Low-Overhead Dynamic Instruction Mix Generation using Hybrid Basic Block Profiling

Andrzej Pawel Nowak, Willy Zwaenepoel
2018
Article de conférence

Résumé

Dynamic instruction mixes form an important part of the toolkits of performance tuners, compiler writers, and CPU architects. Instruction mixes are traditionally generated using software instrumentation, an accurate yet slow method, that is normally limited to user-mode code. We present a new method for generating instruction mixes using the Performance Monitoring Unit (PMU) of the CPU. It has very low overhead, extends coverage to kernel-mode execution, and causes only a very modest decrease in accuracy, compared to software instrumentation. In order to achieve this level of accuracy, we develop a new PMU-based data collection method, Hybrid Basic Block Profiling (HBBP). HBBP uses simple machine learning techniques to choose, on a per basic block basis, between data from two conventional sampling methods, Event Based Sampling (EBS) and Last Branch Records (LBR). We implement a profiling tool based on HBBP, and we report on experiments with the industry standard SPEC CPU2006 suite, as well as with two large-scale scientific codes. We observe an improvement in runtime compared to software instrumentation of up to 76x on the tested benchmarks, reducing wait times from hours to minutes. Instruction attribution errors average 2.1%. The results indicate that HBBP provides a favorable tradeoff between accuracy and speed, making it a suitable candidate for use in production environments.

Official source

https://infoscience.epfl.ch/record/253296?ln=fr

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

The performance monitoring of computer systems is a complex affair, made even more challenging by the increasing gap between hardware and software. Methods that collect and feed data to performance analysis can usually be classified into one of two groups. Methods in the first group (e.g., software instrumentation), can be accurate but have unacceptable overheads and offer no insight into the interaction of the software and the hardware. Methods in the second group (e.g., hardware supported) run in real time and connect code to the response in the architecture, at a considerable decrease in accuracy, with a high degree of specificity to the hardware and therefore with limited actionable information made available. In this work, we focus on the latter group - the collection and analysis of raw performance data sourced from hardware Performance Monitoring Units, built into the majority of processors. The periodic collection of samples (Event Based Sampling), containing code locations, for instance, can be triggered on events such as retired instructions or cache misses. We quantify accuracy improvements to existing methods, propose a better performing method of performance data collection and a new architecture-agnostic method of performance data analysis. In our experiments, we analyze compiled, large-scale production workloads, industry standard benchmarks, and micro-benchmarks reproducing specific architectural behaviors. First, we focus on performance data collection methods and their accuracy, when establishing instruction retirement rates. These vary from least to most advanced and we examine the low level factors influencing accuracy. In particular, we employ a method based on Last Branch Records, a hardware facility present on some Intel architectures, which allows for the sampling of short records of last taken branches leading up to the sample. We compare the most commonly used method, its improved derivatives and the Last Branch Record based method. With respect to the first method, we observe accuracy improvements of up to 18x on synthetic benchmarks and 15x on real applications. Second, to further improve accuracy, we propose a new collection method named Hybrid Basic Block Profiling, fusing those based on Last Branch Records and the widely used Event Based Sampling. We apply simple machine learning techniques to determine the operating criteria. As a demonstration of capability, our method and a profiling tool are used to generate Instruction Mixes, which traditionally require good quality data at the most detailed level of granularity - basic blocks. Compared to software instrumentation, we observe an improvement in runtime of up to 76x, while keeping instruction attribution errors at 2.1% on average. In our third contribution we describe an architecture-agnostic performance analysis methodology called Hierarchical Cycle Accounting. The user is presented with a navigable hierarchy of microarchitectural issues, expressed in a metric that is universal and simple to compare - core cycles. We develop a reference implementation for the Intel Ivy Bridge microarchitecture and provide pointers for implementations on other processor architectures. Our approach and tool are successfully used by non-experts to improve and vectorize complex scientific code. Experts benefit as well - by being able to pinpoint software issues undetected by other methods, and identifying a bug in a family of Intel processors.

EPFL2017

Chargement

Processeur à haut degré de parallélisme basé sur des composantes sérielles

Ralph Hoffmann

Nowaday, the world of processors is still dominated by the RISC architectures, which foundations have been laid down in the 70's. The RISC concept may be summarized by one word : simplicity. With this concept, much simpler architectures are born, in particular from their instruction sets point of view. As a result, these architectures have a more efficient instruction decoding and smaller computing units which in turn provides the ability to have more of these units for the same price, a reduced clock per instruction ratio — a single cycle per instruction ideally — and increased clock frequencies. As time goes by, RISC architectures have shown outstanding performances, on one hand with the spectacular improvements of semiconductor technologies, and on the other hand with the use of large cache memories and with the development of new ways of execution such as speculative execution. Later developments have greatly altered the original spirit of the simplicity of the RISC concept. Current high performance processors are extremely complex and difficult to design. Moreover, semiconductor technologies begin to show serious problems ignored or neglected for a long time. In this thesis, we would like to introduce a new architecture. Two complementary aspects have been taken in consideration : the design of the processor's units (arithmetic operators, control logic, and so on) and the architectural model, the way the processing units are used. The first aspect has been addressed with an unusual and original approach which involves the use of the serial arithmetic principles. Serial arithmetic tends to reduce the hardware cost, enables higher clock rates, and shows interesting properties which will be developped later on. The research of an architectural model is driven by the same wish of simplicity, recurrent in this field, and will inevitably shift the complexity from the processor to the compiler. This simplification relies on two things : first, the way we think of the problem is inspired by the EPIC philosophy (Explicitly Parallel Instruction Computer), the latest offspring of the VLIW model which advocates the delegation of decisions from the hardware to the software. In one sentence, "the compiler decides, the hardware executes". Second, the way we act has been given by the TTA model (Transport Triggered Architecture), the last offspring of the MOVE model which not only removes from the processor the responsability to decide which processing unit to be used for each instruction (it's already the case in a VLIW processor), but also releases the processor from scheduling the data flows. In doing so, the processor is confined to provide a set of computing units, storage units and means of communication, and the compiler decides how to move the data. In this way, a part of the control logic is removed, communication resources may be fine-tuned and the compiler's code-optimization can be enhanced. With the combination of those elements, we wish to get an architecture which exhibits a better computing power/consumption ratio, which is more flexible, easier to developp and to evolve.

EPFL2004

Chargement

Reactive NUCA: Near-Optimal Block Placement and Replication in Distributed Caches

Anastasia Ailamaki, Babak Falsafi, Michael Ferdman

Increases in on-chip communication delay and the large working sets of server and scientific workloads complicate the design of the on-chip last- level cache for multicore processors. The large working sets favor a shared cache design that maximizes the aggregate cache capacity and minimizes off-chip memory requests. At the same time, the growing on-chip communication delay favors core-private caches that replicate data to minimize delays on global wires. Recent hybrid proposals offer lower average latency than conventional designs, but they address the placement requirements of only a subset of the data accessed by the application, require complex lookup and coherence mechanisms that increase latency, or fail to scale to high core counts. In this work, we observe that the cache access patterns of a range of server and scientific workloads can be classified into distinct classes, where each class is amenable to different block placement policies. Based on this observation, we propose Reactive NUCA (R- NUCA), a distributed cache design which reacts to the class of each cache access and places blocks at the appropriate location in the cache. R-NUCA cooperates with the operating system to support intelligent placement, migration, and replication without the overhead of an explicit coherence mechanism for the on-chip last-level cache. In a range of server, scientific, and multi-programmed workloads, R-NUCA matches the performance of the best cache design for each workload, improving performance by 14% on average over competing designs and by 32% at best, while achieving performance within 5% of an ideal cache design.

2009

Chargement

EPFL2017

Processeur à haut degré de parallélisme basé sur des composantes sérielles

Ralph Hoffmann

EPFL2004