYAML::PGOAnalysisMapEntry &E) {
   assert(IO.getContext() && "The IO context is not initialized");
   IO.mapOptional("FuncEntryCount", E.FuncEntryCount);
+  IO.mapOptional("DynamicInstCount", E.DynamicInstCount);
   IO.mapOptional("PGOBBEntries", E.PGOBBEntries);
 }
 

llvm/test/CodeGen/X86/basic-block-address-map-pgo-features.ll

@@ -3,13 +3,14 @@
 ; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=none | FileCheck %s --check-prefixes=CHECK,BASIC,PGO-NONE
 
 ;; Also verify this holds for all PGO features enabled
-; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=func-entry-count,bb-freq,br-prob | FileCheck %s --check-prefixes=CHECK,PGO-ALL,PGO-FEC,PGO-BBF,PGO-BRP
-; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=all | FileCheck %s --check-prefixes=CHECK,PGO-ALL,PGO-FEC,PGO-BBF,PGO-BRP
+; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=func-entry-count,bb-freq,br-prob,dyn-inst-count | FileCheck %s --check-prefixes=CHECK,PGO-ALL,PGO-FEC,PGO-BBF,PGO-BRP
+; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=all | FileCheck %s --check-prefixes=CHECK,PGO-ALL,PGO-FEC,PGO-BBF,PGO-BRP,PGO-DINST
 
 ;; Also verify that pgo extension only includes the enabled feature
 ; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=func-entry-count | FileCheck %s --check-prefixes=CHECK,PGO-FEC,FEC-ONLY
 ; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=bb-freq | FileCheck %s --check-prefixes=CHECK,PGO-BBF,BBF-ONLY
 ; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=br-prob | FileCheck %s --check-prefixes=CHECK,PGO-BRP,BRP-ONLY
+; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=dyn-inst-count | FileCheck %s --check-prefixes=CHECK,PGO-DINST,DINST-ONLY
 
 ; RUN: llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=func-entry-count -basic-block-address-map-skip-bb-entries | FileCheck %s --check-prefixes=SKIP-BB-ENTRIES
 ; RUN: not llc < %s -mtriple=x86_64 -function-sections -unique-section-names=true -basic-block-address-map -pgo-analysis-map=bb-freq -basic-block-address-map-skip-bb-entries 2>&1 | FileCheck %s --check-prefixes=SKIP-BB-ENTRIES-ERROR
@@ -71,8 +72,9 @@ declare i32 @__gxx_personality_v0(...)
 ; CHECK: 	.section	.llvm_bb_addr_map,"o",@llvm_bb_addr_map,.text._Z3bazb{{$}}
 ; CHECK-NEXT:	.byte	2		# version
 ; BASIC-NEXT:	.byte	0		# feature
-; PGO-ALL-NEXT:	.byte	7		# feature
+; PGO-ALL-NEXT:	.byte	39 	 # feature
 ; FEC-ONLY-NEXT:.byte	1		# feature
+; DINST-ONLY-NEXT:.byte	32		# feature
 ; BBF-ONLY-NEXT:.byte	2		# feature
 ; BRP-ONLY-NEXT:.byte	4		# feature
 ; CHECK-NEXT:	.quad	.Lfunc_begin0	# function address
@@ -138,6 +140,8 @@ declare i32 @__gxx_personality_v0(...)
 ; PGO-BRP-NEXT:	.byte	5		# successor BB ID
 ; PGO-BRP-NEXT:	.ascii	"\200\200\200\200\b"	# successor branch probability
 
+; PGO-DINST: .ascii	"\341\f"                        # Dynamic instruction count
+
 ; SKIP-BB-ENTRIES:      .byte	17                              # feature
 ; SKIP-BB-ENTRIES-NEXT:	.quad	.Lfunc_begin0                   # function address
 ; SKIP-BB-ENTRIES-NEXT:	.byte	6                               # number of basic blocks

llvm/test/tools/llvm-objdump/X86/elf-pgoanalysismap.yaml

@@ -43,9 +43,49 @@ Symbols:
 # ENTRYCOUNT: <foo>:
 # ENTRYCOUNT: <BB3> (Entry count: 1000):
 
+## Check the case where we only have dynamic instruction count.
+# RUN: yaml2obj --docnum=2 %s -o %t1
+# RUN: llvm-objdump %t1 -d --symbolize-operands --no-show-raw-insn --no-leading-addr | \
+# RUN:   FileCheck %s --check-prefix=DYNICNT
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name:    .text.foo
+    Type:    SHT_PROGBITS
+    Address: 0x0
+    Flags:   [SHF_ALLOC, SHF_EXECINSTR]
+    Content: '50'
+  - Name:   .llvm_bb_addr_map.foo
+    Type:   SHT_LLVM_BB_ADDR_MAP
+    Link:   .text.foo
+    Entries:
+      - Version: 2
+        Feature: 0x20
+        BBRanges:
+          - BaseAddress: 0x0
+            BBEntries:
+              - ID:            3
+                AddressOffset: 0x0
+                Size:          0x1
+                Metadata:      0x1
+    PGOAnalyses:
+      - DynamicInstCount: 1000
+Symbols:
+  - Name:    foo
+    Section: .text.foo
+    Value:   0x0
+
+# DYNICNT: <foo>:
+# DYNICNT: <BB3> (Dynamic instruction count: 1000):
+
 ## Check the case where we have entry points and block frequency information
 
-# RUN: yaml2obj %s --docnum=2 -o %t2
+# RUN: yaml2obj %s --docnum=3 -o %t2
 # RUN: llvm-objdump %t2 -d --symbolize-operands --no-show-raw-insn --no-leading-addr | \
 # RUN:   FileCheck --match-full-lines --strict-whitespace %s --check-prefix=ENTRYCOUNT-BLOCKFREQ
 # RUN: llvm-objdump %t2 -d --symbolize-operands --pretty-pgo-analysis-map --no-show-raw-insn --no-leading-addr | \
@@ -68,7 +108,7 @@ Sections:
     Link:   .text.foo
     Entries:
       - Version: 2
-        Feature: 0x3
+        Feature: 0x23
         BBRanges:
           - BaseAddress: 0x0
             BBEntries:
@@ -90,6 +130,7 @@ Sections:
                 Metadata:      0x2
     PGOAnalyses:
       - FuncEntryCount: 1000
+        DynamicInstCount: 2000
         PGOBBEntries:
           - BBFreq: 1000
           - BBFreq: 133
@@ -101,21 +142,21 @@ Symbols:
     Value:   0x0
 
 # ENTRYCOUNT-BLOCKFREQ:<foo>:
-# ENTRYCOUNT-BLOCKFREQ:<BB3> (Entry count: 1000, Frequency: 1000):
+# ENTRYCOUNT-BLOCKFREQ:<BB3> (Entry count: 1000, Dynamic instruction count: 2000, Frequency: 1000):
 # ENTRYCOUNT-BLOCKFREQ:<BB1> (Frequency: 133):
 # ENTRYCOUNT-BLOCKFREQ:<BB2> (Frequency: 18):
 # ENTRYCOUNT-BLOCKFREQ:<BB5> (Frequency: 1000):
 
 # ENTRYCOUNT-BLOCKFREQ-PRETTY:<foo>:
-# ENTRYCOUNT-BLOCKFREQ-PRETTY:<BB3> (Entry count: 1000, Frequency: 1.0):
+# ENTRYCOUNT-BLOCKFREQ-PRETTY:<BB3> (Entry count: 1000, Dynamic instruction count: 2000, Frequency: 1.0):
 # ENTRYCOUNT-BLOCKFREQ-PRETTY:<BB1> (Frequency: 0.133):
 # ENTRYCOUNT-BLOCKFREQ-PRETTY:<BB2> (Frequency: 0.018):
 # ENTRYCOUNT-BLOCKFREQ-PRETTY:<BB5> (Frequency: 1.0):
 
 ## Check the case where we have entry points, block frequency, and branch
 ## proabability information.
 
-# RUN: yaml2obj %s --docnum=3 -o %t3
+# RUN: yaml2obj %s --docnum=4 -o %t3
 # RUN: llvm-objdump %t3 -d --symbolize-operands --no-show-raw-insn --no-leading-addr | \
 # RUN:   FileCheck --match-full-lines --strict-whitespace %s --check-prefix=ENTRY-FREQ-PROB
 # RUN: llvm-objdump %t3 -d --symbolize-operands --pretty-pgo-analysis-map --no-show-raw-insn --no-leading-addr | \

llvm/test/tools/llvm-readobj/ELF/bb-addr-map-pgo-analysis-map.test

@@ -15,7 +15,7 @@
 
 ## Check that a malformed section can be handled.
 # RUN: yaml2obj %s -DBITS=32 -DSIZE=24 -o %t2.o
-# RUN: llvm-readobj %t2.o --bb-addr-map 2>&1 | FileCheck --match-full-lines %s -DOFFSET=0x00000018 -DFILE=%t2.o --check-prefix=TRUNCATED
+# RUN: llvm-readobj %t2.o --bb-addr-map 2>&1 | FileCheck --match-full-lines %s -DOFFSET=0x00000014 -DFILE=%t2.o --check-prefix=TRUNCATED
 
 ## Check that missing features can be handled.
 # RUN: yaml2obj %s -DBITS=32 -DFEATURE=0x2 -o %t3.o
@@ -55,6 +55,7 @@
 # CHECK-NEXT:     ]
 # CHECK-NEXT:     PGO analyses {
 # CHECK-NEXT:       FuncEntryCount: 100
+# CHECK-NEXT:       DynamicInstCount: 100
 # CHECK-NEXT:       PGO BB entries [
 # CHECK-NEXT:         {
 # RAW-NEXT:             Frequency: 100
@@ -98,6 +99,7 @@
 # CHECK-NEXT:     ]
 # CHECK-NEXT:     PGO analyses {
 # CHECK-NEXT:       FuncEntryCount: 8888
+# CHECK-NEXT:       DynamicInstCount: 8888
 # CHECK-NEXT:       PGO BB entries [
 # CHECK-NEXT:         {
 # RAW-NEXT:             Frequency: 9000
@@ -173,7 +175,7 @@ Sections:
     Link:   .text
     Entries:
       - Version: 2
-        Feature: 0x7
+        Feature: 0x27
         BBRanges:
           - BaseAddress: [[ADDR=0x11111]]
             BBEntries:
@@ -186,7 +188,7 @@ Sections:
                 Size:          0x4
                 Metadata:      0x15
       - Version: 2
-        Feature: 0x3
+        Feature: 0x23
         BBRanges:
           - BaseAddress: 0x22222
             BBEntries:
@@ -196,6 +198,7 @@ Sections:
                 Metadata:      0x8
     PGOAnalyses:
       - FuncEntryCount: 100
+        DynamicInstCount: 100
         PGOBBEntries:
           - BBFreq:        100
             Successors:
@@ -204,6 +207,7 @@ Sections:
           - BBFreq:        100
             Successors:    []
       - FuncEntryCount: 8888
+        DynamicInstCount: 8888
         PGOBBEntries:
           - BBFreq:        9000
   - Name: dummy_section
@@ -237,4 +241,3 @@ Symbols:
     Section: .text.bar
     Type:    STT_FUNC
     Value:   0x33333
-