Cross module representation of heterogeneous programs

Details Cross module representation of heterogeneous programs Cross module representation of heterogeneous programs

1999-06-30

2003-08-19

Morse, Gregory (Department: 2122)

Data processing: software development, installation, and managem

Software program development tool

Translation of code

C717S141000, C717S146000, C717S148000, C717S149000, C717S152000, C717S153000, C717S154000, C717S159000, C717S161000

Reexamination Certificate

active

06609248

ABSTRACT:

FIELD OF THE INVENTION
This invention relates generally to programming tools, and more particularly to translating code between computer architectures.
COPYRIGHT NOTICE/PERMISSION
A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever. The following notice applies to the software and data as described below and in the drawings hereto: Copyright© 1998, Microsoft Corporation, All Rights Reserved.
BACKGROUND OF THE INVENTION
In a new programming paradigm, a program is now a collection of components. Each component publishes an interface without exposing its inner details. Thus, a component can internally exist in any form: Intel x86 binary, Intel IA-64 binary, Visual Basic (VB) byte codes, Java class files, or any Virtual Machine (VM) binary. A heterogeneous program consists of components in different forms. Heterogeneous programs already exist in some environments: in the Microsoft Windows 32-bit environment, a Visual Basic program is compiled into VB byte codes that can call native-compiled functions in a separate dynamic linked library. Similarly Java class files can call native functions. Intel's IA-64 architecture allows IA-64 code to co-exist with x86 code.
To understand the behavior of a heterogeneous program, all its components, regardless of their form, have to be instrumented and analyzed in the same framework, otherwise, only partial information will be collected. It is important to note that systems that have been ported to several architectures are not sufficient to handle heterogeneous programs. For example, a system for VB byte codes that has been ported to x86, cannot provide a complete execution time analysis of a heterogeneous program consisting of VB byte codes and native x86 because each system operates in isolation on its own input.
Further, a heterogeneous program may consist of heterogeneous components. A heterogeneous component is a single component consisting of routines in different instruction sets. As the interface is well defined, components internally can use any instruction set. Each instruction set has its own advantages such as execution time, portability, and size.
All previous systems have been designed for homogeneous programs: conventional programs consisting of components in the same form. Some systems have been targeted to different architectures, but cannot work with heterogeneous programs. None of these systems can generate a heterogeneous component.
A large number of systems have been developed to help analyze and optimize homogeneous programs. The creation of “Pixie” by MIPS Computers Systems, Inc. in 1986 started a class of basic block counting tools by inserting pre-determined sequence of instructions to record execution frequencies of basic blocks. “Epoxie” extended the technique by using relocations to eliminate dynamic translation overheads. David W. Wall. Systems for late code modification, in Code Generation—Concept, Tools Techniques, pp. 275-293, (Robert Giegrich and Susan L. Graham, eds, 1992). “QPT” further extended the technique by constructing spanning trees to reduce the number of basic blocks that are instrumented. James Larus and Thomas Ball, Rewriting executable files to measure program behavior, Software, Practice and Experience, vol. 24, no. 2, pp 197-218 (1994). “Purify” instruments memory references to detect out-of-bounds memory accesses and memory leaks. Reed Hastings and Bob Joyce, Purify: Fast Detection of Memory Leaks and Access Errors, Proceedings of Winter Usenix Conference, January 1992.
“OM” allowed general transformations to be applied to a binary by converting the binary to an intermediate representation that can be easily manipulated. Amitabh Srivastava and David Wall, A Practical System for Intermodule Code Optimization at Link Time, Journal of Programming Language, 1(1):1-18 (1993). OM has been implemented on MIPS, DEC Alpha and Intel x86 architectures. “EEL” uses a similar technique and provides an editing library for Sun SPARC architectures. James R. Larus and Eric Schnarr, EEL: Machine-Independent Executable Editing, Proceedings of SIGPLAN′ 95 Conference on Programming Language Design and Implementation (1995). “Alto” and “Spike” are optimizers for the DEC Alpha architectures. K. De Bosschere and S. Debray, Alto: a Link-Time Optimizer for the DEC Alpha. Technical Report TR-
96-16
, Computer Science Department, University of Arizona (1996). David W. Goodwin, Interprocedural Dataflow Analysis in an Executable Optimizer, Proceedings of SIGPLAN′ 97 Conference on Programming Language Design and Implementation (1997).
“ATOM” extended OM by providing a flexible instrumentation interface for the DEC Alpha and Intel x86 systems. Amitabh Srivastava and Alan Eustace, ATOM: A System for Building Customized Program Analysis Tools, Proceedings of SIGPLAN′ 94 Conference on Programming Language Design and Implementation (1994). However, ATOM does not allow modifications to a binary. “Etch” provided a similar system for x86 and “BIT” for Java byte codes. T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. Chen, and B. Bershad, Instrumentation and Optimization of Win32/Intel Executables Using Etch, Proceedings of the USENIX Windows NT Workshop (1997). Han Lee and Benjamin Zorn, BIT: A Tool for instrumenting Java bytecodes. Proceedings of the 1997 USENIX Symposium on Internet Technologies and Systems (1997).
None of these systems work on heterogeneous programs. Some of them have been ported to multiple architecture but they provide only a partial view when applied to heterogeneous programs as each implementation operates on its input in isolation. Although OM builds a symbolic representation, the representation was primarily designed for applying arbitrary transformations and is not sufficient to handle heterogeneous programs. None of these systems can generate heterogeneous components. ATOM provides a flexible interface for instrumentation only.
Because a heterogeneous program provides efficiencies that cannot be achieved by a homogeneous program, a mechanism is needed that can convert portions of a homogeneous program into a different instruction set to optimize the execution of the program or produce more compact code. Furthermore, the ability to apply the same mechanism to an existing heterogeneous program to produce further optimization is also desirable.
SUMMARY OF THE INVENTION
The above-mentioned shortcomings, disadvantages and problems are addressed by the present invention, which will be understood by reading and studying the following specification.
An output translator provides for cross module representations of components within a heterogeneous program by enabling a code block in a component to be translated from an platform-neutral intermediate representation of the program into a set of platform-specific instructions that are directed to a different architecture than that for which the code block was originally written. The output translator provides any necessary prologue and/or epilogue code to interface contiguous code blocks that are emitted in different instruction sets. For an architecture that has both short and long forms for instructions, one aspect of the output translator initially assumes the emitted instruction will be in its short form and only substitutes the long form when required. The output translator also enables the substitution of one code block for another, automatically adjusting entry points as required. Changes in the order of code blocks within the component are accommodated while preserving, and optimally optimizing, the original control flow.
Because the architecture of a code block can be changed when it is translated from the intermediate representation, a user can create a more efficient heterogeneous program from a homogeneous program

Affiliated with

Also associated with

Rating

Cross module representation of heterogeneous programs does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Cross module representation of heterogeneous programs, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Cross module representation of heterogeneous programs will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3098829