GCC—An Architectural Overview, Current Status, and Future Directions
Total Page:16
File Type:pdf, Size:1020Kb
Load more
Recommended publications
-
CSE 582 – Compilers
CSE P 501 – Compilers Optimizing Transformations Hal Perkins Autumn 2011 11/8/2011 © 2002-11 Hal Perkins & UW CSE S-1 Agenda A sampler of typical optimizing transformations Mostly a teaser for later, particularly once we’ve looked at analyzing loops 11/8/2011 © 2002-11 Hal Perkins & UW CSE S-2 Role of Transformations Data-flow analysis discovers opportunities for code improvement Compiler must rewrite the code (IR) to realize these improvements A transformation may reveal additional opportunities for further analysis & transformation May also block opportunities by obscuring information 11/8/2011 © 2002-11 Hal Perkins & UW CSE S-3 Organizing Transformations in a Compiler Typically middle end consists of many individual transformations that filter the IR and produce rewritten IR No formal theory for order to apply them Some rules of thumb and best practices Some transformations can be profitably applied repeatedly, particularly if others transformations expose more opportunities 11/8/2011 © 2002-11 Hal Perkins & UW CSE S-4 A Taxonomy Machine Independent Transformations Realized profitability may actually depend on machine architecture, but are typically implemented without considering this Machine Dependent Transformations Most of the machine dependent code is in instruction selection & scheduling and register allocation Some machine dependent code belongs in the optimizer 11/8/2011 © 2002-11 Hal Perkins & UW CSE S-5 Machine Independent Transformations Dead code elimination Code motion Specialization Strength reduction -
Generalized Jump Threading in Libfirm
Institut für Programmstrukturen und Datenorganisation (IPD) Lehrstuhl Prof. Dr.-Ing. Snelting Generalized Jump Threading in libFIRM Masterarbeit von Joachim Priesner an der Fakultät für Informatik Erstgutachter: Prof. Dr.-Ing. Gregor Snelting Zweitgutachter: Prof. Dr.-Ing. Jörg Henkel Betreuender Mitarbeiter: Dipl.-Inform. Andreas Zwinkau Bearbeitungszeit: 5. Oktober 2016 – 23. Januar 2017 KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft www.kit.edu Zusammenfassung/Abstract Jump Threading (dt. „Sprünge fädeln“) ist eine Compileroptimierung, die statisch vorhersagbare bedingte Sprünge in unbedingte Sprünge umwandelt. Bei der Ausfüh- rung kann ein Prozessor bedingte Sprünge zunächst nur heuristisch mit Hilfe der Sprungvorhersage auswerten. Sie stellen daher generell ein Performancehindernis dar. Die Umwandlung ist insbesondere auch dann möglich, wenn das Sprungziel nur auf einer Teilmenge der zu dem Sprung führenden Ausführungspfade statisch be- stimmbar ist. In diesem Fall, der den überwiegenden Teil der durch Jump Threading optimierten Sprünge betrifft, muss die Optimierung Grundblöcke duplizieren, um jene Ausführungspfade zu isolieren. Verschiedene aktuelle Compiler enthalten sehr unterschiedliche Implementierungen von Jump Threading. In dieser Masterarbeit wird zunächst ein theoretischer Rahmen für Jump Threading vorgestellt. Sodann wird eine allgemeine Fassung eines Jump- Threading-Algorithmus entwickelt, implementiert und in diverser Hinsicht untersucht, insbesondere auf Wechselwirkungen mit anderen Optimierungen wie If -
ICS803 Elective – III Multicore Architecture Teacher Name: Ms
ICS803 Elective – III Multicore Architecture Teacher Name: Ms. Raksha Pandey Course Structure L T P 3 1 0 4 Prerequisite: Course Content: Unit-I: Multi-core Architectures Introduction to multi-core architectures, issues involved into writing code for multi-core architectures, Virtual Memory, VM addressing, VA to PA translation, Page fault, TLB- Parallel computers, Instruction level parallelism (ILP) vs. thread level parallelism (TLP), Performance issues, OpenMP and other message passing libraries, threads, mutex etc. Unit-II: Multi-threaded Architectures Brief introduction to cache hierarchy - Caches: Addressing a Cache, Cache Hierarchy, States of Cache line, Inclusion policy, TLB access, Memory Op latency, MLP, Memory Wall, communication latency, Shared memory multiprocessors, General architectures and the problem of cache coherence, Synchronization primitives: Atomic primitives; locks: TTS, ticket, array; barriers: central and tree; performance implications in shared memory programs; Chip multiprocessors: Why CMP (Moore's law, wire delay); shared L2 vs. tiled CMP; core complexity; power/performance; Snoopy coherence: invalidate vs. update, MSI, MESI, MOESI, MOSI; performance trade-offs; pipelined snoopy bus design; Memory consistency models: SC, PC, TSO, PSO, WO/WC, RC; Chip multiprocessor case studies: Intel Montecito and dual-core, Pentium4, IBM Power4, Sun Niagara Unit-III: Compiler Optimization Issues Code optimizations: Copy Propagation, dead Code elimination , Loop Optimizations-Loop Unrolling, Induction variable Simplification, Loop Jamming, Loop Unswitching, Techniques to improve detection of parallelism: Scalar Processors, Special locality, Temporal locality, Vector machine, Strip mining, Shared memory model, SIMD architecture, Dopar loop, Dosingle loop. Unit-IV: Control Flow analysis Control flow analysis, Flow graph, Loops in Flow graphs, Loop Detection, Approaches to Control Flow Analysis, Reducible Flow Graphs, Node Splitting. -
Fast-Path Loop Unrolling of Non-Counted Loops to Enable Subsequent Compiler Optimizations∗
Fast-Path Loop Unrolling of Non-Counted Loops to Enable Subsequent Compiler Optimizations∗ David Leopoldseder Roland Schatz Lukas Stadler Johannes Kepler University Linz Oracle Labs Oracle Labs Austria Linz, Austria Linz, Austria [email protected] [email protected] [email protected] Manuel Rigger Thomas Würthinger Hanspeter Mössenböck Johannes Kepler University Linz Oracle Labs Johannes Kepler University Linz Austria Zurich, Switzerland Austria [email protected] [email protected] [email protected] ABSTRACT Non-Counted Loops to Enable Subsequent Compiler Optimizations. In 15th Java programs can contain non-counted loops, that is, loops for International Conference on Managed Languages & Runtimes (ManLang’18), September 12–14, 2018, Linz, Austria. ACM, New York, NY, USA, 13 pages. which the iteration count can neither be determined at compile https://doi.org/10.1145/3237009.3237013 time nor at run time. State-of-the-art compilers do not aggressively optimize them, since unrolling non-counted loops often involves 1 INTRODUCTION duplicating also a loop’s exit condition, which thus only improves run-time performance if subsequent compiler optimizations can Generating fast machine code for loops depends on a selective appli- optimize the unrolled code. cation of different loop optimizations on the main optimizable parts This paper presents an unrolling approach for non-counted loops of a loop: the loop’s exit condition(s), the back edges and the loop that uses simulation at run time to determine whether unrolling body. All these parts must be optimized to generate optimal code such loops enables subsequent compiler optimizations. Simulat- for a loop. -
Compiler Optimization for Configurable Accelerators Betul Buyukkurt Zhi Guo Walid A
Compiler Optimization for Configurable Accelerators Betul Buyukkurt Zhi Guo Walid A. Najjar University of California Riverside University of California Riverside University of California Riverside Computer Science Department Electrical Engineering Department Computer Science Department Riverside, CA 92521 Riverside, CA 92521 Riverside, CA 92521 [email protected] [email protected] [email protected] ABSTRACT such as constant propagation, constant folding, dead code ROCCC (Riverside Optimizing Configurable Computing eliminations that enables other optimizations. Then, loop Compiler) is an optimizing C to HDL compiler targeting level optimizations such as loop unrolling, loop invariant FPGA and CSOC (Configurable System On a Chip) code motion, loop fusion and other such transforms are architectures. ROCCC system is built on the SUIF- applied. MACHSUIF compiler infrastructure. Our system first This paper is organized as follows. Next section discusses identifies frequently executed kernel loops inside programs on the related work. Section 3 gives an in depth description and then compiles them to VHDL after optimizing the of the ROCCC system. SUIF2 level optimizations are kernels to make best use of FPGA resources. This paper described in Section 4 followed by the compilation done at presents an overview of the ROCCC project as well as the MACHSUIF level in Section 5. Section 6 mentions our optimizations performed inside the ROCCC compiler. early results. Finally Section 7 concludes the paper. 1. INTRODUCTION 2. RELATED WORK FPGAs (Field Programmable Gate Array) can boost There are several projects and publications on translating C software performance due to the large amount of or other high level languages to different HDLs. Streams- parallelism they offer. -
Functional Array Programming in Sac
Functional Array Programming in SaC Sven-Bodo Scholz Dept of Computer Science, University of Hertfordshire, United Kingdom [email protected] Abstract. These notes present an introduction into array-based pro- gramming from a functional, i.e., side-effect-free perspective. The first part focuses on promoting arrays as predominant, stateless data structure. This leads to a programming style that favors compo- sitions of generic array operations that manipulate entire arrays over specifications that are made in an element-wise fashion. An algebraicly consistent set of such operations is defined and several examples are given demonstrating the expressiveness of the proposed set of operations. The second part shows how such a set of array operations can be defined within the first-order functional array language SaC.Itdoes not only discuss the language design issues involved but it also tackles implementation issues that are crucial for achieving acceptable runtimes from such genericly specified array operations. 1 Introduction Traditionally, binary lists and algebraic data types are the predominant data structures in functional programming languages. They fit nicely into the frame- work of recursive program specifications, lazy evaluation and demand driven garbage collection. Support for arrays in most languages is confined to a very resricted set of basic functionality similar to that found in imperative languages. Even if some languages do support more advanced specificational means such as array comprehensions, these usualy do not provide the same genericity as can be found in array programming languages such as Apl,J,orNial. Besides these specificational issues, typically, there is also a performance issue. -
Loop Transformations and Parallelization
Loop transformations and parallelization Claude Tadonki LAL/CNRS/IN2P3 University of Paris-Sud [email protected] December 2010 Claude Tadonki Loop transformations and parallelization C. Tadonki – Loop transformations Introduction Most of the time, the most time consuming part of a program is on loops. Thus, loops optimization is critical in high performance computing. Depending on the target architecture, the goal of loops transformations are: improve data reuse and data locality efficient use of memory hierarchy reducing overheads associated with executing loops instructions pipeline maximize parallelism Loop transformations can be performed at different levels by the programmer, the compiler, or specialized tools. At high level, some well known transformations are commonly considered: loop interchange loop (node) splitting loop unswitching loop reversal loop fusion loop inversion loop skewing loop fission loop vectorization loop blocking loop unrolling loop parallelization Claude Tadonki Loop transformations and parallelization C. Tadonki – Loop transformations Dependence analysis Extract and analyze the dependencies of a computation from its polyhedral model is a fundamental step toward loop optimization or scheduling. Definition For a given variable V and given indexes I1, I2, if the computation of X(I1) requires the value of X(I2), then I1 ¡ I2 is called a dependence vector for variable V . Drawing all the dependence vectors within the computation polytope yields the so-called dependencies diagram. Example The dependence vectors are (1; 0); (0; 1); (¡1; 1). Claude Tadonki Loop transformations and parallelization C. Tadonki – Loop transformations Scheduling Definition The computation on the entire domain of a given loop can be performed following any valid schedule.A timing function tV for variable V yields a valid schedule if and only if t(x) > t(x ¡ d); 8d 2 DV ; (1) where DV is the set of all dependence vectors for variable V . -
Improving the Compilation Process Using Program Annotations
POLITECNICO DI MILANO Corso di Laurea Magistrale in Ingegneria Informatica Dipartimento di Elettronica, Informazione e Bioingegneria Improving the Compilation process using Program Annotations Relatore: Prof. Giovanni Agosta Correlatore: Prof. Lenore Zuck Tesi di Laurea di: Niko Zarzani, matricola 783452 Anno Accademico 2012-2013 Alla mia famiglia Alla mia ragazza Ai miei amici Ringraziamenti Ringrazio in primis i miei relatori, Prof. Giovanni Agosta e Prof. Lenore Zuck, per la loro disponibilità, i loro preziosi consigli e il loro sostegno. Grazie per avermi seguito sia nel corso della tesi che della mia carriera universitaria. Ringrazio poi tutti coloro con cui ho avuto modo di confrontarmi durante la mia ricerca, Dr. Venkat N. Venkatakrishnan, Dr. Rigel Gjomemo, Dr. Phu H. H. Phung e Giacomo Tagliabure, che mi sono stati accanto sin dall’inizio del mio percorso di tesi. Voglio ringraziare con tutto il cuore la mia famiglia per il prezioso sup- porto in questi anni di studi e Camilla per tutto l’amore che mi ha dato anche nei momenti più critici di questo percorso. Non avrei potuto su- perare questa avventura senza voi al mio fianco. Ringrazio le mie amiche e i miei amici più cari Ilaria, Carolina, Elisa, Riccardo e Marco per la nostra speciale amicizia a distanza e tutte le risate fatte assieme. Infine tutti i miei conquilini, dai più ai meno nerd, per i bei momenti pas- sati assieme. Ricorderò per sempre questi ultimi anni come un’esperienza stupenda che avete reso memorabile. Mi mancherete tutti. Contents 1 Introduction 1 2 Background 3 2.1 Annotated Code . .3 2.2 Sources of annotated code . -
Compiler-Based Code-Improvement Techniques
Compiler-Based Code-Improvement Techniques KEITH D. COOPER, KATHRYN S. MCKINLEY, and LINDA TORCZON Since the earliest days of compilation, code quality has been recognized as an important problem [18]. A rich literature has developed around the issue of improving code quality. This paper surveys one part of that literature: code transformations intended to improve the running time of programs on uniprocessor machines. This paper emphasizes transformations intended to improve code quality rather than analysis methods. We describe analytical techniques and specific data-flow problems to the extent that they are necessary to understand the transformations. Other papers provide excellent summaries of the various sub-fields of program analysis. The paper is structured around a simple taxonomy that classifies transformations based on how they change the code. The taxonomy is populated with example transformations drawn from the literature. Each transformation is described at a depth that facilitates broad understanding; detailed references are provided for deeper study of individual transformations. The taxonomy provides the reader with a framework for thinking about code-improving transformations. It also serves as an organizing principle for the paper. Copyright 1998, all rights reserved. You may copy this article for your personal use in Comp 512. Further reproduction or distribution requires written permission from the authors. 1INTRODUCTION This paper presents an overview of compiler-based methods for improving the run-time behavior of programs — often mislabeled code optimization. These techniques have a long history in the literature. For example, Backus makes it quite clear that code quality was a major concern to the implementors of the first Fortran compilers [18]. -
Survey of Compiler Technology at the IBM Toronto Laboratory
An (incomplete) Survey of Compiler Technology at the IBM Toronto Laboratory Bob Blainey March 26, 2002 Target systems Sovereign (Sun JDK-based) Just-in-Time (JIT) Compiler zSeries (S/390) OS/390, Linux Resettable, shareable pSeries (PowerPC) AIX 32-bit and 64-bit Linux xSeries (x86 or IA-32) Windows, OS/2, Linux, 4690 (POS) IA-64 (Itanium, McKinley) Windows, Linux C and C++ Compilers zSeries OS/390 pSeries AIX Fortran Compiler pSeries AIX Key Optimizing Compiler Components TOBEY (Toronto Optimizing Back End with Yorktown) Highly optimizing code generator for S/390 and PowerPC targets TPO (Toronto Portable Optimizer) Mostly machine-independent optimizer for Wcode intermediate language Interprocedural analysis, loop transformations, parallelization Sun JDK-based JIT (Sovereign) Best of breed JIT compiler for client and server applications Based very loosely on Sun JDK Inside a Batch Compilation C source C++ source Fortran source Other source C++ Front Fortran C Front End Other Front End Front End Ends Wcode Wcode Wcode++ Wcode Wcode TPO Wcode Scalarizer Wcode TOBEY Wcode Back End Object Code TOBEY Optimizing Back End Project started in 1983 targetting S/370 Later retargetted to ROMP (PC-RT), Power, Power2, PowerPC, SPARC, and ESAME/390 (64 bit) Experimental retargets to i386 and PA-RISC Shipped in over 40 compiler products on 3 different platforms with 8 different source languages Primary vehicle for compiler optimization since the creation of the RS/6000 (pSeries) Implemented in a combination of PL.8 ("80% of PL/I") and C++ on an AIX reference -
Control Flow Graphs
CONTROL FLOW GRAPHS PROGRAM ANALYSIS AND OPTIMIZATION – DCC888 Fernando Magno Quintão Pereira [email protected] The material in these slides have been taken from the "Dragon Book", Secs 8.5 – 8.7, by Aho et al. Intermediate Program Representations • Optimizing compilers and human beings do not see the program in the same way. – We are more interested in source code. – But, source code is too different from machine code. – Besides, from an engineering point of view, it is better to have a common way to represent programs in different languages, and target different architectures. Fortran PowerPC COBOL x86 Front Back Optimizer Lisp End End ARM … … Basic Blocks and Flow Graphs • Usually compilers represent programs as control flow graphs (CFG). • A control flow graph is a directed graph. – Nodes are basic blocks. – There is an edge from basic block B1 to basic block B2 if program execution can flow from B1 to B2. • Before defining basic void identity(int** a, int N) { int i, j; block, we will illustrate for (i = 0; i < N; i++) { this notion by showing the for (j = 0; j < N; j++) { a[i][j] = 0; CFG of the function on the } right. } for (i = 0; i < N; i++) { What does a[i][i] = 1; this program } do? } The Low Level Virtual Machine • We will be working with a compilation framework called The Low Level Virtual Machine, or LLVM, for short. • LLVM is today the most used compiler in research. • Additionally, this compiler is used in many important companies: Apple, Cray, Google, etc. The front-end Machine independent Machine dependent that parses C optimizations, such as optimizations, such into bytecodes constant propagation as register allocation ../0 %"&'( *+, ""% !"#$% !"#$)% !"#$)% !"#$- Using LLVM to visualize a CFG • We can use the opt tool, the LLVM machine independent optimizer, to visualize the control flow graph of a given function $> clang -c -emit-llvm identity.c -o identity.bc $> opt –view-cfg identity.bc • We will be able to see the CFG of our target program, as long as we have the tool DOT installed in our system. -
GCC Internals
GCC Internals Diego Novillo [email protected] Red Hat Canada CGO 2007 San Jose, California March 2007 Outline 1. Overview 2. Source code organization 3. Internal architecture 4. Passes NOTE: Internal information valid for GCC mainline as of 2007-03-02 11 March 2007 GCC Internals - 2 1. Overview ➢ Major features ➢ Brief history ➢ Development model 11 March 2007 GCC Internals - 3 Major Features Availability – Free software (GPL) – Open and distributed development process – System compiler for popular UNIX variants – Large number of platforms (deeply embedded to big iron) – Supports all major languages: C, C++, Java, Fortran 95, Ada, Objective-C, Objective-C++, etc 11 March 2007 GCC Internals - 4 Major Features Code quality – Bootstraps on native platforms – Warning-free – Extensive regression testsuite – Widely deployed in industrial and research projects – Merit-based maintainership appointed by steering committee – Peer review by maintainers – Strict coding standards and patch reversion policy 11 March 2007 GCC Internals - 5 Major Features Analysis/Optimization – SSA-based high-level global optimizer – Constraint-based points-to alias analysis – Data dependency analysis based on chains of recurrences – Feedback directed optimization – Interprocedural optimization – Automatic pointer checking instrumentation – Automatic loop vectorization – OpenMP support 11 March 2007 GCC Internals - 6 1. Overview ➢ Major features ➢ Brief history ➢ Development model 11 March 2007 GCC Internals - 7 Brief History GCC 1 (1987) – Inspired on Pastel compiler