Flow-sensitive Pointer Analysis Research Articles

Loop-Oriented Pointer Analysis for Automatic SIMD Vectorization

Compiler-based vectorization represents a promising solution to automatically generate code that makes efficient use of modern CPUs with SIMD extensions. Two main auto-vectorization techniques, superword-level parallelism vectorization (SLP) and loop-level vectorization (LLV), require precise dependence analysis on arrays and structs to vectorize isomorphic scalar instructions (in the case of SLP) and reduce dynamic dependence checks at runtime (in the case of LLV). The alias analyses used in modern vectorizing compilers are either intra-procedural (without tracking inter-procedural data-flows) or inter-procedural (by using field-sensitive models, which are too imprecise in handling arrays and structs). This article proposes an inter-procedural L oop-oriented P ointer A nalysis for C, called L pa , for analyzing arrays and structs to support aggressive SLP and LLV optimizations effectively. Unlike field-insensitive solutions that pre-allocate objects for each memory allocation site, our approach uses a lazy memory model to generate access-based location sets based on how structs and arrays are accessed. L pa can precisely analyze arrays and nested aggregate structures to enable SIMD optimizations for large programs. By separating the location set generation as an independent concern from the rest of the pointer analysis, L pa is designed so that existing points-to resolution algorithms (e.g., flow-insensitive and flow-sensitive pointer analysis) can be reused easily. We have implemented L pa fully in the LLVM compiler infrastructure (version 3.8.0). We evaluate L pa by considering SLP and LLV, the two classic vectorization techniques, on a set of 20 C and Fortran CPU2000/2006 benchmarks. For SLP, L pa outperforms LLVM’s BasicAA and ScevAA by discovering 139 and 273 more vectorizable basic blocks, respectively, resulting in the best speedup of 2.95% for 173.applu. For LLV, LLVM introduces totally 551 and 652 static bound checks under BasicAA and ScevAA, respectively. In contrast, L pa has reduced these static checks to 220, with an average of 15.7 checks per benchmark, resulting in the best speedup of 7.23% for 177.mesa.

Time- and space-efficient flow-sensitive points-to analysis

Compilation of real-world programs often requires hours. The term nightly build known to industrial researchers is an artifact of long compilation times. Our goal is to reduce the absolute analysis times for large C codes (of the order of millions of lines). Pointer analysis is one of the key analyses performed during compilation. Its scalability is paramount to achieve the efficiency of the overall compilation process and its precision directly affects that of the client analyses. In this work, we design a time- and space-efficient flow-sensitive pointer analysis and parallelize it on graphics processing units. Our analysis proposes to use an extended bloom filter, called multibloom, to store points-to information in an approximate manner and develops an analysis in terms of the operations over the multibloom. Since bloom filter is a probabilistic data structure, we develop ways to gain back the analysis precision. We achieve effective parallelization by achieving memory coalescing, reducing thread divergence, and improving load balance across GPU warps. Compared to a state-of-the-art sequential solution, our parallel version achieves a 7.8 × speedup with less than 5% precision loss on a suite of six large programs. Using two client transformations, we show that this loss in precision only minimally affects a client’s precision.

Time- and space-efficient flow-sensitive points-to analysis

Compilation of real-world programs often requires hours. The term nightly build known to industrial researchers is an artifact of long compilation times. Our goal is to reduce the absolute analysis times for large C codes (of the order of millions of lines). Pointer analysis is one of the key analyses performed during compilation. Its scalability is paramount to achieve the efficiency of the overall compilation process and its precision directly affects that of the client analyses. In this work, we design a time- and space-efficient flow-sensitive pointer analysis and parallelize it on graphics processing units. Our analysis proposes to use an extended bloom filter, called multibloom, to store points-to information in an approximate manner and develops an analysis in terms of the operations over the multibloom. Since bloom filter is a probabilistic data structure, we develop ways to gain back the analysis precision. We achieve effective parallelization by achieving memory coalescing, reducing thread divergence, and improving load balance across GPU warps. Compared to a state-of-the-art sequential solution, our parallel version achieves a 7.8 × speedup with less than 5% precision loss on a suite of six large programs. Using two client transformations, we show that this loss in precision only minimally affects a client’s precision.

Dead code elimination based pointer analysis for multithreaded programs

Abstract This paper presents a new approach for optimizing multitheaded programs with pointer constructs. The approach has applications in the area of certified code (proof-carrying code) where a justification or a proof for the correctness of each optimization is required. The optimization meant here is that of dead code elimination. Towards optimizing multithreaded programs the paper presents a new operational semantics for parallel constructs like join-fork constructs, parallel loops, and conditionally spawned threads. The paper also presents a novel type system for flow-sensitive pointer analysis of multithreaded programs. This type system is extended to obtain a new type system for live-variables analysis of multithreaded programs. The live-variables type system is extended to build the third novel type system, proposed in this paper, which carries the optimization of dead code elimination. The justification mentioned above takes the form of type derivation in our approach.

Semi-sparse flow-sensitive pointer analysis

Pointer analysis is a prerequisite for many program analyses, and the effectiveness of these analyses depends on the precision of the pointer information they receive. Two major axes of pointer analysis precision are flow-sensitivity and context-sensitivity, and while there has been significant recent progress regarding scalable context-sensitive pointer analysis, relatively little progress has been made in improving the scalability of flow-sensitive pointer analysis. This paper presents a new interprocedural, flow-sensitive pointer analysis algorithm that combines two ideas-semi-sparse analysis and a novel use of BDDs-that arise from a careful understanding of the unique challenges that face flow-sensitive pointer analysis. We evaluate our algorithm on 12 C benchmarks ranging from 11K to 474K lines of code. Our fastest algorithm is on average 197x faster and uses 4.6x less memory than the state of the art, and it can analyze programs that are an order of magnitude larger than the previous state of the art.

General flow-sensitive pointer analysis and call graph construction

General flow-sensitive pointer analysis and call graph construction

On the non-approximability of points-to analysis

Determining points-to sets is an important static-analysis problem. Most of the classic static analyses (used e.g., by compilers or in programming environments) rely on knowing which variables might be used or defined by each expression in a program. In the presence of pointers, the use/def set of an expression like *p = *q can only be determined given (safe) points-to sets for p and q. Previous work has shown that both precise flow-sensitive and precise flow-insensitive pointer analysis is NP-Hard, even when restricted to single-procedure programs with no dynamic memory allocation. In this paper, we show that it is not even possible to compute good approximations to the precise solutions (i.e., to compute points-to sets whose sizes are within a constant factor of the sizes of the precise points-to sets) unless P=NP.

Write barrier removal by static analysis

We present a set of static analyses for removing write barriers in programs that use generational garbage collection. To our knowledge, these are the first analyses for this purpose. Our Intraprocedural analysis uses a flow-sensitive pointer analysis to locate variables that must point to the most recently allocated object, then eliminates write barriers on stores to objects accessed via one of these variables. The Callee Type Extension incorporates information about the types of objects allocated in invoked methods, while the Caller Context Extension incorporates information about the most recently allocated object at call sites that invoke the currently analyzed method. Results from our implemented system show that our Full Interprocedural analysis, which incorporates both extensions, can eliminate the majority of the write barriers in most of the programs in our benchmark set, producing modest performance improvements of up to 7% of the overall execution time. Moreover, by dynamically instrumenting the executable, we are able to show that for all but two of our nine benchmark programs, our analysis is close to optimal in the sense that it eliminates the write barriers for almost all store instructions observed not to create a reference from an older object to a younger object.