diff --git a/compiler-rt/include/profile/MemProfData.inc b/compiler-rt/include/profile/MemProfData.inc --- a/compiler-rt/include/profile/MemProfData.inc +++ b/compiler-rt/include/profile/MemProfData.inc @@ -20,11 +20,10 @@ * \*===----------------------------------------------------------------------===*/ - #ifdef _MSC_VER -#define PACKED(__decl__) __pragma(pack(push,1)) __decl__ __pragma(pack(pop)) +#define PACKED(...) __pragma(pack(push,1)) __VA_ARGS__ __pragma(pack(pop)) #else -#define PACKED(__decl__) __decl__ __attribute__((__packed__)) +#define PACKED(...) __VA_ARGS__ __attribute__((__packed__)) #endif // A 64-bit magic number to uniquely identify the raw binary memprof profile file. @@ -35,6 +34,8 @@ // The version number of the raw binary format. #define MEMPROF_RAW_VERSION 1ULL +#include + namespace llvm { namespace memprof { // A struct describing the header used for the raw binary memprof profile format. @@ -47,14 +48,106 @@ uint64_t StackOffset; }); + // A struct describing the information necessary to describe a /proc/maps // segment entry for a particular binary/library identified by its build id. PACKED(struct SegmentEntry { uint64_t Start; uint64_t End; uint64_t Offset; - uint8_t BuildId[32]; + // This field is unused until sanitizer procmaps support for build ids for + // Linux Elf is implemented. + uint8_t BuildId[32] = {0}; + + SegmentEntry(uint64_t S, uint64_t E, uint64_t O) : + Start(S), End(E), Offset(O) {} + + SegmentEntry(const SegmentEntry& S) { + Start = S.Start; + End = S.End; + Offset = S.Offset; + } + + SegmentEntry& operator=(const SegmentEntry& S) { + Start = S.Start; + End = S.End; + Offset = S.Offset; + return *this; + } + + bool operator==(const SegmentEntry& S) const { + return Start == S.Start && + End == S.End && + Offset == S.Offset; + } +}); + +// A struct representing the heap allocation characteristics of a particular +// runtime context. This struct is shared between the compiler-rt runtime and +// the raw profile reader. The indexed format uses a separate, self-describing +// backwards compatible format. +PACKED(struct MemInfoBlock { + uint32_t alloc_count; + uint64_t total_access_count, min_access_count, max_access_count; + uint64_t total_size; + uint32_t min_size, max_size; + uint32_t alloc_timestamp, dealloc_timestamp; + uint64_t total_lifetime; + uint32_t min_lifetime, max_lifetime; + uint32_t alloc_cpu_id, dealloc_cpu_id; + uint32_t num_migrated_cpu; + + // Only compared to prior deallocated object currently. + uint32_t num_lifetime_overlaps; + uint32_t num_same_alloc_cpu; + uint32_t num_same_dealloc_cpu; + + uint64_t data_type_id; // TODO: hash of type name + + MemInfoBlock() : alloc_count(0) {} + + MemInfoBlock(uint32_t size, uint64_t access_count, uint32_t alloc_timestamp, + uint32_t dealloc_timestamp, uint32_t alloc_cpu, uint32_t dealloc_cpu) + : alloc_count(1), total_access_count(access_count), + min_access_count(access_count), max_access_count(access_count), + total_size(size), min_size(size), max_size(size), + alloc_timestamp(alloc_timestamp), dealloc_timestamp(dealloc_timestamp), + total_lifetime(dealloc_timestamp - alloc_timestamp), + min_lifetime(total_lifetime), max_lifetime(total_lifetime), + alloc_cpu_id(alloc_cpu), dealloc_cpu_id(dealloc_cpu), + num_lifetime_overlaps(0), num_same_alloc_cpu(0), + num_same_dealloc_cpu(0) { + num_migrated_cpu = alloc_cpu_id != dealloc_cpu_id; + } + + void Merge(const MemInfoBlock &newMIB) { + alloc_count += newMIB.alloc_count; + + total_access_count += newMIB.total_access_count; + min_access_count = newMIB.min_access_count < min_access_count ? newMIB.min_access_count : min_access_count; + max_access_count = newMIB.max_access_count < max_access_count ? newMIB.max_access_count : max_access_count; + + total_size += newMIB.total_size; + min_size = newMIB.min_size < min_size ? newMIB.min_size : min_size; + max_size = newMIB.max_size < max_size ? newMIB.max_size : max_size; + + total_lifetime += newMIB.total_lifetime; + min_lifetime = newMIB.min_lifetime < min_lifetime ? newMIB.min_lifetime : min_lifetime; + max_lifetime = newMIB.max_lifetime > max_lifetime ? newMIB.max_lifetime : max_lifetime; + + // We know newMIB was deallocated later, so just need to check if it was + // allocated before last one deallocated. + num_lifetime_overlaps += newMIB.alloc_timestamp < dealloc_timestamp; + alloc_timestamp = newMIB.alloc_timestamp; + dealloc_timestamp = newMIB.dealloc_timestamp; + + num_same_alloc_cpu += alloc_cpu_id == newMIB.alloc_cpu_id; + num_same_dealloc_cpu += dealloc_cpu_id == newMIB.dealloc_cpu_id; + alloc_cpu_id = newMIB.alloc_cpu_id; + dealloc_cpu_id = newMIB.dealloc_cpu_id; + } }); + } // namespace memprof } // namespace llvm diff --git a/compiler-rt/lib/memprof/memprof_allocator.cpp b/compiler-rt/lib/memprof/memprof_allocator.cpp --- a/compiler-rt/lib/memprof/memprof_allocator.cpp +++ b/compiler-rt/lib/memprof/memprof_allocator.cpp @@ -15,11 +15,11 @@ #include "memprof_allocator.h" #include "memprof_mapping.h" -#include "memprof_meminfoblock.h" #include "memprof_mibmap.h" #include "memprof_rawprofile.h" #include "memprof_stack.h" #include "memprof_thread.h" +#include "profile/MemProfData.inc" #include "sanitizer_common/sanitizer_allocator_checks.h" #include "sanitizer_common/sanitizer_allocator_interface.h" #include "sanitizer_common/sanitizer_allocator_report.h" @@ -36,6 +36,42 @@ #include namespace __memprof { +namespace { +using ::llvm::memprof::MemInfoBlock; + +void Print(const MemInfoBlock &M, const u64 id, bool print_terse) { + u64 p; + + if (print_terse) { + p = M.total_size * 100 / M.alloc_count; + Printf("MIB:%llu/%u/%llu.%02llu/%u/%u/", id, M.alloc_count, p / 100, + p % 100, M.min_size, M.max_size); + p = M.total_access_count * 100 / M.alloc_count; + Printf("%llu.%02llu/%llu/%llu/", p / 100, p % 100, M.min_access_count, + M.max_access_count); + p = M.total_lifetime * 100 / M.alloc_count; + Printf("%llu.%02llu/%u/%u/", p / 100, p % 100, M.min_lifetime, + M.max_lifetime); + Printf("%u/%u/%u/%u\n", M.num_migrated_cpu, M.num_lifetime_overlaps, + M.num_same_alloc_cpu, M.num_same_dealloc_cpu); + } else { + p = M.total_size * 100 / M.alloc_count; + Printf("Memory allocation stack id = %llu\n", id); + Printf("\talloc_count %u, size (ave/min/max) %llu.%02llu / %u / %u\n", + M.alloc_count, p / 100, p % 100, M.min_size, M.max_size); + p = M.total_access_count * 100 / M.alloc_count; + Printf("\taccess_count (ave/min/max): %llu.%02llu / %llu / %llu\n", p / 100, + p % 100, M.min_access_count, M.max_access_count); + p = M.total_lifetime * 100 / M.alloc_count; + Printf("\tlifetime (ave/min/max): %llu.%02llu / %u / %u\n", p / 100, + p % 100, M.min_lifetime, M.max_lifetime); + Printf("\tnum migrated: %u, num lifetime overlaps: %u, num same alloc " + "cpu: %u, num same dealloc_cpu: %u\n", + M.num_migrated_cpu, M.num_lifetime_overlaps, M.num_same_alloc_cpu, + M.num_same_dealloc_cpu); + } +} +} // namespace static int GetCpuId(void) { // _memprof_preinit is called via the preinit_array, which subsequently calls @@ -240,7 +276,7 @@ static void PrintCallback(const uptr Key, LockedMemInfoBlock *const &Value, void *Arg) { SpinMutexLock(&Value->mutex); - Value->mib.Print(Key, bool(Arg)); + Print(Value->mib, Key, bool(Arg)); } void FinishAndWrite() { diff --git a/compiler-rt/lib/memprof/memprof_meminfoblock.h b/compiler-rt/lib/memprof/memprof_meminfoblock.h deleted file mode 100644 --- a/compiler-rt/lib/memprof/memprof_meminfoblock.h +++ /dev/null @@ -1,116 +0,0 @@ -#ifndef MEMPROF_MEMINFOBLOCK_H_ -#define MEMPROF_MEMINFOBLOCK_H_ - -#include "memprof_interface_internal.h" // For u32, u64 TODO: Move these out of the internal header. -#include "sanitizer_common/sanitizer_common.h" - -namespace __memprof { - -using __sanitizer::Printf; - -struct MemInfoBlock { - u32 alloc_count; - u64 total_access_count, min_access_count, max_access_count; - u64 total_size; - u32 min_size, max_size; - u32 alloc_timestamp, dealloc_timestamp; - u64 total_lifetime; - u32 min_lifetime, max_lifetime; - u32 alloc_cpu_id, dealloc_cpu_id; - u32 num_migrated_cpu; - - // Only compared to prior deallocated object currently. - u32 num_lifetime_overlaps; - u32 num_same_alloc_cpu; - u32 num_same_dealloc_cpu; - - u64 data_type_id; // TODO: hash of type name - - MemInfoBlock() : alloc_count(0) {} - - MemInfoBlock(u32 size, u64 access_count, u32 alloc_timestamp, - u32 dealloc_timestamp, u32 alloc_cpu, u32 dealloc_cpu) - : alloc_count(1), total_access_count(access_count), - min_access_count(access_count), max_access_count(access_count), - total_size(size), min_size(size), max_size(size), - alloc_timestamp(alloc_timestamp), dealloc_timestamp(dealloc_timestamp), - total_lifetime(dealloc_timestamp - alloc_timestamp), - min_lifetime(total_lifetime), max_lifetime(total_lifetime), - alloc_cpu_id(alloc_cpu), dealloc_cpu_id(dealloc_cpu), - num_lifetime_overlaps(0), num_same_alloc_cpu(0), - num_same_dealloc_cpu(0) { - num_migrated_cpu = alloc_cpu_id != dealloc_cpu_id; - } - - void Print(u64 id, bool print_terse) const { - u64 p; - - if (print_terse) { - p = total_size * 100 / alloc_count; - Printf("MIB:%llu/%u/%llu.%02llu/%u/%u/", id, alloc_count, p / 100, - p % 100, min_size, max_size); - p = total_access_count * 100 / alloc_count; - Printf("%llu.%02llu/%llu/%llu/", p / 100, p % 100, min_access_count, - max_access_count); - p = total_lifetime * 100 / alloc_count; - Printf("%llu.%02llu/%u/%u/", p / 100, p % 100, min_lifetime, - max_lifetime); - Printf("%u/%u/%u/%u\n", num_migrated_cpu, num_lifetime_overlaps, - num_same_alloc_cpu, num_same_dealloc_cpu); - } else { - p = total_size * 100 / alloc_count; - Printf("Memory allocation stack id = %llu\n", id); - Printf("\talloc_count %u, size (ave/min/max) %llu.%02llu / %u / %u\n", - alloc_count, p / 100, p % 100, min_size, max_size); - p = total_access_count * 100 / alloc_count; - Printf("\taccess_count (ave/min/max): %llu.%02llu / %llu / %llu\n", - p / 100, p % 100, min_access_count, max_access_count); - p = total_lifetime * 100 / alloc_count; - Printf("\tlifetime (ave/min/max): %llu.%02llu / %u / %u\n", p / 100, - p % 100, min_lifetime, max_lifetime); - Printf("\tnum migrated: %u, num lifetime overlaps: %u, num same alloc " - "cpu: %u, num same dealloc_cpu: %u\n", - num_migrated_cpu, num_lifetime_overlaps, num_same_alloc_cpu, - num_same_dealloc_cpu); - } - } - - static void printHeader() { - Printf("MIB:StackID/AllocCount/AveSize/MinSize/MaxSize/AveAccessCount/" - "MinAccessCount/MaxAccessCount/AveLifetime/MinLifetime/MaxLifetime/" - "NumMigratedCpu/NumLifetimeOverlaps/NumSameAllocCpu/" - "NumSameDeallocCpu\n"); - } - - void Merge(const MemInfoBlock &newMIB) { - alloc_count += newMIB.alloc_count; - - total_access_count += newMIB.total_access_count; - min_access_count = Min(min_access_count, newMIB.min_access_count); - max_access_count = Max(max_access_count, newMIB.max_access_count); - - total_size += newMIB.total_size; - min_size = Min(min_size, newMIB.min_size); - max_size = Max(max_size, newMIB.max_size); - - total_lifetime += newMIB.total_lifetime; - min_lifetime = Min(min_lifetime, newMIB.min_lifetime); - max_lifetime = Max(max_lifetime, newMIB.max_lifetime); - - // We know newMIB was deallocated later, so just need to check if it was - // allocated before last one deallocated. - num_lifetime_overlaps += newMIB.alloc_timestamp < dealloc_timestamp; - alloc_timestamp = newMIB.alloc_timestamp; - dealloc_timestamp = newMIB.dealloc_timestamp; - - num_same_alloc_cpu += alloc_cpu_id == newMIB.alloc_cpu_id; - num_same_dealloc_cpu += dealloc_cpu_id == newMIB.dealloc_cpu_id; - alloc_cpu_id = newMIB.alloc_cpu_id; - dealloc_cpu_id = newMIB.dealloc_cpu_id; - } - -} __attribute__((packed)); - -} // namespace __memprof - -#endif // MEMPROF_MEMINFOBLOCK_H_ diff --git a/compiler-rt/lib/memprof/memprof_mibmap.h b/compiler-rt/lib/memprof/memprof_mibmap.h --- a/compiler-rt/lib/memprof/memprof_mibmap.h +++ b/compiler-rt/lib/memprof/memprof_mibmap.h @@ -1,7 +1,9 @@ #ifndef MEMPROF_MIBMAP_H_ #define MEMPROF_MIBMAP_H_ -#include "memprof_meminfoblock.h" +#include + +#include "profile/MemProfData.inc" #include "sanitizer_common/sanitizer_addrhashmap.h" #include "sanitizer_common/sanitizer_mutex.h" @@ -9,7 +11,7 @@ struct LockedMemInfoBlock { __sanitizer::StaticSpinMutex mutex; - MemInfoBlock mib; + ::llvm::memprof::MemInfoBlock mib; }; // The MIB map stores a mapping from stack ids to MemInfoBlocks. @@ -17,7 +19,8 @@ // Insert a new MemInfoBlock or merge with an existing block identified by the // stack id. -void InsertOrMerge(const uptr Id, const MemInfoBlock &Block, MIBMapTy &Map); +void InsertOrMerge(const uptr Id, const ::llvm::memprof::MemInfoBlock &Block, + MIBMapTy &Map); } // namespace __memprof diff --git a/compiler-rt/lib/memprof/memprof_mibmap.cpp b/compiler-rt/lib/memprof/memprof_mibmap.cpp --- a/compiler-rt/lib/memprof/memprof_mibmap.cpp +++ b/compiler-rt/lib/memprof/memprof_mibmap.cpp @@ -11,10 +11,12 @@ //===----------------------------------------------------------------------===// #include "memprof_mibmap.h" +#include "profile/MemProfData.inc" #include "sanitizer_common/sanitizer_allocator_internal.h" #include "sanitizer_common/sanitizer_mutex.h" namespace __memprof { +using ::llvm::memprof::MemInfoBlock; void InsertOrMerge(const uptr Id, const MemInfoBlock &Block, MIBMapTy &Map) { MIBMapTy::Handle h(&Map, static_cast(Id), /*remove=*/false, diff --git a/compiler-rt/lib/memprof/memprof_rawprofile.cpp b/compiler-rt/lib/memprof/memprof_rawprofile.cpp --- a/compiler-rt/lib/memprof/memprof_rawprofile.cpp +++ b/compiler-rt/lib/memprof/memprof_rawprofile.cpp @@ -2,7 +2,6 @@ #include #include -#include "memprof_meminfoblock.h" #include "memprof_rawprofile.h" #include "profile/MemProfData.inc" #include "sanitizer_common/sanitizer_allocator_internal.h" @@ -16,6 +15,7 @@ namespace __memprof { using ::__sanitizer::Vector; +using ::llvm::memprof::MemInfoBlock; using SegmentEntry = ::llvm::memprof::SegmentEntry; using Header = ::llvm::memprof::Header; @@ -65,11 +65,8 @@ for (Layout.Reset(); Layout.Next(&segment);) { if (segment.IsReadable() && segment.IsExecutable()) { - SegmentEntry Entry{}; - Entry.Start = segment.start; - Entry.End = segment.end; - Entry.Offset = segment.offset; - memcpy(Entry.BuildId, segment.uuid, sizeof(segment.uuid)); + // TODO: Record segment.uuid when it is implemented for Linux-Elf. + SegmentEntry Entry(segment.start, segment.end, segment.offset); memcpy(Ptr, &Entry, sizeof(SegmentEntry)); Ptr += sizeof(SegmentEntry); NumSegmentsRecorded++; diff --git a/compiler-rt/lib/memprof/tests/rawprofile.cpp b/compiler-rt/lib/memprof/tests/rawprofile.cpp --- a/compiler-rt/lib/memprof/tests/rawprofile.cpp +++ b/compiler-rt/lib/memprof/tests/rawprofile.cpp @@ -3,7 +3,6 @@ #include #include -#include "memprof/memprof_meminfoblock.h" #include "profile/MemProfData.inc" #include "sanitizer_common/sanitizer_common.h" #include "sanitizer_common/sanitizer_procmaps.h" @@ -14,13 +13,13 @@ namespace { -using ::__memprof::MemInfoBlock; using ::__memprof::MIBMapTy; using ::__memprof::SerializeToRawProfile; using ::__sanitizer::MemoryMappedSegment; using ::__sanitizer::MemoryMappingLayoutBase; using ::__sanitizer::StackDepotPut; using ::__sanitizer::StackTrace; +using ::llvm::memprof::MemInfoBlock; using ::testing::_; using ::testing::Action; using ::testing::DoAll; @@ -33,21 +32,21 @@ MOCK_METHOD(void, Reset, (), (override)); }; -u64 PopulateFakeMap(const MemInfoBlock &FakeMIB, uptr StackPCBegin, - MIBMapTy &FakeMap) { +uint64_t PopulateFakeMap(const MemInfoBlock &FakeMIB, uint64_t StackPCBegin, + MIBMapTy &FakeMap) { constexpr int kSize = 5; - uptr array[kSize]; + uint64_t array[kSize]; for (int i = 0; i < kSize; i++) { array[i] = StackPCBegin + i; } StackTrace St(array, kSize); - u32 Id = StackDepotPut(St); + uint32_t Id = StackDepotPut(St); InsertOrMerge(Id, FakeMIB, FakeMap); return Id; } -template T Read(char *&Buffer) { +template T Read(char *&Buffer) { static_assert(std::is_pod::value, "Must be a POD type."); assert(reinterpret_cast(Buffer) % sizeof(T) == 0 && "Unaligned read!"); @@ -86,12 +85,12 @@ FakeMIB.alloc_count = 0x1; FakeMIB.total_access_count = 0x2; - u64 FakeIds[2]; + uint64_t FakeIds[2]; FakeIds[0] = PopulateFakeMap(FakeMIB, /*StackPCBegin=*/2, FakeMap); FakeIds[1] = PopulateFakeMap(FakeMIB, /*StackPCBegin=*/3, FakeMap); char *Ptr = nullptr; - u64 NumBytes = SerializeToRawProfile(FakeMap, Layout, Ptr); + uint64_t NumBytes = SerializeToRawProfile(FakeMap, Layout, Ptr); const char *Buffer = Ptr; ASSERT_GT(NumBytes, 0ULL); @@ -100,10 +99,10 @@ // Check the header. EXPECT_THAT(Read(Ptr), MEMPROF_RAW_MAGIC_64); EXPECT_THAT(Read(Ptr), MEMPROF_RAW_VERSION); - const u64 TotalSize = Read(Ptr); - const u64 SegmentOffset = Read(Ptr); - const u64 MIBOffset = Read(Ptr); - const u64 StackOffset = Read(Ptr); + const uint64_t TotalSize = Read(Ptr); + const uint64_t SegmentOffset = Read(Ptr); + const uint64_t MIBOffset = Read(Ptr); + const uint64_t StackOffset = Read(Ptr); // ============= Check sizes and padding. EXPECT_EQ(TotalSize, NumBytes); @@ -117,7 +116,7 @@ EXPECT_EQ(MIBOffset - SegmentOffset, 64ULL); EXPECT_EQ(MIBOffset, 112ULL); - // We expect 2 mib entry, 8b for the count and sizeof(u64) + + // We expect 2 mib entry, 8b for the count and sizeof(uint64_t) + // sizeof(MemInfoBlock) contains stack id + MeminfoBlock. EXPECT_EQ(StackOffset - MIBOffset, 8 + 2 * (8 + sizeof(MemInfoBlock))); @@ -130,18 +129,19 @@ // ============= Check contents. unsigned char ExpectedSegmentBytes[64] = { - 0x01, 0, 0, 0, 0, 0, 0, 0, // Number of entries - 0x10, 0, 0, 0, 0, 0, 0, 0, // Start - 0x20, 0, 0, 0, 0, 0, 0, 0, // End - 0x10, 0, 0, 0, 0, 0, 0, 0, // Offset - 0x0C, 0x0, 0xF, 0xF, 0xE, 0xE, // Uuid + 0x01, 0, 0, 0, 0, 0, 0, 0, // Number of entries + 0x10, 0, 0, 0, 0, 0, 0, 0, // Start + 0x20, 0, 0, 0, 0, 0, 0, 0, // End + 0x10, 0, 0, 0, 0, 0, 0, 0, // Offset + 0x0, // Uuid (not yet recorded). }; EXPECT_EQ(memcmp(Buffer + SegmentOffset, ExpectedSegmentBytes, 64), 0); // Check that the number of entries is 2. - EXPECT_EQ(*reinterpret_cast(Buffer + MIBOffset), 2ULL); + EXPECT_EQ(*reinterpret_cast(Buffer + MIBOffset), 2ULL); // Check that stack id is set. - EXPECT_EQ(*reinterpret_cast(Buffer + MIBOffset + 8), FakeIds[0]); + EXPECT_EQ(*reinterpret_cast(Buffer + MIBOffset + 8), + FakeIds[0]); // Only check a few fields of the first MemInfoBlock. unsigned char ExpectedMIBBytes[sizeof(MemInfoBlock)] = { @@ -159,9 +159,9 @@ 0); // Check that the number of entries is 2. - EXPECT_EQ(*reinterpret_cast(Buffer + StackOffset), 2ULL); + EXPECT_EQ(*reinterpret_cast(Buffer + StackOffset), 2ULL); // Check that the 1st stack id is set. - EXPECT_EQ(*reinterpret_cast(Buffer + StackOffset + 8), + EXPECT_EQ(*reinterpret_cast(Buffer + StackOffset + 8), FakeIds[0]); // Contents are num pcs, value of each pc - 1. unsigned char ExpectedStackBytes[2][6 * 8] = { @@ -184,7 +184,7 @@ // Check that the 2nd stack id is set. EXPECT_EQ( - *reinterpret_cast(Buffer + StackOffset + 8 + 6 * 8 + 8), + *reinterpret_cast(Buffer + StackOffset + 8 + 6 * 8 + 8), FakeIds[1]); EXPECT_EQ(memcmp(Buffer + StackOffset + 16 + 6 * 8 + 8, ExpectedStackBytes[1], diff --git a/llvm/include/llvm/ProfileData/MemProfData.inc b/llvm/include/llvm/ProfileData/MemProfData.inc --- a/llvm/include/llvm/ProfileData/MemProfData.inc +++ b/llvm/include/llvm/ProfileData/MemProfData.inc @@ -20,11 +20,10 @@ * \*===----------------------------------------------------------------------===*/ - #ifdef _MSC_VER -#define PACKED(__decl__) __pragma(pack(push,1)) __decl__ __pragma(pack(pop)) +#define PACKED(...) __pragma(pack(push,1)) __VA_ARGS__ __pragma(pack(pop)) #else -#define PACKED(__decl__) __decl__ __attribute__((__packed__)) +#define PACKED(...) __VA_ARGS__ __attribute__((__packed__)) #endif // A 64-bit magic number to uniquely identify the raw binary memprof profile file. @@ -35,6 +34,8 @@ // The version number of the raw binary format. #define MEMPROF_RAW_VERSION 1ULL +#include + namespace llvm { namespace memprof { // A struct describing the header used for the raw binary memprof profile format. @@ -47,14 +48,106 @@ uint64_t StackOffset; }); + // A struct describing the information necessary to describe a /proc/maps // segment entry for a particular binary/library identified by its build id. PACKED(struct SegmentEntry { uint64_t Start; uint64_t End; uint64_t Offset; - uint8_t BuildId[32]; + // This field is unused until sanitizer procmaps support for build ids for + // Linux Elf is implemented. + uint8_t BuildId[32] = {0}; + + SegmentEntry(uint64_t S, uint64_t E, uint64_t O) : + Start(S), End(E), Offset(O) {} + + SegmentEntry(const SegmentEntry& S) { + Start = S.Start; + End = S.End; + Offset = S.Offset; + } + + SegmentEntry& operator=(const SegmentEntry& S) { + Start = S.Start; + End = S.End; + Offset = S.Offset; + return *this; + } + + bool operator==(const SegmentEntry& S) const { + return Start == S.Start && + End == S.End && + Offset == S.Offset; + } +}); + +// A struct representing the heap allocation characteristics of a particular +// runtime context. This struct is shared between the compiler-rt runtime and +// the raw profile reader. The indexed format uses a separate, self-describing +// backwards compatible format. +PACKED(struct MemInfoBlock { + uint32_t alloc_count; + uint64_t total_access_count, min_access_count, max_access_count; + uint64_t total_size; + uint32_t min_size, max_size; + uint32_t alloc_timestamp, dealloc_timestamp; + uint64_t total_lifetime; + uint32_t min_lifetime, max_lifetime; + uint32_t alloc_cpu_id, dealloc_cpu_id; + uint32_t num_migrated_cpu; + + // Only compared to prior deallocated object currently. + uint32_t num_lifetime_overlaps; + uint32_t num_same_alloc_cpu; + uint32_t num_same_dealloc_cpu; + + uint64_t data_type_id; // TODO: hash of type name + + MemInfoBlock() : alloc_count(0) {} + + MemInfoBlock(uint32_t size, uint64_t access_count, uint32_t alloc_timestamp, + uint32_t dealloc_timestamp, uint32_t alloc_cpu, uint32_t dealloc_cpu) + : alloc_count(1), total_access_count(access_count), + min_access_count(access_count), max_access_count(access_count), + total_size(size), min_size(size), max_size(size), + alloc_timestamp(alloc_timestamp), dealloc_timestamp(dealloc_timestamp), + total_lifetime(dealloc_timestamp - alloc_timestamp), + min_lifetime(total_lifetime), max_lifetime(total_lifetime), + alloc_cpu_id(alloc_cpu), dealloc_cpu_id(dealloc_cpu), + num_lifetime_overlaps(0), num_same_alloc_cpu(0), + num_same_dealloc_cpu(0) { + num_migrated_cpu = alloc_cpu_id != dealloc_cpu_id; + } + + void Merge(const MemInfoBlock &newMIB) { + alloc_count += newMIB.alloc_count; + + total_access_count += newMIB.total_access_count; + min_access_count = newMIB.min_access_count < min_access_count ? newMIB.min_access_count : min_access_count; + max_access_count = newMIB.max_access_count < max_access_count ? newMIB.max_access_count : max_access_count; + + total_size += newMIB.total_size; + min_size = newMIB.min_size < min_size ? newMIB.min_size : min_size; + max_size = newMIB.max_size < max_size ? newMIB.max_size : max_size; + + total_lifetime += newMIB.total_lifetime; + min_lifetime = newMIB.min_lifetime < min_lifetime ? newMIB.min_lifetime : min_lifetime; + max_lifetime = newMIB.max_lifetime > max_lifetime ? newMIB.max_lifetime : max_lifetime; + + // We know newMIB was deallocated later, so just need to check if it was + // allocated before last one deallocated. + num_lifetime_overlaps += newMIB.alloc_timestamp < dealloc_timestamp; + alloc_timestamp = newMIB.alloc_timestamp; + dealloc_timestamp = newMIB.dealloc_timestamp; + + num_same_alloc_cpu += alloc_cpu_id == newMIB.alloc_cpu_id; + num_same_dealloc_cpu += dealloc_cpu_id == newMIB.dealloc_cpu_id; + alloc_cpu_id = newMIB.alloc_cpu_id; + dealloc_cpu_id = newMIB.dealloc_cpu_id; + } }); + } // namespace memprof } // namespace llvm