N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla, "Object- Oriented Architectural Support for a Java Processor, " Proc. 12th European Conf. Object-Oriented Programming, pp. 430-455, July 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....line size is increased can be explained using method locality and bytecode size information. Prior research on method locality and size distribution [62] showed that 45 of all dynamically invoked methods were either 1 or 9 bytecodes long. Since average bytecode size has been shown to be 1. 8 bytes [78], 45 of all methods can be expected to be less than 16 bytes long. Therefore, unless methods invoked in succession are located contiguously, increasing line sizes beyond 16 bytes (or 32 at the most) cannot capture further useful future references, explaining the data cache behavior of the ....

....dynamic branch predictors and aggressive ILP techniques are relevant. Object oriented architectural support in hardware is another approach taken to improve performance in Java processors, which tries to provide support for manipulating objects, specialized caches and other such optimizations [76, 78]. 5.2 Background and Motivation To motivate the research presented in this chapter, we perform an analysis of the decode and folding stages of the picoJava II processor pipeline in detail. The picoJava II is a stack based processor, since it is a silicon version of the stack based Java Virtual ....

[Article contains additional citation context not shown here]

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla. Object-oriented architectural support for a Java processor. In Proceedings of the 12th European Conference on Object-Oriented Programming, pages 430-- 455, July 1998.

....Inc, and Patriot Scienti c corporations PSC1000 [22] used in the embedded market are examples of architectures providing support for direct execution of Java bytecodes. Hardware support for Java stack processing, object manipulation and method invocation was proposed by Vijaykrishnan et al. [23, 24, 25]. Dynamic translation of bytecodes to the DELFT JAVA RISC instruction set and a link translation bu er to help in dynamic method invocation was proposed by Glossner et al. 26, 27] The MAJC (Microprocessor Architecture for Java) chip [28] proposed by Sun Microsystems uses thread level parallelism ....

N.Vijaykrishnan, N.Ranganathan, and R.Gadekarla, \Object-Oriented Architectural Support for a Java Processor," in ########### ## ######### ### #### ######## ########## ## ############### ###########, pp. 330-354, 1998.

....wide range of embedded applications such as telephony and web tops. The picoJava [10] processor from Sun Microsystems is an example of a Java processor. It is our belief that no one technique will be universally preferred accepted over all platforms in the immediate future. Many previous studies [11, 12, 13, 10, 14] have focussed on enhancing each of the bytecode execution techniques. On the other hand, a three pronged attack at optimizing the runtime system of all techniques would be even more valuable. Many of the proposals for improvements with one technique may be applicable to the others as well. For ....

....interpreted execution can be explained using method locality and bytecode size information. Prior research on method locality and size distribution [41]showed that 45 of all dynamically invoked methods were either 1 or 9 bytecodes long. Since average bytecode size has been shown to be 1. 8 bytes [14], 45 of all methods can be expected to be less than 16 bytes long. Therefore, unless methods invoked in succession are located contiguously, increasing line sizes beyond 16 bytes (or 32 at the most) cannot capture further useful future references, explaining the data cache behavior of the ....

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla, "Object-oriented architectural support for a Java processor," in Proceedings of the 12th European Conference on Object-Oriented Programming, pp. 430--455, July 1998.

....dynamic branch predictors and aggressive ILP techniques are relevant. Object oriented architectural support in hardware is another approach taken to improve performance in Java processors, which tries to provide support for manipulating objects, specialized caches and other such optimizations [19, 20]. Several tradeoffs concerning the design of caches, branch predictors, and exploitation of parallelism using ILP techniques are explored in [21] The dynamic execution time behavior of Java investigated in [22, 23] are also useful in understanding issues related to improving Java performance. ....

....8 bytes into the buffer at one time, where as it can read 7 bytes from the buffer at once. The bytecode ISA has variable length instructions, and most bytecodes consist of a 1 byte opcode, followed by 0, 1 or 2 operands. Since the average length of bytecode instructions is approximately 1. 8 bytes [20], the processor can potentially read more than one instruction in a cycle. The processor can read up to four instructions, depending on the length of the instructions in the buffer. Instruction Instruction Execution control logic data cache controller I O bus and memory interface unit ....

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla, "Object-oriented architectural support for a Java processor, " in Proceedings of the 12th European Conference on Object-Oriented Programming, pp. 430--455, July 1998.

....they may require significantly more resources than interpreters, making them unsuitable for resourceconstrained environments such as hand held devices and embedded systems. They may require several kilobytes of ROM and many more megabytes of RAM [8] than interpreters. Many previous studies [5, 7, 10, 11, 12] have focussed on enhancing each of the bytecode execution techniques. While the results from these studies are suggestive and useful, it is our belief that no one technique will be universally preferred accepted over all platforms in the immediate future. On the other hand, a three pronged attack ....

....line size is increased can be explained using method locality and bytecode size information. Prior research on method locality and size distribution [27] showed that 45 of all dynamically invoked methods were either 1 or 9 bytecodes long. Since average bytecode size has been shown to be 1. 8 bytes [12], 45 of all methods can be expected to be less than 16 bytes long. Therefore, unless methods invoked in succession are located contiguously, increasing line sizes beyond 16 bytes (or 32 at the most) cannot capture further useful future references, explaining the data cache behavior of the ....

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla, "Objectoriented architectural support for a Java processor," in Proceedings of the 12th European Conference on Object-Oriented Programming, pp. 430--455, July 1998.

....in their work that an offline bytecode to native code translator is a more efficient Java execution model for utilizing the caches and branch predictors. Vijaykrishnan et.al. reported the behavior of Java bytecode execution for a Java processor and proposed architectural features based on them [13]. However, there has not been an effort to study the interaction of the Just In Time compilation mode of Java execution with the architectural features. This study is important to provide hints in the development and improvement of HotSpot Java compilers [12] The HotSpot compilers choose between ....

N. Vijaykrishnan, N. Ranganathan and R. Gadekarla, Object-Oriented Architectural Support for a Java Processor, Proc. ECOOP'98, the 12th European Conference on ObjectOriented Programming, 1998.

....we present some of the most important execution characteristics of a recent suite of Java benchmarks, the SPEC JVM98. The benchmarks are analyzed both from a bytecode perspective and at the SPARC machine code level. Although Java performance studies have been performed by Vijaykrishnan et. al [1], Romer et al. 2] and Newhall et al. 3] most of the past studies were based on small, often synthetic Java programs. 2 Experimental Methodology In order to understand the behavior of a Java application it is necessary to first find a mechanism to monitor the stream of bytecodes that are ....

N.Vijaykrishnan, N.Ranganathan, and R.Gadekarla, "Object-Oriented Architectural Support for a Java Processor, " in Proceedings of ECOOP'98, the 12th European Conference on Object-Oriented Programming, 1998.

....of SPEC JVM98 benchmark suite at both bytecode and native code levels. Li et al. 19] investigated the influence of architecture and operating system on the execution of Java applications using complete system simulation. Based on the behaviors of Java applications, Vijaykrishnan et al. [26] proposed architectural support for object manipulation, stack processing and method invocation to enhance the execution speed of Java bytecodes based on characterizing some Java applications. In [18] and [17] the authors investigated the predictability of branches in the control flows of Java ....

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla. Object-oriented architectural support for a Java processor. In ECOOP'98, the 12th European Conference on ObjectOriented Programming, July 1998.

....information to optimize both systems software and architectural support for enhancing the performance of a JVM. In addition, a closer look at the execution profile of the JVM can also give revealing insights that can help to restructure its implementation. To our knowledge, existing studies [1, 2, 3, 4, 5, 6, 7, 8] have been confined to examining JVM profiles from the architectural perspective, and there has been no attempt at understanding the influence of the operating system activities. It is becoming increasingly clear [9, 10, 11, 12] that accurate performance analysis requires an examination of ....

N. Vijaykrishnan, N. Ranganathan and R. Gadekarla, ObjectOriented Architectural Support for a Java Processor, In Proceedings the 12th European Conference on Object-Oriented Programming, pages 430-455,1998.

No context found.

N. Vijaykrishnan, N. Ranganathan, and R. Gadekarla, "Object- Oriented Architectural Support for a Java Processor, " Proc. 12th European Conf. Object-Oriented Programming, pp. 430-455, July 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