K. Ishizaki, M. Kawahito, T. Yasue, M. Takeuchi, T. Ogasawara, T. Suganuma, T. Onodera, H. Komatsu, and T. Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM SIGPLAN '99 Java Grande Conference, pages 119-128, June 12-14, 1999.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....pattern of the matrix A. Any significant processing of this will also overwhelm the cost of a single matrix vector multiplication, so only when many are to be performed is optimization worthwhile. Completely Run time optimization This is the scenario in which just in time (JIT) compilers work [15, 35, 1], as well as the inspector executor model [20] and other dynamic compilation systems [24, 6, 23] In these cases, one has essentially all information about a problem instance, but the least time available to optimize. A standard example of inspector executor is to examine the sparsity pattern of a ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the 1999.

....properties are greatly beneficial for engineering robust Internet services, eliminating many common sources of bugs. Java also provides flexibility in terms of service deployment across multiple architectures. We make use of optimizing Java compilers including OpenJIT [30] and the IBM JIT compiler [27]. Implementing the base platform in Jawa presented two important challenges. The first was the lack of nonblock ing I O mechanisms in the Java core libraries. We overcame this by implementing our own nonblocking I O li brary using native code wrappers to existing system calls, for example ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, Implementation, and Evaluation of Optimizations in a Just-In-Time Compiler. In Proceedings of the ACM 1999.

....compiler in their VM includes implementations of dead code elimination, loop invariant hoisting, common subexpression elimination, constant propagation and a graph coloring global register allocator. Another JVM developed in a commercial environment is one at IBM Tokyo Research Laboratory[17]. It applies a similar set of optimizations as HotSpot does in its JIT. However, it does not use a global register allocator, but use a region based local one instead. They claim Since the JIT compiler requires fast compilation, expensive register allocation algorithms, such as graph coloring, ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In ACM

....However, inter method overhead, i.e. cost on passing an integer to some method or returning it from a method to callee, cannot be eliminated. Figure 2 in the previous section indirectly tells that inter method overhead cannot be ignored. In some cases inter method optimization is also available [7], but it seems dicult to perform such optimization for functions on dynamically typed language, because function s return value has generic type such as Object (or Descriptor with our extention) In our extended Kawa Scheme, for example, even simple integer arithmetic functions may return both ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the 1999 ACM Javagrande Conference, June 1999.

....AWT, which is not supported by the TowerJ implementation. The benchmarking platforms included: an SGI Origin 2000 (300 Mhz R10000) with SGI s JDK 1.1.6, with and without the AOT compiler, SGI s Java2 v1.2.2, and TowerJ 3.3.b1; a Dell Pentium II (350 Mhz) running Red Hat Linux with IBM s JDK 1.1. 8 [6]; and a Sun UltraSPARC (143 MHz) with Sun s Java2 v1.2.2 with the HotSpot 1.0 server compiler [17] 4.1 Comparison of SGI JIT and AOT Compilation Time for SPEC JVM98 JIT compilation of the entire SPEC JVM98 suite takes 2.4 seconds, which is a statistically insignificant addition to the SPEC ....

....that presented in this paper, comparing the JIT and AOT approaches to Java compilation and presenting experimental results, particularly within the context of a single Java execution environment. Significant JIT compilers that have been well documented in the research literature are those from IBM [6], Intel [7] and the CACAO [11] and LaTTe projects [12] Those that have not include Sun s HotSpot compiler [17] and Compaq s Fast JVM [3] Appeal s JRockit [10] virtual machine implements JIT compilation at multiple levels of optimization. IBM Research is also developing a dynamic optimizing ....

Kazuaki Ishizaki et. al. Design, Implementation, and Evaluation of Optimizations in a Just-In-Time Compiler. In Proc. ACM Java Grande Conference 1999, June 1999.

.... to make use of specialized hardware and O S interfaces, such as fast cluster networks [32] asynchronous I O [23] and raw disk access [17] To date, few Java systems have striven to take full advantage of these interfaces, instead focusing on the other aspects of performance, such as compilation [18], garbage collection [1] and thread performance [2] Traditionally, Java applications make use of lowlevel system functionality through the use of native methods, which are written in a language such as C. To bind native method code to the Java application, a native method interface is used, ....

....the (relatively expensive) translation from Java to Jaguar bytecode to a portable front end compiler, minimizing the complexity impact on the back end compilers. 2 Motivation and Background Much previous work has addressed the CPUrelated aspects of Java performance, including compilation [28, 18, 11], thread synchronization [2] and garbage collection [30] However, Java I O performance remains largely uninvestigated. Implementing high performance communication and I O in Java requires two classes of operations which the Java Virtual Machines does not directly support. The first is direct ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM 1999 Java Grande Conference, June 1999.

....a unified set to represent the fields in a class does not lead to a great loss of accuracy. 8 Measurements taken on an IBM ThinkPad 600E PC with a 300Mhz processor and 288MB of main memory. We used the Sun JDK 1.1. 8 VM with the just in time compiler developed at the IBM Tokyo Research Laboratory [22]. None of the benchmarks required more than 200MB of heap space. In light of the improved results of XTA over FTA in some cases, we consider the XTA algorithm the best choice. If heap space is at a premium, FTA offers reduced space consumption in exchange for a slight loss of precision. MTA seems ....

Ishizaki, K., Kawahito, M., Yasue, T., Takeuchi, M., Ogasawara, T., Suganuma, T., Onodera, T., Komatsu, H., and Nakatani, T. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM SIGPLAN JavaGrande Conference (San Francisco, CA, June 1999).

....properties are greatly bene cial for engineering robust Internet services, eliminating many common sources of bugs. Java also provides exibility in terms of service deployment across multiple architectures. We make use of optimizing Java compilers including OpenJIT [30] and the IBM JIT compiler [27]. Implementing the base platform in Java presented two important challenges. The rst was the lack of nonblocking I O mechanisms in the Java core libraries. We overcame this by implementing our own nonblocking I O library using native code wrappers to existing system calls, for example ....

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Komatsu, and Toshio Nakatani. Design, Implementation, and Evaluation of Optimizations in a Just-In-Time Compiler. In Proceedings of the ACM 1999 Java Grande Conference, June 1999.

....for each benchmark. As is shown in the figure, the number of classes is reduced by 3.7 to 88.4 , the number of methods by 12 See www.winzip.com. 13 Measurements taken on a Pentium II 300Mhz PC with 128MB memory, using the just in time compiler developed at the IBM Tokyo Research Laboratory [13]. 10.2 to 89.0 , and the number of fields by 14.0 to 88.8 . The average percentage reductions for classes, methods, and fields are 25.9 , 34.4 , and 49.7 , respectively. 3.4 Evaluation A number of observations can be made about the results reported above. The benchmark for which we measured ....

....are all interactive GUI based applications, so that direct speedup measurements are difficult to conduct. The speedups we measured turned out to be small and somewhat erratic, and highly dependent on the (version of the) used VM and JIT. Using the JIT developed in IBM s Tokyo Research Laboratory [13], we measured speedups of 5.4 , 2.1 ,and 0.4 . We believe that the fact that we measured the relatively large percentage of 5.4 for Jax because Jax spends a relatively small amount of time doing I O, compared to Mockingbird and javac. 4 Related Work Related work falls into a number of different ....

Ishizaki, K., Kawahito, M., Yasue, T., Takeuchi, M., Ogasawara, T., Suganuma, T., Onodera, T., Komatsu, H., and Nakatani, T. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM SIGPLAN JavaGrande Conference (San Francisco, CA, June 1999).

No context found.

K. Ishizaki, M. Kawahito, T. Yasue, M. Takeuchi, T. Ogasawara, T. Suganuma, T. Onodera, H. Komatsu, and T. Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM SIGPLAN '99 Java Grande Conference, pages 119-128, June 12-14, 1999.

No context found.

Kazuaki Ishizaki, Motohiro Kawahito, Toshiaki Yasue, Mikio Takeuchi, Takeshi Ogasawara, Toshio Suganuma, Tamiya Onodera, Hideaki Kamatsu, and Toshio Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In Proceedings of the ACM 1999.

No context found.

K. Ishizaki, M. Kawahito, T. Yasue, M. Takeuchi, T. Ogasawara, T. Suganuma, T. Onodera, H. Komatsu, and T. Nakatani. Design, implementation, and evaluation of optimizations in a just-in-time compiler. In ACM 1999 Java Grande Conference, pages 119--128, June 1999.

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