[SHT_LLVM_BB_ADDR_MAP] Encode and decode callsite offsets in a newly-introduced SHT_LLVM_BB_ADDR_MAP version. #144426
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-objectyaml @llvm/pr-subscribers-llvm-binary-utilities Author: Rahman Lavaee (rlavaee) ChangesRecently, we have been looking at some optimizations targeting individual calls. In particular, we plan to extend the address mapping technique to map to individual callsites. For example, in this piece of code for a basic blocks: We want to emit 0x9 as the call site offset for This PR implements the decode/encode/emit capability. The Codegen change will be implemented in a later PR. Patch is 30.11 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/144426.diff 12 Files Affected:
diff --git a/llvm/include/llvm/Object/ELFTypes.h b/llvm/include/llvm/Object/ELFTypes.h
index 87e4dbe448091..d7a468f1116d7 100644
--- a/llvm/include/llvm/Object/ELFTypes.h
+++ b/llvm/include/llvm/Object/ELFTypes.h
@@ -831,6 +831,7 @@ struct BBAddrMap {
bool BrProb : 1;
bool MultiBBRange : 1;
bool OmitBBEntries : 1;
+ bool CallsiteOffsets : 1;
bool hasPGOAnalysis() const { return FuncEntryCount || BBFreq || BrProb; }
@@ -842,7 +843,8 @@ 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>(CallsiteOffsets) << 5);
}
// Decodes from minimum bit width representation and validates no
@@ -851,7 +853,7 @@ 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 +863,10 @@ struct BBAddrMap {
bool operator==(const Features &Other) const {
return std::tie(FuncEntryCount, BBFreq, BrProb, MultiBBRange,
- OmitBBEntries) ==
+ OmitBBEntries, CallsiteOffsets) ==
std::tie(Other.FuncEntryCount, Other.BBFreq, Other.BrProb,
- Other.MultiBBRange, Other.OmitBBEntries);
+ Other.MultiBBRange, Other.OmitBBEntries,
+ Other.CallsiteOffsets);
}
};
@@ -914,13 +917,18 @@ struct BBAddrMap {
uint32_t Size = 0; // Size of the basic block.
Metadata MD = {false, false, false, false,
false}; // Metdata for this basic block.
+ // Offsets of callsites (end of call instructions), relative to the basic
+ // block start.
+ SmallVector<uint32_t, 1> CallsiteOffsets;
- BBEntry(uint32_t ID, uint32_t Offset, uint32_t Size, Metadata MD)
- : ID(ID), Offset(Offset), Size(Size), MD(MD){};
+ BBEntry(uint32_t ID, uint32_t Offset, uint32_t Size, Metadata MD,
+ SmallVector<uint32_t, 1> CallsiteOffsets)
+ : ID(ID), Offset(Offset), Size(Size), MD(MD),
+ CallsiteOffsets(std::move(CallsiteOffsets)) {}
bool operator==(const BBEntry &Other) const {
return ID == Other.ID && Offset == Other.Offset && Size == Other.Size &&
- MD == Other.MD;
+ MD == Other.MD && CallsiteOffsets == Other.CallsiteOffsets;
}
bool hasReturn() const { return MD.HasReturn; }
diff --git a/llvm/include/llvm/ObjectYAML/ELFYAML.h b/llvm/include/llvm/ObjectYAML/ELFYAML.h
index dfdfa055d65fa..539d56bed102b 100644
--- a/llvm/include/llvm/ObjectYAML/ELFYAML.h
+++ b/llvm/include/llvm/ObjectYAML/ELFYAML.h
@@ -162,6 +162,7 @@ struct BBAddrMapEntry {
llvm::yaml::Hex64 AddressOffset;
llvm::yaml::Hex64 Size;
llvm::yaml::Hex64 Metadata;
+ std::optional<std::vector<llvm::yaml::Hex64>> CallsiteOffsets;
};
uint8_t Version;
llvm::yaml::Hex8 Feature;
diff --git a/llvm/lib/Object/ELF.cpp b/llvm/lib/Object/ELF.cpp
index e6864ca508a54..359dfb14b8c37 100644
--- a/llvm/lib/Object/ELF.cpp
+++ b/llvm/lib/Object/ELF.cpp
@@ -837,7 +837,7 @@ decodeBBAddrMapImpl(const ELFFile<ELFT> &EF,
Version = Data.getU8(Cur);
if (!Cur)
break;
- if (Version > 2)
+ if (Version > 3)
return createError("unsupported SHT_LLVM_BB_ADDR_MAP version: " +
Twine(static_cast<int>(Version)));
Feature = Data.getU8(Cur); // Feature byte
@@ -893,7 +893,32 @@ decodeBBAddrMapImpl(const ELFFile<ELFT> &EF,
? readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr)
: BlockIndex;
uint32_t Offset = readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr);
- uint32_t Size = readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr);
+ // Read the callsite offsets.
+ uint32_t LastCallsiteOffset = 0;
+ SmallVector<uint32_t, 1> CallsiteOffsets;
+ if (FeatEnable.CallsiteOffsets) {
+ if (Version < 3) {
+ return createError(
+ "version should be >= 3 for SHT_LLVM_BB_ADDR_MAP when "
+ "callsite offsets feature is enabled: version = " +
+ Twine(static_cast<int>(Version)) +
+ " feature = " + Twine(static_cast<int>(Feature)));
+ }
+ uint32_t NumCallsites =
+ readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr);
+ CallsiteOffsets.reserve(NumCallsites);
+ for (uint32_t CallsiteIndex = 0;
+ !ULEBSizeErr && Cur && (CallsiteIndex < NumCallsites);
+ ++CallsiteIndex) {
+ LastCallsiteOffset +=
+ readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr);
+ CallsiteOffsets.push_back(LastCallsiteOffset);
+ }
+ if (!Cur || ULEBSizeErr)
+ break;
+ }
+ uint32_t Size = readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr) +
+ LastCallsiteOffset;
uint32_t MD = readULEB128As<uint32_t>(Data, Cur, ULEBSizeErr);
if (Version >= 1) {
// Offset is calculated relative to the end of the previous BB.
@@ -906,7 +931,8 @@ decodeBBAddrMapImpl(const ELFFile<ELFT> &EF,
MetadataDecodeErr = MetadataOrErr.takeError();
break;
}
- BBEntries.push_back({ID, Offset, Size, *MetadataOrErr});
+ BBEntries.push_back(
+ {ID, Offset, Size, *MetadataOrErr, CallsiteOffsets});
}
TotalNumBlocks += BBEntries.size();
}
diff --git a/llvm/lib/ObjectYAML/ELFEmitter.cpp b/llvm/lib/ObjectYAML/ELFEmitter.cpp
index 9ae76a71ede5e..76d679c9ceffb 100644
--- a/llvm/lib/ObjectYAML/ELFEmitter.cpp
+++ b/llvm/lib/ObjectYAML/ELFEmitter.cpp
@@ -1452,7 +1452,7 @@ void ELFState<ELFT>::writeSectionContent(
for (const auto &[Idx, E] : llvm::enumerate(*Section.Entries)) {
// Write version and feature values.
if (Section.Type == llvm::ELF::SHT_LLVM_BB_ADDR_MAP) {
- if (E.Version > 2)
+ if (E.Version > 3)
WithColor::warning() << "unsupported SHT_LLVM_BB_ADDR_MAP version: "
<< static_cast<int>(E.Version)
<< "; encoding using the most recent version";
@@ -1483,6 +1483,13 @@ void ELFState<ELFT>::writeSectionContent(
if (!E.BBRanges)
continue;
uint64_t TotalNumBlocks = 0;
+ bool EmitCallsiteOffsets = FeatureOrErr->CallsiteOffsets;
+ if (!EmitCallsiteOffsets) {
+ for (const ELFYAML::BBAddrMapEntry::BBRangeEntry &BBR : *E.BBRanges)
+ for (const ELFYAML::BBAddrMapEntry::BBEntry &BBE : *BBR.BBEntries)
+ EmitCallsiteOffsets |=
+ BBE.CallsiteOffsets.has_value() && !BBE.CallsiteOffsets->empty();
+ }
for (const ELFYAML::BBAddrMapEntry::BBRangeEntry &BBR : *E.BBRanges) {
// Write the base address of the range.
CBA.write<uintX_t>(BBR.BaseAddress, ELFT::Endianness);
@@ -1500,6 +1507,15 @@ void ELFState<ELFT>::writeSectionContent(
if (Section.Type == llvm::ELF::SHT_LLVM_BB_ADDR_MAP && E.Version > 1)
SHeader.sh_size += CBA.writeULEB128(BBE.ID);
SHeader.sh_size += CBA.writeULEB128(BBE.AddressOffset);
+ if (EmitCallsiteOffsets) {
+ uint32_t NumCallsites =
+ BBE.CallsiteOffsets.has_value() ? BBE.CallsiteOffsets->size() : 0;
+ SHeader.sh_size += CBA.writeULEB128(NumCallsites);
+ if (BBE.CallsiteOffsets.has_value()) {
+ for (uint32_t Offset : *BBE.CallsiteOffsets)
+ SHeader.sh_size += CBA.writeULEB128(Offset);
+ }
+ }
SHeader.sh_size += CBA.writeULEB128(BBE.Size);
SHeader.sh_size += CBA.writeULEB128(BBE.Metadata);
}
diff --git a/llvm/lib/ObjectYAML/ELFYAML.cpp b/llvm/lib/ObjectYAML/ELFYAML.cpp
index 520e956fdab9f..c38f86e4f4f1b 100644
--- a/llvm/lib/ObjectYAML/ELFYAML.cpp
+++ b/llvm/lib/ObjectYAML/ELFYAML.cpp
@@ -1882,6 +1882,7 @@ void MappingTraits<ELFYAML::BBAddrMapEntry::BBEntry>::mapping(
IO.mapRequired("AddressOffset", E.AddressOffset);
IO.mapRequired("Size", E.Size);
IO.mapRequired("Metadata", E.Metadata);
+ IO.mapOptional("CallsiteOffsets", E.CallsiteOffsets);
}
void MappingTraits<ELFYAML::PGOAnalysisMapEntry>::mapping(
diff --git a/llvm/test/tools/llvm-readobj/ELF/bb-addr-map.test b/llvm/test/tools/llvm-readobj/ELF/bb-addr-map.test
index c5d071c11d1de..5d7bc8baa9b25 100644
--- a/llvm/test/tools/llvm-readobj/ELF/bb-addr-map.test
+++ b/llvm/test/tools/llvm-readobj/ELF/bb-addr-map.test
@@ -49,7 +49,8 @@
# CHECK-NEXT: {
# CHECK-NEXT: ID: 2
# CHECK-NEXT: Offset: 0x3
-# CHECK-NEXT: Size: 0x4
+# CHECK-NEXT: Callsite Offsets: [1, 3]
+# CHECK-NEXT: Size: 0x7
# CHECK-NEXT: HasReturn: Yes
# CHECK-NEXT: HasTailCall: No
# CHECK-NEXT: IsEHPad: Yes
@@ -75,7 +76,7 @@
# CHECK-NEXT: HasTailCall: No
# CHECK-NEXT: IsEHPad: No
# CHECK-NEXT: CanFallThrough: Yes
-# CHECK-NEXT: HasIndirectBranch: No
+# CHECK-NEXT: HasIndirectBranch: No
# CHECK-NEXT: }
# CHECK-NEXT: ]
# CHECK-NEXT: }
@@ -143,8 +144,8 @@ Sections:
ShSize: [[SIZE=<none>]]
Link: .text
Entries:
- - Version: 2
- Feature: 0x8
+ - Version: 3
+ Feature: 0x28
BBRanges:
- BaseAddress: [[ADDR=0x11111]]
BBEntries:
@@ -158,6 +159,7 @@ Sections:
AddressOffset: 0x3
Size: 0x4
Metadata: 0x15
+ CallsiteOffsets: [ 0x1 , 0x2 ]
- Version: 2
BBRanges:
- BaseAddress: 0x22222
diff --git a/llvm/test/tools/obj2yaml/ELF/bb-addr-map.yaml b/llvm/test/tools/obj2yaml/ELF/bb-addr-map.yaml
index 8dbf97ef2bc12..861cb94692947 100644
--- a/llvm/test/tools/obj2yaml/ELF/bb-addr-map.yaml
+++ b/llvm/test/tools/obj2yaml/ELF/bb-addr-map.yaml
@@ -14,7 +14,7 @@
# VALID-NEXT: - Name: .llvm_bb_addr_map
# VALID-NEXT: Type: SHT_LLVM_BB_ADDR_MAP
# VALID-NEXT: Entries:
-# VALID-NEXT: - Version: 2
+# VALID-NEXT: - Version: 3
# VALID-NEXT: BBRanges:
## The 'BaseAddress' field is omitted when it's zero.
# VALID-NEXT: - BBEntries:
@@ -30,15 +30,16 @@
# VALID-NEXT: AddressOffset: 0xFFFFFFFFFFFFFFF7
# VALID-NEXT: Size: 0xFFFFFFFFFFFFFFF8
# VALID-NEXT: Metadata: 0xFFFFFFFFFFFFFFF9
-# VALID-NEXT: - Version: 2
-# VALID-NEXT: Feature: 0x8
+# VALID-NEXT: - Version: 3
+# VALID-NEXT: Feature: 0x28
# VALID-NEXT: BBRanges:
# VALID-NEXT: - BaseAddress: 0xFFFFFFFFFFFFFF20
# VALID-NEXT: BBEntries:
-# VALID-NEXT: - ID: 6
-# VALID-NEXT: AddressOffset: 0xA
-# VALID-NEXT: Size: 0xB
-# VALID-NEXT: Metadata: 0xC
+# VALID-NEXT: - ID: 6
+# VALID-NEXT: AddressOffset: 0xA
+# VALID-NEXT: Size: 0xB
+# VALID-NEXT: Metadata: 0xC
+# VALID-NEXT: CallsiteOffsets: [ 0x1, 0x2 ]
--- !ELF
FileHeader:
@@ -50,7 +51,7 @@ Sections:
Type: SHT_LLVM_BB_ADDR_MAP
ShSize: [[SIZE=<none>]]
Entries:
- - Version: 2
+ - Version: 3
Feature: 0x0
BBRanges:
- BaseAddress: 0x0
@@ -67,17 +68,18 @@ Sections:
AddressOffset: 0xFFFFFFFFFFFFFFF7
Size: 0xFFFFFFFFFFFFFFF8
Metadata: 0xFFFFFFFFFFFFFFF9
- - Version: 2
- Feature: 0x8
+ - Version: 3
+ Feature: 0x28
NumBBRanges: [[NUMBBRANGES=<none>]]
BBRanges:
- BaseAddress: 0xFFFFFFFFFFFFFF20
NumBlocks: [[NUMBLOCKS=<none>]]
BBEntries:
- - ID: 6
- AddressOffset: 0xA
- Size: 0xB
- Metadata: 0xC
+ - ID: 6
+ AddressOffset: 0xA
+ Size: 0xB
+ Metadata: 0xC
+ CallsiteOffsets: [ 0x1, 0x2 ]
## Check obj2yaml can dump empty .llvm_bb_addr_map sections.
diff --git a/llvm/test/tools/yaml2obj/ELF/bb-addr-map.yaml b/llvm/test/tools/yaml2obj/ELF/bb-addr-map.yaml
index 709938babffbf..bc12884311e96 100644
--- a/llvm/test/tools/yaml2obj/ELF/bb-addr-map.yaml
+++ b/llvm/test/tools/yaml2obj/ELF/bb-addr-map.yaml
@@ -36,7 +36,8 @@
# Case 4: Specify Entries.
# CHECK: Name: .llvm_bb_addr_map (1)
# CHECK: SectionData (
-# CHECK-NEXT: 0000: 02002000 00000000 0000010B 010203
+# CHECK-NEXT: 0000: 03002000 00000000 0000010B 01020102
+# CHECK-NEXT: 0010: 0203
# CHECK-NEXT: )
# Case 5: Specify Entries and omit the Address field.
@@ -44,28 +45,32 @@
# CHECK: Address:
# CHECK-SAME: {{^ 0x0$}}
# CHECK: SectionData (
-# CHECK-NEXT: 0000: 02000000 00000000 0000010C 010203
+# CHECK-NEXT: 0000: 03000000 00000000 0000010C 010203
# CHECK-NEXT: )
# Case 6: Override the NumBlocks field.
# CHECK: Name: .llvm_bb_addr_map (1)
# CHECK: SectionData (
-# CHECK-NEXT: 0000: 02002000 00000000 0000020D 010203
+# CHECK-NEXT: 0000: 03002000 00000000 0000020D 010203
# CHECK-NEXT: )
# Case 7: Specify empty BBRanges.
# CHECK: Name: .llvm_bb_addr_map (1)
# CHECK: SectionData (
-# CHECK-NEXT: 0000: 020000
+# CHECK-NEXT: 0000: 030000
# CHECK-NEXT: )
# Case 8: Specify empty BBRanges with multi-bb-range.
# CHECK: Name: .llvm_bb_addr_map (1)
# CHECK: SectionData (
-# CHECK-NEXT: 0000: 020800
+# CHECK-NEXT: 0000: 030800
# CHECK-NEXT: )
-
+# Case 9: Specify empty CallsiteOffsets.
+# CHECK: Name: .llvm_bb_addr_map (1)
+# CHECK: SectionData (
+# CHECK-NEXT: 0000: 03202000 00000000 0000010E 01000203
+# CHECK-NEXT: )
--- !ELF
@@ -100,7 +105,7 @@ Sections:
- Name: '.llvm_bb_addr_map (4)'
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
- - Version: 2
+ - Version: 3
BBRanges:
- BaseAddress: 0x0000000000000020
BBEntries:
@@ -108,13 +113,14 @@ Sections:
AddressOffset: 0x00000001
Size: 0x00000002
Metadata: 0x00000003
+ CallsiteOffsets: [0x1, 0x2]
## 5) When specifying the description with Entries, the 'Address' field will be
## zero when omitted.
- Name: '.llvm_bb_addr_map (5)'
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
- - Version: 2
+ - Version: 3
BBRanges:
- BBEntries:
- ID: 12
@@ -127,7 +133,7 @@ Sections:
- Name: '.llvm_bb_addr_map (6)'
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
- - Version: 2
+ - Version: 3
BBRanges:
- BaseAddress: 0x0000000000000020
NumBlocks: 2
@@ -142,7 +148,7 @@ Sections:
- Name: '.llvm_bb_addr_map (7)'
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
- - Version: 2
+ - Version: 3
BBRanges: []
## 8) We can produce a SHT_LLVM_BB_ADDR_MAP section from a multi-bb-range
@@ -150,10 +156,27 @@ Sections:
- Name: '.llvm_bb_addr_map (8)'
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
- - Version: 2
+ - Version: 3
Feature: 0x8
BBRanges: []
+## 9) We can produce a SHT_LLVM_BB_ADDR_MAP section from a description
+## with empty callsite offsets.
+ - Name: '.llvm_bb_addr_map (9)'
+ Type: SHT_LLVM_BB_ADDR_MAP
+ Entries:
+ - Version: 3
+ Feature: 0x20
+ BBRanges:
+ - BaseAddress: 0x0000000000000020
+ BBEntries:
+ - ID: 14
+ AddressOffset: 0x00000001
+ Size: 0x00000002
+ Metadata: 0x00000003
+ CallsiteOffsets: []
+
+
## Check we can't use Entries at the same time as either Content or Size.
# RUN: not yaml2obj --docnum=2 -DCONTENT="00" %s 2>&1 | FileCheck %s --check-prefix=INVALID
# RUN: not yaml2obj --docnum=2 -DSIZE="0" %s 2>&1 | FileCheck %s --check-prefix=INVALID
@@ -175,7 +198,7 @@ Sections:
## Check that yaml2obj generates a warning when we use unsupported versions.
# RUN: yaml2obj --docnum=3 %s 2>&1 | FileCheck %s --check-prefix=INVALID-VERSION
-# INVALID-VERSION: warning: unsupported SHT_LLVM_BB_ADDR_MAP version: 3; encoding using the most recent version
+# INVALID-VERSION: warning: unsupported SHT_LLVM_BB_ADDR_MAP version: 4; encoding using the most recent version
--- !ELF
FileHeader:
@@ -187,4 +210,4 @@ Sections:
Type: SHT_LLVM_BB_ADDR_MAP
Entries:
## Specify unsupported version
- - Version: 3
+ - Version: 4
diff --git a/llvm/tools/llvm-readobj/ELFDumper.cpp b/llvm/tools/llvm-readobj/ELFDumper.cpp
index abaf6077ba9e7..7250d0a129cf5 100644
--- a/llvm/tools/llvm-readobj/ELFDumper.cpp
+++ b/llvm/tools/llvm-readobj/ELFDumper.cpp
@@ -7878,6 +7878,8 @@ void LLVMELFDumper<ELFT>::printBBAddrMaps(bool PrettyPGOAnalysis) {
DictScope BBED(W);
W.printNumber("ID", BBE.ID);
W.printHex("Offset", BBE.Offset);
+ if (!BBE.CallsiteOffsets.empty())
+ W.printList("Callsite Offsets", BBE.CallsiteOffsets);
W.printHex("Size", BBE.Size);
W.printBoolean("HasReturn", BBE.hasReturn());
W.printBoolean("HasTailCall", BBE.hasTailCall());
diff --git a/llvm/tools/obj2yaml/elf2yaml.cpp b/llvm/tools/obj2yaml/elf2yaml.cpp
index c56ed15501b40..53455b8c7580a 100644
--- a/llvm/tools/obj2yaml/elf2yaml.cpp
+++ b/llvm/tools/obj2yaml/elf2yaml.cpp
@@ -899,7 +899,7 @@ ELFDumper<ELFT>::dumpBBAddrMapSection(const Elf_Shdr *Shdr) {
while (Cur && Cur.tell() < Content.size()) {
if (Shdr->sh_type == ELF::SHT_LLVM_BB_ADDR_MAP) {
Version = Data.getU8(Cur);
- if (Cur && Version > 2)
+ if (Cur && Version > 3)
return createStringError(
errc::invalid_argument,
"invalid SHT_LLVM_BB_ADDR_MAP section version: " +
@@ -934,9 +934,19 @@ ELFDumper<ELFT>::dumpBBAddrMapSection(const Elf_Shdr *Shdr) {
++BlockIndex) {
uint32_t ID = Version >= 2 ? Data.getULEB128(Cur) : BlockIndex;
uint64_t Offset = Data.getULEB128(Cur);
+ std::optional<std::vector<llvm::yaml::Hex64>> CallsiteOffsets;
+ if (FeatureOrErr->CallsiteOffsets) {
+ uint32_t NumCallsites = Data.getULEB128(Cur);
+ CallsiteOffsets = std::vector<llvm::yaml::Hex64>(NumCallsites, 0);
+ for (uint32_t CallsiteIndex = 0; Cur && CallsiteIndex < NumCallsites;
+ ++CallsiteIndex) {
+ (*CallsiteOffsets)[CallsiteIndex] = Data.getULEB128(Cur);
+ }
+ }
uint64_t Size = Data.getULEB128(Cur);
uint64_t Metadata = Data.getULEB128(Cur);
- BBEntries.push_back({ID, Offset, Size, Metadata});
+ BBEntries.push_back(
+ {ID, Offset, Size, Metadata, std::move(CallsiteOffsets)});
}
TotalNumBlocks += BBEntries.size();
BBRanges.push_back({BaseAddress, /*NumBlocks=*/{}, BBEntries});
diff --git a/llvm/unittests/Object/ELFObjectFileTest.cpp b/llvm/unittests/Object/ELFObjectFileTest.cpp
index 1073df95c379a..423f92ea07b39 100644
--- a/llvm/unittests/Object/ELFObjectFileTest.cpp
+++ b/llvm/unittests/Object/ELFObjectFileTest.cpp
@@ -531,7 +531,7 @@ TEST(ELFObjectFileTest, InvalidDecodeBBAddrMap) {
// Check that we can detect unsupported versions.
SmallString<128> UnsupportedVersionYamlString(CommonYamlString);
UnsupportedVersionYamlString += R"(
- - Version: 3
+ - Version: 4
BBRanges:
- BaseAddress: 0x11111...
[truncated]
|
rlavaee
changed the title
There was a problem hiding this comment.
Can you describe your motivation for doing this a bit more?
We also needed to do something similar for basic block trace modelling given call instructions are sort of an implicit CFG edge. We just did this by parsing the binary/object files though. It was pretty inexpensive and didn't require any overhead/additional complexity within the BBAddrMap. It sounds like there are additional constraints here that I'm missing though.
I think your understanding is correct. To explain a bit more, we want to map from an address X to the following : Callsite region index #i specifies the address range between the end of the call #i-1 to the end of the call #i.
By "parsing" do you mean disassembly? The approach implemented here does not need any disassembly. And our constraint is no disassembly. |
Yeah. Why is not disassembling a hard requirement? If you're parsing the BBAddrMap I would assume you have access to the rest of the binary and it's not super expensive to disassemble a large binary, at least with our setup when we tried it. |
We would have to support every architecture separately, whereas here it's architecture-independent. We allow disassembly in the llvm-propeller tooling as long as it's a specific architecture-dependent case. What's your specific objection to this approach? We have extended the section multiple times so far (BB-relative offsets, BB ranges). |
Ah, true. I would think
No specific objection, I just want to make sure the additional complexity/version bump has sufficient motivation and isn't something that can just be trivially reconstructed from the binary. The CI failures also look relevant: https://github.com/llvm/llvm-project/actions/runs/15690836465?pr=144426 |
jh7370
reviewed
Jun 17, 2025
| Metadata: 0x00000003 | ||
| CallsiteOffsets: [] | ||
|
|
||
|
|
rlavaee
force-pushed
the
callsite_anchors
branch
from
26f0eba to
01b8f9e
Compare
rlavaee
force-pushed
the
callsite_anchors
branch
from
01b8f9e to
d4ca979
Compare
Correct. We will be using the Codegen intrinsic here too.
Thanks. Fixed the issues. |
rlavaee
force-pushed
the
callsite_anchors
branch
2 times, most recently
from
6632858 to
0b25028
Compare
rlavaee
force-pushed
the
callsite_anchors
branch
from
0b25028 to
1f848b5
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Recently, we have been looking at some optimizations targeting individual calls. In particular, we plan to extend the address mapping technique to map to individual callsites. For example, in this piece of code for a basic blocks:
We want to emit 0x9 as the call site offset for
callq foo(the offset from the block entry to right after the call), so that we know if a sampled address is before the call or after.This PR implements the decode/encode/emit capability. The Codegen change will be implemented in a later PR.