LLVM

The LLVM Compiler Framework and Infrastructure Vikram Adve [email_address] Chris Lattner [email_address] http://llvm.cs.uiuc.edu/ LCPC Tutorial: September 22, 2004

Acknowledgements UIUC Contributors: Tanya Brethour Misha Brukman Cameron Buschardt John Criswell Alkis Evlogimenos Brian Gaeke Ruchira Sasanka Anand Shukla Bill Wendling External Contributors: Henrik Bach Nate Begeman Jeff Cohen Paolo Invernizzi Brad Jones Vladimir Merzliakov Vladimir Prus Reid Spencer Funding: This work is sponsored by the NSF Next Generation Software program through grants EIA-0093426 (an NSF CAREER award) and EIA-0103756. It is also supported in part by the NSF Operating Systems and Compilers program (grant #CCR-9988482), the NSF Embedded Systems program (grant #CCR-0209202), the MARCO/DARPA Gigascale Systems Research Center (GSRC), IBM through the DARPA-funded PERCS project, and the Motorola University Partnerships in Research program.

LLVM Compiler System The LLVM Compiler Infrastructure Provides reusable components for building compilers Reduce the time/cost to build a new compiler Build static compilers, JITs, trace-based optimizers, ... The LLVM Compiler Framework End-to-end compilers using the LLVM infrastructure C and C++ are robust and aggressive: Java, Scheme and others are in development Emit C code or native code for X86, Sparc, PowerPC

Three primary LLVM components The LLVM Virtual Instruction Set The common language- and target-independent IR Internal (IR) and external (persistent) representation A collection of well-integrated libraries Analyses, optimizations, code generators, JIT compiler, garbage collection support, profiling, … A collection of tools built from the libraries Assemblers, automatic debugger, linker, code generator, compiler driver, modular optimizer, …

Tutorial Overview Introduction to the running example LLVM C/C++ Compiler Overview High-level view of an example LLVM compiler The LLVM Virtual Instruction Set IR overview and type-system LLVM C++ IR and important API’s Basics, PassManager, dataflow, ArgPromotion Important LLVM Tools opt, code generator, JIT, test suite, bugpoint Example applications of LLVM

Running example: arg promotion Consider use of by-reference parameters: int callee(const int &X) { return X+1; } int caller() { return callee(4); } Eliminated load in callee Eliminated store in caller Eliminated stack slot for ‘tmp’ int callee(const int *X) { return *X+1; // memory load } int caller() { int tmp; // stack object tmp = 4; // memory store return callee(&tmp); } compiles to int callee(int X) { return X+1; } int caller() { return callee(4); } We want:

Why is this hard? Requires interprocedural analysis: Must change the prototype of the callee Must update all call sites  we must know all callers What about callers outside the translation unit? Requires alias analysis: Reference could alias other pointers in callee Must know that loaded value doesn’t change from function entry to the load Must know the pointer is not being stored through Reference might not be to a stack object!

The LLVM C/C++ Compiler From the high level, it is a standard compiler: Compatible with standard makefiles Uses GCC 3.4 C and C++ parser Distinguishing features: Uses LLVM optimizers, not GCC optimizers .o files contain LLVM IR/bytecode, not machine code Executable can be bytecode (JIT’d) or machine code llvmg++ llvmgcc C file C++ file .o file .o file llvm linker executable Compile Time Link Time

Looking into events at compile-time Dead Global Elimination, IP Constant Propagation, Dead Argument Elimination, Inlining, Reassociation, LICM, Loop Opts, Memory Promotion, Dead Store Elimination, ADCE, … llvmgcc C file .o file llvmg++ C++ file .o file Modified version of G++ Emits LLVM IR as text file Lowers C++ AST to LLVM Modified version of GCC Emits LLVM IR as text file Lowers C AST to LLVM LLVM IR Parser LLVM Verifier 40 LLVM Analysis & Optimization Passes LLVM .bc File Writer C to LLVM Frontend Compile-time Optimizer C++ to LLVM Frontend Compile-time Optimizer “cc1” “cc1plus” “ gccas” “ gccas”

Looking into events at link-time 20 LLVM Analysis & Optimization Passes Optionally “internalizes”: marks most functions as internal, to improve IPO Perfect place for argument promotion optimization! .o file .o file llvm linker executable Native Code Backend Native executable “ llc” C Code Backend C Compiler Native executable “ llc –march=c” “ gcc” Link in native .o files and libraries here LLVM Linker Link-time Optimizer .bc file for LLVM JIT .o file .o file

Goals of the compiler design Analyze and optimize as early as possible: Compile-time opts reduce modify-rebuild-execute cycle Compile-time optimizations reduce work at link-time (by shrinking the program) All IPA/IPO make an open-world assumption Thus, they all work on libraries and at compile-time “ Internalize” pass enables “whole program” optzn One IR (without lowering) for analysis & optzn Compile-time optzns can be run at link-time too! The same IR is used as input to the JIT IR design is the key to these goals!

Goals of LLVM IR Easy to produce, understand, and define! Language- and Target-Independent AST-level IR (e.g. ANDF, UNCOL) is not very feasible Every analysis/xform must know about ‘all’ languages One IR for analysis and optimization IR must be able to support aggressive IPO, loop opts, scalar opts, … high- and low-level optimization! Optimize as much as early as possible Can’t postpone everything until link or runtime No lowering in the IR!

LLVM Instruction Set Overview #1 Low-level and target-independent semantics RISC-like three address code Infinite virtual register set in SSA form Simple, low-level control flow constructs Load/store instructions with typed-pointers IR has text, binary, and in-memory forms for (i = 0; i < N; ++i) Sum(&A[i], &P); loop: %i.1 = phi int [ 0, %bb0 ], [ %i.2, %loop ] %AiAddr = getelementptr float* %A, int %i.1 call void %Sum( float %AiAddr, %pair* %P) %i.2 = add int %i.1, 1 %tmp.4 = setlt int %i.1, %N br bool %tmp.4, label %loop, label %outloop

LLVM Instruction Set Overview #2 High-level information exposed in the code Explicit dataflow through SSA form Explicit control-flow graph (even for exceptions) Explicit language-independent type-information Explicit typed pointer arithmetic Preserve array subscript and structure indexing for (i = 0; i < N; ++i) Sum(&A[i], &P); loop: %i.1 = phi int [ 0, %bb0 ], [ %i.2, %loop ] %AiAddr = getelementptr float* %A, int %i.1 call void %Sum( float %AiAddr, %pair* %P) %i.2 = add int %i.1, 1 %tmp.4 = setlt int %i.1, %N br bool %tmp.4, label %loop, label %outloop

LLVM Type System Details The entire type system consists of: Primitives: void, bool, float, ushort, opaque, … Derived: pointer, array, structure, function No high-level types: type-system is language neutral! Type system allows arbitrary casts: Allows expressing weakly-typed languages, like C Front-ends can implement safe languages Also easy to define a type-safe subset of LLVM See also: docs/LangRef.html

Lowering source-level types to LLVM Source language types are lowered: Rich type systems expanded to simple type system Implicit & abstract types are made explicit & concrete Examples of lowering: References turn into pointers: T&  T* Complex numbers: complex float  { float, float } Bitfields: struct X { int Y:4; int Z:2; }  { int } Inheritance: class T : S { int X; }  { S, int } Methods: class T { void foo(); }  void foo(T*) Same idea as lowering to machine code

LLVM Program Structure Module contains Functions/GlobalVariables Module is unit of compilation/analysis/optimization Function contains BasicBlocks/Arguments Functions roughly correspond to functions in C BasicBlock contains list of instructions Each block ends in a control flow instruction Instruction is opcode + vector of operands All operands have types Instruction result is typed

Our example, compiled to LLVM int callee(const int *X) { return *X+1; // load } int caller() { int T; // on stack T = 4; // store return callee(&T); } internal int %callee(int* %X) { %tmp.1 = load int* %X %tmp.2 = add int %tmp.1, 1 ret int %tmp.2 } int %caller() { %T = alloca int store int 4, int* %T %tmp.3 = call int %callee(int* %T) ret int %tmp.3 } Stack allocation is explicit in LLVM All loads/stores are explicit in the LLVM representation Linker “internalizes” most functions in most cases

Our example, desired transformation internal int %callee(int %X.val) { %tmp.2 = add int %X.val, 1 ret int %tmp.2 } int %caller() { %T = alloca int store int 4, int* %T %tmp.1 = load int* %T %tmp.3 = call int %callee(%tmp.1) ret int %tmp.3 } internal int %callee(int* %X) { %tmp.1 = load int* %X %tmp.2 = add int %tmp.1, 1 ret int %tmp.2 } int %caller() { %T = alloca int store int 4, int* %T %tmp.3 = call int %callee(int* %T) ret int %tmp.3 } Change the prototype for the function Insert load instructions into all callers Update all call sites of ‘callee’ Other transformation (-mem2reg) cleans up the rest int %caller() { %tmp.3 = call int %callee(int 4) ret int %tmp.3 }

LLVM Coding Basics Written in modern C++, uses the STL: Particularly the vector, set, and map classes LLVM IR is almost all doubly-linked lists: Module contains lists of Functions & GlobalVariables Function contains lists of BasicBlocks & Arguments BasicBlock contains list of Instructions Linked lists are traversed with iterators: Function *M = … for (Function::iterator I = M->begin(); I != M->end(); ++I) { BasicBlock &BB = *I; ... See also: docs/ProgrammersManual.html

LLVM Pass Manager Compiler is organized as a series of ‘passes’: Each pass is one analysis or transformation Four types of Pass: ModulePass : general interprocedural pass CallGraphSCCPass : bottom-up on the call graph FunctionPass : process a function at a time BasicBlockPass : process a basic block at a time Constraints imposed (e.g. FunctionPass): FunctionPass can only look at “current function” Cannot maintain state across functions See also: docs/WritingAnLLVMPass.html

Services provided by PassManager Optimization of pass execution: Process a function at a time instead of a pass at a time Example: If F, G, H are three functions in input pgm: “FFFFGGGGHHHH” not “FGHFGHFGHFGH” Process functions in parallel on an SMP (future work) Declarative dependency management: Automatically fulfill and manage analysis pass lifetimes Share analyses between passes when safe: e.g. “DominatorSet live unless pass modifies CFG” Avoid boilerplate for traversal of program See also: docs/WritingAnLLVMPass.html

Pass Manager + Arg Promotion #1/2 Arg Promotion is a CallGraphSCCPass: Naturally operates bottom-up on the CallGraph Bubble pointers from callees out to callers 24: #include "llvm/CallGraphSCCPass.h" 47: struct SimpleArgPromotion : public CallGraphSCCPass { Arg Promotion requires AliasAnalysis info To prove safety of transformation Works with any alias analysis algorithm though 48: virtual void getAnalysisUsage(AnalysisUsage &AU) const { AU.addRequired<AliasAnalysis>(); // Get aliases AU.addRequired<TargetData>(); // Get data layout CallGraphSCCPass::getAnalysisUsage(AU); // Get CallGraph }

Finally, implement runOnSCC (line 65): bool SimpleArgPromotion:: runOnSCC(const std::vector<CallGraphNode*> &SCC) { bool Changed = false, LocalChange; do { // Iterate until we stop promoting from this SCC. LocalChange = false; // Attempt to promote arguments from all functions in this SCC. for (unsigned i = 0, e = SCC.size(); i != e; ++i) LocalChange |= PromoteArguments(SCC[i]); Changed |= LocalChange; // Remember that we changed something. } while (LocalChange); return Changed; // Passes return true if something changed. } Pass Manager + Arg Promotion #2/2 static int foo(int ***P) { return ***P; } static int foo(int P_val_val_val) { return P_val_val_val; }

LLVM Dataflow Analysis LLVM IR is in SSA form: use-def and def-use chains are always available All objects have user/use info, even functions Control Flow Graph is always available: Exposed as BasicBlock predecessor/successor lists Many generic graph algorithms usable with the CFG Higher-level info implemented as passes: Dominators, CallGraph, induction vars, aliasing, GVN, … See also: docs/ProgrammersManual.html

Arg Promotion: safety check #1/4 #1: Function must be “internal” (aka “static”) 88: if (!F || !F->hasInternalLinkage()) return false; #2: Make sure address of F is not taken In LLVM, check that there are only direct calls using F 99: for (Value::use_iterator UI = F->use_begin(); UI != F->use_end(); ++UI) { CallSite CS = CallSite::get(*UI); if (!CS.getInstruction()) // "Taking the address" of F. return false; #3: Check to see if any args are promotable: 114: for (unsigned i = 0; i != PointerArgs.size(); ++i) if (!isSafeToPromoteArgument(PointerArgs[i])) PointerArgs.erase(PointerArgs.begin()+i); if (PointerArgs.empty()) return false; // no args promotable

Arg Promotion: safety check #2/4 #4: Argument pointer can only be loaded from: No stores through argument pointer allowed! // Loop over all uses of the argument (use-def chains). 138: for (Value::use_iterator UI = Arg->use_begin(); UI != Arg->use_end(); ++UI) { // If the user is a load: if (LoadInst *LI = dyn_cast<LoadInst>(*UI)) { // Don't modify volatile loads. if (LI->isVolatile()) return false; Loads.push_back(LI); } else { return false; // Not a load. } }

#5: Value of “*P” must not change in the BB We move load out to the caller, value cannot change! // Get AliasAnalysis implementation from the pass manager. 156: AliasAnalysis &AA = getAnalysis<AliasAnalysis>(); // Ensure *P is not modified from start of block to load 169: if (AA.canInstructionRangeModify(BB->front(), *Load, Arg, LoadSize)) return false; // Pointer is invalidated! Arg Promotion: safety check #3/4 See also: docs/AliasAnalysis.html … … load P … Modifies “*P”?

#6: “*P” cannot change from Fn entry to BB 175: for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) // Loop over predecessors of BB. // Check each block from BB to entry (DF search on inverse graph). for (idf_iterator<BasicBlock*> I = idf_begin(*PI); I != idf_end(*PI); ++I) // Might *P be modified in this basic block? if (AA.canBasicBlockModify(**I, Arg, LoadSize)) return false; Arg Promotion: safety check #4/4 Entry load P load P Entry Modifies “*P”? Modifies “*P”?

Arg Promotion: xform outline #1/4 #1: Make prototype with new arg types: #197 Basically just replaces ‘int*’ with ‘int’ in prototype #2: Create function with new prototype: 214: Function *NF = new Function(NFTy, F->getLinkage(), F->getName()); F->getParent()->getFunctionList().insert(F, NF); #3: Change all callers of F to call NF: // If there are uses of F, then calls to it remain. 221: while (!F->use_empty()) { // Get a caller of F. CallSite CS = CallSite::get(F->use_back());

Arg Promotion: xform outline #2/4 #4: For each caller, add loads, determine args Loop over the args, inserting the loads in the caller 220: std::vector<Value*> Args; 226: CallSite::arg_iterator AI = CS.arg_begin(); for (Function::aiterator I = F->abegin(); I != F->aend(); ++I, ++AI) if (!ArgsToPromote.count(I)) // Unmodified argument. Args.push_back(*AI); else { // Insert the load before the call. LoadInst *LI = new LoadInst(*AI, (*AI)->getName()+".val", Call); // Insertion point Args.push_back(LI); }

Arg Promotion: xform outline #3/4 #5: Replace the call site of F with call of NF // Create the call to NF with the adjusted arguments. 242: Instruction *New = new CallInst(NF, Args, "", Call); // If the return value of the old call was used, use the retval of the new call. if (!Call->use_empty()) Call->replaceAllUsesWith(New); // Finally, remove the old call from the program, reducing the use-count of F. Call->getParent()->getInstList().erase(Call); #6: Move code from old function to new Fn 259: NF->getBasicBlockList().splice(NF->begin(), F->getBasicBlockList());

Arg Promotion: xform outline #4/4 #7: Change users of F’s arguments to use NF’s 264: for (Function::aiterator I = F->abegin(), I2 = NF->abegin(); I != F->aend(); ++I, ++I2) if (!ArgsToPromote.count(I)) { // Not promoting this arg? I->replaceAllUsesWith(I2); // Use new arg, not old arg. } else { while (!I->use_empty()) { // Only users can be loads. LoadInst *LI = cast<LoadInst>(I->use_back()); LI->replaceAllUsesWith(I2); LI->getParent()->getInstList().erase(LI); } } #8: Delete old function: 286: F->getParent()->getFunctionList().erase(F);

LLVM tools: two flavors “ Primitive” tools: do a single job llvm-as: Convert from .ll (text) to .bc (binary) llvm-dis: Convert from .bc (binary) to .ll (text) llvm-link: Link multiple .bc files together llvm-prof: Print profile output to human readers llvmc: Configurable compiler driver Aggregate tools: pull in multiple features gccas/gccld: Compile/link-time optimizers for C/C++ FE bugpoint: automatic compiler debugger llvm-gcc/llvm-g++: C/C++ compilers See also: docs/CommandGuide/

opt tool: LLVM modular optimizer Invoke arbitrary sequence of passes: Completely control PassManager from command line Supports loading passes as plugins from .so files opt -load foo.so -pass1 -pass2 -pass3 x.bc -o y.bc Passes “register” themselves: 61: RegisterOpt<SimpleArgPromotion> X("simpleargpromotion", "Promote 'by reference' arguments to 'by value'"); From this, they are exposed through opt: > opt -load libsimpleargpromote.so –help ... -sccp - Sparse Conditional Constant Propagation -simpleargpromotion - Promote 'by reference' arguments to 'by -simplifycfg - Simplify the CFG ...

Running Arg Promotion with opt Basic execution with ‘opt’: opt -simpleargpromotion in.bc -o out.bc Load .bc file, run pass, write out results Use “-load filename.so” if compiled into a library PassManager resolves all dependencies Optionally choose an alias analysis to use: opt –basicaa –simpleargpromotion (default) Alternatively, –steens-aa, –anders-aa, –ds-aa, … Other useful options available: -stats : Print statistics collected from the passes -time-passes : Time each pass being run, print output

Example -stats output (gccas 176.gcc) ===-------------------------------------------------------------------------=== ... Statistics Collected ... ===-------------------------------------------------------------------------=== 23426 adce - Number of instructions removed 1663 adce - Number of basic blocks removed 5052592 bytecodewriter - Number of bytecode bytes written 57489 cfgsimplify - Number of blocks simplified 4186 constmerge - Number of global constants merged 211 dse - Number of stores deleted 15943 gcse - Number of loads removed 54245 gcse - Number of instructions removed 253 inline - Number of functions deleted because all callers found 3952 inline - Number of functions inlined 9425 instcombine - Number of constant folds 160469 instcombine - Number of insts combined 208 licm - Number of load insts hoisted or sunk 4982 licm - Number of instructions hoisted out of loop 350 loop-unroll - Number of loops completely unrolled 30156 mem2reg - Number of alloca's promoted 2934 reassociate - Number of insts with operands swapped 650 reassociate - Number of insts reassociated 67 scalarrepl - Number of allocas broken up 279 tailcallelim - Number of tail calls removed 25395 tailduplicate - Number of unconditional branches eliminated ..........................

Example -time-passes (gccas 176.gcc) ===-------------------------------------------------------------------------=== ... Pass execution timing report ... ===-------------------------------------------------------------------------=== ---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name --- 16.2400 ( 23.0%) 0.0000 ( 0.0%) 16.2400 ( 22.9%) 16.2192 ( 22.9%) Global Common Subexpression Elimination 11.1200 ( 15.8%) 0.0499 ( 13.8%) 11.1700 ( 15.8%) 11.1028 ( 15.7%) Reassociate expressions 6.5499 ( 9.3%) 0.0300 ( 8.3%) 6.5799 ( 9.3%) 6.5824 ( 9.3%) Bytecode Writer 3.2499 ( 4.6%) 0.0100 ( 2.7%) 3.2599 ( 4.6%) 3.2140 ( 4.5%) Scalar Replacement of Aggregates 3.0300 ( 4.3%) 0.0499 ( 13.8%) 3.0800 ( 4.3%) 3.0382 ( 4.2%) Combine redundant instructions 2.6599 ( 3.7%) 0.0100 ( 2.7%) 2.6699 ( 3.7%) 2.7339 ( 3.8%) Dead Store Elimination 2.1600 ( 3.0%) 0.0300 ( 8.3%) 2.1900 ( 3.0%) 2.1924 ( 3.1%) Function Integration/Inlining 2.1600 ( 3.0%) 0.0100 ( 2.7%) 2.1700 ( 3.0%) 2.1125 ( 2.9%) Sparse Conditional Constant Propagation 1.6600 ( 2.3%) 0.0000 ( 0.0%) 1.6600 ( 2.3%) 1.6389 ( 2.3%) Aggressive Dead Code Elimination 1.4999 ( 2.1%) 0.0100 ( 2.7%) 1.5099 ( 2.1%) 1.4462 ( 2.0%) Tail Duplication 1.5000 ( 2.1%) 0.0000 ( 0.0%) 1.5000 ( 2.1%) 1.4410 ( 2.0%) Post-Dominator Set Construction 1.3200 ( 1.8%) 0.0000 ( 0.0%) 1.3200 ( 1.8%) 1.3722 ( 1.9%) Canonicalize natural loops 1.2700 ( 1.8%) 0.0000 ( 0.0%) 1.2700 ( 1.7%) 1.2717 ( 1.7%) Merge Duplicate Global Constants 1.0300 ( 1.4%) 0.0000 ( 0.0%) 1.0300 ( 1.4%) 1.1418 ( 1.6%) Combine redundant instructions 0.9499 ( 1.3%) 0.0400 ( 11.1%) 0.9899 ( 1.4%) 0.9979 ( 1.4%) Raise Pointer References 0.9399 ( 1.3%) 0.0100 ( 2.7%) 0.9499 ( 1.3%) 0.9688 ( 1.3%) Simplify the CFG 0.9199 ( 1.3%) 0.0300 ( 8.3%) 0.9499 ( 1.3%) 0.8993 ( 1.2%) Promote Memory to Register 0.9600 ( 1.3%) 0.0000 ( 0.0%) 0.9600 ( 1.3%) 0.8742 ( 1.2%) Loop Invariant Code Motion 0.5600 ( 0.7%) 0.0000 ( 0.0%) 0.5600 ( 0.7%) 0.6022 ( 0.8%) Module Verifier …

Analyze tool: Visualize analysis results Print most LLVM data structures Dominators, loops, alias sets, CFG, call graph, … Converts most LLVM data structures to ‘dot’ graphs

LLC Tool: Static code generator Compiles LLVM  native assembly language Currently for X86, Sparc, PowerPC (others in alpha) llc file.bc -o file.s -march=x86 as file.s –o file.o Compiles LLVM  portable C code llc file.bc -o file.c -march=c gcc –c file.c –o file.o Targets are modular & dynamically loadable: llc –load libarm.so file.bc -march=arm

The LLVM Code Generator Target independent: Driven by an algorithm independent target description Data layout, Register, Instruction, Scheduling, … Basic code generator layout: All passes are replaceable e.g. Trivial to change and add register allocators Targets can add custom passes e.g. X86 has special support for FP stack See also: docs/CodeGenerator.html Instruction Selection LLVM .s file Machine SSA Opts Register Allocator Instr Sched Code Emission Exposes all target-specific details about a function (calling conventions, etc) Scheduling, Peephole, ? 4 algorithms available today llc -regalloc=foo Target Independent Target Specific (generated) Target Specific (by hand for now)

Porting LLVM to a new target LLVM targets are very easy to write: Anecdotal evidence suggests 1 week for a basic port … for someone familiar with the target machine and compilers in general, but not with LLVM LLVM targets are written with “tablegen” tool Simple declarative syntax Designed to factor out redundancy in target desc Some C++ code is still required Primarily in the instruction selector Continuing work to improve this See also: docs/TableGenFundamentals.html and WritingAnLLVMBackend.html

LLI allows direct execution of .bc files E.g.: lli grep.bc -i foo *.c LLI uses a Just-In-Time compiler if available: Uses same code generator as LLC Optionally uses faster components than LLC Emits machine code to memory instead of “.s” file JIT is a library that can be embedded in other tools Otherwise, it uses the LLVM interpreter: Interpreter is extremely simple and very slow Interpreter is portable though! LLI Tool: LLVM Execution Engine

C and C++ Program Test Suite Large collection of programs and benchmarks: Standard suites (e.g. SPEC 95/2000, Olden, Ptrdist, McCat, Stanford, Freebench, Shootout…) Individual programs: sgefa, siod, sim, pi, povray, … Proprietary suites (e.g. SPEC) require suite source Consistent build environment: Easy add hooks to build for profiling/instrumentation Easy to get performance numbers from entire test suite Entire test suite is checked every night: Hosted on Linux,Solaris,FreeBSD on X86,Sparc & PPC See also: docs/TestingGuide.html

Integrated Debugging Tools Extensive assertions throughout code Find problems as early as possible (close to source) LLVM IR Verifier: Checks modules for validity Checks type properties, dominance properties, etc. Automatically run by opt Problem found?: print an error message and abort LLVM IR Leak Detector Efficient and simple “garbage collector” for IR objects Ensure IR objects are deallocated appropriately

The Bugpoint automated bug finder Simple idea: automate ‘binary’ search for bug Bug isolation: which passes interact to produce bug Test case reduction : reduce input program Optimizer/Codegen crashes: Throw portion of test case away, check for crash If so, keep going Otherwise, revert and try something else Extremely effective in practice Simple greedy algorithms for test reduction Completely black-box approach See also: docs/Bugpoint.html

Debugging Miscompilations Optimizer miscompilation: Split testcase in two, optimize one. Still broken? Keep shrinking the portion being optimized Codegen miscompilation: Split testcase in two, compile one with CBE, broken? Shrink portion being compiled with non CBE codegen Code splitting granularities: Take out whole functions Take out loop nests Take out individual basic blocks

How well does this thing work? Extremely effective: Can often reduce a 100K LOC program and 60 passes to a few basic blocks and 1 pass in 5 minutes Crashes are found much faster than miscompilations no need to run the program to test a reduction Interacts with integrated debugging tools Runtime errors are detected faster Limitations: Program must be deterministic … or modified to be so Finds “a” bug, not “the” bug

Use Case 1: Edge or Path Profiling Goal: Profiling Research or PGO Implementation: FunctionPass : LLVM-to-LLVM transformation Instrumentation: Use CFG, intervals, dominators Code generation: Use C or any native back end Profile feedback: Use profile query interface Core extensions needed: none Major LLVM Benefits Language-independence, CFG, very simple IR

Use Case 2: Alias Analysis Goal: Research on new alias analysis algorithms Implementation: ModulePass : Whole-program analysis pass on LLVM Use type information; SSA; heap/stack/globals Compare SimpleAA , Steensgard’s , Andersen’s , DSA Evaluate many clients via AliasAnalysis interface Core extensions needed: none Major LLVM Benefits Language-independence, type info, SSA, DSA, IPO AliasAnalysis interface with many pre-existing clients

Use Case 3: LDS Prefetching Goal: Prefetching linked data structures Implementation: ModulePass : Link-time LLVM-to-LLVM transformation Code transformations : use type info, loop analysis, unrolling, prefetch insertion Data transformations (e.g,. adding history pointers): use strong type info from DSA, IPO Core extensions needed: Prefetch operation: add as intrinsic (in progress) Major LLVM Benefits Language-independence, type info, DSA, IPO

Use Case 4: Language Front end Goal: Use LLVM to implement a new language Implementation: Parser (say to AST), Semantic checking AST-to-LLVM translator Core extensions needed: depends High-level type system is omitted by design Major LLVM Benefits Low-level, but powerful type system Very simple IR to generate (e.g., compare GCC RTL) Extensive global and IP optimization framework JIT engine, native back-ends, C back-end

Use Case 5: JIT Compiler Goal: Write JIT compiler for a bytecode language Implementation: Extend the LLVM JIT framework Simple JIT : Fast translation from bytecode to LLVM (then use LLVM JIT + GC) Optimizing JIT : Language-specific optimizations + fast translation (then use LLVM optimizations, JIT, GC) Core extensions needed: none in general Major LLVM Benefits Compact, typed, language-independent IR Existing JIT framework and GC

Use Case 6: Architecture Research Goal: Compiler support for new architectures Implementation: Add new machine description (or modify one) Add any new LLVM-to-LLVM transformations Core extensions needed: depends on goals Imminent features: modulo sched; vector ops Major LLVM Benefits Low-level, typed, machine-independent IR Explicit register/memory architecture Aggressive mid-level and back-end compiler framework Full-system evaluation: applications, libraries, even OS

Five point LLVM Review Extremely simple IR to learn and use 1-to-1 correspondence between .ll, .bc, and C++ IR Very positive user reactions Powerful and modular optimizer Easy to extend, or just use what is already there Clean and modular code generator Easy to retarget, easy to replace/tweak components Many “productivity tools” (bugpoint, verifier) Get more done, quicker! Active dev community, good documentation Mailing lists, IRC, doxygen, extensive docs

Get started with LLVM! Download latest release or CVS: http://llvm.cs.uiuc.edu/releases/ Follow the “Getting Started Guide”: http://llvm.cs.uiuc.edu/docs/GettingStarted.html Walks you through install and setup Lots of other docs available in “docs” directory Join us on mailing lists and IRC Happy hacking!

LLVM

More Related Content

What's hot

Similar to LLVM

Recently uploaded

In this document

LLVM