Ali-Reza Adl-Tabatabai, Michal Cierniak, Guei-Yuan Lueh, Vishesh M. Parikh, and James M. Stichnoth. Fast, eective code generation in a just-in-time Java compiler. In Proc. of the SIGPLAN'98 Conference on Programming Language Design and Implementation, June 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....runtime system and give experimental results on a symmetric shared memory multiprocessor. Softspec requires only local program information and does not rely on any global analysis. Its simplicity means it could execute entirely at runtime and target program binaries. Runtime translation [1, 8, 14] and runtime optimization [4] techniques are becoming prevalent. Softspec can be readily incorporated into such frameworks. The paper is organized as follows. The next section gives an overview of our technique. Section 3 describes the core algorithm in detail. Section 5 gives experimental ....

.... access has a consistent stride (i.e. the strides between the rst and second and between the second and third pro led iterations are the same) the runtime system then determines how many profile loop for (i=0; i 3; i ) f profile address[i] 0] a[i] a[i] i; profile address[i][1] = b[i j] profile address[i] 2] b[i] profile address[i] 3] p; b[i j] b[i] p; g Figure 4: The pro le loop created from the sequential loop in Figure 2. iterations can be parallelized before an inter iteration dependence is encountered. If the strides are not consistent then no ....

Ali-Reza Adl-Tabatabai, Michal Cierniak, Guei-Yuan Lueh, Vishesh M. Parikh, and James M. Stichnoth. Fast, eective code generation in a just-in-time Java compiler. In Proc. of the SIGPLAN'98 Conference on Programming Language Design and Implementation, June 1998.

....and related technologies. To date, much of this research has concentrated in two main areas: Static analysis of Java class files, for purposes such as optimisation [14] or compression [2] The performance of the bytecode interpreter, yielding techniques such as Just In Time (JIT) e.g. [1, 8]) and hotspot centered compilation (see [9] for a survey) The platform independent bytecode analysis presented in this paper describes the bytecode as it is interpreted, without the interference of JIT compilation or any machinespecific issues. This type of analysis can help to clarify the ....

Ali-Reza Adl-Tabatabai, Michal Cierniak, Guei-Yuan Lueh, Vishesh M. Parikh, and James M. Stichnoth. Fast, e#ective code generation in a Just-In-Time Java compiler. In ACM SIGPLAN Conference on Programming Language Design and Implementation, pages 280--290, Montreal, Canada, June 1998.

....since execution must pause waiting for compilation to complete. For this research, we use an open source, dual compiler system called the Open Runtime Platform (ORP) which was recently released by the Intel Corporation [19] The rst compiler (O1) provides very fast translation of Java programs [1] and incorporates a few very basic bytecode optimizations that improve execution performance. The second (O3) compiler performs a small number of commonly used optimizations on bytecode and an intermediate form to produce improved code quality and execution time. O3 optimization algorithms were ....

A. Adl-Tabatabai, M. Cierniak, G. Lueh, V. Parikh, and J. Stichnoth. Fast,Eective Code Generation in a Just-In-Time Java Compiler. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation, October 2000.

....libraries that they can build on. 6 Julia Lawall. Personal communication, 2000. 2. 6 Summary and Interdependencies Some of the key features of the systems we discussed and MetaOCaml can be summarized as follows: System L R A I P Fabius ML subset LL E C 1 C C LL(2) E C 2 DyC C LL I C [1] Tempo C HL LL I C 2 MetaML SML HL E I MetaOCaml OCaml HL E C Where the columns are: L Language chosen as basis; A Annotation language: Explicit or Implicit; R Representation: High level (HL) or low level (LL) In the case of C, two di erent LLs were studied; I Implementation: ....

....translation presented in this section. The essence of the compilation function can be illustrated as follows. Assume the OCaml compiler takes a language with the syntax: e 2 O : x j e e j x:e j e; e j C e 7 This also seems to be part of the appeal of the JIT approach in Java systems [1, 6] Where x is a variable, e1 e2 is an application, x:e is a lambda abstraction, and e1 ; en is a tuple of length n, and C e is an application of a value constructor (or attachment of a union type tag) C to the result of the expression e. The idea will be to represent the staging manipulations ....

Adl-Tabatabai, A.-R., Cierniak, M., Lueh, G.-Y., Parikh, V. M., and Stichnoth, J. M. Fast, eective code generation in a just-in-time Java compiler. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation (1998), pp. 280-290.

....static analysis, control transfers, optimization, compilers 1 Introduction Binary modi cation has become a major research eld during the last decade. Three major applications of binary modi cation are: i) binary translation, becoming increasingly popular with Just In Time compilation for Java [1], ii) instrumentation of binaries [2] for debugging purposes, and (iii) whole program optimization at link time. Link time optimiza The work described in this paper is funded by the Fund for Scienti c Research Flanders under grant 3G001998. tion systems are used for di erent purposes such ....

A. Adl-Tabatabai, M. Cierniak, G. Lueh, V.M. Parikh, and J.M. Stichnoth. Fast, eective code generation in a just-in-time java compiler. In ACM Sigplan Notices, volume 33, pages 280-290, 1998.

....natively compiled code, the microprocessor hardware performs the fetch and decode actions. Therefore natively compiled code executes much faster compared to software translated code. Instead of dynamically interpreting each bytecode at runtime, optimized JVMs use a Just In Time (JIT) compiler [2, 3, 4] to compile Java methods (i.e. functions) to native code. JIT compilers compile bytecodes into nativecode at run time, thereby adding compilation time to the total execution time. Initial JIT compilers were quick and dirty, generating unoptimized code for any method which was invoked. CurrentJava ....

A.-R.Adl-Tabatabai, M.Ciernaki, G.-Y.Lueh, V.M.Parikh, and J.MStichnoth, \Fast, EectiveCode Generation in a Just-In-Time Java Compiler," in ########### ## ####################### ######## ###### ### ##############, pp. 280-290, 1998.

....of a method. JITs range from the very naive, that use templates to replace each bytecode with a xed sequence of native code instructions (early versions of Ka e[5] did this) to the very sophisticated that perform register allocation, instruction scheduling and other scalar optimizations (e.g. [8, 23, 29, 33]) JITs face two major problems. First, they strive to generate good code in very little time, as compile time is lost to the running application. Second, the code of compiled method resides in memory; this augments the pressure on the memory manager and garbage collector. Recent virtual machines ....

Ali-Reza Adl-Tabatabai, Micha l Cierniak, Guei-Yuan Lueh, Vishesh M. Parikh, and James M. Stichnoth. Fast, eective code generation in a Just-in-Time Java compiler. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation, pages 280-290. ACM Press, 1998.

....which translates Java bytecode into native machine code at runtime, so that a Java program runs in the form of machine code without interpretation. JIT compilation provides better performance than interpretation, since it executes a program as machine code with various code optimizations applied [1, 4, 8, 9, 15]. However, the existence of exception handlers often prevents JIT compilers from aggressively optimizing code [6, 10, 14] Our exception handling mechanism removes all such overheads so that most opportunities for optimizations may be exploited, and also improves the exception handling speed for ....

....code optimizations on normal ow if exception handlers exist [10, 14] We propose methods to remove such ineciencies due to the existence of exception handlers. 3. 1 On demandTranslation of Exception Handlers So far most JIT compilers translate exception handlers when translating a method [1, 10, 14]. Instead we determined not to translate an exception handler together with the other parts of a method, because we know that an exception is unlikely to be used. The on demand translation of exception handlers has two major bene ts, one of which is to reduce JIT compilation time, and the other ....

A.-R. Adl-Tabatabai, M. Cierniak, G.-Y. Lueh, V. M. Parikh, and J. M. Stichnoth. Fast, eective code generation in a just-in-time Java compiler. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation, pages 280-290, 1998.

.... across the boundaries of independently distributed software components and thereby reduce the performance penalty paid for modularization.This approach has been validated in previous work and is now common practice [Deutsch and Schi#man 1984; Franz 1994; Holzle 1994; Holzle and Ungar 1996; Adl Tabatabai et al. 1998]. This paper presents a system that goes another step further, achieving a yet again better match between the software and the hardware by re evaluating the bindings between software and hardware at regular intervals instead of permanently fixing them prior to execution. To this e#ect, a profiler ....

.... of garbage collection, and minimizing the overhead of thread synchronization; among them are the Smalltalk 80 system [Deutsch and Schi#man 1984] the Self 93 system [Holzle 1994; Holzle and Ungar 1996] the Continuous Program Optimization: A Case Study 35 HotSpot system, the Intel VTune system [Adl Tabatabai et al. 1998], and the Jalapeno system [Alpern et al. 1999; Alpern et al. 1999] Typical optimizations performed by these systems include run time type feedback [Holzle and Ungar 1994] message inlining [Dean and Chambers 1994] message splitting [Chambers 1992] polymorphic inline caches [Holzle et al. 1991] ....

Adl-Tabatabai, A.-R., Cierniak, M., Lueh, C.-Y., Parikh,V.M., and Stichnoth, J. M. 1998. Fast, E#ective Code Generation in a Just-in-Time Java Compiler. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation (PLDI). Montreal, Canada, 280--290.

....but are expensive to support; applications written in Java are often much slower than their counterparts written in C or C . To use these features without having to pay a great performance penalty, sophisticated optimizations and runtime systems are required. For example, Just In Time compilers [1], adaptive compilers such as Hotspot TM and Way Ahead Of Time Java compilers[18, 17] are three approaches used to improve performance. Our approach is to statically optimize Java bytecode. There are several reasons for optimizing at the bytecode level. Firstly, the optimized bytecode can then ....

Ali-Reza Adl-Tabatabai, Michal Cierniak, Guei-Yuan Lueh, Vishesh M. Parikh, and James M. Stichnoth. Fast and eective code generation in a just-in-time Java compiler. ACM SIGPLAN Notices, 33(5):280290, May 1998.

....a large e#ort should be made. 3 BCS: ByteCode Specialization We propose ByteCode Specialization (BCS) which performs run time specialization on a virtual machine and e#ciently translates specialized code into optimized native code by using Just In Time (JIT) compilation techniques (e.g. [4]) The characteristics of BCS are: e#cient specialization: Similar to RTS, BCS constructs specializers that directly generate specialized programs in compiled code. e#cient specialized programs: Thanks to optimizations after code generation, specialized programs have less overheads than the RTS ....

A.-R. Adl-Tabatabai, M. Cierniak, G.-Y. Lueh, V. M. Parikh, and J. M. Stichnoth. Fast, e#ective code generation in a just-in-time Java compiler. In PLDI'98 [3], pages 280--290.

....3 the software to the hardware can be solved in a straightforward manner simply by deferring native code generation at least until load time. This approach has been validated in previous work and is now common practice [Deutsch and Schi# man 1984; Franz 1994; Holzle 1994; Holzle and Ungar 1996; Adl Tabatabai et al. 1998]. Besides enabling a closer match between software and hardware, load time code generation also makes it possible to perform optimizations across the boundaries of independently distributed software components and hence can reduce the performance penalty paid for modularization. Interestingly ....

....aspects of the individual components of our system that is the optimizer, the profiler, and the replacer. The design of code generating loaders and other just in time compilers has already been documented elsewhere [Deutsch and Schi#man 1984; Franz 1994; Holzle 1994; Holzle and Ungar 1996; Adl Tabatabai et al. 1998] and our implementation provides no meaningful new insights on this topic. 3. THE OPTIMIZER SUBSYSTEM The optimizer is composed of three main parts: the optimization manager, a history database, and a set of optimization components that can be dynamically added, removed, and exchanged. Figure 2 ....

Adl-Tabatabai, A.-R., Cierniak, M., Lueh, C.-Y., Parikh,V.M., and Stichnoth, J. M. 1998. Fast, E#ective Code Generation in a Just-in-Time Java Compiler. In Proceedings of the ACM SIGPLAN'98 Conference on Programming Language Design and Implementation (PLDI), Montreal, Canada, pp. 280--290.

No context found.

A.-R. Adl-Tabatabai et al. Fast, e#ective code generation in a just-in-time Java compiler. In Proc. PLDI'98, pages 280#290, Montreal, Canada, June 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC