High-level adaptive program optimization with ADAPT

Abstract

Compile-time optimization is often limited by a lack of target machine and input data set knowledge. Without this information, compilers may be forced to make conservative assumptions to preserve correctness and to avoid performance degradation. In order to cope with this lack of information at compile-time, adaptive and dynamic systems can be used to perform optimization at runtime when complete knowledge of input and machine parameters is available. This paper presents a compiler-supported high-level adaptive optimization system. Users describe, in a domain specific language, optimizations performed by stand-alone optimization tools and backend compiler flags, as well as heuristics for applying these optimizations dynamically at runtime. The ADAPT compiler reads these descriptions and generates application-specific runtime systems to apply the heuristics. To facilitate the usage of existing tools and compilers, overheads are minimized by decoupling optimization from execution. Our system, ADAPT, supports a range of paradigms proposed recently, including dynamic compilation, parameterization and runtime sampling. We demonstrate our system by applying several optimization techniques to a suite of benchmarks on two target machines. ADAPT is shown to consistently outperform statically generated executables, improving performance by as much as 70%.

Automated empirical tuning of scientific codes for performance and power consumption

“…The flaws in the search algorithm can limit the effectiveness of collectively tuning for both performance and power consumption, as demonstrated by the less-than-ideal tuning result for gemm in Figure 10. Our tuning approach, however, is independent of specific search algorithms and can easily collaborate other existing alternative generic search algorithms [6,13,21].…”

“…The focus of this paper is to explore the configuration space of a large number of POET-implemented compiler optimizations for both performance and energy efficiency. Instead of using an existing generic search technique [6,13,21], we have developed a transformation-aware search engine to explore the optimization space.…”

Dynamic parallelization and mapping of binary executables on hierarchical platforms

“…Subsequent invocations of such regions use the optimized copy in the software cache, which over time hold all the frequently executed code. The ADAPT system by Voss et al [21] is similar, obtaining performance increases with a system that called standard static compilers on remote machines (changing optimization flags and strategies) during program execution.…”

Empirical Optimization for a Sparse Linear Solver: A Case Study

“…The resulting code is complex and virtually impossible to port and maintain. In ADAPT, (27) researchers have defined new languages used exclusively to specify adaptation policies, triggering events and performance metrics. Programmers develop their base application textually independently from their adaptation specification.…”