Successfully identified regression in *llvm* in CI configuration tcwg_bmk_llvm_tx1/llvm-master-aarch64-spec2k6-O3_LTO. So far, this commit has regressed CI configurations: - tcwg_bmk_llvm_tx1/llvm-master-aarch64-spec2k6-O3_LTO

Culprit: <cut> commit 5a9b25e15b4f9707245bb9d1ab56ab99b725505a Author: Sam Clegg sbc@chromium.org Date: Fri May 14 16:25:04 2021 -0700

[lld][WebAssembly] Refactor input chunk class hierarchy. NFC

The main motivation for this refactor is to remove the subclass relationship between the InputSegment and MergeInputSegment and SyntenticMergedInputSegment so that we can use the merging classes for debug sections which are not data segments.

In the process of refactoring I also remove all the virtual functions from the class hierarchy and try to reuse techniques used in the ELF linker (see `lld/ELF/InputSections.h`).

Differential Revision: https://reviews.llvm.org/D102546 </cut>

Results regressed to (for first_bad == 5a9b25e15b4f9707245bb9d1ab56ab99b725505a) # reset_artifacts: -10 # build_abe binutils: -9 # build_abe stage1 -- --set gcc_override_configure=--disable-libsanitizer: -8 # build_abe linux: -7 # build_abe glibc: -6 # build_abe stage2 -- --set gcc_override_configure=--disable-libsanitizer: -5 # build_llvm true: -3 # true: 0 # benchmark -O3_LTO -- artifacts/build-5a9b25e15b4f9707245bb9d1ab56ab99b725505a/results_id: 1 # 462.libquantum,libquantum_base.default regressed by 113

from (for last_good == ed9d70781bbd1dc858b9f6913729fcc49a1e4bdf) # reset_artifacts: -10 # build_abe binutils: -9 # build_abe stage1 -- --set gcc_override_configure=--disable-libsanitizer: -8 # build_abe linux: -7 # build_abe glibc: -6 # build_abe stage2 -- --set gcc_override_configure=--disable-libsanitizer: -5 # build_llvm true: -3 # true: 0 # benchmark -O3_LTO -- artifacts/build-ed9d70781bbd1dc858b9f6913729fcc49a1e4bdf/results_id: 1

Artifacts of last_good build: https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... Results ID of last_good: tx1_64/tcwg_bmk_llvm_tx1/bisect-llvm-master-aarch64-spec2k6-O3_LTO/2499 Artifacts of first_bad build: https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... Results ID of first_bad: tx1_64/tcwg_bmk_llvm_tx1/bisect-llvm-master-aarch64-spec2k6-O3_LTO/2539 Build top page/logs: https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a...

Configuration details:

Reproduce builds: <cut> mkdir investigate-llvm-5a9b25e15b4f9707245bb9d1ab56ab99b725505a cd investigate-llvm-5a9b25e15b4f9707245bb9d1ab56ab99b725505a

git clone https://git.linaro.org/toolchain/jenkins-scripts

mkdir -p artifacts/manifests curl -o artifacts/manifests/build-baseline.sh https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... --fail curl -o artifacts/manifests/build-parameters.sh https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... --fail curl -o artifacts/test.sh https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... --fail chmod +x artifacts/test.sh

# Reproduce the baseline build (build all pre-requisites) ./jenkins-scripts/tcwg_bmk-build.sh @@ artifacts/manifests/build-baseline.sh

# Save baseline build state (which is then restored in artifacts/test.sh) mkdir -p ./bisect rsync -a --del --delete-excluded --exclude /bisect/ --exclude /artifacts/ --exclude /llvm/ ./ ./bisect/baseline/

cd llvm

# Reproduce first_bad build git checkout --detach 5a9b25e15b4f9707245bb9d1ab56ab99b725505a ../artifacts/test.sh

# Reproduce last_good build git checkout --detach ed9d70781bbd1dc858b9f6913729fcc49a1e4bdf ../artifacts/test.sh

cd .. </cut>

History of pending regressions and results: https://git.linaro.org/toolchain/ci/base-artifacts.git/log/?h=linaro-local/c...

Artifacts: https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a... Build log: https://ci.linaro.org/job/tcwg_bmk_ci_llvm-bisect-tcwg_bmk_tx1-llvm-master-a...

Full commit (up to 1000 lines): <cut> commit 5a9b25e15b4f9707245bb9d1ab56ab99b725505a Author: Sam Clegg sbc@chromium.org Date: Fri May 14 16:25:04 2021 -0700

[lld][WebAssembly] Refactor input chunk class hierarchy. NFC

The main motivation for this refactor is to remove the subclass relationship between the InputSegment and MergeInputSegment and SyntenticMergedInputSegment so that we can use the merging classes for debug sections which are not data segments.

In the process of refactoring I also remove all the virtual functions from the class hierarchy and try to reuse techniques used in the ELF linker (see `lld/ELF/InputSections.h`).

Differential Revision: https://reviews.llvm.org/D102546 --- lld/wasm/Driver.cpp | 4 +- lld/wasm/InputChunks.cpp | 99 ++++++++++------ lld/wasm/InputChunks.h | 254 +++++++++++++++++++---------------------- lld/wasm/InputFiles.cpp | 6 +- lld/wasm/InputFiles.h | 2 +- lld/wasm/OutputSections.cpp | 2 +- lld/wasm/OutputSegment.cpp | 27 +++-- lld/wasm/OutputSegment.h | 4 +- lld/wasm/SymbolTable.cpp | 2 +- lld/wasm/SymbolTable.h | 3 +- lld/wasm/Symbols.cpp | 6 +- lld/wasm/Symbols.h | 6 +- lld/wasm/SyntheticSections.cpp | 2 +- lld/wasm/Writer.cpp | 10 +- 14 files changed, 214 insertions(+), 213 deletions(-)

diff --git a/lld/wasm/Driver.cpp b/lld/wasm/Driver.cpp index a0ba04c4fa1e..6fec9d5b2278 100644 --- a/lld/wasm/Driver.cpp +++ b/lld/wasm/Driver.cpp @@ -800,8 +800,8 @@ static void splitSections() { // before calling finalizeContents(). LLVM_DEBUG(llvm::dbgs() << "splitSections\n"); parallelForEach(symtab->objectFiles, [](ObjFile *file) { - for (InputSegment *seg : file->segments) { - if (auto *s = dyn_cast<MergeInputSegment>(seg)) + for (InputChunk *seg : file->segments) { + if (auto *s = dyn_cast<MergeInputChunk>(seg)) s->splitIntoPieces(); } }); diff --git a/lld/wasm/InputChunks.cpp b/lld/wasm/InputChunks.cpp index f73e3362ae8e..b7d1aa25e9c8 100644 --- a/lld/wasm/InputChunks.cpp +++ b/lld/wasm/InputChunks.cpp @@ -57,8 +57,37 @@ StringRef InputChunk::getComdatName() const { return file->getWasmObj()->linkingData().Comdats[index]; }

+uint32_t InputChunk::getSize() const { + if (const auto *ms = dyn_cast<SyntheticMergedChunk>(this)) + return ms->builder.getSize(); + + if (const auto *f = dyn_cast<InputFunction>(this)) { + if (config->compressRelocations && f->file) { + return f->getCompressedSize(); + } + } + + return data().size(); +} + +uint32_t InputChunk::getInputSize() const { + if (const auto *f = dyn_cast<InputFunction>(this)) + return f->function->Size; + return getSize(); +} + // Copy this input chunk to an mmap'ed output file and apply relocations. void InputChunk::writeTo(uint8_t *buf) const { + if (const auto *f = dyn_cast<InputFunction>(this)) { + if (file && config->compressRelocations) + return f->writeCompressed(buf); + } else if (const auto *ms = dyn_cast<SyntheticMergedChunk>(this)) { + ms->builder.write(buf + outSecOff); + // Apply relocations + ms->relocate(buf + outSecOff); + return; + } + // Copy contents memcpy(buf + outSecOff, data().data(), data().size());

@@ -73,7 +102,7 @@ void InputChunk::relocate(uint8_t *buf) const { LLVM_DEBUG(dbgs() << "applying relocations: " << toString(this) << " count=" << relocations.size() << "\n"); int32_t inputSectionOffset = getInputSectionOffset(); - auto tombstone = getTombstone(); + uint64_t tombstone = getTombstone();

for (const WasmRelocation &rel : relocations) { uint8_t *loc = buf + rel.Offset - inputSectionOffset; @@ -149,6 +178,14 @@ void InputChunk::writeRelocations(raw_ostream &os) const { } }

+uint64_t InputChunk::getTombstone() const { + if (const auto *s = dyn_cast<InputSection>(this)) { + return s->tombstoneValue; + } + + return 0; +} + void InputFunction::setFunctionIndex(uint32_t index) { LLVM_DEBUG(dbgs() << "InputFunction::setFunctionIndex: " << getName() << " -> " << index << "\n"); @@ -235,7 +272,7 @@ void InputFunction::calculateSize() { uint32_t start = getInputSectionOffset(); uint32_t end = start + function->Size;

- auto tombstone = getTombstone(); + uint64_t tombstone = getTombstone();

uint32_t lastRelocEnd = start + functionSizeLength; for (const WasmRelocation &rel : relocations) { @@ -259,10 +296,7 @@ void InputFunction::calculateSize() {

// Override the default writeTo method so that we can (optionally) write the // compressed version of the function. -void InputFunction::writeTo(uint8_t *buf) const { - if (!file || !config->compressRelocations) - return InputChunk::writeTo(buf); - +void InputFunction::writeCompressed(uint8_t *buf) const { buf += outSecOff; uint8_t *orig = buf; (void)orig; @@ -270,7 +304,7 @@ void InputFunction::writeTo(uint8_t *buf) const { const uint8_t *secStart = file->codeSection->Content.data(); const uint8_t *funcStart = secStart + getInputSectionOffset(); const uint8_t *end = funcStart + function->Size; - auto tombstone = getTombstone(); + uint64_t tombstone = getTombstone(); uint32_t count; decodeULEB128(funcStart, &count); funcStart += count; @@ -294,26 +328,30 @@ void InputFunction::writeTo(uint8_t *buf) const { LLVM_DEBUG(dbgs() << " total: " << (buf + chunkSize - orig) << "\n"); }

-uint64_t InputSegment::getOffset(uint64_t offset) const { - if (const MergeInputSegment *ms = dyn_cast<MergeInputSegment>(this)) { - LLVM_DEBUG(dbgs() << "getOffset(merged): " << getName() << "\n"); +uint64_t InputChunk::getOffset(uint64_t offset) const { + return outSecOff + offset; +} + +uint64_t InputChunk::getSegmentOffset(uint64_t offset) const { + if (const auto *ms = dyn_cast<MergeInputChunk>(this)) { + LLVM_DEBUG(dbgs() << "getSegmentOffset(merged): " << getName() << "\n"); LLVM_DEBUG(dbgs() << "offset: " << offset << "\n"); LLVM_DEBUG(dbgs() << "parentOffset: " << ms->getParentOffset(offset) << "\n"); assert(ms->parent); - return ms->parent->getOffset(ms->getParentOffset(offset)); + return ms->parent->getSegmentOffset(ms->getParentOffset(offset)); } return outputSegmentOffset + offset; }

-uint64_t InputSegment::getVA(uint64_t offset) const { - return (outputSeg ? outputSeg->startVA : 0) + getOffset(offset); +uint64_t InputChunk::getVA(uint64_t offset) const { + return (outputSeg ? outputSeg->startVA : 0) + getSegmentOffset(offset); }

// Generate code to apply relocations to the data section at runtime. // This is only called when generating shared libaries (PIC) where address are // not known at static link time. -void InputSegment::generateRelocationCode(raw_ostream &os) const { +void InputChunk::generateRelocationCode(raw_ostream &os) const { LLVM_DEBUG(dbgs() << "generating runtime relocations: " << getName() << " count=" << relocations.size() << "\n");

@@ -324,7 +362,7 @@ void InputSegment::generateRelocationCode(raw_ostream &os) const { ? WASM_OPCODE_I64_ADD : WASM_OPCODE_I32_ADD;

- auto tombstone = getTombstone(); + uint64_t tombstone = getTombstone(); // TODO(sbc): Encode the relocations in the data section and write a loop // here to apply them. for (const WasmRelocation &rel : relocations) { @@ -382,7 +420,7 @@ void InputSegment::generateRelocationCode(raw_ostream &os) const {

// Split WASM_SEG_FLAG_STRINGS section. Such a section is a sequence of // null-terminated strings. -void MergeInputSegment::splitStrings(ArrayRef<uint8_t> data) { +void MergeInputChunk::splitStrings(ArrayRef<uint8_t> data) { LLVM_DEBUG(llvm::dbgs() << "splitStrings\n"); size_t off = 0; StringRef s = toStringRef(data); @@ -405,15 +443,15 @@ void MergeInputSegment::splitStrings(ArrayRef<uint8_t> data) { // // Note that this function is called from parallelForEach. This must be // thread-safe (i.e. no memory allocation from the pools). -void MergeInputSegment::splitIntoPieces() { +void MergeInputChunk::splitIntoPieces() { assert(pieces.empty()); // As of now we only support WASM_SEG_FLAG_STRINGS but in the future we // could add other types of splitting (see ELF's splitIntoPieces). - assert(segment->Data.LinkingFlags & WASM_SEG_FLAG_STRINGS); + assert(flags & WASM_SEG_FLAG_STRINGS); splitStrings(data()); }

-SegmentPiece *MergeInputSegment::getSegmentPiece(uint64_t offset) { +SectionPiece *MergeInputChunk::getSectionPiece(uint64_t offset) { if (this->data().size() <= offset) fatal(toString(this) + ": offset is outside the section");

@@ -421,36 +459,25 @@ SegmentPiece *MergeInputSegment::getSegmentPiece(uint64_t offset) { // In that case we need to do a binary search of the original section piece // vector. auto it = partition_point( - pieces, [=](SegmentPiece p) { return p.inputOff <= offset; }); + pieces, [=](SectionPiece p) { return p.inputOff <= offset; }); return &it[-1]; }

// Returns the offset in an output section for a given input offset. // Because contents of a mergeable section is not contiguous in output, // it is not just an addition to a base output offset. -uint64_t MergeInputSegment::getParentOffset(uint64_t offset) const { +uint64_t MergeInputChunk::getParentOffset(uint64_t offset) const { // If Offset is not at beginning of a section piece, it is not in the map. // In that case we need to search from the original section piece vector. - const SegmentPiece *piece = getSegmentPiece(offset); + const SectionPiece *piece = getSectionPiece(offset); uint64_t addend = offset - piece->inputOff; return piece->outputOff + addend; }

-uint32_t SyntheticMergedDataSegment::getSize() const { - return builder.getSize(); -} - -void SyntheticMergedDataSegment::writeTo(uint8_t *buf) const { - builder.write(buf + outSecOff); - - // Apply relocations - relocate(buf + outSecOff); -} - -void SyntheticMergedDataSegment::finalizeContents() { +void SyntheticMergedChunk::finalizeContents() { // Add all string pieces to the string table builder to create section // contents. - for (MergeInputSegment *sec : segments) + for (MergeInputChunk *sec : chunks) for (size_t i = 0, e = sec->pieces.size(); i != e; ++i) if (sec->pieces[i].live) builder.add(sec->getData(i)); @@ -461,7 +488,7 @@ void SyntheticMergedDataSegment::finalizeContents() { // finalize() fixed tail-optimized strings, so we can now get // offsets of strings. Get an offset for each string and save it // to a corresponding SectionPiece for easy access. - for (MergeInputSegment *sec : segments) + for (MergeInputChunk *sec : chunks) for (size_t i = 0, e = sec->pieces.size(); i != e; ++i) if (sec->pieces[i].live) sec->pieces[i].outputOff = builder.getOffset(sec->getData(i)); diff --git a/lld/wasm/InputChunks.h b/lld/wasm/InputChunks.h index 5bf96df695d5..47ed7ccaadcb 100644 --- a/lld/wasm/InputChunks.h +++ b/lld/wasm/InputChunks.h @@ -40,40 +40,68 @@ public: enum Kind { DataSegment, Merge, - MergedSegment, + MergedChunk, Function, SyntheticFunction, Section };

- Kind kind() const { return sectionKind; } + StringRef name; + StringRef debugName; + + StringRef getName() const { return name; } + StringRef getDebugName() const { return debugName; } + Kind kind() const { return (Kind)sectionKind; }

- virtual uint32_t getSize() const { return data().size(); } - virtual uint32_t getInputSize() const { return getSize(); }; + uint32_t getSize() const; + uint32_t getInputSize() const;

- virtual void writeTo(uint8_t *buf) const; + void writeTo(uint8_t *buf) const; void relocate(uint8_t *buf) const;

ArrayRef<WasmRelocation> getRelocations() const { return relocations; } void setRelocations(ArrayRef<WasmRelocation> rs) { relocations = rs; }

- uint64_t getOffset(uint64_t offset) const { return outSecOff + offset; } - virtual StringRef getName() const = 0; - virtual StringRef getDebugName() const = 0; - virtual uint32_t getComdat() const = 0; + // Translate an offset in the input section to an offset in the output + // section. + uint64_t getOffset(uint64_t offset) const; + // For data segments, translate and offset into the input segment into + // an offset into the output segment + uint64_t getSegmentOffset(uint64_t offset) const; + uint64_t getVA(uint64_t offset = 0) const; + + uint32_t getComdat() const { return comdat; } StringRef getComdatName() const; - virtual uint32_t getInputSectionOffset() const = 0; + uint32_t getInputSectionOffset() const { return inputSectionOffset; }

size_t getNumRelocations() const { return relocations.size(); } void writeRelocations(llvm::raw_ostream &os) const; + void generateRelocationCode(raw_ostream &os) const; + + bool isTLS() const { + // Older object files don't include WASM_SEG_FLAG_TLS and instead + // relied on the naming convention. + return flags & llvm::wasm::WASM_SEG_FLAG_TLS || name.startswith(".tdata") || + name.startswith(".tbss"); + }

ObjFile *file; OutputSection *outputSec = nullptr; + uint32_t comdat = UINT32_MAX; + uint32_t inputSectionOffset = 0; + uint32_t alignment; + uint32_t flags; + + // Only applies to data segments. + uint32_t outputSegmentOffset = 0; + const OutputSegment *outputSeg = nullptr;

// After assignAddresses is called, this represents the offset from // the beginning of the output section this chunk was assigned to. int32_t outSecOff = 0;

+ uint8_t sectionKind : 3; + // Signals that the section is part of the output. The garbage collector, // and COMDAT handling can set a sections' Live bit. // If GC is disabled, all sections start out as live by default. @@ -83,14 +111,15 @@ public: unsigned discarded : 1;

protected: - InputChunk(ObjFile *f, Kind k) - : file(f), live(!config->gcSections), discarded(false), sectionKind(k) {} - virtual ~InputChunk() = default; - virtual ArrayRef<uint8_t> data() const = 0; - virtual uint64_t getTombstone() const { return 0; } + InputChunk(ObjFile *f, Kind k, StringRef name, uint32_t alignment = 0, + uint32_t flags = 0) + : name(name), file(f), alignment(alignment), flags(flags), sectionKind(k), + live(!config->gcSections), discarded(false) {} + ArrayRef<uint8_t> data() const { return rawData; } + uint64_t getTombstone() const;

ArrayRef<WasmRelocation> relocations; - Kind sectionKind; + ArrayRef<uint8_t> rawData; };

// Represents a WebAssembly data segment which can be included as part of @@ -104,64 +133,34 @@ protected: class InputSegment : public InputChunk { public: InputSegment(const WasmSegment *seg, ObjFile *f) - : InputChunk(f, InputChunk::DataSegment), segment(seg) { - alignment = segment->Data.Alignment; - flags = segment->Data.LinkingFlags; - } - - InputSegment(uint32_t alignment, uint32_t flags) - : InputChunk(nullptr, InputChunk::DataSegment), alignment(alignment), - flags(flags) {} - - static bool classof(const InputChunk *c) { - return c->kind() == DataSegment || c->kind() == Merge || - c->kind() == MergedSegment; - } - - void generateRelocationCode(raw_ostream &os) const; - - StringRef getName() const override { return segment->Data.Name; } - StringRef getDebugName() const override { return StringRef(); } - uint32_t getComdat() const override { return segment->Data.Comdat; } - uint32_t getInputSectionOffset() const override { - return segment->SectionOffset; + : InputChunk(f, InputChunk::DataSegment, seg->Data.Name, + seg->Data.Alignment, seg->Data.LinkingFlags), + segment(seg) { + rawData = segment->Data.Content; + comdat = segment->Data.Comdat; + inputSectionOffset = segment->SectionOffset; }

- // Translate an offset in the input section to an offset in the output - // section. - uint64_t getOffset(uint64_t offset) const; - - uint64_t getVA(uint64_t offset = 0) const; - - bool isTLS() const { - // Older object files don't include WASM_SEG_FLAG_TLS and instead - // relied on the naming convention. - return flags & llvm::wasm::WASM_SEG_FLAG_TLS || - getName().startswith(".tdata") || getName().startswith(".tbss"); - } + InputSegment(StringRef name, uint32_t alignment, uint32_t flags) + : InputChunk(nullptr, InputChunk::DataSegment, name, alignment, flags) {}

- const OutputSegment *outputSeg = nullptr; - uint32_t outputSegmentOffset = 0; - uint32_t alignment = 0; - uint32_t flags = 0; + static bool classof(const InputChunk *c) { return c->kind() == DataSegment; }

protected: - ArrayRef<uint8_t> data() const override { return segment->Data.Content; } - const WasmSegment *segment = nullptr; };

-class SyntheticMergedDataSegment; +class SyntheticMergedChunk;

// Merge segment handling copied from lld/ELF/InputSection.h. Keep in sync // where possible.

-// SegmentPiece represents a piece of splittable segment contents. +// SectionPiece represents a piece of splittable segment contents. // We allocate a lot of these and binary search on them. This means that they // have to be as compact as possible, which is why we don't store the size (can // be found by looking at the next one). -struct SegmentPiece { - SegmentPiece(size_t off, uint32_t hash, bool live) +struct SectionPiece { + SectionPiece(size_t off, uint32_t hash, bool live) : inputOff(off), live(live || !config->gcSections), hash(hash >> 1) {}

uint32_t inputOff; @@ -170,13 +169,17 @@ struct SegmentPiece { uint64_t outputOff = 0; };

-static_assert(sizeof(SegmentPiece) == 16, "SectionPiece is too big"); +static_assert(sizeof(SectionPiece) == 16, "SectionPiece is too big");

// This corresponds segments marked as WASM_SEG_FLAG_STRINGS. -class MergeInputSegment : public InputSegment { +class MergeInputChunk : public InputChunk { public: - MergeInputSegment(const WasmSegment *seg, ObjFile *f) : InputSegment(seg, f) { - sectionKind = Merge; + MergeInputChunk(const WasmSegment *seg, ObjFile *f) + : InputChunk(f, Merge, seg->Data.Name, seg->Data.Alignment, + seg->Data.LinkingFlags) { + rawData = seg->Data.Content; + comdat = seg->Data.Comdat; + inputSectionOffset = seg->SectionOffset; }

static bool classof(const InputChunk *s) { return s->kind() == Merge; } @@ -188,7 +191,7 @@ public:

// Splittable sections are handled as a sequence of data // rather than a single large blob of data. - std::vector<SegmentPiece> pieces; + std::vector<SectionPiece> pieces;

// Returns I'th piece's data. This function is very hot when // string merging is enabled, so we want to inline. @@ -201,52 +204,43 @@ public: }

// Returns the SectionPiece at a given input section offset. - SegmentPiece *getSegmentPiece(uint64_t offset); - const SegmentPiece *getSegmentPiece(uint64_t offset) const { - return const_cast<MergeInputSegment *>(this)->getSegmentPiece(offset); + SectionPiece *getSectionPiece(uint64_t offset); + const SectionPiece *getSectionPiece(uint64_t offset) const { + return const_cast<MergeInputChunk *>(this)->getSectionPiece(offset); }

- SyntheticMergedDataSegment *parent = nullptr; + SyntheticMergedChunk *parent = nullptr;

private: void splitStrings(ArrayRef<uint8_t> a); };

-// SyntheticMergedDataSegment is a class that allows us to put mergeable +// SyntheticMergedChunk is a class that allows us to put mergeable // sections with different attributes in a single output sections. To do that we -// put them into SyntheticMergedDataSegment synthetic input sections which are +// put them into SyntheticMergedChunk synthetic input sections which are // attached to regular output sections. -class SyntheticMergedDataSegment : public InputSegment { +class SyntheticMergedChunk : public InputChunk { public: - SyntheticMergedDataSegment(StringRef name, uint32_t alignment, uint32_t flags) - : InputSegment(alignment, flags), name(name), - builder(llvm::StringTableBuilder::RAW, 1ULL << alignment) { - sectionKind = InputChunk::MergedSegment; - } + SyntheticMergedChunk(StringRef name, uint32_t alignment, uint32_t flags) + : InputChunk(nullptr, InputChunk::MergedChunk, name, alignment, flags), + builder(llvm::StringTableBuilder::RAW, 1ULL << alignment) {}

static bool classof(const InputChunk *c) { - return c->kind() == InputChunk::MergedSegment; + return c->kind() == InputChunk::MergedChunk; }

- uint32_t getSize() const override; - - StringRef getName() const override { return name; } - - uint32_t getComdat() const override { return segments[0]->getComdat(); } - - void writeTo(uint8_t *buf) const override; - - void addMergeSegment(MergeInputSegment *ms) { + void addMergeChunk(MergeInputChunk *ms) { + comdat = ms->getComdat(); ms->parent = this; - segments.push_back(ms); + chunks.push_back(ms); }

void finalizeContents();

-protected: - std::vector<MergeInputSegment *> segments; - StringRef name; llvm::StringTableBuilder builder; + +protected: + std::vector<MergeInputChunk *> chunks; };

// Represents a single wasm function within and input file. These are @@ -254,44 +248,39 @@ protected: class InputFunction : public InputChunk { public: InputFunction(const WasmSignature &s, const WasmFunction *func, ObjFile *f) - : InputChunk(f, InputChunk::Function), signature(s), function(func), - exportName(func && func->ExportName.hasValue() - ? (*func->ExportName).str() - : llvm::Optionalstd::string()) {} + : InputChunk(f, InputChunk::Function, func->SymbolName), signature(s), + function(func), exportName(func && func->ExportName.hasValue() + ? (*func->ExportName).str() + : llvm::Optionalstd::string()) { + inputSectionOffset = function->CodeSectionOffset; + rawData = + file->codeSection->Content.slice(inputSectionOffset, function->Size); + debugName = function->DebugName; + comdat = function->Comdat; + } + + InputFunction(StringRef name, const WasmSignature &s) + : InputChunk(nullptr, InputChunk::Function, name), signature(s) {}

static bool classof(const InputChunk *c) { return c->kind() == InputChunk::Function || c->kind() == InputChunk::SyntheticFunction; }

- void writeTo(uint8_t *buf) const override; - StringRef getName() const override { return function->SymbolName; } - StringRef getDebugName() const override { return function->DebugName; } llvm::Optional<StringRef> getExportName() const { return exportName.hasValue() ? llvm::Optional<StringRef>(*exportName) : llvm::Optional<StringRef>(); } void setExportName(std::string exportName) { this->exportName = exportName; } - uint32_t getComdat() const override { return function->Comdat; } uint32_t getFunctionInputOffset() const { return getInputSectionOffset(); } uint32_t getFunctionCodeOffset() const { return function->CodeOffset; } - uint32_t getSize() const override { - if (config->compressRelocations && file) { - assert(compressedSize); - return compressedSize; - } - return data().size(); - } - uint32_t getInputSize() const override { return function->Size; } uint32_t getFunctionIndex() const { return functionIndex.getValue(); } bool hasFunctionIndex() const { return functionIndex.hasValue(); } void setFunctionIndex(uint32_t index); - uint32_t getInputSectionOffset() const override { - return function->CodeSectionOffset; - } uint32_t getTableIndex() const { return tableIndex.getValue(); } bool hasTableIndex() const { return tableIndex.hasValue(); } void setTableIndex(uint32_t index); + void writeCompressed(uint8_t *buf) const;

// The size of a given input function can depend on the values of the // LEB relocations within it. This finalizeContents method is called after @@ -301,14 +290,14 @@ public:

const WasmSignature &signature;

-protected: - ArrayRef<uint8_t> data() const override { - assert(!config->compressRelocations); - return file->codeSection->Content.slice(getInputSectionOffset(), - function->Size); + uint32_t getCompressedSize() const { + assert(compressedSize); + return compressedSize; }

const WasmFunction *function; + +protected: llvm::Optionalstd::string exportName; llvm::Optional<uint32_t> functionIndex; llvm::Optional<uint32_t> tableIndex; @@ -320,51 +309,38 @@ class SyntheticFunction : public InputFunction { public: SyntheticFunction(const WasmSignature &s, StringRef name, StringRef debugName = {}) - : InputFunction(s, nullptr, nullptr), name(name), debugName(debugName) { + : InputFunction(name, s) { sectionKind = InputChunk::SyntheticFunction; + this->debugName = debugName; }

static bool classof(const InputChunk *c) { return c->kind() == InputChunk::SyntheticFunction; }

- StringRef getName() const override { return name; } - StringRef getDebugName() const override { return debugName; } - uint32_t getComdat() const override { return UINT32_MAX; } - - void setBody(ArrayRef<uint8_t> body_) { body = body_; } - -protected: - ArrayRef<uint8_t> data() const override { return body; } - - StringRef name; - StringRef debugName; - ArrayRef<uint8_t> body; + void setBody(ArrayRef<uint8_t> body) { rawData = body; } };

// Represents a single Wasm Section within an input file. class InputSection : public InputChunk { public: InputSection(const WasmSection &s, ObjFile *f) - : InputChunk(f, InputChunk::Section), section(s), tombstoneValue(getTombstoneForSection(s.Name)) { + : InputChunk(f, InputChunk::Section, s.Name), + tombstoneValue(getTombstoneForSection(s.Name)), section(s) { assert(section.Type == llvm::wasm::WASM_SEC_CUSTOM); + comdat = section.Comdat; + rawData = section.Content; }

- StringRef getName() const override { return section.Name; } - StringRef getDebugName() const override { return StringRef(); } - uint32_t getComdat() const override { return section.Comdat; } + static bool classof(const InputChunk *c) { + return c->kind() == InputChunk::Section; + }

-protected: - ArrayRef<uint8_t> data() const override { return section.Content; } + const uint64_t tombstoneValue;

- // Offset within the input section. This is only zero since this chunk - // type represents an entire input section, not part of one. - uint32_t getInputSectionOffset() const override { return 0; } - uint64_t getTombstone() const override { return tombstoneValue; } +protected: static uint64_t getTombstoneForSection(StringRef name); - const WasmSection &section; - const uint64_t tombstoneValue; };

} // namespace wasm diff --git a/lld/wasm/InputFiles.cpp b/lld/wasm/InputFiles.cpp index 2cb5d59286e3..c65b05109c3d 100644 --- a/lld/wasm/InputFiles.cpp +++ b/lld/wasm/InputFiles.cpp @@ -464,9 +464,9 @@ void ObjFile::parse(bool ignoreComdats) {

// Populate `Segments`. for (const WasmSegment &s : wasmObj->dataSegments()) { - InputSegment *seg; + InputChunk *seg; if (shouldMerge(s)) { - seg = make<MergeInputSegment>(&s, this); + seg = make<MergeInputChunk>(&s, this); } else seg = make<InputSegment>(&s, this); seg->discarded = isExcludedByComdat(seg); @@ -568,7 +568,7 @@ Symbol *ObjFile::createDefined(const WasmSymbol &sym) { return symtab->addDefinedFunction(name, flags, this, func); } case WASM_SYMBOL_TYPE_DATA: { - InputSegment *seg = segments[sym.Info.DataRef.Segment]; + InputChunk *seg = segments[sym.Info.DataRef.Segment]; auto offset = sym.Info.DataRef.Offset; auto size = sym.Info.DataRef.Size; if (sym.isBindingLocal()) diff --git a/lld/wasm/InputFiles.h b/lld/wasm/InputFiles.h index dab44cd6830c..b720b889adcc 100644 --- a/lld/wasm/InputFiles.h +++ b/lld/wasm/InputFiles.h @@ -136,7 +136,7 @@ public: std::vector<uint32_t> tableEntries; std::vector<uint32_t> tableEntriesRel; std::vector<bool> keptComdats; - std::vector<InputSegment *> segments; + std::vector<InputChunk *> segments; std::vector<InputFunction *> functions; std::vector<InputGlobal *> globals; std::vector<InputEvent *> events; diff --git a/lld/wasm/OutputSections.cpp b/lld/wasm/OutputSections.cpp index 6d91708d326a..780c8e4b2776 100644 --- a/lld/wasm/OutputSections.cpp +++ b/lld/wasm/OutputSections.cpp @@ -178,7 +178,7 @@ void DataSection::finalizeContents() { log("Data segment: size=" + Twine(segment->size) + ", startVA=" + Twine::utohexstr(segment->startVA) + ", name=" + segment->name);

- for (InputSegment *inputSeg : segment->inputSegments) { + for (InputChunk *inputSeg : segment->inputSegments) { inputSeg->outputSec = this; inputSeg->outSecOff = segment->sectionOffset + segment->header.size() + inputSeg->outputSegmentOffset; diff --git a/lld/wasm/OutputSegment.cpp b/lld/wasm/OutputSegment.cpp index 899308e87c94..bf3e40c968a0 100644 --- a/lld/wasm/OutputSegment.cpp +++ b/lld/wasm/OutputSegment.cpp @@ -19,7 +19,7 @@ namespace lld {

namespace wasm {

-void OutputSegment::addInputSegment(InputSegment *inSeg) { +void OutputSegment::addInputSegment(InputChunk *inSeg) { alignment = std::max(alignment, inSeg->alignment); inputSegments.push_back(inSeg); size = llvm::alignTo(size, 1ULL << inSeg->alignment); @@ -33,16 +33,16 @@ void OutputSegment::addInputSegment(InputSegment *inSeg) {

// This function scans over the input segments. // -// It removes MergeInputSegments from the input section array and adds +// It removes MergeInputChunks from the input section array and adds // new synthetic sections at the location of the first input section // that it replaces. It then finalizes each synthetic section in order // to compute an output offset for each piece of each input section. void OutputSegment::finalizeInputSegments() { LLVM_DEBUG(llvm::dbgs() << "finalizeInputSegments: " << name << "\n"); - std::vector<SyntheticMergedDataSegment *> mergedSegments; - std::vector<InputSegment *> newSegments; - for (InputSegment *s : inputSegments) { - MergeInputSegment *ms = dyn_cast<MergeInputSegment>(s); + std::vector<SyntheticMergedChunk *> mergedSegments; + std::vector<InputChunk *> newSegments; + for (InputChunk *s : inputSegments) { + MergeInputChunk *ms = dyn_cast<MergeInputChunk>(s); if (!ms) { newSegments.push_back(s); continue; @@ -51,15 +51,14 @@ void OutputSegment::finalizeInputSegments() { // A segment should not make it here unless its alive assert(ms->live);

- auto i = - llvm::find_if(mergedSegments, [=](SyntheticMergedDataSegment *seg) { - return seg->flags == ms->flags && seg->alignment == ms->alignment; - }); + auto i = llvm::find_if(mergedSegments, [=](SyntheticMergedChunk *seg) { + return seg->flags == ms->flags && seg->alignment == ms->alignment; + }); if (i == mergedSegments.end()) { LLVM_DEBUG(llvm::dbgs() << "new merge section: " << name << " alignment=" << ms->alignment << "\n"); - SyntheticMergedDataSegment *syn = - make<SyntheticMergedDataSegment>(name, ms->alignment, ms->flags); + SyntheticMergedChunk *syn = + make<SyntheticMergedChunk>(name, ms->alignment, ms->flags); syn->outputSeg = this; mergedSegments.push_back(syn); i = std::prev(mergedSegments.end()); @@ -67,7 +66,7 @@ void OutputSegment::finalizeInputSegments() { } else { LLVM_DEBUG(llvm::dbgs() << "adding to merge section: " << name << "\n"); } - (*i)->addMergeSegment(ms); + (*i)->addMergeChunk(ms); }

for (auto *ms : mergedSegments) @@ -75,7 +74,7 @@ void OutputSegment::finalizeInputSegments() {

inputSegments = newSegments; size = 0; - for (InputSegment *seg : inputSegments) { + for (InputChunk *seg : inputSegments) { size = llvm::alignTo(size, 1ULL << seg->alignment); LLVM_DEBUG(llvm::dbgs() << "outputSegmentOffset set: " << seg->getName() << " -> " << size << "\n"); diff --git a/lld/wasm/OutputSegment.h b/lld/wasm/OutputSegment.h index 10e28a4a8836..1d5cf9145241 100644 --- a/lld/wasm/OutputSegment.h +++ b/lld/wasm/OutputSegment.h @@ -22,7 +22,7 @@ class OutputSegment { public: OutputSegment(StringRef n) : name(n) {}

- void addInputSegment(InputSegment *inSeg); + void addInputSegment(InputChunk *inSeg); void finalizeInputSegments();

bool isTLS() const { return name == ".tdata"; } @@ -35,7 +35,7 @@ public: uint32_t sectionOffset = 0; uint32_t alignment = 0; uint64_t startVA = 0; - std::vector<InputSegment *> inputSegments; + std::vector<InputChunk *> inputSegments;

// Sum of the size of the all the input segments uint32_t size = 0; diff --git a/lld/wasm/SymbolTable.cpp b/lld/wasm/SymbolTable.cpp index 56438c7b5d99..7e794a1bd431 100644 --- a/lld/wasm/SymbolTable.cpp +++ b/lld/wasm/SymbolTable.cpp @@ -365,7 +365,7 @@ Symbol *SymbolTable::addDefinedFunction(StringRef name, uint32_t flags, }

Symbol *SymbolTable::addDefinedData(StringRef name, uint32_t flags, - InputFile *file, InputSegment *segment, + InputFile *file, InputChunk *segment, uint64_t address, uint64_t size) { LLVM_DEBUG(dbgs() << "addDefinedData:" << name << " addr:" << address << "\n"); diff --git a/lld/wasm/SymbolTable.h b/lld/wasm/SymbolTable.h index 36c776be30f5..bee18ce9c6bd 100644 --- a/lld/wasm/SymbolTable.h +++ b/lld/wasm/SymbolTable.h @@ -54,8 +54,7 @@ public: Symbol *addDefinedFunction(StringRef name, uint32_t flags, InputFile *file, InputFunction *function); Symbol *addDefinedData(StringRef name, uint32_t flags, InputFile *file, - InputSegment *segment, uint64_t address, - uint64_t size); + InputChunk *segment, uint64_t address, uint64_t size); Symbol *addDefinedGlobal(StringRef name, uint32_t flags, InputFile *file, InputGlobal *g); Symbol *addDefinedEvent(StringRef name, uint32_t flags, InputFile *file, diff --git a/lld/wasm/Symbols.cpp b/lld/wasm/Symbols.cpp index 00fe754e3eaa..24b7e10dc559 100644 --- a/lld/wasm/Symbols.cpp +++ b/lld/wasm/Symbols.cpp @@ -160,8 +160,8 @@ void Symbol::markLive() { // (splittable) sections, each piece of data has independent liveness bit. // So we explicitly tell it which offset is in use. if (auto *d = dyn_cast<DefinedData>(this)) { - if (auto *ms = dyn_cast<MergeInputSegment>(c)) { - ms->getSegmentPiece(d->value)->live = true; + if (auto *ms = dyn_cast<MergeInputChunk>(c)) { + ms->getSectionPiece(d->value)->live = true; } } c->live = true; @@ -301,7 +301,7 @@ void DefinedData::setVA(uint64_t value_) {

uint64_t DefinedData::getOutputSegmentOffset() const { LLVM_DEBUG(dbgs() << "getOutputSegmentOffset: " << getName() << "\n"); - return segment->getOffset(value); + return segment->getSegmentOffset(value); }

uint64_t DefinedData::getOutputSegmentIndex() const { diff --git a/lld/wasm/Symbols.h b/lld/wasm/Symbols.h index 8bc3be0d2247..ff6eb0ae7ccb 100644 --- a/lld/wasm/Symbols.h +++ b/lld/wasm/Symbols.h @@ -278,8 +278,8 @@ protected: class DefinedData : public DataSymbol { public: // Constructor for regular data symbols originating from input files. - DefinedData(StringRef name, uint32_t flags, InputFile *f, - InputSegment *segment, uint64_t value, uint64_t size) + DefinedData(StringRef name, uint32_t flags, InputFile *f, InputChunk *segment, + uint64_t value, uint64_t size) : DataSymbol(name, DefinedDataKind, flags, f), segment(segment), value(value), size(size) {}

@@ -298,7 +298,7 @@ public: uint64_t getOutputSegmentIndex() const; uint64_t getSize() const { return size; }

- InputSegment *segment = nullptr; + InputChunk *segment = nullptr; uint64_t value = 0;

protected: diff --git a/lld/wasm/SyntheticSections.cpp b/lld/wasm/SyntheticSections.cpp index 4439761df9f9..b10ecc04b4b2 100644 --- a/lld/wasm/SyntheticSections.cpp +++ b/lld/wasm/SyntheticSections.cpp @@ -569,7 +569,7 @@ void LinkingSection::writeBody() { continue; StringRef comdat = inputSegments[0]->getComdatName(); #ifndef NDEBUG - for (const InputSegment *isec : inputSegments) + for (const InputChunk *isec : inputSegments) assert(isec->getComdatName() == comdat); #endif if (!comdat.empty()) diff --git a/lld/wasm/Writer.cpp b/lld/wasm/Writer.cpp index 08149bc605a2..055972560b81 100644 --- a/lld/wasm/Writer.cpp +++ b/lld/wasm/Writer.cpp @@ -460,7 +460,7 @@ void Writer::populateTargetFeatures() { }

// Find TLS data segments - auto isTLS = [](InputSegment *segment) { + auto isTLS = [](InputChunk *segment) { return segment->live && segment->isTLS(); }; tlsUsed = tlsUsed || @@ -816,7 +816,7 @@ void Writer::assignIndexes() { out.tableSec->assignIndexes(); }

-static StringRef getOutputDataSegmentName(const InputSegment &seg) { +static StringRef getOutputDataSegmentName(const InputChunk &seg) { // We always merge .tbss and .tdata into a single TLS segment so all TLS // symbols are be relative to single __tls_base. if (seg.isTLS()) @@ -852,7 +852,7 @@ OutputSegment *Writer::createOutputSegment(StringRef name) {

void Writer::createOutputSegments() { for (ObjFile *file : symtab->objectFiles) { - for (InputSegment *segment : file->segments) { + for (InputChunk *segment : file->segments) { if (!segment->live) continue; StringRef name = getOutputDataSegmentName(*segment); @@ -919,7 +919,7 @@ void Writer::combineOutputSegments() { combined->initFlags = WASM_DATA_SEGMENT_IS_PASSIVE; } bool first = true; - for (InputSegment *inSeg : s->inputSegments) { + for (InputChunk *inSeg : s->inputSegments) { if (first) inSeg->alignment = std::max(inSeg->alignment, s->alignment); first = false; @@ -1201,7 +1201,7 @@ void Writer::createApplyDataRelocationsFunction() { raw_string_ostream os(bodyContent); writeUleb128(os, 0, "num locals"); for (const OutputSegment *seg : segments) - for (const InputSegment *inSeg : seg->inputSegments) + for (const InputChunk *inSeg : seg->inputSegments) inSeg->generateRelocationCode(os);

writeU8(os, WASM_OPCODE_END, "END"); </cut>