|
|
|
|
|
|
|
Data Note 4. The data in Tables 3.19 and 3.20 are taken from Rossmann [248, 247], again based upon a series of monitor and trace studies for the R/M architecture. |
|
|
|
|
|
|
|
|
Reliability. The data presented are for a reasonably comprehensive study of various benchmark and trace materials; however, no software has been used to reorganize code to reduce the interlock effects. Further, one would expect that the interlock effects for the L/S architecture would be somewhat diminished because of the occurrence of additional instructions in the dynamic instruction stream. Similarly, one would expect that the R+M architecture would show an even greater effect from these interlocks. No direct data is available for either of these architectures. |
|
|
|
|
|
|
|
|
Stability. Interlock distance is a strong function of application and of the availability of optimizing software. |
|
|
|
|
|
|
|
|
Data Note 5. The data presented in Tables 3.22 and 3.21 are also from Rossmann [248, 247]. |
|
|
|
|
|
|
|
|
Reliability. The reliability of the use of variable and operand instruction is a strong function of the application and the compiler. |
|
|
|
|
|
|
|
|
Stability. Significant differences are expected as a function of a particular application. |
|
|
|
|
|
|
|
|
3.8 Annotated Bibliography |
|
|
|
|
|
|
|
|
Among the earliest papers published on program and machine behavior is the work of Knuth [167] and Gibson [103]. System/360 has been thoroughly studied by Connors, among others [55]. Clark [52] has a particularly useful analysis of the VAX architecture. Lunde [187] and Tanenbaum [277] are particularly noteworthy earlier references on high-level machine behavior. The design target data presented in this chapter is largely the work of Huck [136]. This data has been validated against other data based on Pascal and C programs using simulation tools developed at Stanford. (See Mitchell and Flynn [201].) |
|
|
|
|
|
|
|
|
Machine and Program Behavior |
|
|
|
 |
|
|
|
|
D. Knuth. An Empirical Study of Fortran Programs. Technical Report STANCS-70186, Computer Science Department, Stanford University, Stanford, CA, 1970. |
|
|
|
|
|
|
|
|
J. C. Gibson. The Gibson Mix. Technical Report TR 00.2043, IBM Systems Development Division, Poughkeepsie, NY, 1970. |
|
|
|
|
|
|
|
|
D. W. Clark. Measurement and analysis of instruction use in the VAX 11/780. Proceedings of the 9th Annual Symposium on Computer Architecture, pages 917, IEEE Computer Society Press, Washington, DC, April 1982. |
|
|
|
|
|
|
|
|
A. Lunde. Empirical valuation of some features of instruction set processor architectures. Communications of the ACM, 20(3):141200, 1977. |
|
|
|
|
|