diff --git a/llvm/include/llvm/Analysis/MemoryProfileInfo.h b/llvm/include/llvm/Analysis/MemoryProfileInfo.h new file mode 100644 --- /dev/null +++ b/llvm/include/llvm/Analysis/MemoryProfileInfo.h @@ -0,0 +1,112 @@ +//===- llvm/Analysis/MemoryProfileInfo.h - memory profile info ---*- C++ -*-==// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file contains utilities to analyze memory profile information. +// +//===----------------------------------------------------------------------===// + +#ifndef LLVM_ANALYSIS_MEMORYPROFILEINFO_H +#define LLVM_ANALYSIS_MEMORYPROFILEINFO_H + +#include "llvm/IR/Constants.h" +#include "llvm/IR/InstrTypes.h" +#include "llvm/IR/Metadata.h" +#include "llvm/IR/Module.h" +#include + +namespace llvm { +namespace memprof { + +// Allocation type assigned to an allocation reached by a given context. +// More can be added but initially this is just noncold and cold. +// Values should be powers of two so that they can be ORed, in particular to +// track allocations that have different behavior with different calling +// contexts. +enum class AllocationType : uint8_t { None = 0, NotCold = 1, Cold = 2 }; + +/// Return the allocation type for a given set of memory profile values. +AllocationType getAllocType(uint64_t MaxAccessCount, uint64_t MinSize, + uint64_t MinLifetime); + +/// Build callstack metadata from the provided list of call stack ids. Returns +/// the resulting metadata node. +MDNode *buildCallstackMetadata(ArrayRef CallStack, LLVMContext &Ctx); + +/// Returns the stack node from an MIB metadata node. +MDNode *getMIBStackNode(const MDNode *MIB); + +/// Returns the allocation type from an MIB metadata node. +AllocationType getMIBAllocType(const MDNode *MIB); + +/// Class to build a trie of call stack contexts for a particular profiled +/// allocation call, along with their associated allocation types. +/// The allocation will be at the root of the trie, which is then used to +/// compute the minimum lists of context ids needed to associate a call context +/// with a single allocation type. +class CallStackTrie { +private: + struct CallStackTrieNode { + // Allocation types for call context sharing the context prefix at this + // node. + uint8_t AllocTypes; + // Map of caller stack id to the corresponding child Trie node. + std::map Callers; + CallStackTrieNode(AllocationType Type) + : AllocTypes(static_cast(Type)) {} + }; + + // The node for the allocation at the root. + CallStackTrieNode *Alloc; + // The allocation's leaf stack id. + uint64_t AllocStackId; + + void deleteTrieNode(CallStackTrieNode *Node) { + if (!Node) + return; + for (auto C : Node->Callers) + deleteTrieNode(C.second); + delete Node; + } + + // Recursive helper to trim contexts and create metadata nodes. + bool buildMIBNodes(CallStackTrieNode *Node, LLVMContext &Ctx, + std::vector &MIBCallStack, + std::vector &MIBNodes, + bool CalleeHasAmbiguousCallerContext); + +public: + CallStackTrie() : Alloc(nullptr), AllocStackId(0) {} + ~CallStackTrie() { deleteTrieNode(Alloc); } + + bool empty() const { return Alloc == nullptr; } + + /// Add a call stack context with the given allocation type to the Trie. + /// The context is represented by the list of stack ids (computed during + /// matching via a debug location hash), expected to be in order from the + /// allocation call down to the bottom of the call stack (i.e. callee to + /// caller order). + void addCallStack(AllocationType AllocType, ArrayRef StackIds); + + /// Add the call stack context along with its allocation type from the MIB + /// metadata to the Trie. + void addCallStack(MDNode *MIB); + + /// Build and attach the minimal necessary MIB metadata. If the alloc has a + /// single allocation type, add a function attribute instead. The reason for + /// adding an attribute in this case is that it matches how the behavior for + /// allocation calls will be communicated to lib call simplification after + /// cloning or another optimization to distinguish the allocation types, + /// which is lower overhead and more direct than maintaining this metadata. + /// Returns true if memprof metadata attached, false if not (attribute added). + bool buildAndAttachMIBMetadata(CallBase *CI); +}; + +} // end namespace memprof +} // end namespace llvm + +#endif diff --git a/llvm/lib/Analysis/CMakeLists.txt b/llvm/lib/Analysis/CMakeLists.txt --- a/llvm/lib/Analysis/CMakeLists.txt +++ b/llvm/lib/Analysis/CMakeLists.txt @@ -98,6 +98,7 @@ MemoryBuiltins.cpp MemoryDependenceAnalysis.cpp MemoryLocation.cpp + MemoryProfileInfo.cpp MemorySSA.cpp MemorySSAUpdater.cpp ModelUnderTrainingRunner.cpp diff --git a/llvm/lib/Analysis/MemoryProfileInfo.cpp b/llvm/lib/Analysis/MemoryProfileInfo.cpp new file mode 100644 --- /dev/null +++ b/llvm/lib/Analysis/MemoryProfileInfo.cpp @@ -0,0 +1,226 @@ +//===-- MemoryProfileInfo.cpp - memory profile info ------------------------==// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file contains utilities to analyze memory profile information. +// +//===----------------------------------------------------------------------===// + +#include "llvm/Analysis/MemoryProfileInfo.h" +#include "llvm/Support/CommandLine.h" + +using namespace llvm; +using namespace llvm::memprof; + +#define DEBUG_TYPE "memory-profile-info" + +// Upper bound on accesses per byte for marking an allocation cold. +cl::opt MemProfAccessesPerByteColdThreshold( + "memprof-accesses-per-byte-cold-threshold", cl::init(10.0), cl::Hidden, + cl::desc("The threshold the accesses per byte must be under to consider " + "an allocation cold")); + +// Lower bound on lifetime to mark an allocation cold (in addition to accesses +// per byte above). This is to avoid pessimizing short lived objects. +cl::opt MemProfMinLifetimeColdThreshold( + "memprof-min-lifetime-cold-threshold", cl::init(200), cl::Hidden, + cl::desc("The minimum lifetime (s) for an allocation to be considered " + "cold")); + +AllocationType llvm::memprof::getAllocType(uint64_t MaxAccessCount, + uint64_t MinSize, + uint64_t MinLifetime) { + if (((float)MaxAccessCount) / MinSize < MemProfAccessesPerByteColdThreshold && + // MinLifetime is expected to be in ms, so convert the threshold to ms. + MinLifetime >= MemProfMinLifetimeColdThreshold * 1000) + return AllocationType::Cold; + return AllocationType::NotCold; +} + +MDNode *llvm::memprof::buildCallstackMetadata(ArrayRef CallStack, + LLVMContext &Ctx) { + std::vector StackVals; + for (auto Id : CallStack) { + auto *StackValMD = + ValueAsMetadata::get(ConstantInt::get(Type::getInt64Ty(Ctx), Id)); + StackVals.push_back(StackValMD); + } + return MDNode::get(Ctx, StackVals); +} + +MDNode *llvm::memprof::getMIBStackNode(const MDNode *MIB) { + assert(MIB->getNumOperands() == 2); + // The stack metadata is the first operand of each memprof MIB metadata. + return cast(MIB->getOperand(0)); +} + +AllocationType llvm::memprof::getMIBAllocType(const MDNode *MIB) { + assert(MIB->getNumOperands() == 2); + // The allocation type is currently the second operand of each memprof + // MIB metadata. This will need to change as we add additional allocation + // types that can be applied based on the allocation profile data. + auto *MDS = dyn_cast(MIB->getOperand(1)); + assert(MDS); + if (MDS->getString().equals("cold")) + return AllocationType::Cold; + return AllocationType::NotCold; +} + +static std::string getAllocTypeAttributeString(AllocationType Type) { + switch (Type) { + case AllocationType::NotCold: + return "notcold"; + break; + case AllocationType::Cold: + return "cold"; + break; + default: + assert(false && "Unexpected alloc type"); + } + llvm_unreachable("invalid alloc type"); +} + +static void addAllocTypeAttribute(LLVMContext &Ctx, CallBase *CI, + AllocationType AllocType) { + auto AllocTypeString = getAllocTypeAttributeString(AllocType); + auto A = llvm::Attribute::get(Ctx, "memprof", AllocTypeString); + CI->addFnAttr(A); +} + +static bool hasSingleAllocType(uint8_t AllocTypes) { + const unsigned NumAllocTypes = countPopulation(AllocTypes); + assert(NumAllocTypes != 0); + return NumAllocTypes > 1; +} + +void CallStackTrie::addCallStack(AllocationType AllocType, + ArrayRef StackIds) { + bool First = true; + CallStackTrieNode *Curr = nullptr; + for (auto StackId : StackIds) { + // If this is the first stack frame, add or update alloc node. + if (First) { + First = false; + if (Alloc) { + assert(AllocStackId == StackId); + Alloc->AllocTypes |= static_cast(AllocType); + } else { + AllocStackId = StackId; + Alloc = new CallStackTrieNode(AllocType); + } + Curr = Alloc; + continue; + } + // Update existing caller node if it exists. + auto Next = Curr->Callers.find(StackId); + if (Next != Curr->Callers.end()) { + Curr = Next->second; + Curr->AllocTypes |= static_cast(AllocType); + continue; + } + // Otherwise add a new caller node. + auto *New = new CallStackTrieNode(AllocType); + Curr->Callers[StackId] = New; + Curr = New; + } + assert(Curr); +} + +void CallStackTrie::addCallStack(MDNode *MIB) { + MDNode *StackMD = getMIBStackNode(MIB); + assert(StackMD); + std::vector CallStack; + CallStack.reserve(StackMD->getNumOperands()); + for (auto &MIBStackIter : StackMD->operands()) { + auto *StackId = mdconst::dyn_extract(MIBStackIter); + assert(StackId); + CallStack.push_back(StackId->getZExtValue()); + } + addCallStack(getMIBAllocType(MIB), CallStack); +} + +static MDNode *createMIBNode(LLVMContext &Ctx, + std::vector &MIBCallStack, + AllocationType AllocType) { + std::vector MIBPayload( + {buildCallstackMetadata(MIBCallStack, Ctx)}); + MIBPayload.push_back( + MDString::get(Ctx, getAllocTypeAttributeString(AllocType))); + return MDNode::get(Ctx, MIBPayload); +} + +// Recursive helper to trim contexts and create metadata nodes. +// Caller should have pushed Node's loc to MIBCallStack. Doing this in the +// caller makes it simpler to handle the many early returns in this method. +bool CallStackTrie::buildMIBNodes(CallStackTrieNode *Node, LLVMContext &Ctx, + std::vector &MIBCallStack, + std::vector &MIBNodes, + bool CalleeHasAmbiguousCallerContext) { + // Trim context below the first node in a prefix with a single alloc type. + // Add an MIB record for the current call stack prefix. + if (hasSingleAllocType(Node->AllocTypes)) { + MIBNodes.push_back( + createMIBNode(Ctx, MIBCallStack, (AllocationType)Node->AllocTypes)); + return true; + } + + // We don't have a single allocation for all the contexts sharing this prefix, + // so recursively descend into callers in trie. + if (!Node->Callers.empty()) { + bool NodeHasAmbiguousCallerContext = Node->Callers.size() > 1; + bool AddedMIBNodesForAllCallerContexts = true; + for (auto &Caller : Node->Callers) { + MIBCallStack.push_back(Caller.first); + AddedMIBNodesForAllCallerContexts &= + buildMIBNodes(Caller.second, Ctx, MIBCallStack, MIBNodes, + NodeHasAmbiguousCallerContext); + // Remove Caller. + MIBCallStack.pop_back(); + } + if (AddedMIBNodesForAllCallerContexts) + return true; + // We expect that the callers should be forced to add MIBs to disambiguate + // the context in this case (see below). + assert(!NodeHasAmbiguousCallerContext); + } + + // If we reached here, then this node does not have a single allocation type, + // and we didn't add metadata for a longer call stack prefix including any of + // Node's callers. That means we never hit a single allocation type along all + // call stacks with this prefix. This can happen due to recursion collapsing + // or the stack being deeper than tracked by the profiler runtime, leading to + // contexts with different allocation types being merged. In that case, we + // trim the context just below the deepest context split, which is this + // node if the callee has an ambiguous caller context (multiple callers), + // since the recursive calls above returned false. Conservatively give it + // non-cold allocation type. + if (!CalleeHasAmbiguousCallerContext) + return false; + MIBNodes.push_back(createMIBNode(Ctx, MIBCallStack, AllocationType::NotCold)); + return true; +} + +// Build and attach the minimal necessary MIB metadata. If the alloc has a +// single allocation type, add a function attribute instead. Returns true if +// memprof metadata attached, false if not (attribute added). +bool CallStackTrie::buildAndAttachMIBMetadata(CallBase *CI) { + auto &Ctx = CI->getContext(); + if (hasSingleAllocType(Alloc->AllocTypes)) { + addAllocTypeAttribute(Ctx, CI, (AllocationType)Alloc->AllocTypes); + return false; + } + std::vector MIBCallStack; + MIBCallStack.push_back(AllocStackId); + std::vector MIBNodes; + assert(!Alloc->Callers.empty() && "addCallStack has not been called yet"); + buildMIBNodes(Alloc, Ctx, MIBCallStack, MIBNodes, + /*CalleeHasAmbiguousCallerContext=*/true); + assert(MIBCallStack.size() == 1 && + "Should only be left with Alloc's location in stack"); + CI->setMetadata(LLVMContext::MD_memprof, MDNode::get(Ctx, MIBNodes)); + return true; +} diff --git a/llvm/unittests/Analysis/CMakeLists.txt b/llvm/unittests/Analysis/CMakeLists.txt --- a/llvm/unittests/Analysis/CMakeLists.txt +++ b/llvm/unittests/Analysis/CMakeLists.txt @@ -38,6 +38,7 @@ LoopInfoTest.cpp LoopNestTest.cpp MemoryBuiltinsTest.cpp + MemoryProfileInfoTest.cpp MemorySSATest.cpp MLModelRunnerTest.cpp PhiValuesTest.cpp diff --git a/llvm/unittests/Analysis/MemoryProfileInfoTest.cpp b/llvm/unittests/Analysis/MemoryProfileInfoTest.cpp new file mode 100644 --- /dev/null +++ b/llvm/unittests/Analysis/MemoryProfileInfoTest.cpp @@ -0,0 +1,362 @@ +//===- MemoryProfileInfoTest.cpp - Memory Profile Info Unit Tests-===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// + +#include "llvm/Analysis/MemoryProfileInfo.h" +#include "llvm/AsmParser/Parser.h" +#include "llvm/IR/Instructions.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/Module.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/SourceMgr.h" +#include "gtest/gtest.h" +#include + +using namespace llvm; +using namespace llvm::memprof; + +extern cl::opt MemProfAccessesPerByteColdThreshold; +extern cl::opt MemProfMinLifetimeColdThreshold; + +namespace { + +class MemoryProfileInfoTest : public testing::Test { +protected: + std::unique_ptr makeLLVMModule(LLVMContext &C, const char *IR) { + SMDiagnostic Err; + std::unique_ptr Mod = parseAssemblyString(IR, Err, C); + if (!Mod) + Err.print("MemoryProfileInfoTest", errs()); + return Mod; + } + + // This looks for a call that has the given value name, which + // is the name of the value being assigned the call return value. + CallBase *findCall(Function &F, const char *Name = nullptr) { + for (auto &BB : F) + for (auto &I : BB) + if (auto *CB = dyn_cast(&I)) + if (!Name || CB->getName() == Name) + return CB; + return nullptr; + } +}; + +// Test getAllocType helper. +// Basic checks on the allocation type for values just above and below +// the thresholds. +TEST_F(MemoryProfileInfoTest, GetAllocType) { + // Long lived with more accesses per byte than threshold is not cold. + EXPECT_EQ( + getAllocType(/*MaxAccessCount=*/MemProfAccessesPerByteColdThreshold + 1, + /*MinSize=*/1, + /*MinLifetime=*/MemProfMinLifetimeColdThreshold * 1000 + 1), + AllocationType::NotCold); + // Long lived with less accesses per byte than threshold is cold. + EXPECT_EQ( + getAllocType(/*MaxAccessCount=*/MemProfAccessesPerByteColdThreshold - 1, + /*MinSize=*/1, + /*MinLifetime=*/MemProfMinLifetimeColdThreshold * 1000 + 1), + AllocationType::Cold); + // Short lived with more accesses per byte than threshold is not cold. + EXPECT_EQ( + getAllocType(/*MaxAccessCount=*/MemProfAccessesPerByteColdThreshold + 1, + /*MinSize=*/1, + /*MinLifetime=*/MemProfMinLifetimeColdThreshold * 1000 - 1), + AllocationType::NotCold); + // Short lived with less accesses per byte than threshold is not cold. + EXPECT_EQ( + getAllocType(/*MaxAccessCount=*/MemProfAccessesPerByteColdThreshold - 1, + /*MinSize=*/1, + /*MinLifetime=*/MemProfMinLifetimeColdThreshold * 1000 - 1), + AllocationType::NotCold); +} + +// Test buildCallstackMetadata helper. +TEST_F(MemoryProfileInfoTest, BuildCallStackMD) { + LLVMContext C; + MDNode *CallStack = buildCallstackMetadata({1, 2, 3}, C); + ASSERT_EQ(CallStack->getNumOperands(), 3u); + unsigned ExpectedId = 1; + for (auto &Op : CallStack->operands()) { + auto *StackId = mdconst::dyn_extract(Op); + EXPECT_EQ(StackId->getZExtValue(), ExpectedId++); + } +} + +// Test CallStackTrie::addCallStack interface taking allocation type and list of +// call stack ids. +// Check that allocations with a single allocation type along all call stacks +// get an attribute instead of memprof metadata. +TEST_F(MemoryProfileInfoTest, Attribute) { + LLVMContext C; + std::unique_ptr M = makeLLVMModule(C, + R"IR( +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-pc-linux-gnu" +define i32* @test() { +entry: + %call1 = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40) + %0 = bitcast i8* %call1 to i32* + %call2 = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40) + %1 = bitcast i8* %call2 to i32* + ret i32* %1 +} +declare dso_local noalias noundef i8* @malloc(i64 noundef) +)IR"); + + Function *Func = M->getFunction("test"); + + // First call has all cold contexts. + CallStackTrie Trie1; + Trie1.addCallStack(AllocationType::Cold, {1, 2}); + Trie1.addCallStack(AllocationType::Cold, {1, 3, 4}); + CallBase *Call1 = findCall(*Func, "call1"); + Trie1.buildAndAttachMIBMetadata(Call1); + + EXPECT_FALSE(Call1->hasMetadata(LLVMContext::MD_memprof)); + EXPECT_TRUE(Call1->hasFnAttr("memprof")); + EXPECT_EQ(Call1->getFnAttr("memprof").getValueAsString(), "cold"); + + // Second call has all non-cold contexts. + CallStackTrie Trie2; + Trie2.addCallStack(AllocationType::NotCold, {5, 6}); + Trie2.addCallStack(AllocationType::NotCold, {5, 7, 8}); + CallBase *Call2 = findCall(*Func, "call2"); + Trie2.buildAndAttachMIBMetadata(Call2); + + EXPECT_FALSE(Call2->hasMetadata(LLVMContext::MD_memprof)); + EXPECT_TRUE(Call2->hasFnAttr("memprof")); + EXPECT_EQ(Call2->getFnAttr("memprof").getValueAsString(), "notcold"); +} + +// Test CallStackTrie::addCallStack interface taking allocation type and list of +// call stack ids. +// Test that an allocation call reached by both cold and non cold call stacks +// gets memprof metadata representing the different allocation type contexts. +TEST_F(MemoryProfileInfoTest, ColdAndNotColdMIB) { + LLVMContext C; + std::unique_ptr M = makeLLVMModule(C, + R"IR( +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-pc-linux-gnu" +define i32* @test() { +entry: + %call = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40) + %0 = bitcast i8* %call to i32* + ret i32* %0 +} +declare dso_local noalias noundef i8* @malloc(i64 noundef) +)IR"); + + Function *Func = M->getFunction("test"); + + CallStackTrie Trie; + Trie.addCallStack(AllocationType::Cold, {1, 2}); + Trie.addCallStack(AllocationType::NotCold, {1, 3}); + + CallBase *Call = findCall(*Func, "call"); + Trie.buildAndAttachMIBMetadata(Call); + + EXPECT_FALSE(Call->hasFnAttr("memprof")); + EXPECT_TRUE(Call->hasMetadata(LLVMContext::MD_memprof)); + MDNode *MemProfMD = Call->getMetadata(LLVMContext::MD_memprof); + ASSERT_EQ(MemProfMD->getNumOperands(), 2u); + for (auto &MIBOp : MemProfMD->operands()) { + MDNode *MIB = dyn_cast(MIBOp); + MDNode *StackMD = getMIBStackNode(MIB); + ASSERT_NE(StackMD, nullptr); + ASSERT_EQ(StackMD->getNumOperands(), 2u); + auto *StackId = mdconst::dyn_extract(StackMD->getOperand(0)); + ASSERT_EQ(StackId->getZExtValue(), 1u); + StackId = mdconst::dyn_extract(StackMD->getOperand(1)); + if (StackId->getZExtValue() == 2u) + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::Cold); + else { + ASSERT_EQ(StackId->getZExtValue(), 3u); + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::NotCold); + } + } +} + +// Test CallStackTrie::addCallStack interface taking allocation type and list of +// call stack ids. +// Test that an allocation call reached by multiple call stacks has memprof +// metadata with the contexts trimmed to the minimum context required to +// identify the allocation type. +TEST_F(MemoryProfileInfoTest, TrimmedMIBContext) { + LLVMContext C; + std::unique_ptr M = makeLLVMModule(C, + R"IR( +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-pc-linux-gnu" +define i32* @test() { +entry: + %call = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40) + %0 = bitcast i8* %call to i32* + ret i32* %0 +} +declare dso_local noalias noundef i8* @malloc(i64 noundef) +)IR"); + + Function *Func = M->getFunction("test"); + + CallStackTrie Trie; + // We should be able to trim the following two and combine into a single MIB + // with the cold context {1, 2}. + Trie.addCallStack(AllocationType::Cold, {1, 2, 3}); + Trie.addCallStack(AllocationType::Cold, {1, 2, 4}); + // We should be able to trim the following two and combine into a single MIB + // with the non-cold context {1, 5}. + Trie.addCallStack(AllocationType::NotCold, {1, 5, 6}); + Trie.addCallStack(AllocationType::NotCold, {1, 5, 7}); + + CallBase *Call = findCall(*Func, "call"); + Trie.buildAndAttachMIBMetadata(Call); + + EXPECT_FALSE(Call->hasFnAttr("memprof")); + EXPECT_TRUE(Call->hasMetadata(LLVMContext::MD_memprof)); + MDNode *MemProfMD = Call->getMetadata(LLVMContext::MD_memprof); + ASSERT_EQ(MemProfMD->getNumOperands(), 2u); + for (auto &MIBOp : MemProfMD->operands()) { + MDNode *MIB = dyn_cast(MIBOp); + MDNode *StackMD = getMIBStackNode(MIB); + ASSERT_NE(StackMD, nullptr); + ASSERT_EQ(StackMD->getNumOperands(), 2u); + auto *StackId = mdconst::dyn_extract(StackMD->getOperand(0)); + EXPECT_EQ(StackId->getZExtValue(), 1u); + StackId = mdconst::dyn_extract(StackMD->getOperand(1)); + if (StackId->getZExtValue() == 2u) + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::Cold); + else { + ASSERT_EQ(StackId->getZExtValue(), 5u); + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::NotCold); + } + } +} + +// Test CallStackTrie::addCallStack interface taking memprof MIB metadata. +// Check that allocations annotated with memprof metadata with a single +// allocation type get simplified to an attribute. +TEST_F(MemoryProfileInfoTest, SimplifyMIBToAttribute) { + LLVMContext C; + std::unique_ptr M = makeLLVMModule(C, + R"IR( +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-pc-linux-gnu" +define i32* @test() { +entry: + %call1 = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40), !memprof !0 + %0 = bitcast i8* %call1 to i32* + %call2 = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40), !memprof !3 + %1 = bitcast i8* %call2 to i32* + ret i32* %1 +} +declare dso_local noalias noundef i8* @malloc(i64 noundef) +!0 = !{!1} +!1 = !{!2, !"cold"} +!2 = !{i64 1, i64 2, i64 3} +!3 = !{!4} +!4 = !{!5, !"notcold"} +!5 = !{i64 4, i64 5, i64 6, i64 7} +)IR"); + + Function *Func = M->getFunction("test"); + + // First call has all cold contexts. + CallStackTrie Trie1; + CallBase *Call1 = findCall(*Func, "call1"); + MDNode *MemProfMD1 = Call1->getMetadata(LLVMContext::MD_memprof); + ASSERT_EQ(MemProfMD1->getNumOperands(), 1u); + MDNode *MIB1 = dyn_cast(MemProfMD1->getOperand(0)); + Trie1.addCallStack(MIB1); + Trie1.buildAndAttachMIBMetadata(Call1); + + EXPECT_TRUE(Call1->hasFnAttr("memprof")); + EXPECT_EQ(Call1->getFnAttr("memprof").getValueAsString(), "cold"); + + // Second call has all non-cold contexts. + CallStackTrie Trie2; + CallBase *Call2 = findCall(*Func, "call2"); + MDNode *MemProfMD2 = Call2->getMetadata(LLVMContext::MD_memprof); + ASSERT_EQ(MemProfMD2->getNumOperands(), 1u); + MDNode *MIB2 = dyn_cast(MemProfMD2->getOperand(0)); + Trie2.addCallStack(MIB2); + Trie2.buildAndAttachMIBMetadata(Call2); + + EXPECT_TRUE(Call2->hasFnAttr("memprof")); + EXPECT_EQ(Call2->getFnAttr("memprof").getValueAsString(), "notcold"); +} + +// Test CallStackTrie::addCallStack interface taking memprof MIB metadata. +// Test that allocations annotated with memprof metadata with multiple call +// stacks gets new memprof metadata with the contexts trimmed to the minimum +// context required to identify the allocation type. +TEST_F(MemoryProfileInfoTest, ReTrimMIBContext) { + LLVMContext C; + std::unique_ptr M = makeLLVMModule(C, + R"IR( +target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" +target triple = "x86_64-pc-linux-gnu" +define i32* @test() { +entry: + %call = call noalias dereferenceable_or_null(40) i8* @malloc(i64 noundef 40), !memprof !0 + %0 = bitcast i8* %call to i32* + ret i32* %0 +} +declare dso_local noalias noundef i8* @malloc(i64 noundef) +!0 = !{!1, !3, !5, !7} +!1 = !{!2, !"cold"} +!2 = !{i64 1, i64 2, i64 3} +!3 = !{!4, !"cold"} +!4 = !{i64 1, i64 2, i64 4} +!5 = !{!6, !"notcold"} +!6 = !{i64 1, i64 5, i64 6} +!7 = !{!8, !"notcold"} +!8 = !{i64 1, i64 5, i64 7} +)IR"); + + Function *Func = M->getFunction("test"); + + CallStackTrie Trie; + ASSERT_TRUE(Trie.empty()); + CallBase *Call = findCall(*Func, "call"); + MDNode *MemProfMD = Call->getMetadata(LLVMContext::MD_memprof); + for (auto &MIBOp : MemProfMD->operands()) { + MDNode *MIB = dyn_cast(MIBOp); + Trie.addCallStack(MIB); + } + ASSERT_FALSE(Trie.empty()); + Trie.buildAndAttachMIBMetadata(Call); + + // We should be able to trim the first two and combine into a single MIB + // with the cold context {1, 2}. + // We should be able to trim the second two and combine into a single MIB + // with the non-cold context {1, 5}. + + EXPECT_FALSE(Call->hasFnAttr("memprof")); + EXPECT_TRUE(Call->hasMetadata(LLVMContext::MD_memprof)); + MemProfMD = Call->getMetadata(LLVMContext::MD_memprof); + ASSERT_EQ(MemProfMD->getNumOperands(), 2u); + for (auto &MIBOp : MemProfMD->operands()) { + MDNode *MIB = dyn_cast(MIBOp); + MDNode *StackMD = getMIBStackNode(MIB); + ASSERT_NE(StackMD, nullptr); + ASSERT_EQ(StackMD->getNumOperands(), 2u); + auto *StackId = mdconst::dyn_extract(StackMD->getOperand(0)); + EXPECT_EQ(StackId->getZExtValue(), 1u); + StackId = mdconst::dyn_extract(StackMD->getOperand(1)); + if (StackId->getZExtValue() == 2u) + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::Cold); + else { + ASSERT_EQ(StackId->getZExtValue(), 5u); + EXPECT_EQ(getMIBAllocType(MIB), AllocationType::NotCold); + } + } +} + +} // end anonymous namespace