R. H. Saavedra and A. J. Smith. Performance characterization of optimizing compilers. IEEE Trans. on Software Engineering, 21(7):615--628, July 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....involves a pixel and its neighborhood. For these reasons we focus our attention on the structure of RISC CPUs. To overcome the inefficiencies that prevent the CPU from delivering a level of performance close to its peak it is necessary to perform a quantitative analysis of the problem [5] 4] 9][12]. In fact, to provide solutions apt to improve the efficiency of RISC architectures in IPPR tasks, it is necessary to identify the sources of inefficiencies and to measure the contribution of each of them to the global slowdown with respect to ideal performance. These solutions concern compiler ....

Saavedra R. H. and. Smith A. J, Performance Characterization of Optimizing Compilers, IEEE Transactions on Software Engineering, vol.21, no. 7, pp. 615-628, July 1995.

....More details are given in Appendix 9. The redundancy of the characteristics must be avoided. 7.2.2. Number of measurements and loops Due to the precision of the clock frequency the programs have been run several times. The problem was to know how much time they need to be executed. Saavedra in [15] made some tests in order to find the necessary magnitude of the number of loops. He ran several experiments using different values for the number of loops in several machines. Each test has been run at least 0.1, 0.2, 0.5, 1.0, 2.0 and 4.0 seconds and he obtained measurements for 5, 10, and 20 ....

Saavedra R., Performance Characterization of Optimizing Compilers, Computer Science Division, University of California Berkeley, August 1992.

....platforms. A crucial component within Fortran RED is performance prediction. Our work on performance prediction in the context of Fortran RED has been based on our previous work on training sets [5] This work is very similar to the micro benchmarking approach developed by Saavedra and Smith [36, 35]. The importance of performance prediction for optimizing compilers has been recognized by many researchers and several proposals have been published in the literature, such as [40, 33, 10, 29] However, accurate and cost efficient performance prediction, in particular for superscalar ....

R.H. Saavedra and A.J. Smith. Performance characterization of optimizing compilers. IEEE Transactions on Software Engineering, 21(7):615--628, July 1995.

....it in its own way. So, this model is limited to a specific compiler. Besides, compilers optimise the code in order to try to speed up the generated code. So, in addition to fix the choice of the compiler, the optimization level must also be fixed. The most important optimizations are described in [SBS95] and in [Dow93] ffl Code motion: The compiler identifies expressions or statements which are invariant in a loop and moves them out of the loop. ffl Constant folding: Replace constants by their values and evaluate the resulting expressions at compiling time. ffl Subroutine inlining: ....

R.H. Saavedra-Barrera and A.J. Smith. performance characterization of optimizing compilers. IEEE transactions on software engineering, Vol.21(No.7), July 1995.

....on a large number of machines. Our results were successful in accurately predicting inconsistent machine performance, i.e. that machine A is faster than B for program x, but slower for program y. Both of these studies assumed that programs were compiled and executed without optimization. In [Saav92c] we extended our model to include the effect of (scalar) compiler optimization. It is very difficult to predict which optimizations will be performed by a compiler and also to predict their performance impact. We found, surprisingly, that we could model the performance improvement due to ....

Saavedra, R.H. and Smith, A.J., "Performance Characterization of Optimizing Compilers, submitted for publication, USC Tech. Rept. No. USC-CS-92-525, also UC Berkeley, Tech. Rept. No. UCB/CSD 92/699, August 1992.

....In this section we compare different machine characterizations under various levels of compiler optimization. We ran the system characterizer using different optimization levels on three high performance workstations. The complete results, including those without optimization, can be found in [23]. Table 2 shows a set of thirteen parameters which were synthesized from the basic measurements. The vector of reduced parameters can be used to characterize a machine and to compute the degree of similarity between machines. We can also use a graphical representation of performance called the ....

....appear smaller than they are, it is clear from the figure that the predictions even at the maximum optimization level are quite good. Summaries of the predictive errors, by machine and program, are presented in Tables 3 and 4. The complete execution times and relative errors can be found in [23]) The RMS error, shown in Tables 3 and 4, is the square root of the average of the square of the individual errors. As expected the magnitude of the error increases with the optimization level, but this increase is relatively small with an average error of less than 11 . Note that the average ....

[Article contains additional citation context not shown here]

Saavedra, R.H. and Smith, A.J., Performance Characterization of Optimizing Compilers, University of Southern California Technical Report No. USC-CS-92-525, August 1992.

....on a large number of machines. Our results were successful in accurately predicting inconsistent machine performance, i.e. that machine A is faster than B for program x, but slower for program y. Both of these studies assumed that programs were compiled and executed without optimization. In [Saav92c] we extended our model to include the effect of (scalar) compiler optimization. It is very difficult to predict which optimizations will be performed by a compiler and also to predict their performance impact. We found, however, that we could model the performance improvement due to optimization ....

Saavedra, R.H. and Smith, A.J., "Performance Characterization of Optimizing Compilers, submitted for publication, USC Tech. Rept. No. USC-CS-92-525, also UC Berkeley, Tech. Rept. No. UCB/CSD 92/699, August 1992.

No context found.

R. H. Saavedra and A. J. Smith. Performance characterization of optimizing compilers. IEEE Trans. on Software Engineering, 21(7):615--628, July 1995.

No context found.

Saavedra, Rafeal H., and Alan Jay Smith, "Performance Characterizations of Optimizing Compilers," IEEE Transactions on Software Engineering, July 1995, pp. 615-627.

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