Romer, T., Voelker, G., Lee, D., Wolman, A., Wong, W., Levy, H., Bershad, B., Chen, B., "Instrumentation and Optimization of Win32/Intel Executables Using Etch," Proceedings of the USENIX Windows NT Workshop, pp. 1--8, Seattle, WA (August 1997).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the code as it was loaded into memory. This approach is the one taken by the originators of sandboxing [Wahbe93] and has real technical advantages but, on the x86, thornier implementation issues as well. Modification of x86 binary code is a difficult problem, but some tools exist (e.g. Etch [Romer97]) that would simplify the task. 1. Up to now we have not made any major modifications to the tool set; to a large extent, we use the standard GNU versions of the compiler, assembler, and linker. 81 Another way of solving this problem would be to revisit the choice of safety technology. For ....

Romer, T., Voelker, G., Lee, D., Wolman, A., Wong, W., Levy, H., Bershad, B., Chen, B., "Instrumentation and Optimization of Win32/Intel Executables Using Etch," Proceedings of the USENIX Windows NT Workshop, pp. 1--8, Seattle, WA (August 1997).

....a linktime optimizer; our experiences indicate that, precisely for this reason, link time optimization can be useful for improving program performance even if the compiler carries out whole program optimization. The systems that are the closest to ours are the OM [36, 37] Spike [12] and Etch [31] link time optimizers. The actions carried out by these systems are conceptually very similar to ours (as they must be) though they differ in details. Spike and Etch are intended for executables running under Windows, on Compaq Alpha and Intel x86 processors respectively. Spike carries out three ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", 1997 USENIX Windows NT Workshop (to appear).

....application data structures. The second type of change is most useful for either performance steering, or other debugging applications. Our API is designed to support both of these uses. Our approach differs from other post compiler instrumentation tools such as EEL[11] ATOM[15] or Etch[14] that permit code to be inserted into a binary before it starts to execute. Often times, the specific code to be inserted may not be known until runtime. If the user is unsure what type of instrumentation they will require, they have only two alternatives. First, they could include all possible ....

....tools include a dynamic trace logger, a tool to gather statistics from hardware performance monitors, and a tool to help with understanding the performance of MPI message passing programs. 6. Related Work One area of research that is similar to our Dyninst work is the area of binary editing tools[11, 14, 15]. Binary editing permits code to be inserted into an application binary either after it has been compiled, or sometimes after it has been linked. These tools are useful because they avoid the need to recompile the application program. However, they still require the instrumentation code to remain ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. H. Levy, and B. Bershad, "Instrumentation and optimization of Win32/Intel executables using Etch," Proceedings of the USENIX Windows NT Workshop. Aug. 1997, Seattle, WA, USA, pp. 1-7.

....purchases a new processor, the appropriate reoptimizer may be packaged with it, allowing for immediate reoptimization of the user s executables for the new processor. 1. 1 Related Work Other systems have been proposed and developed that attempt to tackle similar or related problems including Etch [1], Morph [2] and Fx32 [3] A summary of these other efforts can be found in [4] Other proposed systems have suggested that the final binary be constructed from an intermediate representation at run time. This would allow for customization for the target model because the model specific portion ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. Bershad, and B. Chen, "Instrumentation and optimization of Win32/Intel executables using etch," in Proceedings of the USENIX Windows NT Workshop, August 1997, pp. 1--7.

....techniques, such as profile based optimization. This technique, which can move infrequently executed basic blocks out of line and lay out more frequently interacting basic blocks contiguously, could improve branch misprediction and BTB miss behavior, as well as L1 instruction cache miss behavior [24]. However, there is a limit to the savings this technique can offer. Furthermore, processors must be able to efficiently execute code, even if it has not been aggressively optimized. At the least, these miss ratios suggest that this workload requires a much larger BTB, and perhaps a different ....

T. Romer, et al. "Instrumentation and optimization of Win32/Intel executables using Etch," In Proc. of the USENIX Windows NT Workshop, pages 1-7, August 1997.

....amount of information about memory references across a variety of benchmark programs. This work was supported in part by the National Science Foundation under grant CCR 9502826 1 Introduction Recent years have seen increasing interest in reasoning about and manipulating executable files [5, 14, 19, 24, 26, 29, 30, 32]. When working with an executable file, we typically have information about the entire program including, potentially, library functions that is usually not available at compile time. Because of this, code manipulation and optimization at this level offers benefits that are difficult or ....

....the results of alias analysis into our instruction scheduler as well as a number other optimizations, and expect to have more extensive experimental results for the utility of this information shortly. 5 Related Work While a number of systems have been described for link time code optimization [5, 14, 15, 26, 29, 30, 32], to the best of our knowledge none of these carry out any alias analysis on the executable files they process. There is an extensive body of work on pointer alias analysis of various kinds (see, for example, 2, 3, 4, 6, 8, 10, 11, 12, 13, 17, 18, 20, 21, 22, 23, 25, 27, 28, 31, 33, 34] The ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", 1997 USENIX Windows NT Workshop (to appear).

....spend between 11 and 24 of their time on tag checking [17] We expect that the overall conclusion that programs spend a significant amount of time in tag manipulation holds for Scheme programs as well. Most of the prior work on link time optimization has focused on imperative languages [7, 12, 15, 16, 19]. The differences in runtime characteristics between Scheme and C programs, as discussed above, can have a significant effect on the extent to which systems designed for executables resulting from (human written) C programs will be effective on code generated from Scheme programs. The reasons for ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", Proc. 1997 USENIX Windows NT Workshop.

....spend between 11 and 24 of their time on tag checking [17] We expect that the overall conclusion that programs spend a significant amount of time in tag manipulation holds for Scheme programs as well. Most of the prior work on link time optimization has focused on imperative languages [7, 12, 15, 16, 19]. The differences in runtime characteristics between Scheme and C programs, as discussed above, can have a significant effect on the extent to which systems designed for executables resulting from (humanwritten) C programs will be effective on code generated from Scheme programs. The reasons for ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", Proc. 1997 USENIX Windows NT Workshop.

....greater. To accurately assess the effects of different design tradeoffs, application traces which contain not only the application code execution, but the operating system, library, etc. will need to be captured. Most software based tracing environments only capture the application code execution [10 , 18, 23, 30]. Most hardware based tracing systems allow for the capture of the entire execution [17] but the cost of such systems may be very high. A few software based tracing system allow for the collection of operating system execution [8, 25, 27] We discuss these systems next. ATOM was developed by ....

....of providing this level of trace capability, and its importance as we move to studying applications run on NT based platforms. Lee et al. have recently reported on the characteristics of NT based execution on the Intel x86 platform [19] To capture these traces they used the Etch toolset [23]. While Etch is able to instrument most NT binaries, it is not able to instrument some DLLs, nor any of the NT kernel image. Lee et al. attempted to compare the characteristics of commercial desktop applications (e.g. MS Word, PowerPoint, Netscape) with execution from the SPEC 95 benchmark ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy and B. Bershad, "Instrumentation and Optimization of Win32/Intel Executables Using Etch," Proc. of USENIX Windows NT Workshop, August 1997, pp. 1-8.

....this goal. Experimental results indicate that it is nevertheless able to provide a reasonable amount of information about memory references across a variety of benchmark programs. 1 Introduction Recent years have seen increasing interest in reasoning about and manipulating executable files [5, 15, 20, 25, 27, 30, 31, 33]. When working with an executable file, we typically have information about the entire program including, potentially, library functions that is usually not available at compile time. Because of this, code manipulation and optimization at this level offers benefits that are difficult or ....

....inferred as optimizable. To some extent, imprecision in our analysis, arising from the sources discussed in Section 5.1, also affected the number of memory operations deemed suitable for optimization. 6 Related Work While a number of systems have been described for link time code optimization [5, 15, 16, 27, 30, 31, 33], to the best of our knowledge, any alias analysis carried out by these systems is limited to fairly simple local analyses. There is an extensive body of work on pointer alias analysis of various kinds (see, for example, 3, 4, 6, 9, 11, 12, 13, 14, 18, 19, 21, 22, 23, 24, 26, 28, 29, 32, 34, ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", 1997 USENIX Windows NT Workshop (to appear).

....not amenable to optimization by these tools. Machine level global optimization is discussed also by Johnson and Miller [23] but unlike alto, this system does not carry out interprocedural analysis and optimizations. The systems that are the closest to ours are the OM [34, 35] Spike [8] and Etch [28] link time optimizers. The actions carried out by these systems are conceptually very similar to ours (as they must be) though they differ in details. Spike and Etch are intended for executables running under Windows, on DEC Alpha and Intel x86 processors respectively. Spike carries out three ....

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. N. Bershad, and J. B. Chen, "Instrumentation and Optimization of Win32/Intel Executables", 1997 USENIX Windows NT Workshop (to appear).

No context found.

T. Romer, et al. "Instrumentation and optimization of Win32/Intel executables using Etch," Proc. of the USENIX Windows NT Workshop, pages 1-7, August 1997.

No context found.

Romer, T., G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy, B. Bershad, B. Chen, "Instrumentation and Optimization of Win32/Intel Executables Using Etch." Proc. of the USENIX Windows NT Workshop, 1997, pp. 1 -- 7.

No context found.

T. Romer, G. Voelker, D. Lee, A. Wolman, W. Wong, H. Levy and B. Bershad, "Instrumentation and Optimization of Win32/Intel Executables Using Etch," Proc. of USENIX Windows NT Workshop, August 1997, pp. 1-8.

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